site effects associated with the 2010 maule

DOI 10.4461/GFDQ.2015.38.01

Geogr. Fis. Dinam. Quat. 38 (2015), 3-13, 14 figg., 1 tab.

ARTURO BELMONTE (*), EDILIA JAQUE (**), JORGE QUEZADA (***), ALFONSO FERNÁNDEZ (**), CECILIA DONOSO (****) & CARLOS CARTEAU (*****)

SITE EFFECTS ASSOCIATED WITH THE 2010 MAULE EARTHQUAKE IN ZONES CHARACTERIZED BY THE PRESENCE OF WETLANDS IN THE BIOBIO REGION, CHILE

ABSTRACT: BELMONTE A, JAQUE E., QUEZADA J., FERNANDÉZ A., DONOSO C. & CARTEAU C. Site effects associated with the 2010 maule earthquake in zones characterized by the presence of wetlands in the Biobio Region,Chile. (IT ISSN 0391-9838, 2015). After 2010 Maule (Chilean) earthquake (Mw=8.8, February 27th) some residential areas located along the coastal border in southern Chile and close to water bodies and wetlands showed extensive damage in housing, streets, electric, water, and gas lines. That was the case of the neighborhoods Villa Las Araucarias (Arauco) and Bayona (San Pedro), situated 100170 km south of the epicenter zone, and established on top of a porous artificial filling and compacted soils. Conversely, in other sites soils and infrastructure remained intact. Here, we present a study that aimed to describe and assess the effects of the Maule earthquake on urban sites. We tried to disentangle the differences between damaged and undamaged zones by identifying modes of damage as well as by detecting geophysical anomalies associated with such a divergent behavior. For this reason, we also analyzed three undamaged zones: Colcura (Lota), Laguna Grande (San Pedro) and Los Canelos (San Pedro). Results suggest that the damaged locations behaved as examples of the liquefaction phenomenon, triggered by a large earthquake. The implications of this finding are discussed according to existing Chilean regulation in terms of the so-called VS30 parameter as a quality factor for soils. We also discuss different aspects related to the relationship between geophysical methodologies applied here, visual observations and geological/soil interpretation. ———————— (*) Departamento de Geofísica, Facultad de Ciencias Físicas y Matemáticas, Universidad de Concepción, Casilla 160-C, Concepción, Chile, [email protected] (**) Departamento de Geografía, Facultad de Arquitectura, Urbanismo y Geografía, Universidad de Concepción, Casilla 160-C, Concepción, Chile, [email protected], [email protected] (***) Departamento de Ciencias de la Tierra, Facultad de Ciencias Químicas, Universidad de Concepción, Chile, [email protected] (****) Servicio Nacional de Geología y Minería (Sernageomin), Av. Santa María 0104, Providencia, Santiago, Chile, [email protected] (*****) Consorcio Eólico, Av. Collao 1485, Concepción, Chile, carlos. [email protected] We want to thank a lot to Villa Las Araucarias’ neighbors who welcomed us from the first moment and helped us to carry out the present work. Furthermore we wish to thank to Vicerrectoría de de Investigación de la Universidad de Concepción by financing geophysical measurements in frame of DIUC Project 211.012.011-1.0.

KEY WORDS: Soils, Wetland, Maule Earthquake, Electrical, Seismic and ReMi Methods, Chile. RESUMEN: BELMONTE A, JAQUE E., QUEZADA J., FERNÁNDEZ A., DONOSO C. & CARTEAU C. Efectos de Sitio asociados con el terremoto del Maule 2010 en zonas caracterizadas por la presencia de humedales en la región del BíoBío, Chile. (IT ISSN 0391-9838, 2015). Durante el terremoto del Maule (27/02/2010, Mw=8.8), el área centro sur de la Región del Biobío, Chile y en particular algunas zonas residenciales ubicadas a lo largo del borde costero y cerca de cuerpos de agua y humedales, mostraron grandes daños en viviendas, calles, instalaciones de electricidad, agua y líneas de gas. Ese fue el caso de los barrios Villa Las Araucarias (Arauco) y Bayona (San Pedro), situados a 100 y 170 km al sur de la zona del epicentro, establecidas sobre un relleno artificial poroso y suelos compactados. Por el contrario, en otros tipos de suelos la infraestructura permanecieron intactas. A continuación, presentamos un estudio que tuvo como objetivo describir y evaluar los efectos del terremoto de Maule en sitios urbanos; tratamos de explicar las diferencias entre las zonas dañadas y no dañadas mediante la identificación de modos de daños, así como mediante la detección de anomalías geofísicas asociadas con tal comportamiento divergente. Por esta razón, también se analizó tres zonas no dañadas: Colcura (Lota), Laguna Grande (San Pedro) y Los Canelos (San Pedro). Los resultados sugieren que los lugares dañados se comportaron como ejemplos del fenómeno de licuefacción, provocada por un terremoto de gran magnitud. Las implicaciones de este hallazgo se analizan de acuerdo a la regulación chilena existente en términos del llamado parámetro VS30 como un factor de calidad de los suelos. También se discuten diferentes aspectos relacionados con la relación entre metodologías geofísicas aplicadas aquí, observaciones visuales y la interpretación geológica y de suelos. PALABRAS CLAVE: Los suelos, Humedales, Maule Terremoto, Eléctricos, Sísmica y Métodos ReMi, Chile.

INTRODUCTION Local geology can explain damage in small areas hit by earthquakes. Observations of site effects after a main seismic shock can provide insights into why some areas are strongly affected and how they could be qualified under geophysical measurements. The February 27th, 2010 Chilean earthquake struck an extensive area in the central-south region of Chile (fig. 1) 3

FIG. 1 - Location map of the 2010 Maule Earthquake (Mw=8.8) rupture area, including the epicenter and the focal mechanism (indicated by the socalled white and black “beach ball”). Black dots represent the actual rupture extent, the epicenter and the focal mechanism Cities (green squares), volcanoes (purple triangles), geomorphological units (Coastal and Andean Cordillera) as well as direction of Nazca Plate with respect to South American Plate (black arrows) are also included. Concepción and Arauco are located in the southern section of the earthquake rupture plane. The earthquake mechanism is associated with a thrust event.

FIG. 2 - Geological map of BioBio Region coastal border. Villa Las Araucarias (Arauco) as well as Bayona (San Pedro) and Los Canelos (San Pedro) are shown, all located in areas featured by non-consolidated sediments. Colcura (Lota), where metamorphic basement crops out, is also included. The BioBio River can be seen down to its river mouth. The Laja River has its river mouth about 60 km south-east of the BioBio river mouth. Intrusive and metamorphic outcrops as well as volcanic rocks are related to the so-called Cordillera de la Costa which south of Concepción is known as Cordillera de Nahuelbuta.

where a number of events (M>8) have left their mark since the Spaniards arrived in Chile: 1575, 1657, 1751, and 1835. This region was considered a well-known seismic gap zone (Barrientos, 1994; Quezada & alii, 2010). Although the destructive capacity of a main shock is often proportional to its magnitude and distance from the rupture zone (Kramer, 1996; Verdugo & alii, 2010; Bertalot, 2011), the largest level of destruction observed after the 2010 main shock was distributed along sites characterized by 10 to 15 year-old artificial fillings situated close to and/ or on the top of wetlands (in Spanish humedal) whose origin dates from the Holocene – late Pleistocene time-span and whose peculiarity consists of high levels of water saturation. A high water-saturated soil is more susceptible to liquefaction in cases of Holocene sandy deposits (Youd & Hoose, 1977) and sediments of similar grain-size, packed in layers thicker than 1 m. Furthermore, some man-made artificial fillings subjected to compacting technics have also proven to be highly susceptible to the liquefaction phenomenon. Quake shaking produces an increase of water pressure on pores, associated to soils with loamy-sandy content, reducing effective stress and shear resistance of sand (Obando, 2009). Therefore, material behaves as liquid giving rise to vertical and horizontal motion, which turns into displacements and/or large settlements. This process is known as induced liquefaction and it is one of the main causes of high urban seismic risk. A relevant and necessary condition for

a site to show such behavior is the presence both of high water content in the subsurface (extending more than 3-5 m deep) and low-grade compacting. According to Verdugot& alii. (2010), low water table levels, a usual condition during the dry season, prevented more intense and extensive damage during and after the Maule Earthquake. Our main objective is to describe and assess the effects of the Maule earthquake on urban sites. In our work, we link both geology and soil response to the main shock observed effects with geophysical parameters such as conductivity/resistivity and seismic waves propagation velocity.

4

GEOLOGIC AND GEOMORPHIC FRAMEWORK The study area is located south of the Bio Bio River. Here, a northward, open embayment known as Arauco Gulf characterizes the Chilean coastal border. Block faulting produced a geomorphological structure characterized by a sandy plain at the top of Mesozoic-Cenozoic unit downthrown in a graben-like structure. Such a structure is bordered to the east by the fault scarp of the Coastal Cordillera and to the west by a similar type of steep slopes known as Arauco Peninsula (Galli, 1967). The littoral zone of Arauco Gulf is a Coastal Plain, a flat area composed by Holocene sands (Kaizuka & alii, 1973; Islat& alii, 2012). Its extension varies between a hundred

of meters to four kilometers with maximum heights of a few meters. The eastern border of the Coastal Plain (the Coastal Cordillera) is composed in their westernmost part by Pleistocene marine terraces (Kaizuka & alii, 1973). These terraces are also exposed along the border of the Coastal Plain located in the southernmost shoreline where the city of Arauco is located. A fossil cliff of 40-70 m height separates the younger Pleistocene terrace of Coastal Cordillera from the Holocene Coastal Plain. Metamorphic and sedimentary rocks constitute the basement of the Coastal, emerging as outcrops along the Coastal Cordillera (fig. 2). Metamorphic rocks are phylites formed in the Carboniferous age. A high degree of weathering affects such rocks generating red clay. Normal faulting with NE-SW strike affects rocks located on the Arauco Peninsula and the Coastal Cordillera, generating tilted blocks. This morphostructural setting favoured the creation of a coastal-alluvial plain constituted by thick sand deposits over which present-day wetland systems developed. Holocene sands belonging to the Coastal Plain have different composition. From Biobio River until Coronel (including San Pedro) they are black. Galli (1967) named these unconsolidated deposits as Huachipato Formation, characterized by fine to medium sands with clasts of basaltic composition. These sands came from the volcanic activity located at the high Río Laja Valley during the Late Pleistocene and Holocene. They were first carried down to the palaeo-coastline through the Laja and Bio Bio Rivers (Ilabaca, 1989; Mardones & Jaque, 1998; Mardones, 2005) and then reworked and deposited southwards in the direction of the Arauco Gulf. South Coronel city including the Arauco area, the sands are fine with light colors and came from small rivers from Coastal Cordillera and reworked by the sea forming succesive beaches. The Biobio region in southern Chile presents about 17 large size wetlands and another 30 of smaller sizes along the coastal border (Gonzales & Victoriano, 2005). Between the cities of Talcahuano and Arauco wetland areas have been covered by urban expansion that has both occupied their surfaces and transformed the surficial sediments in those sites. The most important wetlands in the study are Los Batros and Arauco. Los Batros wetland is located at San Pedro be-

tween the Coastal Cordillera and Biobio River. Its extension reaches up to 2 km from Laguna Grande Lake to Biobio River (fig. 2). Here several branched natural fluvial streams have been filled with sand for building residential areas. The Arauco wetland area is located along the Laraquete-Arauco Plain, associated to the Carampangue River, at the eastern border of Arauco city. The Arauco wetland area reaches up to 4 km2 (figs. 2 and 3). Las Araucarias place are located above this wetland over an unconsolidated, man-made filling deposit. STUDY SITES During inter-seismic cycles (i.e. normal times for the population) any new residential area tends to cover previous landscapes and landforms, making more difficult to estimate the potential risk associated with the way that soil will behave under strong seismic shaking. That is the case for Villa Las Araucarias in Arauco and Bayona in San Pedro, among others. In this study, we focused in these two areas as example of damaged locations. Villa Las Araucarias was built in 1995 after a new urban zoning plan had been created in 1988, allowing occupation of a section of an extensive wetland area (Carampangue) (figs. 2 and 3). A similar zoning plan permitted the urban expansion and construction of residential areas in Bayona. We also studied locations apparently unaffected by the 2010 earthquake, because they are situated over more consolidated sites. More specifically, we analyzed Los Canelos and Laguna Grande in San Pedro as well as Colcura in Lota (midway between Arauco and San Pedro). Los Canelos is situated above an apparently well-compacted filling soil, out of the perimeter of the San Pedro wetland. Laguna Grande and Colcura lie on a weathered metamorphic basement. METHODOLOGY Shortly after the February 27th Earthquake, on March 17th, we carried out field visits to Bayona and Villa Las Araucarias. From these two locations, we performed de-

Fig. 3 - Specific location of Bayona, Los Canelos and Laguna Grande in San Pedro (left) and Villa Las Araucarias in Arauco (right) are shown. In both pictures it is possible to see wetland areas related to suburban zones (modified Google Map Images).

5

tailed analysis in Villa Las Araucarias, where we dug two soil pits, allowing us to obtain information on the type, quality, and compactness of soil. In Bayona, we could record field observations and photographs only. We also determined electrical and seismic profiles: Vertical Electric Sounding (VES/Schlumberger), Dipole-Dipole, Refraction Seismic, and ReMi Method (see Appendix 1). For resistivity and velocity measurements, in addition to the Villa Las Araucarias site, we sampled three other sites to characterize areas with no earthquake damage. (Donoso, 2011, Carteau, 2013): Colcura (Lota), Laguna Grande (San Pedro) and Los Canelos (San Pedro). We aim to establish a reference for geophysical parameters in wetland areas as well as to quantify the expected contrast. RESULTS FIELD OBSERVATIONS Places under observation reach dimensions of a few hundreds of meters per side. Sites without damage must be understood as places where houses, play-ground squares, pavement streets and services resulted with minor to no damage after main shock. Field observations concentrated

in Villa Las Araucarias and Bayona. They enabled us to recognize deformations and damages that we could associate as closely related to liquefaction phenomena. Theoretically, liquefaction phenomena triggered by earthquakes can affect buildings, housing, bridges, streets and pipes in the following different ways: settlement, lateral spread, cracking along streets, sand boils and stability failures. Field observations show clearly evidence for such effects after the main shock, with exception of sand boils which can be expected in more saturated soils than those present in the summer season. We distinguish: (a) irregular failure concentrated along streets and houses with extensions of 20 to 40 cm, and consequences such as asphalt blocks removed; (b) lateral spread towards existing filling’s free faces which generated extensive cracking up to 20 cm thick inside the houses, located around two meters from the filling´s border; (c) systematic cracking along streets; (d) inside the houses, cement floor (0.5-1 meter thick) settled up to 10 cm with respect to parking and/or gardens. Figure 4 shows pictures where these features are evident. FIELD TEST AND FILLING CHARACTERISTICS Field tests were carried out on two soil pits caved in Villa Las Araucarias, for both anthropic filling and wetland mate-

FIG. 4 - In Villa Las Araucarias, (A) failure and removed blocks, (B, C) lateral spread outside and inside houses. In Bayona, (D) an uplifted sewer cylinder terrain due to soil settlement, (E) irregular settlement, (F) floor uprising within a house, (G) extensive lateral spread along a recreational sector, (F) abrupt longitudinal uprising.

6

FIG. 5 - On the left a description of the profile observed in one of the soil pits in Villa Las Araucarias neighborhood is shown. On the right the corresponding soil pit is also shown.

rial showing differences in resistance when pressure was exerted on the material. While pressure on a handful of filling disintegrated the sample into particles of different grain size, the same exercise on wetland muddy material deformed malleably without producing any fracture or relevant humidity release. This simple field test indicates that both materials behaved in different ways during the main shock, thus being a possible cause of fissures generated up to the surface. Soil pit itself revealed a filling characterized almost entirely of clay, some sand and wood pieces. Observed filling extends down to a depth of 160 cm, below which mud associated to wetland material extends for another 120 cm down to reach water apparently linked to wetland (fig. 5).

free face or downhill. Such deformations can vary from a few centimeters up to 2 m. These are consequence of inappropriate artificial filling. In fact, man-made comprehensive filling should be the best soil stabilizer. For residential areas the appropriate grain mixture should be formed by 0 to 40 mm diameter grain size, that is, sand, clay (in 40-50% proportion) and gravel. Pieces of wood in artificial fillings are forbidden, because their porous decomposition negatively influences the degree of compacting. APPLICATION OF GEOPHYSICAL METHODS GEOELECTRIC SURVEY

TYPE OF DEFORMATION A sketch trying to recreate observations on deformation is shown in Figure 6. The shallower layer can be fractured into blocks that are separated through fissures which open and close during quake shaking. Settlement of structures and housing is associated to the redistribution of porous fluid during and after quake shaking, generating material that compacts itself. This effect produces both concentration of deformation at structures based on superficial foundations and damage in piping systems. Consequently, large cracking is observed after the main shock has occurred. Lateral spread effects produce the breaking up of shallow layers into blocks which move progressively towards a so-called

FIG. 6 - In blue color a sketch giving a possible form to deformations observed in surface at Villa Las Araucarias neighborhood. This sketch is deduced from field and soil pits observations. Red line shows the original state (before the main shock).

A total of three VES profiles and one Dipole-Dipole profile were carried out, both in Villa Las Araucarias, Colcura and Laguna Grande (San Pedro). Our aim is to establish evidence of some expected contrast in resistivity according to their geologic/soil properties compared to the main area of study in Villa Las Araucarias (fig. 7). Three clear changes of slope can be detected from the plot log10 ρ [Ohm-m] v/s log10 depth [m] (fig. 8), which can be associated with two layers and infinite half-space. In Villa Las Araucarias the first meter of depth is linked to resistivity values of about 60 [Ohm-m], then a decrease down to 10 [Ohm-m] is observed for the next 3 m of depth associated possibly with the presence of water and finally a strong increased resistivity ranging up to orders of 103 [Ohm-m] is clearly marked. In spite of finding similar shaped curves for other two VES profiles in Colcura and Laguna Grande, the lowest part extends no lower than 150-200 [Ohm-m]. The first 5 m of the Dipole-Dipole profile in Villa Las Araucarias reveals resistivity values ranging from 1 – 200 [Ohm-m]. Minimum values coincide with the place where the soil pit was dug. At depths of 5 to 6 meters, resistivity increases apparently due to the presence of sandstone or sand associated with Tubul Formation. SEISMIC REFRACTION AND REFRACTION MICROTREMOR (REMI) In Los Canelos one ReMi and Seismic Refraction line consisting of 23 geophones (14 Hz) every 4 m was displayed 7

FIG. 7 - Sites prospected with electrical profiles: VES (Vertical Electrical Sounding or Schlumberger) and Dipole-Dipole (modified Google Images). Red thick lines indicate geo-electrical profiles arrangement which reached lengths up to 60-80 m.

FIG. 8 - Resistivity results and stratigraphic interpretation are shown here. From top to down: (I) Schlumberger profiles (VES) results plotted as log10 ρ [Ohm-m] v/s log10 Depth [m] for 3 field sites: Villa Las Araucarias, Colcura and Laguna Grande, (II) 2-D Dipole-Dipole resistivity section for Villa Las Araucarias profile and (III) stratigraphic interpretation for all three sites.

8

FIG. 9 - On the left location for seismic refraction profile in Los Canelos (San Pedro); one the right ReMi and seismic refraction profiles displayed in Villa Las Araucarias (Arauco) are shown (modified Google Images).

FIG. 10 - Los Canelos. On the left seismic refraction travel times are shown together distinct interpretation curves represented by different colors. On the right hand 1-D velocity models is shown (VP[m/s] v/s h[m] plot). In red color the average value which deviation is marked by blue color lines.

FIG. 11 - Los Canelos. On the left hand spectral ratio and chosen dispersion curve is shown. On the right hand inversion results for six best iterations (S-wave velocity [m/s] v/s depth [m]) are exposed as 1-D S-wave velocity models. Vs30 is the S-wave velocity average for first 30 meters (df=0.03125 Hz, dp=0.0001 s/m and np=101).

along 88 m length while that in Villa Las Araucarias two ReMi and Seismic Refraction lines consisting of 17 geophones (4.5 and 14 Hz) every 4 m were carried out along 64 m length (fig. 9). In Los Canelos, seismic refraction travel times were adjusted to a model with two layers above an infinite halfspace. Maximum, media and minimum resulting model velocities are shown in Figure 10. For ReMi profiles main noise source can be assigned to vehicular traffic. Combining 3 recordings it was possible to obtain a clear descending signal. Several inversions were carried out for a frequency range 3 to 26 Hz. Three first layers were constrained to P-wave velocity obtained from seismic refraction experiment. Density was taken as 2000 kg/

m3 and a Poisson coefficient between 0 and 0.5. In Figure 11 spectral ratio (see Appendix 1) and 1-D S-wave velocity model are shown. A thin first layer is associated to vegetation cover (Vp ~ 240 m/s; Vs ~ 180 m/s), only detectable with seismic refraction experiment. The next two lower layers are associated to typical values for both dry and wet sands (Vp ~ 430 m/s; Vs ~ 220 – 340 m/s). These sediments can be classified as fine to thick silty clay, which would also be coherent with the geologically known BioBio sands in San Pedro. Lowering of S-wave velocity around 10 m depth (Vs ~ 150 m/s) could be associated with a prior vegetation cover inserted into the BioBio sands sequence. VS30 parameter is estimated as VS30 ~ 211 m/s, which according to modified NCh433 of96 (Chil9

FIG. 12 - Villa Las Araucarias. On the left and right respectively plots related with VILLA 1 and VILLA 2 are exposed. On top seismic refraction travel times and fit curves. On the middle 1-layer velocity model using delay time method is sketched. On the lower part resulting 1-D velocity models are shown. Here red color lines indicates an average value; deviation is shown by blue color line.

ean regulation for Seismic Building Design, 1996) would be related with a soil type D, that is, moderately dense and solid (see also Appendix 1). In Villa Las Araucarias, two profiles could be arranged: VILLA 1 and VILLA 2. In both cases two shots in the borders and one shot in the mid-section allowed to using the so-called delay time method for seismic refraction profiles. Figure 12 presents travel time data as well as one layer with a half-space model. For ReMi profiles main sources of noise area associated to vehicular traffic and near construction as well as forest activity. For VILLA 1 and VILLA 2 a descending trend for dispersion selected curve can be observed in range of 5 to 16 Hz. Firsts layers were constrained to P-wave velocity obtained from seismic refraction experiment. In the same way as for Los Canelos profiles, density was taken as 2000 kg/m3 and a Poisson coefficient variable between 0 and 0.5. Integrating VILLA 1 and VILLA 2 (fig. 13), it is possible to identify a first layer reaching down to 4-5 meters with velocities associated with vegetable covering and clay (Vp ~ 300-500 m/s; Vs ~ 120-140 m/s). As we know from soil pits observations, this layer contains 2 sub-layers: artificial filling and a (wet) mud containing clay and sand. Apparently, a second layer with Vs ~ 210-280 m/s is thicker in VILLA 2 than VILLA 1, which could be related with wetland distribution below the artificial filling. A third layer below around 20 meters depth can be assigned with Vs ~ 10

400 m/s associated possibly with sands. A fourth layer can only be seen in VILLA 1 with Vs ~ 600 m/s and could be related with a more rigid material as the Tubul Formation sandstone. In both profiles VS30 ~ 230-280 m/s which is associated with soils classification D. DISCUSSION Field observations and geophysical results appear to lead to some global expected relationships such as wetland areas tend to present both low resistivity and low S-wave velocity values, that is, artificial fillings above wetlands present low soil quality factor in light of geophysical parameters. However, some particular aspects look contradictory as well as new. The first aspect is related with seismic measurements. They show that differences between sites like Villa Las Araucarias and Los Canelos do not diverge as much as expected in light of damage observed on each place. According to VS30 estimations, both soils fall in same category, namely moderately dense and solid type D. Then we estimated average velocity in the whole range from 0 to 30 m depth (fig. 14). Although at 30 m depth average velocity is smaller in Los Canelos than in Villa Las Araucarias, the opposite situation occurs in the superficial layer. That is, extreme low average S-wave velocities characterize shallow range of depths (z < 10 meters). In comparison to places without damage,

FIG. 13 - Villa Las Aracuarias. On the left hand spectral Ratio, chosen dispersion curves and 1-D S-wave velocity model for VILLA 1 are shown. Respectively on the right same plots and graphs for VILLA 2 are displayed. Vs30 is the S-wave velocity average for first 30 meters. (df=0.0333 Hz, dp=0.0001 s/m and np=101).

like Los Canelos, such low S-wave velocity would clearly be in correlation with consequences observed in Villa Las Araucarias after 2010 Earthquake. This implies that the VS30 parameter may not be an accurate soil-quality index. Although the modified Chilean law (NCh433 of96) should describe places like Villa Las Araucarias as a “collapsible soil” assigning it into a category of “special type of soil”, urbanism and building expansion tends to cover original landscapes with new human settlements making not obvious the identification of such classification soil type.

A second aspect is related to the very thin low S-wave velocity “layer” observed in Los Canelos. Although it is not possible to generalize this observation, the fact that Los Canelos itself does not lie over a much more compacted soil according to S-wave velocity arises the question whether it would be possible that this low velocity anomaly attenuates incident waves in comparison to the permanent increasing velocity distribution with depth as in Villa Las Araucarias. It is known and demonstrable- for Love waves- that interfaces separating 2 layers with velocity increasing with depth are expected to show amplification in amplitude for incident waves coming from the bottom. This aspect requires further research. Regarding to resistivity values obtained a goo, and expecte, correlation appears between low resistivity (10-30 TABLE. 1

FIG. 14 - For each profile (Los Canelos and Villa Las Araucarias), S-wave average velocities for distinct range of depth are shown. At 30 meters depth we reach the corresponding Vs30 value.

TYPE

DESCRIPTION

VS30[m/s]

A

Rock, compacted soil.

≥ 900

B

Soft or fractured rock, very dense and solid soil.

≥ 500

C

Dense and solid soil.

≥ 350

D

Moderately dense and solid soil.

≥ 180

E

Weak of soft soil.

< 180

F

Special soil.

-

11

[Ohm-m]) and areas damaged. The first meter of depth in Villa Las Araucarias appears related with ρ ~ 60 (Ohm-m) and then a second layer down to 4 m with resistivity values around 10-30 (Ohm-m). These first depths appear to be related with extreme low S-wave velocities (Vs ~120-140 m/s) suggesting (VILLA 1 profile) a material particularly not very dense with presence of water. The increase of S-wave velocity below 4 m depth up to 400 m/s shows a correlation with resistivity value of order 3. In comparison with sites where a more compacted soil gives form to subsurface (mainly constituted by metamorphic basement complex), orders are 1 to 2 greater compared to muddy-sandy soils. Such higher orders of magnitude for resistivity characterize well almost the whole electrical profiles in sites like Colcura and Laguna Grande. Although no electrical profile was performed in Los Canelos and no seismic profiles were performed in Colcura and Laguna Grande, it is possible to link them since they did not show damage after Maule Earthquake. CONCLUSIONS Resistivity values as low as 10 (Ohm-m) - in first 5 to 4 m depth - clearly indicate a close relation with both (1) low S-wave velocities and (2) severe damage caused by strong motion. So, the logs from Villa Las Araucarias indicate its link to a layer with high amount of water and in consequence with a weaker geotechnical soil material. Here the first meter depth is formed by a clay layer and then 4 m of clay with water. Below this depth sandstones and sand appear to be present. Adding visual observations, hand-man test and linking it with role of wetland areas in both Villa Las AraucariasArauco and Bayona-San Pedro, it is possible to affirm that both locations behaved as examples of the liquefaction phenomenon triggered by a large earthquake. Verdugo (2010) had already recognized the liquefaction phenomenon in and around Concepción, indicating that replacing soft and weak soil with compacted sandy filling did not conduct to prevent liquefaction. Non-cohesive soil foundations saturated with water added to seismic strong motion lead to land faulting and cracking. Apparently in the first 3 to 5 m depth two mechanisms appear to have main control over observed damage: (1) cracking and settlement appear to be produced due to decoupling of the upper from the lower layer. A 2 to 3 m thick layer constitutes a highly susceptible zone to liquefy and therefore it provides conditions to cause systematic fissures in the upper crust causing the asphalt and cement floor to become disaggregated into blocks; (2) lateral spread of fillings border where a free face (without any resistance to horizontal displacement) induces horizontal opening and cracking. A reason for such effects is often related to a decoupling process between a deeper layer susceptible to liquefaction and a compact superficial layer, in this case, the artificial filling. Geophysical evidence shows a very low S-wave velocity as well as a decoupling in resistivity values in first 5 m depth. Conversely, a more homogeneous S-wave velocity value in shallow depths, even with low value as observed in 12

Los Canelos, could result in a more efficient system to dissipate strong motion energy resulting in no observed damage.

REFERENCES BARRIENTOS S. (1994) - Large Events, Seismic Gaps and Stress Diffusion in Central Chile, Tectonics of the Southern Central Andes. Springer Verlag, Berlin, Heidelberg. BEEKMAN A.N. (2008) - A comparison of experimental ReMi measurements with various source, array and site conditions. Master’s thesis, University of Arkansas, USA. BERTALOT D. (2011) - An overview on field and experimental evidences concerning seismic liquefaction induced settlement of buildings with shallow foundations. Obras y Proyectos, 10, 36-45. CARTEAU C. (2013) - Caracterización de suelo en Villa Las Araucarias (Arauco) mediante refracción de microtremores (ReMi). Tesis para Optar el grado de Geofísico, Departamento de Geofísica, Universidad de Concepción, 128 pp. DONOSO C. (2011) - Caracterización de Suelos en las localidades de Arauco, Colcura, San Pedro de la Paz y Concepción mediante Prospección Eléctrica. Tesis para Optar al grado de Geofísico, Departamento de Geofísica, Universidad de Concepción, 153 pp. GALLI C. (1967) - Geología urbana y suelo de fundación de Concepción y Talcahuano, Chile. Departamento de Geología y Mineralogía, Instituto Central de Química. Universidad de Concepción. 248 pp. GONZALES A. & VICTORIANO P. (2005) - Aves de los humedales costeros de la zona de Concepción y alrededores. In: SMITH-RAMÍREZ C., ARMESTO J.J. & VALDOVINOS C. (eds.) “Historia, biodiversidad y ecología de los bosques costeros de Chile”. Santiago Editorial Universitaria, p. 485-497. ILABACA P. (1989) - Evolución de la Costa de Concepción: El caso de las bahías de Concepción y San Vicente. Revista de Biología Pesquera, 18, 29-35. ISLA I., QUEZADA J., Martínez C., FERNÁNDEZ A. & JAQUE E. (2012) - Maximum Holocene highstand for the coastal plains of the Arauco Gulf, Biobío Region: A sea-level curve for Chile. Journal of Costal Research, 28, 1, 102-111. KAIZUKA S., MATSUDA T., NOGAMI M. & YONEKURA N. (1973) - Quaternary tectonic and recent seismic cristal movements in the Arauco península and its environs, central Chile, Geographical Reports Tokio, Metropolitan Univiversity, 8, 1-49. KRAMER S.L. (1996) - Geotechnical Earthquake Engineering. University of Washington, Prentice Hall, 653 pp. LIRA E. (2009) - Problema de n1 capas y un semiespacio: Filtro de Gosh. Facultad de Ciencias Físicas y Matemáticas, U de Chile, 1-7. LOUIE J.N. (2001) - Faster, better: Shear-wave velocity to 100 meters depth from refraction microtremors arrays. Bullettin Seismological Society of America, 91, 347-364. MARDONES M. (2005) - La Cordillera de la Costa: Caracterización FísicoAmbiental y Regiones Morfoestructurales. In: SMITH-RAMÍREZ C., ARMESTO J.J. & VALDOVINOS C. (eds.) “Historia, biodiversidad y ecología de los bosques costeros de Chile”. Santiago Editorial Universitaria, 39-59. MARDONES M. & JAQUE E. (1998) - Geomorfología del Valle del Río Laja. Primer Congreso de Ciencias de la Tierra, 115-121. NCh433 of. 96 (1996) - Norma Chilena Oficial “Diseño Sísmico de Edificios”. División de Normas del Instituto de Nacional de Normalización (INN). Decreto No. 172 del 05 de Diciembre de 1996 del Ministerio de Vivienda y Urbanismo. OBANDO T. (2009) - Modelación geomecánica y temporal de la licuefacción en suelos de minas no metálicas. Estudio Caso: Ciudad de Managua

(Nicaragua). Tesis Doctoral, Ed. Universidad Internacional de Andalucía UNIA (Huelva, España), Huelva, 900 pp. QUEZADA J., JAQUE E., BELMONTE A., FERNÁNDEZ A., VÁSQUEZ D. & MARTÍNEZ C. (2010) - Movimientos co-sísmicos verticales y cambios geomorfológicos generados durante el terremoto Mw=8.8 del 27 de Febrero de 2010 en el centro-sur de Chile. Revista Geográfica del Sur, 2, 11-44. VERDUGO R., VILLALOBOS F., YASUDA S., KONAGAI K., SUGANO T., OKAM., TOBITA T. & TORRES A. (2010) - Description and Analysis of Geotechnical Aspects Associated to the Large 2010 Chile Earthquake. Obras y Proyectos, 8, 25-36. MURA

WATHELET M. (2005) - Array recordings of ambient vibrations: Surfacewave inversión. PhD thesis, Université de Liège, Bélgium. WATHELET M. (2008) - An improved neighborhood algorithm: Parameter conditions and dynamic scaling. Geophysical Research Letters, 35, L09301. YOUD T. L. & HOOSE S.N. (1977) - Liquefaction susceptibility and geologic setting. In Proceedings, 6th World Conference on Earthquake Engineering, New Delhi, India, 2189-2194. YOUD T.L. (1984b) - Geologic Effects – Liquefaction and Associated Ground Failure. In Proceedings of the Geologic and Hydrologic Hazards Training Program, Open – File Report 84-760, U.S. Geological Survey, Menlo Park, California.

APPENDIX 1 GEOELECTRICAL APPLICATION Geoelectrical prospecting allows to exploring resistivity from sub-surface in a broad range. Resistivity values [Ohmmeter] can vary up to 15 orders of magnitude. Conduction is defined as electrolytic type in sense that water presence into rocks porosity allows electrolytic conduction. In present work resistivity values ranged around 100 and 103 [Ohm-m]. Vertical Electrical Sounding [VES] - also known as Schlumberger Sounding- and Dipole-Dipole profile were conducted on field. Measurements were carried out using a transmitter with power up to 400 Volts to generate currents of about 200 mA [milli-Amperes]. Profiles distances AB extended up to 60 meters for dipole-dipole and 100 meters for VES. Spacing between electrodes was adjusted to 2 meters. Distance between voltage difference electrodes in VES was adjusted to 0.5 meters. VES data set was processed using IPI2DINV and IPI2WIN together a matlab routine (Lira, 2009). Dipole-dipole data set were processed using RES2DINV. Inversion models were adjusted with an error up to 20%. Details can be found in Donoso (2011). ReMi-SEISMIC REFRACTION APPLICATION Measurements were carried out with a GEODE (Geometrics) seismograph by using 17 and 23 vertical geophones which natural frequencies are respectively 4.5 and 14 Hz for ReMi and seismic refraction-. Seismic refraction source was simulated by hitting a metal base with an 8-kilogramms hammer. Sampling gap was fixed to 0.25 millisec. (ms) and registering time to 0.25 – 0.5 sec. For ReMi method sampling gap was fixed to 2 ms for registering times of 30-32 sec. Measurements were repeated twice. Profiles lengths reached up between 64 and 80 meters with spacing among geophones of 4 m.

Data processing was done by using a pair of routines developed in frame of Carteau thesis work (2012) together the use of GEOPSY software. These routines were tested with matlab routines provided generously by A. Beekman (Beekman, 2008). Main difference between Beekman’s routines and the one developed by Carteau (2012) is the way for calculating slant-stack transform. Inversion process from dispersion curve to S-wave vertical velocity accepts different possible solutions, so it is relevant to count with previous geological information as initial model. In this work such process is carried out using a free-distributed software (Dinver) created by Marc Wathelet (Wathelet, 2005; Wathelet, 2008). Here the so-called Neighbourhood algorithm is implemented. Details related with theory and data processing for ReMi method are exposed in Carteau (2012). The spectral analysis required for ReMi analysis is based mainly in Louie (2011) which establish as control parameter the spectral ratio R(p,f).

np is the number of steps for p discretization. is the potency spectra. is the ratio between the potency spectra of one or several registers for a particular (selected) value of slowness and frequency and the average of potency spectra along slowness in such selected frequency. VS30 PARAMETER Before 2010 Maule Earthquake Seismic Chilean Norm NCh433 Of96 used to classify soils into 4 categories from high to low quality. Here geotechnical and geophysical parameters are combined to establish such classification. One of these parameters has been often the P-wave velocity (Vp). On November 2th, 2011 a relevant change was established by Chilean government. It became obligation to know apart of Vp – the S-wave propagation velocity average for the first 30 meters depth. This parameter is known as VS30. VS30 is calculated according to equation (A1) where Zi is the ith layer potency.

(A1) This new classification establishes the following categorizing (tab. 1) for the so-called VS30 parameter. Theory and observations indicate that soils behavior against strong motion is clearly better in rock outcrop than finer and softer soils. F-type soils are those showing singularities in their mechanical behavior and it is requirement a geotechnical study: sands, mud, organic soils among others. (ms received 30 September 2014; accepted 30 April 2015) 13

DOI 10.4461/GFDQ.2015.38.02

Geogr. Fis. Dinam. Quat. 38 (2015), 15-24, 6 figg., 1 tab.

Davide DAGNINO (*) & Carlo MONTANARI (*)

SEDIMENTS OF SMALL LAKES ALONG CREEKS IN NORTHERN APENNINES (NW-ITALY) AS EVIDENCE OF ANCIENT SLOPE INSTABILITY

ABSTRACT: DAGNINO D. & MONTANARI C., Sediments of small lakes along creeks in Northern Apennines (NW-Italy) as evidence of ancient slope instability. (IT ISSN 0391-9838, 2015). In the last decades the erosion of some creeks in the Ligurian Apennines has revealed sediments of stagnant waters rich in plant remains. These small “fossil lakes” represent the effects of phases of slope instability that caused barrages from landslides at different times, since at least 10,000 BP until the 19th century. This paper presents a review of studies in the central Ligurian mountains concerning the formation of temporary ponds along streams, which is documented for different sites in this area for prehistoric times, the Middle Ages and later. It also aims to deepen by means of biostratigraphyc analysis - the knowledge of a site along the creek Rio Dell’Orso, near the village of Senarega (710 m, Scrivia Valley, NW-Italy). There, erosion has exposed sediments of a disappeared lake, containing logs, branches and pollen. A small basin has formed, after a landslide that had dammed the stream in prehistoric times, for an estimated period of about three centuries. Radiocarbon dating allowed to attribute the deposits to a period around 4000 BP. More than 40 branches and trunks were identified and pollen analysis was carried out. Among the macroremains, silver fir (Abies) clearly prevails in number and size; also Fagus and Fraxinus were identified. Even within the pollen assemblages, unfortunately rather badly preserved, the silver fir is dominant in all layers. These data are consistent with what is known about the Holocene history of the Ligurian Apennines: also the upper Scrivia Valley was long characterized by forests dominated by silver fir with beech, which were still widespread throughout the region until the Roman period and the Middle Ages.

Nell’Appennino Ligure, negli ultimi decenni, l’erosione di alcuni torrenti ha messo in luce depositi di acque stagnanti più o meno ricchi di resti vegetali. Queste tracce di “laghetti fossili” rappresentano gli effetti di fasi di instabilità dei versanti che hanno causato frane di sbarramento in epoche diverse, da 10.000 BP al XIX secolo. Questo contributo offre una panoramica dello stato dell’arte per i rilievi della Liguria centrale e approfondisce, su basi biostratigrafiche, le conoscenze sul sito di Senarega (710 m, alta Valle Scrivia, Genova): presso questo piccolo borgo medievale, l’erosione del Rio dell’Orso ha messo in luce un antico deposito lacustre costituito da circa un metro di sedimenti contenenti tronchi, rami, polline. Datazioni radiocarboniche permettono di riferire questi sedimenti ad un periodo intorno a 4000 BP. Un piccolo bacino deve essersi formato in seguito ad una frana che ha sbarrato il corso del torrente in epoca preistorica, per un periodo stimato di circa tre secoli. Più di 40 tra rami e tronchi sono stati identificati e sono state condotte analisi polliniche di quattro livelli. Tra i macroresti di specie arboree, Abies prevale nettamente per numero e dimensioni; altre essenze identificate sono Fagus e Fraxinus. Anche nell’ambito dei complessi pollinici, piuttosto mal conservati, l’abete bianco risulta prevalente in tutti i livelli. Questi dati concordano con ciò che si sa della storia olocenica dell’Appennino Ligure per la cronologia di attribuzione: anche questa valle, come la vicina Val Vobbia, più a SW quelle del gruppo del M. Beigua e tutto l’Appennino settentrionale, è stata a lungo caratterizzata da abetine con faggio che nell’Età Romana e nel Medioevo erano ancora ampiamente diffuse in tutta la regione. Anche il fenomeno della formazione di laghetti temporanei lungo i torrenti è abbastanza documentato per queste aree, dalla preistoria al Medioevo e anche più recentemente.

KEY WORDS: Biostratigraphical archives; Palaeolakes; Landslides; Archaeobotany; N-Apennines, NW-Italy.

TERMINI CHIAVE: Archivi biostratigrafici; Laghetti fossili; Paleofrane; Archeobotanica; Appennino Ligure; Abies.

RIASSUNTO: DAGNINO D. & MONTANARI C., Sedimenti di laghetti scomparsi lungo torrenti dell’Appennino settentrionale come tracce di instabilità dei versanti. (IT ISSN 0391-9838, 2015).

INTRODUCTION

———————— (*) Dipartimento di Scienze della Terra, dell’Ambiente e della Vita (DISTAV) - Polo Botanico Hanbury, Università di Genova, Corso Dogali, 1m – 16136, Genova, Italy. ([email protected]) The authors are grateful to Dr. S. Pedemonte and Prof. G.C. Cortemiglia for showing some of the sites studied and collaboration, to Dr. C. Bellini for the revision of the English manuscript and to two anonymous reviewers, whose comments have allowed improving the paper quality.

Events of environmental instability (e.g. slope instability, landslide, alluvial phenomena) seem to be increasing in Italy over recent years and in some cases their effects are catastrophic. These events are usually attributed to recent climate changes, which can in turn be ascribed both to natural climate cycles and human activity. Similar episodes have surely occurred even in the mid-late Holocene, a period 15

FIG. 1 - Location of the sites where evidence of disappeared small lakes are known in the Ligurian Apennines: 1) Senarega; 2) Ponte di Zan; 3) Gordenella; 4) Daglio.

characterized by very high climatic variability (Mayewski & alii, 2004). For ancient times, when no written documentation is available, methods of palaeoecology can be used besides the classical geologic-geomorphologic approach (e.g. Magaldi & alii., 2007). Palaeoenvironmental reconstructions, based both on geological and biological evidence, can establish correlations within natural phenomena, geographical and geomorphological features, human activities, trends, recurrent events. Landslides are obvious signs of slope instability and therefore they are important subjects of study by geologists, geomorphologists and applied geologists. Actually, also biostratigraphic research and particularly archaeobotany proved to be a rich source of information on past environmental dynamics, especially with regard to the dating of the events and the reconstruction of the environment in which they occurred. The evidence of past slope instability often consists of ancient silty-clayey lacustrine sediment interposed within recent coarse alluvial cover, along riverbeds; these small “fossil lakes”, that represent barrages of the watercourse, are often rich in plant micro- and macro-remains, so they are very interesting palaeoenvironmental archives. In this paper we review the state of the art about a few cases of palaeoenvironmental reconstruction from sediments of landslide-dammed palaeolakes in the central Ligurian Apennines (NW Italy); furthermore, a case study that until now has been presented only to a limited audience is illustrated in detail (Dagnino & Montanari, 2012). GEOLOGIC AND GEOMORPHOLOGIC SETTING The upper Scrivia Valley (Ligurian Apennines, NW Italy) is characterized by turbidites both calcareous and pelitic-arenaceous from Upper Cretaceous (Calcari di 16

Monte Antola and Flysch di Busalla, respectively) and late and post-orogenic deposits of the Tertiary Piedmont Basin from the Oligocene (conglomerates and sandstones), named Conglomerati di Savignone (Giammarino & alii., 2002). The hydrographic network shows a rejuvenation in the present sediments of the riverbeds (Cortemiglia & Pedemonte, 2005), due to erosion that highlights a deepening trend of the main stream and of some secondary basins (Vobbia, Gordenella, Agnellasca, Brevenna). This process has brought to light some ancient lacustrine deposits caused from landslides which had dammed the riverbeds at different times, at least from 10,000 BP until the 19th century (fig. 1) (Montanari & alii, 1985; Cortemiglia & Pedemonte, 2001, 2005; Cortemiglia & alii, 2004). Also the examination of historical documents reveals a number of landslides that caused damage to buildings and farmland in this area (Pedemonte & alii, 1995), and still today the upper Scrivia Valley shows a lot of landslides, variables for their size, even not always in relation to water courses (http://geoportale.regione.liguria.it//geoviewer/pages/apps/repertorio/repertorio. html?id=492). REVIEW OF PREVIOUS STUDIES IN THE UPPER SCRIVIA VALLEY (NW-ITALY) In upper Scrivia Valley several sites have been recorded in which fine lacustrine sediments are deposited in torrential riverbeds after the formation of palaeolakes due to landslides (fig. 1). Along the Vobbia creek, a tributary of T. Scrivia, at an altitude of c. 400 m a.s.l. near the “Ponte di Zan” and the medieval castle Castello della Pietra, sediments of a disappeared lake have been described and studied both for their geologic-sedimentologic features and bi-

ostratigraphic ones (Montanari & alii, 1985) Palynological and macrofossil analyses have highlighted the persistence of forests dominated by silver fir (Abies), beech and deciduous meso-thermophilous broadleaves along a profile of about 3 meters of sandy silt. A dating from wood remains (4461 ± 50 uncal. BC, Montanari & alii, 1985) had attributed these sediments to the Neolithic period, in agreement with the regional paleoenvironmental history known at that time. In the same paper, also a study of recent pollen deposition was carried out in the surroundings of the site, in order to refine the interpretation of the palaeo-palynological data: this has shown good congruence between recent pollen deposition and the composition of current vegetation (Montanari, 1989; Guido & Montanari, 1991; Guido & alii, 1997-99). Further investigation in the same site (Cortemiglia & Pedemonte, 2001), after the flood of 1999, has shown a conifer trunk about 16 metres long dated to 11,190 cal. BP (cal. 9250 BC - Beta 143343-Rsd), embedded in coeval clayey sediments (11,200 cal. BP, 9250 cal. BC - Beta 145740-Rsd), with apparent varves. This dating is much older than the previous one, suggesting that the fossil conifer woodland dates back to the Late-Glacial period. Therefore, some doubts arise about the accuracy of the more recent date, if we admit a single landslide for this site; nevertheless they both agree with the Late-Glacial and Holocene history of the silver fir in this area (see below). As the authors under-

line, the sliding-surface of the palaeo-landslide is visible at present on the left river slope, downstream of the palaeolake sediments (fig. 2). Taking into account the difference of altitude, there are several subsequent studies that seem to agree with this chronology: at Mogge di Ertola (Aveto Valley, c. 1100 m a.s.l.), pollen, wood and charcoal evidence of silver fir dated since 10,000 cal. BP testify its spread even at higher altitudes at the beginning of the post-glacial period (Guido & alii, 2013), and the same happened elsewhere in the Ligurian Apennines (Branch, 2004). Near the hamlet of Gorreto (Gordenella Valley, Alessandria, fig. 1) at an altitude of 520 meters a.s.l., the fluvial erosion has exposed silty sediments deposited as a result of damming from landslide, which took place in the Middle Ages (Cortemiglia & alii, 2004; Cortemiglia & Pedemonte, 2010). The thickness of the lacustrine sediments is at least 350 cm, which were deposited during three centuries. Preliminary wood and pollen analysis has demonstrated the existence of an open forest cover, mainly composed of deciduous mesothermophilous broadleaves (Ostrya, Quercus decid., Corylus, Fraxinus, Alnus) and beech (Fagus). Sweet chestnut (Castanea) was probably cultivated in the surroundings. Among the conifers, Abies and Pinus were recorded. Based on two conventional radiocarbon datings, performed the first at the top of the sediments (800±50 BP, cal. BP 790-660 (95% prob.), Beta 178801) and the second 80.5

FIG. 2 - Vobbia Valley. Just downstream of the Castello della Pietra (on the right bank, in the box), is clearly visible in the left bank a sliding surface probably connected to the ancient landslide that blocked the stream around 13,000 BP (in the circle on the right).

17

TABLE. 1 - Main features of the lacustrine sediment samples from Senarega Sample

Depth (cm)

Dry weight (g)

Colour (Munsell soil color charts)

S2

240

2,97

10YR 8/1 10YR 3/2

Grey clay, with pointed small pebbles included

S1

210

1,56

7,5YR 6/2 7,5YR 3/2

Brown clay

S3

165

2,26

10YR 7/2 10YR 4/2

Brown clay (embedding the dated trunks)

S4

125

3,61

10YR 5/1 10YR 3/1

Grey sand

cm lower (870±60 BP, cal. BP 930-680 (95% prob.), Beta 178800), we can locate the end of the sedimentation in the High Middle Ages, in a climatic phase just prior to the modest cooling of the Little Ice Age. Also in this case, clues of the ancient landslide which dammed the course of the creek have been identified. Near the village of Daglio (Agnellasca Valley, Alessandria), in the mid-nineteenth century, a pond now disappeared was formed twice along the course of the creek Carreghino, at an altitude of c. 580 m a.l.m., due to landslides (fig. 1). At the confluence of the Rio Robè, the toponym “Cappella del Lago” (literally “Lake Chapel”) and a small aedicule, in addition to a large detachment niche, are nowadays the sole witnesses of a temporary basin which probably lasted only a few years (Pedemonte & alii, 1995). Another report concerns a small pond which would have formed around the 9th century along the creek Seminella (Busalla) (Cortemiglia & Pedemonte, 2010).

FIG. 3 - Location of the studied site, along the Rio dell’Orso, near Senarega; the area of the finding of lacustrine sediments is included in the rectangle.

18

Description

NEW DATA FROM THE SITE OF SENAREGA STUDY AREA The site of Senarega was reported for the first time by Cortemiglia & Pedemonte (2005) and later it was presented from a palaeobotanic point of view to a limited audience (Dagnino & Montanari, 2012). It belongs to the Brevenna Valley, a secondary basin of Scrivia Valley, that is mainly characterized by limestone substrate (Giammarino & alii., 2002). The studied lacustrine sediments are located to the east of the village of Senarega, in the riverbed of the Rio dell’Orso, a right side tributary of the T. Brevenna, at an altitude of 710 m a.s.l., and they extend for a little more than 60 m in a stretch, where the creek flows towards the North (fig. 1 and fig. 3). FEATURES OF THE SEDIMENTS The lacustrine sediments are about 1 m thick, and the layers are inclined by 37° with respect to the current bed.

FIG. 4 - Stratigraphic column of the sediment outcropping (left) and a photograph of a representative section of the column (right) in which the deepest layer (S2) is clearly visible, characterized by coarse material embedded in a fine matrix and probably connected to the landslide event which resulted in the formation of the pond.

The top of the sequence is placed at a depth of 120 cm with respect to the surface of the current pebble riverbed. Starting from the top, a sandy layer 15 cm thick is observed; the bulk of the sediment appears instead as a rhythmic deposit of siltyclay layers 1-2 cm thick, alternating with thin deposits rich in organic matter 0.5-1 cm thick; therefore, several varves can be easily recognized locally (fig. 4). In the main part of this package of sediments (from 135 to 200 cm depth) numerous plants macroremains crop out; they are tree trunks and branches, driven parallel to the sediment layers. The deepest part (200 to 218 cm depth) instead is constituted solely by silt deposit. At the base of the pack of lake sediments a colluvial slope deposit is found, consisting of eterometric (1-10 cm) sharp limestone stones, that are embedded in a clay-silt matrix. This layer has an average thickness of 50 cm (218 to 260 cm depth). Three calibrated radiocarbon dating are available; they were performed on wood samples of silver fir that were located at the top of lacustrine sediments, at 35 and 55 cm deep respectively. These data allowed to date the upper layer of the sediment (from 120 to 175 cm depth) to a period between 2480 BC and 2620 BC, whereas the total duration of the phase of lacustrine sedimentation is estimated to be about 300 years (Cortemiglia & Pedemonte, 2005; Cortemiglia, 2012).

A lot of plant macroremains were found along the part of the riverbed which showed lacustrine sediments. They consisted of trunks and branches, even rather large (a few metres), brought to light by the fluvial erosion. All wood macroremains embedded at least partially in the sediment and visible at the time of the survey (2011) were numbered, photographed and sampled; therefore, the collection must be considered as bulk, unlike that for pollen analysis; however, as mentioned above, the trunks are prevailingly contained in the layer S3 (fig. 4); overall we have collected and examined 46 wood samples (each of few cm3). Moreover, some lightly burned wood remains were observed.

MICROREMAINS ANALYSIS

RESULTS

Four sediment samples for pollen analysis were taken from the section, approximately 100 g each (from the

The pollen analysis has focused on samples S2, S1 and S3 (fig. 5 and fig. 6); sample S4 is omitted in the diagram, being

deepest, named S2, S1, S3 and S4; tab. 1; fig. 4). For the extraction and concentration of palynomorphs the standard methodology has been employed (treatment with HCl, HF, NaOH, Na hexametaphosphate, Moore & Webb, 1983). The pollen sum for the calculation of pollen percentages includes only plants from dry environments (TLP, Total Land Plants), in order to avoid possible over-representation from wetland plants. The pollen concentration was calculated as FPA (Absolute Pollen Frequency), i.e. in relation to the dry weight of the sub-samples used (Accorsi & Rodolfi, 1975). MACROREMAINS ANALYSIS

19

FIG. 5 - Pollen diagram: histograms represent percentages of selected pollen taxa, divided into three levels examined (S3: 165 cm depth; S1: 210 cmdepth; S2: 240 cm depth). The last histogram refers to the percentages of tree, shrubs and herbs.

very poor in pollen. In all the samples, however, the pollen grains were scarce and poorly preserved, making it difficult their identification, in some cases. The corrosion of pollen can be attributed to fluctuations in the water level of the river, resulting in exposure to air and oxidation of the grains contained in the sediments. Overall, we identified 926 grains (588 of which are terrestrial angiosperms - Total Land Plants, TLP). The modest pollen concentration (FPA) allows to obtain statistically reliable information for those samples for which the counts have achieved at least a few hundreds of grains (S1, FPA = 6560; S3, FPA = 10326), while the results must be considered indicative for S2 (FPA = 1935). The pollen analysis has allowed the identification of 30 taxa, consisting of 16 trees/shrubs and 14 herbs (the latter comprising three hygro-hydrophilous taxa and the group of Pteridophyta, which were excluded from the pollen sum). In each levels the percentage of tree species were greater than 70%, and the most abundant species was Abies; this is confirmed by the finding, during the pollen analysis, of a lot of microscopic wood fragments with anatomic features corresponding to silver fir, and by macroremains. The analysis also included non-pollen palynomorphs (NPP), fairly abundant in the samples (fig. 6): we found some fungal spores, among which Ustulina deusta type was very abundant and also Alternaria and Cirrenalia were found. Microcharcoals, unlike the pollen grains, were very abundant in all levels, in particular those belonging to the class size between 10 and 50 µm. The level richest in microcharcoal is undoubtedly the deepest, S2. 20

Xilological analysis has been carried out on 46 wood samples. They were attributed to three taxa, namely Abies (24 samples), Fagus (14 samples) and Fraxinus (6 samples); only two samples remained unidentified, due to their poor state of preservation. PALEOENVIRONMENTAL RECONSTRUCTION The pollen analysis has shown that the vegetation of the valley was to be a forest, with a clear dominance of conifers (Abies and, in lower amount, Pinus cf. sylvestris). Deciduous trees always maintained a secondary role, and were represented mainly by Quercus deciduous and Corylus; nevertheless, considering its under-representation in the pollen record (Montanari, 1989; Guido & Montanari, 1991; Guido & alii., 1997-99; Guido & Montanari, 2004), Fagus had to be one of the main deciduous trees. The forest vegetation was probably rather closed, given the scarcity of herbaceous species, consisting mostly of Poaceae; the abundance of Cyperaceae and Juncaceae indicates the occurrence of hydrophilous vegetation along the shores of the palaeobasin. Ferns (Filicales) were very abundant, and often are the main herbaceous component; the recurrent discovery of sporangia suggests that ferns were present locally or along the waterway. Xilological analysis detected the dominance of Abies, confirming what has been observed with the palynological analysis. Based on macroremains, other abundant species in the forest had to be Fagus and Fraxinus. The latter, however, is found more as wood macroremain than in pollen spectra: a certain pollen under-representation can be expected for Fraxinus ornus (flowering ash), which, unlike F. excelsior

FIG. 6 - Diagram of pollen concentration: the histograms represent the concentration of pollen grains, the NPPs and microcharcoal, divided into three levels examined (S3 - 165 cm depth; S1 - 210 cm depth; S2 - 240 cm depth).

(ash), has predominantly entomophilous pollination (Guido & alii. 1997-99); nevertheless, on the basis of wood anatomy, the two species cannot be distinguished. The overall picture, then, refers to a fir woodland with beech, which lasted throughout the sedimentary series. Although the overall palaeoenvironmental image is rather homogeneous in the three levels, the deepest one (S2) was significantly richer in microcharcoal and poorer in pollen; this level refers to a silt deposit with abundant angular small pebbles which is probably an evidence of the material deposited during the landslide responsible for the formation of the ancient pond. So the abundance of microcharcoals may be an indication of fire that would have partially destroyed the fir-woodland, exposing the soil to erosion and probably contributing to cause the landslide; however, the small size of the carbon particles and the fluvial context imply the possibility of fluvial transport from afar. Moreover, in the same level, ascospores of Ustulina deusta are markedly more abundant: this fungal parasite causes soft rot in many tree species, including all of those found in the analysis, such as Abies alba (Van Geel & Aptroot, 2006; Menozzi & alii, 2010; Guido & alii, 2013). Though its find is very common in the Holocene deposits, higher percentages are in some cases related to the presence of many infected trees (Van Geel, 1978; Van Geel & alii, 1981; Menozzi & alii, 2010). So this parasitic fungus could have contributed to the weakening of the silver fir before the formation of the pond (Montanari & Guido, 2011). In the other two levels the presence of microcharcoals is still good, although lower than in the deepest one; the simultaneous presence of abundant microcharcoals and the

presence of burned wood may be related to the fire assumed for the deepest level. DISCUSSION The events that have led or accompanied the formation of these small temporary ponds along several rivers of Liguria can be easily framed in the regional Holocene history. At present, we have a lot of data, mostly palynological, but also relating to macroremains (wood, charcoal) that bear witness to a complex history, either climatic and biological and of human exploitation of environmental resources. The relationships that can be suggested between the formation of these ephemeral ponds and paleoenvironmental events are mainly to landslides, forest fires and anthropogenic management of vegetation and geomorphology. Interestingly, the palaeoenvironmental reconstructions of the sites described above are in agreement with each other and with the new case study of Senarega. For example, despite the most recent dating of the Vobbia Valley sediments is doubtful and the altitude is significantly lower (400 against 730 m a.s.l.), the palaeoenvironmental picture obtained from pollen analysis is similar, except for the lower amount of beech, in Vobbia pollen spectra. Extensive palynological studies carried out in other parts of Ligurian Apennines (Cruise, 1990; Branch, 2004; Guido & alii, 2004; Molinari & alii, 2008; Cruise & alii, 2009; Guido & alii, 2009; Guido & alii, 2013) have repeatedly documented the role of silver fir as dominant woodland species in sediments related to ages between 11,500 and 2000 years 21

BP. However, in most recent sediments the presence of Abies is increasingly rare. The comparison with the present vegetation, consisting of thermophilous deciduous forests which, at these low altitudes, lack altogether both Abies and Fagus, can suggest a cooler and wetter climatic context than the current one, for the periods in which the ponds were depositing these sediments. Silver fir can thrive even a few metres above sea level (Giacobbe, 1950) and there are a lot of evidences that suggest a wider range of habitats occupancy in the past for this species (Montanari & alii, 1997; Colombaroli & alii, 2007; Mariotti Lippi & alii, 2007; Bellini & alii, 2009; Guido & alii, 2013). Nonetheless, silver fir has dramatically reduced its distribution in the midlate Holocene. Some authors placed the beginning of Abies disappearence around 6000 BP for different regions of Europe (Tinner & alii, 1999; Keller & alii, 2002; Wick & Möhl, 2006; Colombaroli & alii, 2007; Bellini & alii, 2009; Savelli & alii, 2013). However, this process took place gradually, so there are abundant remains of silver fir in more recent sediments (e.g. Senarega site), and up to the Middle Ages (Branch & alii, 2003; Guido & alii, 2003; Guido & alii, 2013). It was suggested that some local extinctions of silver fir were due to increasing of fire activity (Tinner & alii, 2000; Colombaroli & alii, 2007) and human activity (Maggi, 2000; Carcaillet & Muller, 2005), also in very recent times (Vrška & alii, 2008). Nevertheless, even climate may have played an important role in the decline of silver fir; in particular it was detected a climate trend towards aridification and warming over the last three millennia (Sadori & alii, 2013), but some authors demonstrated an earlier beginning of the aridification phenomena, such as 5500 BP ( Jalut & alii, 2009) and 7700 BP (Peyron & alii, 2011). However, sites like these small streams, with a particularly humid mesoclimate, seem less useful for general climatic comparisons. Elsewhere, in central Italy, a case of a watercourse dammed by an ancient landslide has been recently detected and studied with similar methods, for the midHolocene (Savelli & alii, 2013); in that case, however, the sediments were much more thick and the lake lasted over seven thousand years, owing to the formation of a stable dam of calcareous breccia (8990-8550 cal BP); the cause of the slope instability was attributed there to an increase in precipitation at a continental scale. Several investigations on the climatic implications of landslide recurrence have been described in literature on various European regions, among which, in particular, are Austria, Great Britain, Italy, Norway, Poland, Spain and Switzerland (Soldati & alii, 2004). Although some situations of increased landslide activity in the mid-late Holocene are attributed to human activity (e.g. Dapples & alii, 2002; Polemio & Lonigro, 2013), it is possible to think about a correlation between climate instability and landslides triggering during the Holocene. This time interval, indeed, was characterized by very high climate variability (Zak & alii, 2002; Mayewski & alii, 2004; Drysdale & alii, 2006). 22

CONCLUSIONS The information from palaeoecology investigations proved once again to be useful to complete the geological hypotheses concerning palaeo-landslides by providing both indirect and direct datings, and palaeoenvironmental pictures which are basic to interpret possible causes and evidence of ancient slope instability (e.g. vegetation cover, climate changes, human activities). Moreover, this knowledge can act as an important tool for studying and forecasting current slope instability. Obviously, there are different scales of analysis, from the strictly local phenomena up to the regional and continental levels. The examples shown here belong to a local level, to which the environmental effects are not always easily correlated to global events (e.g. global climate change). Moreover, even the more detailed studies often leave open the question of if and how far human activities have a responsibility in this regard. The site of Senarega is another good example recording the geomorphological and vegetation dynamics that characterized the Ligurian Apennines and much of southern Europe in the mid-late Holocene, also contributing to illustrate the history of the postglacial expansion and subsequent decline of silver fir (Abies). This site also provides a further evidence of how the process of damming of rivers due to landslides, resulting in the formation of small lakes today disappeared, have been a reality quite common during the Holocene rather than isolated episodes.

REFERENCES ACCORSI C.A. & RODOLFI G. (1975) - Primi risultati sullo studio di un suolo calcimorfo delle Alpi Apuane in relazione ad analisi palinologiche e microbiologiche. Bollettino della Società Italiana di Scienze del Suolo, 9, 35-51. BELLINI C., MARIOTTI LIPPI M. & MONTANARI C. (2009) - The Holocene landscape history of the NW Italian coasts. The Holocene, 19, 11611172. BRANCH N. (2004) - Late Würm Lateglacial and Holocene environmental history of the Ligurian Apennines, Italy. In Balzaretti, Pearce & Watkins (Eds.), Ligurian Landscapes, studies in archaeology, geography & history. Accordia Specialist Studies on Italy,10. London, 7-69. BRANCH N., GUIDO M.A., MENOZZI B.I., PLACEREANI S. & MONTANARI C. (2003) - Prime analisi polliniche per il sito «Moggia di Pian Brogione» (Casanova di Rovegno – GE). Archeologia Postmedievale (2002), 6, 125-131. CARCAILLET C. & MULLER S.D. (2005) - Holocene tree-limit and distribution of Abies alba in the inner French Alps: anthropogenic or climatic changes? Boreas, 34, 468-476. COLOMBAROLI D., MARCHETTO A. & TINNER W. (2007) - Long-term interactions between Mediterranean climate, vegetation and fire regime at Lago di Massaciuccoli (Tuscany, Italy). Journal of Ecology, 95, 755770. CORTEMIGLIA G.C. (2012) - Lineamenti generali della storia climatica del territorio alessandrino (Piemonte, Italia). Atti della Società Toscana

di Scienze Naturali, Memorie, Serie A, 117-119, 5-16. doi: 10.2424/ ASTSN.M.2012.23. CORTEMIGLIA G.C., MENOZZI B.I., MONTANARI C., PEDEMONTE S. & PLACEREANI S. (2004) - Ritrovamento di un deposito lacustre di età medievale in val Gordenella (Piemonte-Italia). In Novitate, 19, II, 38, 59-66. CORTEMIGLIA G.C. & PEDEMONTE S. (2001) - Ritrovamento nell’Appennino ligure-piemontese (Val Vobbia – Genova) di varve lacustri appartenenti alla cronozona Preboreale. Il Quaternario, 14 (1), 23-30. CORTEMIGLIA G.C. & PEDEMONTE S. (2005) - Il deposito lacustre sub-boreale di Senarega ritrovato lungo il tratto ligure della valle Scrivia (Genova – Italia). In Novitate, 39 (1), 55-58. CORTEMIGLIA G.C. & PEDEMONTE S. (2010) - Ritrovamenti lacustri antichi nel bacino della Scrivia. Un’opportunità per la geologia, per la botanica e per l’archeologia. In Novitate, 50 (2),75-80. CRUISE G.M. (1990) - Pollen stratigraphy of two Holocene peat sites in the Ligurian Apennines, northern Italy. Review of Palaeobotany and Palynology , 63 (3-4), 299-313. CRUISE G.M., MACPHAIL R.I., LINDERHOLM J., MAGGI R. & MARSHALL P.D. (2009) - Lago di Bargone, Liguria, N Italy: a reconstruction of Holocene environmental and land-use history. The Holocene 19 (7), 987-1003. DAGNINO D. & MONTANARI C. (2012) - Un laghetto fossile in Val Brevenna (Genova): indagini paleobotaniche. Atti del Congresso A.I.Ar. 2012 Modena. Modena, 22-24 Febbraio, 2012, Patron Editore, Bologna, CD Rom, 330-341. ISBN/ISSN: 9788855531665

GUIDO M.A., MENOZZI B.I, BELLINI C., PLACEREANI S. & MONTANARI C. (2013) - A palynological contribution to the environmental archaeology of a Mediterranean mountain wetland (North west Apennines, Italy). The Holocene, 23 (11), 1517-1527. GUIDO M.A., MARIOTTI LIPPI M., MENOZZI B.I., PLACEREANI S., MONTANARI C. (2004) - Il paesaggio vegetale montano della Liguria centro-occidentale nell’Età del Ferro: area del Monte Beigua (Savona). In: De Marinis R.C. & Spadea G. (a cura di), I Liguri,1. Un antico popolo europeo tra Alpi e Mediterraneo. Skira, Ginevra-Milano, 91-95. JALUT G., DEDOUBAT J.J., FONTUGNE M. & OTTO T. (2009) - Holocene circum-Mediterranean vegetation changes: Climate forcing and human impact. Quaternary International, 200, 4–18. KELLER F., LISCHKE H., MATHIS T., MöHL A., WICK, L., AMMANN, B. AND KIENNAST F. (2002) - Effects of climate, fire and humans on forest dynamics: forest simulations compared to the palaeoecological record. Ecological Modelling, 152 (2), 109–127. MAGALDI D., LOR A. & PERONI P. (2007) - Mapping slope instability ground factors in the National Park of Gran Sasso- Monti della Laga (Abruzzo, Central Italy): results and perspectives. Geografia Fisica e Dinamica Quaternaria, 30, 55-62. MAGGI R. (2000) - Aspetti di archeologia del territorio in Liguria: la formazione del paesaggio dal Neolitico all’Età del Bronzo. Annali dell’Istituto ‘Alcide Cervi’, 19, 143–162. MARIOTTI LIPPI M., GUIDO M.A., MENOZZI B.I., BELLINI C. & MONTANARI C. (2007) - The Massaciuccoli Holocene pollen sequence and the vegetation history of the coastal plains by the Mar Ligure (Tuscany and Liguria, Italy). Vegetation History and Archaeobotany, 16, 267- 277.

DAPPLES F., LOTTER A.F., VAN LEEUWEN J.F.N., VAN DER KNAAP W.O., DIMITRIADIS S. & OSWALD D. (2002) - Paleolimnological evidence for increased landslide activity due to forest clearing and landuse since 3600 cal BP in the western Swiss Alps. Journal of Paleolimnology, 27, 239-248.

MAYEWSKY P.A., ROHLING E.E., STAGER J.C., KARLŽN W., MAASCH K.A., MEEKER L.D., MEYERSON E.A., GASSE F., VAN KREVELD S., HOLMGREN K., LEE-THORP J., ROSQVIST G., RACK F., STAUBWASSER M., SCHNEIDER R.R. & STEIG E.J. (2004) - Holocene climate variability. Quaternary Research, 62, 243-255.

DRYSDALE R., ZANCHETTA G., HELLSTROM J., MAAS R., FALLICK A., PICKETT M., CARTWRIGHT I. & PICCINI L. (2006) - Late Holocene drought responsible for the collapse of Old World civilizations is recorded in an Italian cave flowstone. Geology, 34 (2), 101-104.

MENOZZI B.I., ZOTTI M. & MONTANARI C. (2010) - A non-pollen palynomorphs contribution to the local environmental history in the Ligurian Apennines: a preliminary study. Vegetation History and Archaeobotany, 19, 503-512.

GIACOBBE A. (1950) - L’ecologia dell’Abete bianco appenninico, III, Caratteri floristici e climatici dell’Abete bianco. Archivio Botanico Italiano, 26, 186-221.

MOLINARI C., PAROLA C. & GUIDO M.A. (2008) - Palynological reconstruction of long-term vegetation changes in the Ligurian Apennines, northern Italy. In Terra Nostra – Schriften der GeoUnion Alfred-Wegener Stiftung 2008/2 – 12th International Palynological Congress (IPC-XII) and 8th International Organisation of Palaeobotany Conference (IOPC-VIII), August 30-september 5, 2008 in Bonn, Germany, Abstract Vol. 2, 195.

GIAMMARINO S., GIGLIA G., CAPPONI G., CRISPINI L. & PIAZZA M. (2002) - Carta Geologica della Liguria. Litografia Artistica Cartografica. Firenze. GUIDO M.A. & MONTANARI C. (1991) - Pollen assemblages in surface samples and vegetation relationships in the woods of Liguria northern Italy preliminary results. Archivio Botanico Italiano, 67(1-2), 54-75. GUIDO M.A. & MONTANARI C. (2004) - Immagine pollinica attuale di fitocenosi forestali nell’Appennino nord-occidentale. Colloques Phytosociologiques (Vegetazione postglaciale passata e presente, Camerino, 1998), 28, 147-161. GUIDO M.A., MONTANARI C. & MENOZZI B. (1997-99) - Deposizione pollinica attuale in relazione alla copertura vegetale nell’Alta Val di Vara. Memorie della Accademia Lunigianese di Scienze “Giovanni Cappellini”, 57-58-59, (1997-1998-1999), 295-311. GUIDO M.A., SCIPIONI S. & MONTANARI C. (2003) - Il paesaggio colturale nei dintorni di Casanova di Rovegno (Ge) dal VII-VIII sec. d.C.: dati archeobotanici per l’area di Pian delle Groppere. Archeologia Postmedievale (2002), 6, 117- 123. GUIDO M., MOLINARI C., MONTANARI C. (2009) - Primi dati palinologici per la storia ambientale tardo-pleistocenica della Liguria orientale. In: Di Marzio P., Fortini P. & Scippa G.S. (a cura di), Le scienze botaniche nella cultura e sviluppo economico del territorio - 104° Congresso Naz. Società Botanica Italiana, Campobasso, 16-19 settembre 2009, Abstract, 272.

MONTANARI C. (1989) - Recent pollen spectra in two small mountain basins of the Ligurian Apennines (northern Italy). Grana, 28, 305-315. MONTANARI C. & GUIDO M.A. (2011) - Who has wiped out the silver fir (Abies Alba Mill.)? Clues of pathogens in the Ligurian Apennines. Bolletino dei Musei e degli Istituti Biologici, Universitˆ di Genova, 73, 213; (Available at: www.bmib.it). MONTANARI C., GUIDO, M.A., CORNARA L. & PLACEREANI S. (1997) - Tracce polliniche di boschi neolitici di abete bianco in Val Bisagno (area urbana di Genova). Biogeographia, 19, 133-143. MONTANARI C., GUIDO M.A. & PETRILLO M. (1985) - Tracce di un bosco di Abete bianco in Val Vobbia messe in luce dall’analisi pollinica. Archivio Botanico e Biogeografico Italiano, 61(3-4), 169-184. MOORE P.D. & WEBB J.A. (1983) - An illustrate guide to Pollen Analysis. Hodder & Stoughton. London, 133 pp. PEDEMONTE S., SERRATTO G. & SERRATTO M. (1995) - Frane, dissesti idrogeologici ed evoluzione del territorio in Val Borbera (AL) e località vicine. In Novitate, 19, 27-35. PEYRON O., GORING S., DORMOY I., KOTTHOFF U., PROSS J., DE BEAULIEU J.L., DRESCHER-SCHNEIDER R., VANNI RE B. & MAGNY M. (2011) - Holocene seasonality changes in the central Mediterranean region reconstructed

23

from the pollen sequences of Lake Accesa (Italy) and Tenaghi Philippon (Greece). The Holocene, 21(1), 131-146. POLEMIO M. & LONIGRO T. (2013) - Climate Variability and Landslide Occurrence in Apulia (Southern Italy), Lanslide Science and Practice, 4, 37-41. SADORI L., ORTU E., PEYRON O., ZANCHETTA G., VANNI RE B., DESMET M. & MAGNY M. (2013) - The last 7 millennia of vegetation and climate changes at Lago di Pergusa (central Sicily, Italy). Climate of the Past, 9, 1969-1984. doi:10.5194/cp-9-1969-2013. SAVELLI D., TROIANI F., BRUGIAPAGLIA E., CALDERONI G., CAVITOLO P., DIGNANI A., ORTU E., TEODORI S., VENERI F. & NESCI O. (2013) - The landslide-dammed paleolake of Montelago (North Marche Apennines, Italy): geomorphological evolution and paleoenvironmental outlines. Geografia Fisica e Dinamica Quaternaria, 36, 267-287.

VAN GEEL B. (1978) - A palaeoecological study of Holocene peat bog section in Germany and the Netherlands. Review of Palaeobotany and Palynology, 25, 1-120. VAN GEEL B. & APTROOT A. (2006) - Fossil ascomycetes in Quaternary deposits. Nova Hedwigia, 82, 313-329. VAN GEEL B., BOHNCKE S.J.P. & DEE H. (1981) - A palaeocological study of upper Late Glacial and Holocene sequence from “De Borchert”, The Netherlands. Review of Palaeobotany and Palynology, 31, 367-448. VRŠKA T., ADAM D., HORT L., KOLÁRˇ T. & JANÍK D. (2008) - European beech (Fagus sylvatica L.) and silver fir (Abies alba Mill.) rotation in the Carpathians – A developmental cycle or a linear trend induced by man? Forest Ecology and Management (2009), 258, 347–356.

SOLDATI M., CORSINI A. & PASUTO A. (2004) - Landslides and climate change in the Italian dolomites since the Late glacial. Catena, 55, 141-161.

WICK L. & MöHL A. (2006) - The mid-Holocene extinction of silver fir (Abies alba) in the southern Alps: a consequence of forest fires? Palaeobotanical records and forest simulations. Vegetation History and Archaeobotany, 15, 435-444.

TINNER W., HUBSCHMID P., WEHRLI M., AMMANN B. & CONEDERA M. (1999) Long-term forest fire ecology and dynamics in southern Switzerland. Journal of Ecology, 87, 273-289.

ZAK K., LOZEK V., KADLEC J., HLADIKOVA J. & CILEK V. (2002) - Climateinduced changes in Holocene calcareous tufa formations, Bohemian Karst, Czech Republic. Quaternary International, 91, 137-152.

TINNER W., CONEDERA M., GOBET E., HUBSCHMID P., WEHRLI M. & AMMANN B. (2000) - A palaeoecological attempt to classify fire sensitivity of trees in the southern Alps. The Holocene, 10, 565-574.

24

(Ms received 30 June 2014; accepted 15 January 2015).

DOI 10.4461/GFDQ.2015.38.03

Geogr. Fis. Dinam. Quat. 38 (2015), 25-33, 4 figg., 2 tab.

DAVIDE FUGAZZA(*), ANTONELLA SENESE(*), ROBERTO SERGIO AZZONI(*), CLAUDIO SMIRAGLIA(*), MASSIMO CERNUSCHI(**), DAVIDE SEVERI(**) & GUGLIELMINA ADELE DIOLAIUTI(*)

HIGH-RESOLUTION MAPPING OF GLACIER SURFACE FEATURES. THE UAV SURVEY OF THE FORNI GLACIER (STELVIO NATIONAL PARK, ITALY) ABSTRACT: D. FUGAZZA, A.SENESE, R.S. AZZONI, C. SMIRAGLIA, M. CERD. SEVERI & G. A. DIOLAIUTI. High resolution mapping of glacier surface features. The UAV survey of the Forni Glacier (Stelvio National Park, Italy) (IT ISSN 0391-9838, 2015).

NUSCHI,

Fast, reliable and accurate methods for glacier mapping are necessary for understanding glacier dynamics and evolution and assessing their response to climate change. Conventional semi-automatic approaches are based on medium-resolution satellite images, but their use can cause significant loss of accuracy when analyzing small glaciers, which are predominant in the Alps. In this paper, we present a semi-automatic segmentation approach based on very high-resolution visible RGB images acquired from a UAV (Unmanned Aerial Vehicle) survey of the Forni Glacier, in the Italian Alps, using an off-the-shelf digital camera. The method has the ability to map large-scale morphological features, i.e. bare ice and medial moraines, with better accuracy than methods relying on medium-resolution satellite imagery, with only slight misclassification at the margins. By using segmentation, we also mapped small-scale morphologies not discernible on satellite images, including epiglacial lakes and snow patches, in a semi-automatic way. On a small portion of the eastern ablation tongue, featuring homogeneous illumination conditions, we also investigated in finer detail the occurrence of fine and sparse debris and tested a texture filter technique for mapping crevasses, which showed promising results. Our analyses confirm that the glacier is un———————— (*) Dipartimento di Scienze della Terra, Università degli Studi di Milano. (**) Agricola 2000 S.c.p.A., Milano. (*) corresponding authors: [email protected], guglielmina. [email protected] The research was performed under the umbrella of an agreement between the Università degli Studi di Milano and Sanpellegrino SpA brand Levissima. Moreover the Agricola 2000 S.c.p.A supported this study taking part with their UAV to the field investigations and also participating to the Lab analysis. The authors kindly acknowledge the Stelvio National Park managers and staff for their help and support. This work was also performed in the framework of the PRIN project 2010/2011 (2010AYKTAB_006), local leader C. Smiraglia. The results of this research also represent a contribution to the development of the updated guidelines of the New Italian Geomorphological Map under the umbrella of the AIGEO working group devoted to this issue.

dergoing intense dynamic processes, including darkening of the ablation tongue and increased surface instability, and show the potential of UAVs to revolutionize glaciological studies. We suggest that by using a combination of different payloads, mapping of glacier features via UAVs could reach high levels of accuracy and speed, making them useful tools for glacier inventories and geomorphological maps. KEY WORDS: High resolution mapping, Remote sensing, UAV (Unmanned Aerial Vehicle), Alpine glaciers, Forni Glacier, Italy. RIASSUNTO: D. FUGAZZA, A. SENESE, R.S. AZZONI, C. SMIRAGLIA, M. CERD. SEVERI & G. A. DIOLAIUTI. Cartografia ad alta risoluzione della superficie glaciale e delle sue forme. Il rilievo tramite drone del Ghiacciaio dei Forni (Parco Nazionale dello Stelvio, Italia). (IT ISSN 0391-9838, 2015). NUSCHI,

La cartografia delle superfici glaciali richiede l’utilizzo di metodi veloci, precisi e affidabili, per meglio comprendere le dinamiche glaciali, i processi evolutivi e la risposta dei ghiacciai ai cambiamenti climatici. L’ uso di approcci semi-automatici tradizionali, basato su immagini satellitari alle medie risoluzioni, può causare notevoli perdite di precisione nell’ analisi di corpi glaciali di piccole dimensioni, che costituiscono la maggior parte dei glacialismo alpino. In questo studio, presentiamo un nuovo approccio semi-automatico basato sulla segmentazione di immagini ad altissima risoluzione acquisite tramite fotocamera digitale durante un rilievo del Ghiacciaio dei Forni (Parco Nazionale dello Stelvio, Italia), effettuato con un drone (Aeromobile a Pilotaggio Remoto). Con questo metodo è stato possibile cartografare forme a grande scala, ovvero ghiaccio esposto e morene mediane, con maggiore precisione rispetto ai metodi basati su immagini satellitari a media risoluzione, nonostante alcuni trascurabili errori di classificazione ai margini del ghiacciaio. L’uso della segmentazione ha permesso anche la cartografia semi-automatica di forme a piccola scala, non visibili da immagini satellitari, inclusi laghi epiglaciali e nevai. Su una porzione ridotta della lingua d’ablazione orientale, caratterizzata da condizioni di illuminazione uniforme nelle immagini, abbiamo inoltre indagato con maggiore dettaglio la presenza di detrito fine e sparso e sperimentato un filtro d’immagine basato sulle texture per cartografare i crepacci, con risultati molto incoraggianti. Le analisi forniscono una conferma degli intensi processi dinamici a cui è sottoposto il ghiacciaio, come la sempre maggiore presenza di detrito sulla lingua d’ablazione e un incremento di instabilità superficiale, e dimostrano le potenzialità dei droni per innovare gli studi glaciologici. Attraverso la combinazione di diversi

25

sensori, la cartografia delle superfici glaciali tramite droni potrebbe in futuro raggiungere elevati livelli di precisione e rapidità, risultando così uno strumento utile per i catasti glaciali e la cartografia geomorfologica. TERMINI CHIAVE: Cartografia ad alta risoluzione, Telerilevamento, Drone - APR (Aeromobile a Pilotaggio Remoto), Ghiacciai alpini, Ghiacciaio dei Forni, Italia.

INTRODUCTION Retreat of glaciers worldwide since the Little Ice Age is one of the clearest clues of climate change. In the Alps, glacier shrinkage has been particularly severe since the 1980s and if the trend continues, many Alpine glaciers could disappear during the 21st century, with a serious impact on the energy and water supply (EEA, 2012). During its active phase, glacier recession is also followed by geomorphological changes such as an increase in supraglacial debris cover and formation of epiglacial lakes (Diolaiuti & Smiraglia, 2010). Debris cover affects ice albedo and plays an important role in the glacier energy balance (Oerlemans & alii, 2009), while ice-contact lakes can increase ice ablation via calving processes (Benn & alii, 2012). Fast and reliable methods for glacier mapping are necessary to study the evolution of glaciers and assess their response to climate change. For this purpose, several semiautomatic approaches employing medium-resolution satellite imagery such as ASTER and Landsat (15-30 m of pixel resolution) have been developed, using a combination of optical, thermal and/or morphometric data (for a review see Shukla & alii, 2010). Accuracy of medium-resolution methods, however, decreases when analyzing changes in smaller glaciers (Paul & alii, 2013). This precludes mapping of small-scale geomorphological features, including crevasses, snow patches and small epiglacial lakes, and of fine and sparse debris, which is widespread on most glaciers and likewise has an impact on the albedo (Azzoni & alii, 2014). In theory, images with higher resolution could provide a more detailed discrimination of supraglacial features. High-resolution satellite imagery however is available at a high cost to the end users, and so far it has mainly been employed as a means of validating lower resolution methods, substituting fieldwork. In glacier inventories, orthomosaics from aerial surveys are also employed for mapping glacier boundaries via manual delineation, as in the recently developed Swiss Glacier inventory (Fischer & alii, 2014). However, this process is slow and further information is usually needed to identify glacier margins on highly debris-covered glaciers. A semi-automatic approach, developed by Knoll & Kerschner (2009) for the Tyrol Inventory, was based on Airborne Laser Scanning (ALS) surveys, but whenever it was applied over debris-covered glaciers manual correction was required. ALS surveys are also expensive, and this limits their application on a global scale. 26

More recently, unmanned aerial vehicles (UAV) have been introduced to the field of glacier studies. These enable low cost on-demand inspection of relatively wide areas of interest, with the ability to carry different types of payload (e.g. a digital camera, near-infrared or thermal sensor) and to survey an area with a better resolution than commercial satellites. UAV-derived orthomosaics have been used to analyze specific features such as cryoconite holes and granules, in combination with ground photography (Hodson & alii, 2007), supraglacial lakes (Immerzeel & alii, 2014), and crevasses, mapped by Ryan & alii (2015) via manual delineation. In this paper, we assess the feasibility of mapping several glacier features (i.e. snow, buried- and bare-ice, epiglacial lakes, and crevasses) of the Forni Glacier from a UAV high-resolution orthomosaic produced using an off-the-shelf digital camera with a semi-automatic approach. In addition, we investigate in more detail these glacier features analyzing a smaller area along the Forni Glacier ablation tongue featuring homogeneous illumination conditions. Finally, to evaluate the advantages provided by this methodology we compare it to classification based on medium-resolution data acquired by Landsat 8. STUDY AREA The Forni Glacier is part of the Ortles-Cevedale group, Lombardy Alps, and of the Stelvio National Park (fig. 1). It is the widest Italian valley glacier (ca. 11.36 km2; D’Agata & alii, 2014), with a northerly aspect and an elevation range between 3670 and 2600 m. It is a composite glacier, formed by three ice streams joining into a tongue that extends for ca. 2 km. Two medial moraines are located on the tongue, originating from rock outcrops located below the snow line and nourished by supra-glacial debris derived from valley rocks (Smiraglia, 1989). The ablation tongue is mostly debris-free, although a recent increase in fine and sparse debris deposition due to the ongoing glacier shrinkage has been reported (Diolaiuti & Smiraglia, 2010; Senese & alii, 2012a). Other considerable morphological changes have lately taken place, including the appearance of debris cones, epiglacial ponds and the formation of neo-moraines. Owing to this diverse range of glacial features, together with the ease of access and the long series of volume, area and length dataset, the Forni Glacier is an ideal ground for evaluating our methodology. METHODS The methodology proposed in this study aims at classifying both large-scale and small-scale features on the surface of the Forni Glacier ablation tongue. We consider as large-scale features the main glacier body and the medial moraines. These are much larger than a Landsat pixel, and

FIG. 1 - (A) Map of the Forni Glacier from aerial orthophoto acquired in 2003: medial moraines and supraglacial debris are indicated in light-grey (1), and nunataks are highlighted in dark-grey (2). (B) Area imaged by the UAV in August 2014: the yellow dot displays the UAV launch and landing location, while the area investigated in higher detail is shown as the red box.

thus clearly identifiable on both Landsat and UAV images. It follows that small-scale features are those smaller than a Landsat pixel, and we considered among them debris-rich ice, snow-patches, epiglacial lakes and crevasses.

DATA SOURCES The UAV used in this study was SwingletCAM, built by SenseFly. SwingletCAM is a fixed-wing UAV, with a weight of only 0.5 kg including GPS receiver, altimeter, wind speed sensor and a Canon IXUS 125 HS digital camera, automatically triggered for picture acquisition. The camera has a 16 megapixel CMOS sensor, capable of 4608 by 3456 pixels resolution with a 4:3 aspect ratio, and captures JPEG format images in the visible light range. The device operates in full-auto mode, including self-adjustment of aperture, ISO and shutter speed for the given light condition. In our survey, this resulted in ISO values in the range 100–500 and shutter speeds of 1/125–1/640 s with an aperture of f/2.7. The survey took place on 28 August 2014, at approximately 8:20 AM. This time was chosen i) to avoid direct illumination of the glacier surface and thus a higher solar reflection which could saturate the pictures, and ii) to minimize wind speed that can cause staggering of the UAV and consequently produce blurred pictures (katabatic winds can occur along the Forni ablation tongue, see Senese & alii, 2012b; Gambelli & alii, 2014). Launch and landing location was set beside Rosole Lake at 2493 m a.s.l. (the yellow

dot in fig. 1b), close to Branca refuge and on the way to the glacier, and the maximum flight height was 250 m. The UAV was flown on the central and eastern parts of the tongue (see fig. 1b), for a total flight time of less than 30 minutes. The upper basins were not investigated due to UAV flight limitations, including battery life and line-of-sight operation and considering that reaching higher elevation in a short time could have entailed significant radial distortion in the acquired pictures. Images were processed via bundle block adjustment, including geometric correction to compensate for radial distortion in Menci APS software, to produce an orthomosaic with a ground resolution of 0.15 m x 0.15 m and a DEM with slightly lower resolution of 0.60 m x 0.60 m, to avoid gaps and artefacts generated by the stereomatching process. In the absence of ground control points (GCPs), we estimated positional accuracy of the orthophoto and DEM to be about 3 m in the x and y coordinates, following Küng & alii (2011). To compare this high-resolution image with a medium-resolution one, we obtained from USGS a Landsat 8 OLI/TIRS image acquired on 14/9/2014. This was the closest date of a Landsat overpass where the ablation tongue was free from cloud cover and shadows. We converted DN values (i.e. quantized and calibrated scaled Digital Numbers) to units of reflectance and brightness temperature using the coefficients found in the scene metadata (USGS, 2013). Geometric accuracy of the Landsat 8 OLI/TIRS sensors is estimated to be approximately 12 m (Storey & alii, 2014). 27

EXTRACTION OF GLACIER LARGE-SCALE FEATURES BASED ON THE UAV ORTHOMOSAIC Concerning separability in the spectral domain, the medial moraines clearly stand out due to overall low values in all RGB channels of the orthomosaic, an effect that is also visible to a lesser degree on the Landsat image, caused by absorption of light at visible wavelengths by meltwater. Other features however are less distinct. In particular, the proglacial stream (i.e. Frodolfo stream) has similar spectral features as the glacier itself, with colors varying from a clear blue to almost saturation of the sensor, depending on water turbidity. Fine and sparse debris located on the tongue and on snow patches is also spectrally similar to proglacial debris. In order to isolate large-scale features from the rest of the scene, we adopted a segmentation approach, identifying connected segments where pixels show similar spectral characteristics. This is in fact the recommended approach when dealing with high-resolution imagery, where objects to be identified are much larger than pixel size (Blaschke, 2010). We adopted different strategies to discriminate between features: the blue channel is ideal for identifying the glacier surface and the red channel for identifying the medial moraines.

EXTRACTION OF GLACIER LARGE-SCALE FEATURES BASED ON LANDSAT IMAGERY To classify large-scale features of the Forni Glacier ablation tongue based on the Landsat OLI/TIRS image acquired on 14 September 2014, we employed optical and thermal information. First, we performed a supervised classification of the whole glacier surface using a stack of Landsat 8 bands, excluding band 1 (used for coastal aerosol detection), 8 (panchromatic) and 9 (cirrus detection), and including both thermal bands (i.e. 10 and 11). In particular, we defined training areas for the classification process by using the UAV image as a reference. Finally, by means of the shapefile created from the UAV orthomosaic, we selected the glacier ablation tongue on the Landsat image. This allowed a direct comparison between the two different approaches for identifying large-scale features: segmentation based on the UAV orthomosaic and supervised classification based on the Landsat OLI/TIRS image.

CLASSIFICATION OF THE GLACIER SMALL-SCALE BASED ON THE UAV ORTHOMOSAIC

FEATURES

Once the glacier area was identified via segmentation, we attempted to further classify the different glacier surfaces by i) segmentation (i.e. snow, exposed ice, debris-rich ice, moraines, and water pond) 28

ii) partitioning into different classes based on RGB values (i.e. debris free or debris covered ice) iii) Gabor filters (i.e. crevasses). Concerning segmentation, given the differences in water color, no single spectral threshold could be used to extract all of the lakes, so they were identified separately. The largest lakes exhibit a greenish hue, likely due to a combination of suspended sediments and populations of algae that have colonized them (see also Boggero & alii, 2014). One example is the epiglacial pond on the eastern tongue, extracted based on the condition green > red > blue, whereas smaller ponds were identified via the blue channel. Snow patches featured different RGB values depending on clean and dirty snow, thus requiring two separate segmentation steps to identify them. Concerning the second approach, an underlying assumption was made that over a glacier surface lower RGB values represent an increasing amount of fine debris deposition (in case of snow or ice), a wet surface (whenever water ponds occur) or shadows. In fact, localized shadow effects produced by surface topography at the time of image acquisition made it difficult to apply a single classification approach over the whole glacier surface. Over large areas, two pixels can feature a similar RGB value even if characterized by different surface conditions, e.g. one can be shadowed and the other one can feature a debris-covered surface. We therefore generated a hillshading map using the DEM extracted from the UAV survey (see section on Data Sources) and solar zenith and azimuth angles (Dozier & Frew, 1990) in order to identify an area with relatively uniform shadowing. We found that a limited part of the eastern tongue revealed these characteristics (red box in fig. 1b). Finally, we performed a decision-tree classification based on spectral properties in the RGB domain. This approach was chosen over supervised classification given the difficulty inherent in selecting homogeneous training areas. Concerning crevasses, previous work on automatic delineation by Johannesson & alii (2011) was based on the local curvature of the ice surface obtained from a LiDAR digital elevation model (DEM). The photogrammetric DEM obtained from the UAV survey however was of insufficient resolution to test this method. In fact, most crevasses on the surface of the Forni Glacier that can be located on the orthomosaic are less than 0.50 m wide, whereas the resolution of the DEM was slightly coarser (0.60 x 0.60 m). Besides, photogrammetric DEM generation is known to be prone to errors in crevassed areas (Barrand & alii, 2009). The orthomosaic was therefore chosen to perform the detection. Our approach was based on Gabor filters, an image analysis technique for texture extraction that emphasizes linear features in an image, with tunable parameters of orientation, wavelength and frequency. In a glaciological context, Gabor filters have been used by Brenning & alii (2012) to map rock

glacier flow patterns. Here we used a bank of filters with 8 different orientations and wavelengths between 5 and 10 m. The analysis was performed using OpenCV implementation of Gabor filters, written in the Python programming language. RESULTS AND DISCUSSION

IDENTIFICATION OF LARGE-SCALE FEATURES Firstly, we identified the two large-scale features: the glacier area and the debris-ice transition zones (i.e. moraines). Based on the UAV orthomosaic, the glacier is easily separated from the rest of the scene (fig. 2) and when compared against visual assessment the approach identifies the terminus correctly, although misclassification occurs at the margins of the medial moraines. On the western part of the glacier tongue, a transition zone from glacier ice to medial moraine to lateral moraine is well visible (see also fig. 2) and the glacier margin is quite difficult to delimit correctly here. Slight overestimation of the glacier boundaries also occurs at the right margin of the eastern ice stream, because of fine debris located below a rock spur outside the glacier. Inclusion of morphometric parameters (such as slope and curvature) from the DEM in the classification process does not significantly improve the results. In fact, a DEM obtained via stereo-matching needs several images from different view angles and this does not occur at the margins. Therefore, the DEM quality decreases significantly along these areas (Küng & alii, 2011). The segmentation process produces gaps in the final glacier and moraine segments due to very different RGB values compared to the neighboring pixels. On the upper parts of the moraines, these are mostly caused by isolated larger dry clasts not affected by meltwater, and thus possessing higher reflectance. Conversely on the actual glacier surface and on the terminal part of the eastern moraine, gaps appear as crevasses or at the debris/ice interface. If the final purpose is to produce a glacier map for inventories, it is recommended that these gaps should be removed via vector analysis. In this study however, they were included in the classification process. Isolated clasts can be considered as part of the medial moraine, and we included them in this category by performing neighborhood analysis, applying a 5x5 pixels neighborhood smoothing filter on unclassified data surrounded by medial moraine pixels. Conversely, gaps surrounded by glacier ice pixels were automatically assigned to the class debris-rich ice. This can leave clasts misclassified if larger than the filter size, and this issue could be solved using another segmentation step or individual correction to refine the overall classification.

To evaluate the advantages provided by the UAV orthoimage, we compared it to the classification based on medium-resolution data acquired by Landsat 8 (fig. 2 and tab. 1). Figure 2 shows the comparison between segmentation applied on UAV orthoimage (dark blue, yellow colours) and supervised classification using a combination of optical and thermal bands applied on Landsat 8 OLI/ TIRS image (cyan, red colours). The higher resolution semi-automatic approach (based on the UAV orthoimage), even if using optical data alone, identifies the glacier surface better than the lower resolution method (based on the Landsat 8 image). In fact, through visual assessment we found that the latter misses the debris-covered sections of the glacier terminus and the terminal part of the western medial moraine. Another means of validating the accuracy of these approaches could consist in sampling reference points on the ground. However, this appears not necessary since the resolution of the UAV image is 0.15 m, much finer than the size of the large-scale features. To better discriminate the medial moraines on the Landsat image, thermal bands could be excluded, because of their coarseness compared to optical bands. This way, however, relatively clean parts of the glacier would be classified as debris-covered. Table 1 reports the areas of glacier features calculated from the final high- and medium-resolution classification maps: the total area of medial moraine pixels is very similar. Conversely, the medium-resolution-based approach underestimates the exposed ice area, as can be seen by looking at the eastern ice stream on Figure 2 and by considering the glacier ice area (Tab. 1). As a result, the total area occupied by the glacier in the Landsat based classification is also slightly lower. To evaluate the area estimation accuracy, we followed the methodology suggested by Vögtle & Shilling (1997) and applied among others by D’Agata & alii (2014), buffering the glacier perimeter to account for the sharpness of glacier limits, using a value of half the pixel size as the linear resolution error (O’Gorman, 1996). The uncertainty in the total area estimation from pixel resolution alone associated with a medium-resolution based classification is significantly larger (12 % compared to 0.002%), giving the limited size of the Forni glacier. Over short time scales, estimation of areal changes using medium-resolution tools like Landsat would thus be subject to an important noise component. A direct comparison between the two approaches is only possible when considering large-scale features, since the small-scale ones, i.e. snow patches, epiglacial lakes and crevasses are smaller than a single pixel on the Landsat scene. Although fine debris is also not visible on the Landsat images, a shift towards lower reflectance values for ice is observable downglacier, especially closer to the terminus. 29

FIG. 2 - (Comparison between areas of bare ice and glacier perimeter derived from UAV-based (dark blue and yellow, respectively) and Landsat 8-based (light blue and red, respectively) classification. The base layer is the UAV image

IDENTIFICATION OF SMALL-SCALE FEATURES BASED ON THE UAV ORTHOMOSAIC

Visual inspection of the entire UAV orthomosaic reveals many features not discernible on lower resolution images, namely: i) snow patches prominent on the upper parts (rela-

tive to the study area) of the eastern and central ice streams; ii) fine and sparse debris patches and debris bands located on the glacier surface; iii) the debris cone on the central part of the tongue; iv) the supraglacial pond located on the eastern part of the tongue and the small lake at the take-off site (Rosole Lake), as well as smaller scattered proglacial and epiglacial ponds; and v) longitudinal and transverse crevasses widespread on the glacier surface, especially the eastern ice stream. The final map produced by the segmentation approach is shown in Figure 3. Slight misclassification errors occurred for dirty snow, as some patches were missed by the segmentation approach because of insufficient contrast with the surrounding terrain. For this reason, snow patches are displayed in purple in Figure 3, not distinguishing between clean and dirty ones, and they cover 25132 m2 (Tab. 3). Unlike snow, the ice surfaces were well detected, and labelled differently whether partially or completely debris-covered (yellow and brown pixels, respectively) or debris-free (cyan pixels). As expected along the ablation tongue in August, the glacier ice (cyan pixels) covers a wider area compared to snow (45.66% and 2.58%, respectively, Tab. 2). Like the medial moraines, the small neo-moraines are correctly identified by the segmentation approach. In addition, the terminal part of the eastern medial moraine results correctly characterized by a mix of supraglacial debris (brown pixels), debris-rich ice (yellow pixels) and epiglacial lakes (blue pixels). In fact, recently the glacier has undergone very intense dynamic processes, compromising surface stability and causing enlargement of crevasses especially along the eastern tongue, and collapse of other areas. An example of the latter is indeed found on the terminal part of the eastern medial moraine (visible as cyan pixels in fig. 3). As an evidence of the darkening phenomena reported by previous studies (Diolaiuti & Smiraglia, 2010; D’Agata & alii, 2014), the supraglacial debris covers a similar area with respect to the glacier ice (i.e. ca. 46%, Tab. 2). Moreover, the large pond located at the margin of the eastern ice stream is correctly identified compared to visual assessment, with an area of 309.23 m2, and the two smaller ponds at the frontal glacier margin are properly recognized as well. The total area of epiglacial lakes covers 841 m2 (blue pixels in fig. 3, see also Tab. 2).

TABLE 1 - Comparison between areas resulting from high- and medium-resolution classification of the Forni Glacier ablation tongue.

Category

UAV-based classification

Landsat 8 -based classification

Area (%)

Area (m )

Area (%)

Area (m2 and %)

512870 ± 0.006%

53.93

496568 ± 24%

52.48

-16302 or – 3.2%

Moraines

448036 ± 0.006%

46.07

449571 ± 24%

47.52

+1535 or + 0.34%

Total

960906 ± 0.002%

100.00

946139 ± 12%

100.00

-14767 or – 1.5%

30

2

Δ area (Landsat 8 -UAV)

Area (m )

Glacier ice

2

FIG. 3 - Segmentation-based classification of UAV orthoimage of the ablation tongue of the Forni Glacier.

TABLE 2 - Area values covered by small-scale features resulting from highresolution classification of the Forni Glacier ablation tongue UAV-based classification Category

Area (m2)

Area (%)

Snow

25132

2.58

Debris-Free Ice

453841

46.66

Debris-Rich Ice

33058

3.40

Moraines

448036

46.07

Lakes

841

0.09

FIG. 4 - (A) Original UAV orthoimage and (B) classified test area for fine debris occurrence, from exposed ice to supraglacial debris (red color scale). Crevasses are also shown in black and the large epiglacial lake in blue. With the numbers 1 to 3 are displayed examples of exposed ice, epiglacial lake and crevasses, respectively.

Finally, we applied the Gabor filter over the whole UAV orthoimage (not shown). The largest number of crevasses is recognized when using a low wavelength value of 5 m, consistent with the spacing of crevasses on the glacier surface. However, a significant number of false positives is also produced, mostly due to shadows projected by larger clasts that appear as linear features. Overall, the approach showed promising results. As suggested by Brenning & alii (2012), further research aimed at improving the classification should combine different methodologies, including other texture measures or terrain parameters (i.e. local curvature) to filter out unwanted shadow effects. A more detailed analysis was performed over an area with relatively uniform shadowing along the eastern ablation tongue (47544 m2, fig. 4). There, in addition to the 31

delineation of water ponds and crevasses, fine debris cover is more deeply analyzed. In this area, debris supply comes mostly from macrogelivation and rock degradation processes (Azzoni & alii, 2014) and a transition from supraglacial debris (i.e. buried-ice) to exposed ice (i.e. bare-ice) is well discerned away from the margin towards the center (lower part of both the images in fig. 4). Here, the darkening phenomena are even more evident compared to the picture showing the entire tongue (fig. 3): the bare-ice pixels are 7.49% of the total excluding water pond and crevasses. The area is also heavily crevassed and here application of the Gabor filter is particularly successful (black pixels in fig. 4). This is a consequence of uniform illumination restricting the range of RGB values, although some sections are still missed because of lack of contrast. As expected the water pond is also well identified. The analysis shown in Figure 4 is a further proof of the intensity of the recent glacier dynamics. In fact, in spite of crevasses covering only 2.79% of the total investigated area, the surface appears very fragmented due to their wide spatial distribution. In addition, the water pond was likely created by a significant collapse. CONCLUSIONS The methodology presented here shows that it is possible to provide high-resolution mapping of a glacier using a semi-automatic approach based on uncalibrated RGB data, obtained with a low cost on-demand UAV survey, with an accuracy comparable or higher than when using mediumresolution satellite data. Large-scale features (i.e. exposed glacier surface and medial moraines) could be easily identified using this approach, with only slight misclassification at the margins of the medial moraines. This suggests that for the purpose of generating glacier inventories a segmentation step based on high-resolution orthoimagery (from UAV or conventional aerial surveys) could speed up delineation of glacier boundaries especially on debris free glaciers. Another significant advantage of a high-resolution approach is the decrease in uncertainty related to pixel size, potentially enabling estimation of areal changes over short periods. Classification of debris-covered glaciers with different lithology and morphology of debris cover remains to be tested, although the capability of UAVs to carry different payloads means that different data sets, including near infrared, thermal and morphometric data could be used to improve accuracy of the results. The inclusion of these data sets could also lead to faster algorithms. In fact, compared with coarser resolution approaches, the methodology we proposed requires considerable user input for threshold selection and manual checking of the segments. An important improvement over medium-resolution based classification is the ability to map smaller geomorphological features, including snow coverage, epiglacial lakes 32

and crevasses. For the latter, a new approach was presented here based on Gabor filters, which showed promising results. Continuous mapping of these features can inform the local communities, as these features can represent a hazard, and give important clues as to the evolution of the glacier and the magnitude of the stresses acting upon the surface. Our findings represent a further proof of the general darkening phenomena occurring at the Alpine debris-free glaciers (Oerlemans & alii, 2009; Diolaiuti & Smiraglia, 2010) and in particular at the Forni Glacier tongue (D’Agata & alii, 2014). Indeed, we found that the bare-ice surfaces cover a similar area compared to the buried-ice ones. In addition, the intensity of recent glacier dynamics and the resulting surface instability is evident in the mapped morphologies such as crevasses (2.79% of the total area investigated in detail) and water ponds. UAV-based remote sensing is still explorative, and at present there is a lack of standard procedures for correct calibration of the data, when they are used to estimate physical quantities (e.g. reflectance) rather than for qualitative assessment such as mapping. As an example, in our research we found the shadowing effect of local topography and geometry of the illumination to be important factors in driving the final RGB values of the orthomosaic. This prevented further classification of fine debris glacier-wide, and although these issues have been tackled for spaceborne sensors, the algorithms will have to be adapted for the platform shift. Another critical issue of UAV-based remote sensing, also noted by Whitehead & alii (2013), is that limitations imposed by battery life and line-of-sight operation currently restrict flight time, so that for instance an entire survey of the Forni Glacier would have required multiple flights. However, this is the largest valley glacier in the Italian Alps, where instead smaller glaciers are predominant. Since small glaciers are more likely to be inaccurately classified when using lower resolution data and have been shown to react strongly to climate change (e.g. D’Agata & alii, 2014; Diolaiuti & alii, 2012), the introduction and improvement of low-cost mapping methods can make an important contribution to Alpine glaciological and geomorphological studies.

REFERENCES AZZONI R.S., SENESE A., ZERBONI A., MAUGERI M., SMIRAGLIA C. & DIOLAIUTI G. (2014) - A novel integrated method to describe dust and fine supraglacial debris and their effects on ice albedo: the case study of Forni Glacier, Italian Alps. The Cryosphere Discuss, 8, 3171-3206 BARRAND N.E., MURRAY T., JAMES T.D., BARR S.L. & MILLS J.P. (2009) - Optimizing photogrammetric DEMs for glacier volume change assessment using laser-scanning derived ground-control points. Journal of Glaciology, 55, 106-116 BENN D.I., BOLCH T., HANDS K., GULLEY J., LUCKMAN A., NICHOLSON L.I., QUINCEY D., THOMPSON S., TOUMI R. & WISEMAN S. (2012) - Response of debris-covered glaciers in the Mount Everest region to recent warm-

ing, and implications for outburst flood hazards. Earth Science Reviews, 114,156-174 BLASCHKE T. (2010) - Object-based image analysis for remote sensing. ISPRS Journal of Photogrammetry and Remote Sensing, 65, 2-16 BOGGERO A., ROGORA M., MUSAZZI S., ZAUPA S., LAMI A., SALERNO F., GUZZELLA L., GAMBELLI S., GUYENNON N., VIVIANO G., THAKURI S. & TARTARI G. (2013) - Un Mondo D’acqua In Alta Quota. Le acque del Parco Nazionale dello Stelvio, un laboratorio a cielo aperto per lo studio dei cambiamenti climatici. Ed. Associazione Comitato Ev-K2CNR, 91 pp. BRENNING A., LONG S. & FIEGUTH P. (2012) - Detecting rock glacier flow structures using Gabor filters and IKONOS imagery. Remote Sensing of Environment, 125, 227-237 D’AGATA C., BOCCHIOLA D., MARAGNO D., SMIRAGLIA C. & DIOLAIUTI G. (2014) - Glacier shrinkage driven by climate change during half a century (1954–2007) in the Ortles-Cevedale group (Stelvio National Park, Lombardy, Italian Alps). Theory of Applied Climatology, 116, 169-190 DIOLAIUTI G., BOCCHIOLA D., D’AGATA C., & SMIRAGLIA C. (2012) - Evidence of climate change impact upon glaciers’ recession within the Italian alps: the case of Lombardy glaciers. Theoretical and Applied Climatology, 109, 429-445 DIOLAIUTI G. & SMIRAGLIA C. (2010) - Changing glaciers in a changing climate: how vanishing geomorphosites have been driving deep changes in mountain landscapes and environments. Géomorphologie : relief, processus, environnement, 2, 131-152 DOZIER J. & FREW J. (1990) - Rapid Calculation of Terrain Parameters For Radiation Modeling From Digital Elevation Data. IEEE Transactions on Geoscience and Remote Sensing, 28, 963-969 EEA (EUROPEAN ENVIRONMENT AGENCY) (2012) - Climate change, impacts and vulnerability in Europe 2012. http://www.eea.europa.eu/publications/climate-impacts-and-vulnerability-2012/at_download/file FISCHER M., HUSS M., BARBOUX C. & HOELZLE M. (2014) - The new Swiss Glacier Inventory SGI2010: relevance of using high-resolution source data in areas dominated by very small glaciers. Arctic, Antarctic and Alpine Research, 46, 933–945 GAMBELLI S., SENESE A., D’AGATA C., SMIRAGLIA C. & DIOLAIUTI G. (2014) – Distribution of the surface Energy budget: preliminary anlysis on the incoming solar radiation. The case study of the Forni Glacier (Italy). Geografia Fisice e Dinamica Quaternaria, 37, 15-22. HODSON A., ANESIO A.M., NG F., WATSON R., QUIRK J., IRVINE-FYNN T., DYE A., CLARK C., MCCLOY P., KOHLER J. & SATTLER B. (2007) - A glacier respires: Quantifying the distribution and respiration CO2 flux of cryoconite across an entire Arctic supraglacial ecosystem. Journal of Geophysical research, 112, 1-9 IMMERZEEL W.W., KRAAIJENBRINK P.D.A., SHEA M. J., SHRESTHA A.B., PELLICCIOTTI F., BIERKENS M.F.P. & DE JONG S.M. (2014) - High-resolution monitoring of Himalayan glacier dynamics using unmanned aerial vehicles. Remote Sensing, 150, 93-103 JOHANNESSON T., BJORNSSON H., PALSSON F., SIGURDSSON O. & ÞORSTEINSSON Þ. (2011) - LiDAR mapping of the Snæfellsjökull ice cap, western Iceland. Jokull, 61, 19-32

KNOLL C. & KERSCHNER H. (2009) - A glacier inventory for South Tyrol, Italy, based on airborne laser-scanner data. Annals of Glaciology, 50, 46-52 KÜNG O., STRECHA C., BEYELER A., ZUFFEREY J.C., FLOREANO D., FUA P. & GERVAIX F. (2011) - The Accuracy Of Automatic Photogrammetric Techniques On Ultra-Light Uav Imagery. 38, International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Conference on Unmanned Aerial Vehicle in Geomatics, Zurich, Switzerland OERLEMANS J., GIESEN R.H. & VAN DEN BROEKE M.R. (2009) - Retreating alpine glaciers: increased melt rates due to accumulation of dust (Vadret da Morteratsch, Switzerland). Journal of Glaciology, 55, 729-736 O’GORMAN L. (1996) - Subpixel precision of straight-edged shapes for registration and measurement. IEEE Transactions on Pattern Analysis and Machine Intelligence, 18, 746-751 PAUL F., BARRAND N.E., BAUMANN S., BERTHIER E., BOLCH T., CASEY K., FREY H., JOSHI S.P., KONOVALOV V., LE BRIS N., MOLG N., NOSENKO G., NUTH C., POPE A., RACOVITEANU A., RASTNER P., RAUP B., SCHARRER K., STEFFEN S. & WINSVOLD S. (2013) - On the accuracy of glacier outlines derived from remote-sensing data. Annals of Glaciology, 54, 171-182 RYAN J.C., HUBBARD A.L., BOX J.E., TODD J., CHRISTOFFERSEN P., CARR J.R., HOLT T.O. & SNOOKE N. (2015) - UAV photogrammetry and structure from motion to assess calving dynamics at Store Glacier, a large outlet draining the Greenland ice sheet. The Cryosphere, 9, 1-11 SENESE A., DIOLAIUTI G., MIHALCEA C. & SMIRAGLIA C. (2012a) – Energy and mass balance of Forni Glacier (Stelvio National Park, Italian Alps) from a 4-year meteorological data record. Arctic, Antarctic, and Alpine Research, 44, 122-134 SENESE A., DIOLAIUTI G., VERZA G.P. & SMIRAGLIA C. (2012b) – Surface energy budget and melt amount for the years 2009 and 2010 at the Forni Glacier (Italian Alps, Lombardy). Geografia Fisica e Dinamica Quaternaria, 35, 69-77 SHUKLA A., ARORA M.K. & GUPTA R. P. (2010) - Synergistic approach for mapping debris-covered glaciers using optical–thermal remote sensing data with inputs from geomorphometric parameters. Remote Sensing of Environment, 114, 1378-1387 SMIRAGLIA C. (1989) - The Medial moraines of Ghiacciaio dei Forni, Valtellina, Italy: morphology and sedimentology. Journal of glaciology, 35, 81-84 STOREY J., CHOATE M. & LEE K. (2014) - Landsat 8 Operational Land Imager On-Orbit Geometric Calibration and Performance. Remote sensing, 6, 11127-11152 USGS (UNITED STATES GEOLOGICAL SURVEY) (2013) - Using the USGS Landsat 8 Product. Available at: http://landsat.usgs.gov/Landsat8_Using_Product.php, Last Accessed 17/5/2013 VÖGTLE T. & SCHILLING K. J. (1999) - Digitizing Maps. In: BÄHR H.-P. & VÖGTLE T. (eds.), «GIS for Environmental Monitoring», Stuttgart, Schweizerbart, 201-216. WHITEHEAD K., MOORMAN B.J. & HUGENHOLTZ C.H. (2013). Brief Communication: Low-cost, on-demand aerial photogrammetry for glaciological measurement. The Cryosphere, 7, 1879-1884 (Ms received 1 March 2015; Accepted 1 June 2015)

33

Geogr. Fis. Dinam. Quat. 38 (2015), 35-40, 8 figg., 1 tab.

DOI 10.4461/GFDQ.2015.38.04

SHIPING JIN (*,**), LIANGSUO SHU (*,**), SY HUANG (*), QINQIN ZHANG (*), ZHILING WU (*) & CHENGBANG HU (*)

CHINA’S ANTI-SEASON ICE CAVES AND THEIR MECHANISM

ABSTRACT: JIN S., SHU L., HUANG S., ZHANG Q., WU Z. & HU C., China’s anti-season ice caves and their mechanism. (IT ISSN 0391-9838, 2015). A dynamic ice cave (DIC) is a natural phenomenon: the ice in the caves freezes in midsummer or late spring rather than in winter. It is created by the unique local geological structure. DICs are important tourism resources; actually, during recent years, they have been developed in varying degrees. Due to tourism exploitation or other unknown factors, some DICs in China are suffering a recession. To reveal the mechanism of DICs’ operation can help with their protection and save them from destructive exploitation. In this paper, the geographical distribution of China’s DICs is given in the first part. Then the influencing factors of the nature ventilation of DIC, the transition mechanism between two seasons of DIC, the interesting relation between the temperatures of warm cave and outside air was analyzed. KEY WORDS: Ice cave, Cave climate, Impact of tourism exploitation, China.

warm caves when the ones with lower altitude blow cold air in winter be called cold caves or ice caves. The cycles of DIC have two different links, cold-cave season and warm-cave season (Byun, 2004). The cold-cave season, usually extending from early April to late September, was named for the cold wind blowing out from cold caves. A warm-cave season locates between two cold-cave seasons. It was named for the characteristic warm air with white steam blowing out from warm caves. In previous works, beside Chimneyeffect, many other theories have been proposed to explain DIC including evaporation effect, radiative cooling and the adiabatic expansion theory, the limitations of which were discussed by Byun (2011). In this paper, after a brief introduction of China’s dynamic ice caves, the influencing factors of the nature ventilation of DIC were analyzed. The transition mechanism, which had almost not been mentioned in the previous references, was also discussed. DISTRIBUTION OF CHINA’S DYNAMIC ICE CAVE

INTRODUCTION Dynamic ice cave was named by Luetscher based on their origin (Luetscher, 2004). A DIC usually has two or more entrances located at different elevation. The ones with higher altitude blow warm air in winter and usually be called ———————— (*) School of Energy and Power Engineering, Huazhong University of Science & Technology, 1037 Luoyu Road, Wuhan, China. (**) Innovation Research Institute, Huazhong University of Science & Technology, 1037 Luoyu Road, Wuhan, China ([email protected]). Many thanks to Prof. Hi-Ryong Byun for the sharing of original observational data of Ice Valley and Prof. Tanaka H.L, and Mrs. Lucy Serody for their selfless help.

China’s DIC distribution is really concentrated: No. 1 to 3 are located in the great Changbai Mountains region, in the northeast of China, No. 4 in the Yan Mountains region and No. 5 in Taihang Mountains region, No. 6 to 9 in a small region beside the Yangtze river; the last one located beside Jinsha River (the name of Yangtze river’s upper reaches, as shown in fig. 1 and their Latitudes and Longitudes were shown in table 1). Lack of space forbids the description of every DIC in detail, so only part of them will be introduced below. Ren ice cave (Li Jinrong, 1992) is located in a slope near a Village Chuanyinggou in the eastern mountainous region of Liaoning Province. This area is about 1,000 meters by 20 meters. Early in the 20th century, a local family named Ren found this strange phenomenon when they built their own house. Taking advan35

tage of this DIC, they constructed a small storage room, which survives to this day. In summer, outdoor temperature is increased by the warm sunshine; however, ground temperature under the slope falls. According to tests by Ren, the temperature in the small storage room goes down to –10°C in hot summer, and the temperature in the rock crevices is much lower, reaching –15°C. From May to early August, rain water flowing into the storage room freezes into icicles; during the cold winter, the storage room is as warm as spring. The family of Ren also assarted two little land areas around the fumaroles spurting hot air on the hill, and then built a greenhouse with plastic film and branches of fraxinus and oak. The vegetables grew very well in the green house when the minimum temperature of outdoor dived to –30°C; what is more surprising is that even the branches of fraxinus and oak sprouted and foliated. Ren Hongfu tested the temperature in the greenhouse repeatedly: the air temperature maintained at 17 °C, and the ground temperature stayed at 15°C. The Chengde Miracle Well (Nie, 2006) is located in a little village named Shuanglin in the Wuling Mountain of Chengde. This region was the royal tombs area of the Qing Dynasty, a forbidden place to common people before the end of the Qing Dynasty in 1911. It is a famous sightseeing district for scenic mountains, dense forests, great canyons and clear streams. Due to the changes in climate and geology or other unknown factors, the summer ice recorded by Li Daoyuan in the valley about 15 centuries before is no longer visible. However, a well dug by the local villagers proves his record to a certain extent. The depth of the dry well is 8.6 m, which is ice cold in summer but warm in winter. Cui Shilin (2011) led a geologic survey of the Chengde Miracle Well on June 20, 2011. They found the well is located in a north-south valley, which is high in the north and opens to the south. Caused by the enormous uplift stress fractures of volcanic activity, the layer of dolomite limestone broke into large or small slates. There are a lot of cracks between the slates, from which cold air flows out in summer and warm air is released in winter. The detection results of acoustical frequency electric showed a high underground porosity in this area. The steady burning of the candle proved that the gas flowing out from the crevice had adequate oxygen. It was –6°C at the well bottom while the outdoor temperature was 30.5°C. Ice Back and Taiji Mountain (Guo, 1985) are both located in Linzhou, Henan Province. I visited Ice back on September 20, 2012. Ice back is located in a slope of Olitic Limestone. The slope is a typical collapse accumulation (in some place, the collapse is still developing). The water from the spring converges into stream. The slope filled with gaps from which cold wind blows every summer. During early summer, the temperature of the cold wind is below zero and water freezes in the gaps. To increase the income from tourism, Ice Back was developed as a beauty spot. The former little cave turned into a deep manmade cave with a door only open when there are some visitors. Such a huge human intervention has a very big risk of weakening the cave. As most tourists 36

FIG. 1 - Geographical distribution of China’a dynamic ice cave.

are attracted by the Great Taihang Gorge and the famous Hongqi Canal nearby, it is not a crowd interest places. Taiji Mountain (Guo, 1985) is located on the side of the Valley of Peach Flower, 10 kilometers south of Ice Back. In Taiji Mountain, there are thousands of little caves, from which cold wind blows during summer. Therefore, all butterflies fly high above the grass to avoid freezing. In winter, no snow covers the areas around the caves even when there is an accumulation of thick snow nearby. THE POWER OF AIR CIRCULATION AND THE TEMPERATURE RESPONDING PHENOMENON The air circulation of a dynamic cave is the result of the «Chimney-effect» (Thury, 1868). The air flowing in the talus is heated or cooled by convection heat transfer with the rocks. This temperature rise (or drop) changes the air density and generates a natural ventilation by breaking the gravity balance of the inside air and outside air. This natural ventilation usually referred to as «Chimney-effect». The natural ventilation pressure (Pn) can be calculated approximately use the fellow equation _ Pn = r g z (1) where = r_= r_0 – r_i is the equivalent difference between the average density of the outside air and the inside air, g is the acceleration of gravity, Dz is the altitude intercept between the entrance and exit. For a DIC, its pipeline characterization can be regarded as invariable. Therefore, the flow velocity (µ) will change with Pn.

1 2 q–u=_ µ +z+( 2

)

(2)

where q is the heat exchange amount of air, Du is its internal energy change. The heat convection between air and rocks is governed by the Newton cooling law, q=h t

warm cave decreased with a very little rate about 1.3°C per month(Byun, 2011). Therefore the temperature of the rocks (Trock) in the talus can be deduced as invariable in the period of one week and Pn is decided by To. The temperature responding phenomenon was first recorded by Tanaka at Nakayama Wind-hole (Tanaka, 2000) and reappeared in the field investigations of Ice Valley (Byun, 2004). During the warm-cave season, if the outside

(3)

where Dt is the temperature between rocks and air, h is convective heat-transfer coefficient and obeys the Dittus-Boelter equation (Winterton, 1998), Nu = 0.023 Re0.8 Prn

(4)

In the warm-cave process n = 0.4; in the cooling process n = 0.3. hl Nu = λ

(5)

µl Re = ν

(6)

ν Pr = a

(7)

FIG. 3 - The temperature of warm cave (Tw), cold cave (Tc) and outside air (To , rappresented by the measured values at mountain top) of Ice Valley from September 2003 to September 2004 (after Byun, 2011).

where Nu is the Nusselt number, Re is the Reynolds number, Pr is the Prandtl number, λ is the coefficient of thermal conductivity, l is the qualitative size (in this case, it is the equivalent diameter of the runner in the rocks), µ is the flowing velocity, ν is the kinematical viscosity, a is the thermal diffusivity of air. The equations from (1) to (7) describe the relation between temperatures of outside air and exit (cold cave or warm cave). It can be summarized by fig. 2. Taking warm-cave season as example, the air sucked from cold cave is heated by the rocks in the talus and blow out from the warm cave. The temperature of outside air (To) and the temperature of the rocks (Trock) are the two most significant affecting factors of Pn. For a Korea DIC, Ice Valley in this season, the temperature of cold cave change dramatically with when that of the outside air the temperature of

FIG. 4 - The diagram of summer mountain-valley breeze and natural ventilation in the talus: the left one is the valley breeze during the day, the right one is the mountain breeze at night; the dashed line represent the natural ventilation in the talus.

FIG. 2 - The relation schema between temperature of outside air and exit (cold cave or warm cave)

air temperature rises, the temperature of wind from warm cave drops. This drop is big when the outer air temperature is above zero and small when it is below zero (as shown in fig. 3, fig. 5 and fig. 7). This temperature responding phenomenon can be explained using the relation schema in fig. 1 with equation (1) to equation (7). During the warmcave season, Pn reaches higher in colder weather (equation (1)). As the gas velocity µ increases with Pn (equation (2)) 37

The transition mechanism, which has almost not been mentioned in the previous references, is very important to uncover the secret of DIC. The transition mechanism includes the influence of intrinsic factors and extrinsic factors. At the end of the cold-cave season, the rocks of the talus lose most of their cooling potential and the temperature of outside air is much lower than that at the height of summer. These intrinsic factors result in a weakening of Pn, which has a marked effect on the cold-cave season. Under the in-

fluence of intrinsic factors, the cold-cave season reaches an unstable critical state gradually. Then, the extrinsic factors play their roles in state transition at a vital moment. A similar situation takes place at the end of the warm-cave season. Large temperature changes, the mountain-valley breeze and rainfall are possible extrinsic factors that may play significant roles in the transition. Mountain-valley breeze is a localized climate created by the heat budget of valley air: during the day, the valley air is heated by the sun and rises up, causing a warm, upslope valley wind; at night, mountain air cools rapidly and flows down, causing a cold, downslope mountain wind (Christopherson, 1992). From fig. 4, we can see the effect of mountain-valley breezes on the cycle of DICs in summer. During the day, the valley breeze is contrary to the direction of the natural ventilation in the talus, so it impedes the suction and discharge of air, weakening the natural ventilation. At night, the mountain breeze is in the same direction as the natural ventilation in the talus, so it makes it easy for air to flow into the talus and has an injection action on the outflow of air, enhancing the natural ventilation. As the temperature difference between the outside air and the talus rock during the day is larger than at night, so from day to night, the natural ventilation is declining. Therefore, the mountain-valley breeze helps the cycle to run more smoothly by reducing the peak and filling the valley. In summer, the air blowing from the cold cave is cold and heavy; therefore it drops below the outside air to form a stable layered structure, which can protect itself from the disturbance of the valley breeze (Tanaka, 2000). However, when autumn comes, with decreasing temperature of outside air and weakening cooling potential, the air blowing from the cold cave is gradually warmed up. The temperature gradient of layered structure declines subsequently, which destroys the stability of the layered structure. The weakened layered structure can no longer protect itself from the disturbance of the valley breeze, and then the critical moment of transition comes. After the up-slope valley breeze turns the cold-cave season to the warm-cave season, the following down-slope mountain breeze may reverse this action. Because of the seesaw battle between the valley breeze and the mountain breeze, there will be perturbations and oscillation, or even regressions to the former cycle more than

FIG. 5 - The transition from cold-cave season to warm-cave season, typical temperature opposite of warm-cave season in ellipses and temperature follow of cold-cave season in rectangle. (after Byu, 2011)

FIG. 6 - The diagram of winter mountain-valley breeze: the left one is valley breeze during the day, the right one is mountain breeze at night; the dashed line represent the natural ventilation in the talus.

TABLE 1 - China’s DICs from high latitude to low No.

Name

Latitude and Longitude

1

White Mountain ice cave

41° 50’N, 126° 20’E

2

Ren ice cave

41° 0’N, 125° 29’E

3

Qianshan ice cave

40° 59’N, 123° 7’E

4

Chengde Miracle Well

40° 21’N, 117° 30’E

5

Ice Back and Taiji Mountain

36° 13’N, 113° 43’E

6

Shennongjia ice cave

31° 38’N, 110° 34’E

7

Three Gorges ice cave

31° 31’N, 109° 04’E

8

Zigui ice cave

31° 5’N, 110° 31’E

9

Wufeng Baiyi ice cave

30° 15’N, 110° 33’E

10

Yan Mountain ice cave

27° 27’N, 103° 13’E

and the heat transfers by convection will be better under a bigger (equation (3)-equation (7)), so higher Pn means warmer air spurting out from the warm cave. However once the outside air is cold enough to ensure the air flowing in the talus is fully-heated to the temperature of the rocks, a colder outside air can result in a higher Pn, a larger µ but is helplessness to increase Tw (or to decrease Tc during coldcave season). This has been proved by Tanaka and Mizuho’s observational study of summertime ice at the Nakayama windhole (Tanaka, 2000). On 30 June 1998, the temperature of outside air reached a maximum at 12:00, and the wind velocity reached its maximum at 16:00, however Tc was almost unchanged all the day. THE TRANSITION MECHANISMS BETWEEN COLD-CAVE SEASONS AND WARM-CAVE SEASONS OF DICS

38

FIG. 7 - The successful transition from warm-cave season to coldcave season, typical temperature opposite of warm-cave season in ellipses and temperature follow of cold-cave season in rectangle. The right one is the detail with enlarged scale from 20:00 of April 9 to 23:00 of April 10 (after Byun, 2011).

once before the new cycle is strong enough to maintain its own stability. From September 15 to 17, the temperature of warm cave changed with the outside air but with a much smaller changing scope, which is a typical characteristic of a DIC during warm-cave season. Then a large temperature drop triggered the transition. However, there was no typical characteristic of warm-cave season until September 30. The typical characteristic includes the temperature opposite between warm-cave and outside air and the nearly immutability of the temperature of warm cave when the outside air is cold enough. It took two weeks for Ice Valley to finish the transition from cold-cave season to warm-cave season (as shown in fig. 5). In winter, as the direction of natural ventilation changes, the roles of valley breeze and mountain breeze also reverse: the valley breeze enhances the natural ventilation and the mountain breeze weakens it (as shown in fig. 6). When the arm-cave season reaches its unstable critical state under the influence of intrinsic factors, the mountain breeze plays an important role in the transition. The situation is similar to the transition from the cold-cave season to the warmcave season. The transition at February 19 was reversed 3 days later by a dramatic temperature drop of 15°C in 24 hours (as shown in fig. 8). Both the two transition from warm-cave season to cold-cave season took place in night, when the nature ventilation can be weaken by a reverse mountain breeze.

season (or a warm-cave season), the continual decreasing temperature difference between outside air and talus rocks results in a recession of natural ventilation and makes DIC reach an unstable critical state. Two possible extrinsic factors large temperature changes, the mountain-valley breeze was analyzed and compared with the observed temperature of Ice Valley by Byun (2011). The mechanism of DICs can be used in the buildings of mountain areas. This kind of building can provide air conditioning in summer by taking advantage of unique local geology and climate. However, to describe the cycle of a DIC quantitatively, more in situ observations need to be carried out.

DISCUSSION AND CONCLUSION In this work, we give an explanation of the interesting emperature responding phenomenon of DIC during warmcave season after an comprehensive analysis of the interaction of To, Tw and Pn. Before the air fully-heated to the temperature of the rocks during warm-cave season, a colder outside air means a higher Pn, a larger µ, and a warmer air spurting out from the warm cave. However, once the air has been fully-heated, the air spurting out will keep unchanged. The transition mechanism, which has almost not been mentioned in the previous references, is also discussed. The transition mechanism includes the influence of intrinsic factors and extrinsic factors. At the end of a cold-cave

FIG. 8 - A failure transition from warm-cave season to cold-cave season (after Byun, 2011)

39

REFERENCES

LI JINRONG (1992) - Attractions of Henren. Liangning: local history office of Henren Country. 183-185.

BYUN H.R., Choi K.S., Kim K.H. & Tanaka H.L. (2004) - The characteristics and thermal mechanism of the warm wind hole found at the IceValley in Mt Jae-Yak. Journal of Korean Meteorological Society, 40:453-465.

LUETSCHER M. & JEANNIN P.Y. (2004) - A process-based classification of alpine ice caves. Theoretical and Applied Karstology, 17(5), 5-10.

BYUN H.R, Tanaka H.L., POM-yong Choi & Do-Woo Kim (2011) - Seasonal reversal at Miryang Eoreumgol (Ice Valley), Korea: observation and monitoring. Theoretical and Applied Climatology, 106, 403-415.

TANAKA H.L, Mizuho Yokoi & Nohara Daisuke (2000) - Observational study of Summertime ice at the Nakayama wind-hole in Shimogo, Fukushima. Science Reports, Institute of Geoscience, University of Tsukuba, Section A (Geographical Sciences), 21, 1-21.

CHRISTOPHERSON R.W. (1992) - Geosystems: An Introduction to PhysicalGeography. Macmillan Publishing Company, 155 pp. CUI (2011) - http://kejiao.cntv.cn/C30488/classpage/video/20110821 / 100574.shtml, 2011-8-30/2011-10-1. GUO ZIMING (Ed.) (1985) - The historical of Li Country. Anyang: Historical Committee of Li Country. 110-115.

40

NIE SHUFENG & GAO JIANHUA (Eds.) (2006) - Anecdotes of Hebei. Tourism Education Press, Beijing, 138 pp.

THURY M. (1861) - Etude des Glacières naturelles. Archives des Sciences de la Bibliothèque Universelle, Genève, 1-59. WINTERTON R.H.S. (1998) - Where did the Dittus and Boelter equation come from? International Journal of Heat and Mass Transfer, 41(4), 809-810. (Ms. received 31 January 2013; accepted 15 January 2014)

DOI 10.4461/GFDQ.2015.38.05

Geogr. Fis. Dinam. Quat. 38 (2015), 41-53, 7 figg., 7 tab.

MARCO PERESANI(*) & CRISTIANO NICOSIA (**,***)

COMPARATIVE STUDY OF TWO LATE PLEISTOCENE SEQUENCES WITH PALEOSOLS AND AEOLIAN DEPOSITS AT THE SOUTHERN ALPINE FORELAND

ABSTRACT: PERESANI M. & NICOSIA C., Comparative study of two Late Pleistocene sequences with paleosols and aeolian deposits at the Southern Alpine foreland. (IT ISSN 0391-9838, 2015). Two pedostratigraphic sequences located between the Euganean and Berici hills (Veneto region, north-eastern Italy) were investigated. In such sequences, morphogenic and pedogenic processes could be ascribed to the Late Pleistocene climatic evolution, based on pedostratigraphic characteristics and on archaeological finds. The series prove the degradation of the vegetation cover along the local slopes, followed by the truncation of paleosols due to widespread hillwash phenomena. The latter result in the concentration of coarser elements along an erosional surface, with successive displacement due to gelifluction. This process indicates climatic change towards glacial conditions, with an open environment that is characterized also by loess sedimentation. KEYWORDS: Paleosols, Loess, Late Pleistocene, Berici Hills, Euganean Hills, Northern Italy. RIASSUNTO: PERESANI M. & NICOSIA C., Studio comparative di due sequenze del Pleistocene Superiore con paleosuoli e depositi eolici nell’avanpaese delle Alpi meridionali. (IT ISSN 0391-9838, 2015). Due serie pedostratigrafiche sono state analizzate nell’area posta tra le pendici occidentali dei Colli Euganei ed i Colli Berici (Veneto). Le due sequenze documentano eventi morfogenetici e pedogenetici che, sulla base delle caratteristiche pedo-stratigrafiche e del contenuto archeologico, possono essere inquadrate nell’evoluzione climatica del Pleistocene Superiore. Le serie documentano la degradazione delle coperture ———————— (*) Università di Ferrara, Dipartimento di Studi Umanistici, Sezione di Scienze Preistoriche e Antropologiche, Corso Ercole I d’Este 32, I-44100 Ferrara (Italy); [email protected] (**) Université Libre de Bruxelles, Centre de Recherches en Archéologie et Patrimoine CP 175 - 50, avenue F.D. Roosevelt, B-1050 Bruxelles (Belgium); [email protected] (***) Geoarchaeology & Soil Micromorphology Consultant, Via Cilento 10, I-36100 Vicenza (Italy) The Soprintendenza per i Beni Archeologici del Veneto is thanked for authorizing the excavations at Monte Versa. The Rotary Club Abano Terme Montegrotto provided support for this study. The authors are grateful to two anonymous reviewers for constructive suggestions and to E.B. Modrall for revision of the manuscript.

vegetazionali dei versanti, la conseguente troncatura dei paleosuoli per dilavamento diffuso ed accumulo di elementi grossolani sulla superficie di erosione, ulteriori processi di geliflusso che delineano una progressiva degradazione del clima in senso glaciale. Gli effetti di quest’ultima sulla riduzione della copertura vegetale si fanno sempre più evidenti fino a raggiungere le condizioni di ambiente aperto correlate anche alla deposizione del loess sui rilievi calcarei. TERMINI CHIAVE: Paleosuoli, Loess, Pleistocene Superiore, Colli Berici, Colli Euganei, Italia settentrionale.

INTRODUCTION In the Po plain and in the bordering hill complexes, different archives provide data on Middle and Late Pleistocene climate evolution (Cremaschi, 1987; Amorosi & alii, 2004; Massari & alii, 2004; Fontana & alii, 2014; Ferraro, 2009; Ridente & alii, 2009; Scardia & alii, 2010; Ravazzi & alii, 2012), occasionally with millennial-scale resolution (Monegato & alii, 2007; Pini & alii, 2010). Notwithstanding an increasing amount of contexts studied between the Alpine fringe and the northern Adriatic Sea (Fontana 2006; Fontana & alii, 2008; Pini & alii, 2009; 2010), some areas remain marginally investigated. Among these, we can include the Euganean and Berici hills, two isolated hill complexes that emerge from the Quaternary alluvial sediments deposited by the Bacchiglione, Brenta, Adige and other minor rivers (Zangheri, 1988-89; Mozzi, 2005; Fontana & alii, 2008; Monegato & alii, 2011; Fontana & alii, 2014). Especially in the Euganean hills, the effects of Late Pleistocene climate variability on landforms and slope evolution are still poorly known. These are in fact generally inferred on the basis of supra-regional evidences (Piccoli & alii, 1981), gathered in areas that were not directly affected by glacial advance or by fluvial morphogenesis, such as the central-oriental Alpine fringe. Sedimentary and palaeopedological slope sequences can be of key importance for the definition of the 41

palaeo-environmental and climatic changes that took place in the Mediterranean area between the Middle and Late Pleistocene (Busacca & Cremaschi, 1998). Here, the main morphogenic processes include aeolian sedimentation, deposition of colluvium, solifluction, cryoclastic phenomena in caves and rock shelters, as well as pedogenesis (see Ferrarese & Sauro, 2005; Sauro, 2002; Cremaschi, 1987; 1990a; 1990b; 1990c). Aeolian dust sedimentation has been traced over the Northern Adriatic region until southern Dalmatia (Cremaschi, 1990a), including the margin of the Po plain and the pre-Alpine plateaus, where loess blankets thinner

than one meter can be found. These are characterized by a marked granulometric sorting, by trends reflecting the distance from deflation areas, by the dominance of quartz as well as micas, and by mineralogical compositions reflecting those of the sediments of the main Po plain rivers. A new contribution to this scenario is provided by the two pedostratigraphic sequences presented in this paper. The exposures ar quence, some of the processes will be also put in relation to an archaeological record of the Middle Palaeolithic (see Peresani, 2000-2001). PHYSICAL SETTING OF THE EUGANEAN HILLS AND SURROUNDING AREAS

FIG. 1 - Position of the studied profiles between the Euganean and the Berici Hills (Based on: Regione del Veneto, 1990; Piccoli & alii, 1981; Mozzi, 2005)

42

The Euganean hills, and the ridges extending westward towards the foothill of the Berici, are isolated hills emerging from the floodplain with an area of more than 100 km². They occur as a rough ellipse with a N-S major axis, a main central nucleus and isolated outcrops located to its W and E (fig. 1). They include three main morphological domains. First, a group of lowlands located around the hill complex (Galzignano formation; see Cucato & alii, 2011). These occur 3 to 5 m below the surrounding floodplain and host palustrine environments. Their formation dates to the Last Glacial Maximum (LGM), when the Brenta megafan reached its maximum extension (Mozzi, 2005; Fontana & alii, 2008). The remaining portion of the floodplain, south of the Euganean and Berici hills, is the outcome of the Adige river activity. Sedimentary accretion led to the burial of the Euganean foothill and to the damming of many Euganean valleys. The alluvial sedimentation is also responsible for the rugged outline of the contact between the foothill and the plain and for the occurrence of scattered isolated hills or groups of hills. The second morphological domain occurs between 100 m and 200 m a.s.l., where calcareous marls and marls prevail. These give rise to slightly undulating topographic profiles, with smooth ridges and short plateaus. This domain comprises slightly dipping surfaces occurring at different heights, interpreted originally as the remains of ancient piedmont surfaces (Schlarb, 1961), often difficult to identify on the ground (Donà, 1964). Their origin is nevertheless linked to structural or lithologic causes, but a review of their formation model appears still necessary (see Piccoli & alii, 1981; Zanferrari & alii, 1980). The existence of surfaces unrelated to structural control or to selective erosion is in fact well established. These occur on volcanic bedrocks, colluvia, strongly weathered stream deposits but also on thick blankets of alterites formed on acid volcanic bedrocks. The degree of weathering suggests a Middle or even Early Pleistocene age (Cucato & alii, 2011). The third morphological domain occurs in the highest areas (400-600 m a.s.l.), and comprises basic and acid volcanic rocks formed as a consequence of two distinct volcanic cycles in the Upper Eocene and in the Lower Oligocene (Piccoli & alii, 1981). Here, the relief is characterized by marked topographic gradients, and conical and pyramidal morphologies are dominant.

Limited outcrops of Upper Cretaceous limestone occur along the foot of the eastern scarp of the Berici hills and in small isolated hills between the Berici and Euganean hills, among which the hills around the town of Albettone, where one of the studied profiles is located (Monte San Giorgio – fig. 1). These hills have an elevation of 70-100 m a.s.l. and are the relicts of a planation surface, possibly of Miocene age (Sauro, 2002). These upper Cretaceous limestones were deposited in a deep-sea environment and occur in lenticular stratified sets. They are termed “Scaglia Rossa” formation (Mietto, 1988; Cucato & alii, 2011). The local landscape appears therefore rather articulated and is characterized by abrupt passages from one morphological style to the next. Beside slope processes, which contributed to the degradation of relief, the role of surface water erosion and mass wasting phenomena must be also noted. These processes led to the formation of gentle and small fans at the mouth of the valleys. No surface karst features are present. PRESENT-DAY CLIMATE The meteorological station for the studied portion of the Euganean hills is located on Monte Venda (580 m a.s.l.; data series from 1923 to 1962 from ESAV, 1996). Here, the average annual temperature is 10.8 °C, with July being on average the warmest month (20.2 °C) and January the coldest (1.4 °C). Average rainfall is between 700 and 900 mm/ yr, with absolute maximum values between 1200 and 1400 mm (spring and autumn) and minima between 450 and 600 mm (winter and summer). The annual average potential evapotranspiration in these higher parts of the hills is 668 mm, with an average excess of precipitation of ca. 200 mm. Climate data for Albettone (24 m a.s.l) are derived from Veneto Agricoltura (2010) and are based on 1976-2005 data series. Mean annual temperature is 13.0 °C, with July being on average the warmest month (23.6° C) and January the coldest (2.4° C). Average rainfall is 925 mm/yr with maxima in spring and autumn and minima in winter and summer. Potential evapotranspiration amounts to 950-1000 mm/yr (slight water deficit). According to the method of Thornthwaite & Mather (1957), in the study area the climate is humid (see also ARPAV 2005). For the classification system of Soil Taxonomy (Soil Survey Staff, 2010) the soil temperature regime is mesic with a predominant udic soil moisture regime. MATERIALS AND METHODS

THE STUDIED SEQUENCES Monte Versa Monte Versa is a calcareous hill that makes up the SW end of the Monte Venda – Monte Vendevolo ridge. The hill is made up by calcareous marls of the Maiolica, Scaglia Variegata Alpina, and Scaglia Rossa formations (see Cucato &

alii, 2011), all of which exhibit planar parallel stratification and occur in decimetre-thick units. The Maiolica formation outcrops in the lower portion of the SW hillside, with a plunging angle of about 10° and NW-SE direction. The Scaglia Variegata Alpina and Scaglia Rossa formations outcrop in the middle and upper portion of the hillside and make up the remaining part of the hill up to the Monte Vendevolo foothill to the north-east. Here, a belt of Euganean marls, tuffs and basaltic hyaloclastites outcrops (fig. 1). The hilltop (136 m a.s.l.) is smooth and rounded. Beyond it, the hill is markedly asymmetrical with respect to the NESW axis, with the SE side being steeper than the NW one. The first break in slope is where the Monte Versa sequence is located. The foot of the hill is buried by a silty-clay deposit which gets thicker until as it reaches the alluvial plain. In the upper portion of the hillside, an outcrop of ca. 1 m of thickness with slight planar parallel stratification can be observed. This is composed by an alternating sequence of gravels and silty gravels, varying from light brown (7.5YR 6/4) to light yellowish brown (10YR 6/4). The skeletal grains comprise subangular to subrounded gravels of Scaglia Rossa with frequent angular red chert fragments. Occasional flattened clasts of Scaglia Rossa show traces of secondary fracturing. Gravels are in general dominant on the fine fraction, even in the clast-supported silty gravel layers. They show variations in size, however, never exceed 5 cm. Layers are cm-thick and locally cemented, giving rise to veins oriented parallel to the topographic surface above. The overall thickness of these deposits is not known, as the contact with the underlying bedrock was not reached. The carbonatic matrix contains a subordinate amount of unweathered loess. This sequence can be compared with other deposits observed on the Euganean hills and interpreted as the result of colluvia and sheet flows which incorporated the weathering mantle formed on Scaglia Rossa and pre-existing soils, with intercalated solifuction deposits. The cemented horizons can be associated with the calcic horizons of a well-developed soil, now truncated, covered by more recent colluvia (Po synthem – see Cucato & alii, 2011). The pedo-sedimentary sequence that is analyzed in detail below rests on top of these deposits (defined horizon 4Cck). Albettone Similarly to Monte Versa, also Monte San Giorgio is a hill (101 m a.s.l.) formed on the calcareous marl substrate of the Scaglia Rossa formation, with plano-parallel stratifications and decimetre-thick beds (see fig. 1). The studied sequence was exposed along the SE hillside, characterized by a slope angle of ca. 10°. The foothill is linked to the alluvial plain by a colluvial cone. The south-eastern foot of the Monte San Giorgio hill is articulated in a series of ridges and smoothed valley that descend towards the floodplain.

LABORATORY METHODS The field description of profiles was carried out according to the guidelines of FAO (2006). Munsell colours were determined on the field on wet soil. 43

Sands (2000 mm - 63 mm) were quantified by wet sieving, while silts and clay by hydrometric analyses. Chemical analyses were performed on the 2mm

Pedofeatures7 Fe nod.

Mn nod.

Calc. inf.

Monte Versa Ap

**

*** m/s AB

SiCl

*

5%

7.5YR strong brown

cs, ps, gs

30/70

dsp

5-10%

*

2Bt1

**

***

s AB/ SB

SiCl

*

5%

7.5YR strong brown

cs, gs, ps

35/65

dsp

2-3.5%

**

2Bt2

*

***

s AB/ SB

SiCl

*

3.5-5%

7.5YR strong brown

cs, gs, ps

25/75

op

3.5-5%

**

3Bts1

*

*

*** s P/ AB

Cl/SiCl

**

1-2%

5YR yellowish red cs, ps, gs

20/80

op

5%

***

3Bts2

*

*

***

Cl/SiCl

*

66% bare soil

5

Infrastructure

Buildings and paved roads

Soil characteristics The calculation of the runoff volumes for single precipitation events is based on relevant parameters measured in the field and in the laboratory or estimated using remote sensing techniques. We measured soil texture for topsoil at 61 locations with a simple finger test method, according to 58

by EcoTech, Germany) that measures 6 parameters: wind speed and direction, precipitation, air temperature, relative humidity as well as air pressure. Data are measured with high temporal resolution of 10 min intervals and have been collected from March 2010 up to present.

where Q is steady-state flow; a is radius of the infiltration area; ku is hydraulic conductivity as a function of water tension; and h is water tension. Accordingly, hydraulic conductivities for different tensions can be calculated using Equations 4 and 5, after Umwelt-Geräte-Technik (2005):

Soil infiltration

k(h1) =

With the hood infiltrometer (Umwelt-Geräte-Technik GmbH, Germany), which is a type of tension infiltrometer (Schwärzel & Punzel, 2007), water infiltrability is determined for the soil surface. The unsaturated conductivity was calculated following Wooding (1968) (Equation 2) and (Umwelt-Geräte-Technik, 2005) (Equation 3): Q= a2ku (1+ 4

a

)

( )

ln Q1 = Q2 h1− h2

( k(h2) =

Q2 a2 1+ 4

(

(2)

(3)

Q1 a2 1+ 4

)

(4)

)

(5)

The capillary theory can be used to estimate the size of pores excluded from the transmission of infiltration at different pressure heads. Pore diameters can be predicted from Equation 6 (Sauer & Logsdon, 2002):

TABLE 2 - Soil Properties of the Soil Profilesa. Soil Depth

Particle size composition % Sand

Silt

Class USDA

Corg (%)

Munsell soil colour

5 YR 3/1

Clay 03 Haplic Vertisol

10 cm

42·3

36·8

20·9

L

0·77

20 cm

35·0

28·7

36·3

CL

0·94

5 YR 2.5/1

30 cm

34·7

25·3

40·0

CL

1·23

7.5 YR 2.5/1

40 cm

35·1

24·2

40·7

C

1·03

5 YR 2.5/1

50 cm

31·2

27·3

41·5

C

0·93

7.5 YR 2.5/1

60 cm

31·1

24·5

44·4

C

1·05

7.5 YR 2.5/1

10 cm

41·2

39·1

19·7

L

1·41

7.5 YR 4/3

20 cm

39·5

30·0

30·5

CL

1·27

7.5 YR 2.5/3

40 cm

37·7

26·0

36·3

CL

0·81

2.5 YR 2.5/2

10 cm

54·3

34·1

11·6

SL

1·32

10 YR 3/4

20 cm

53·1

35·5

11·4

SL

0·95

7.5 YR 4/6

30 cm

51·4

36·4

12·2

L

0·90

5 YR 4/4

40 cm

51·5

37·3

11·2

L

0·65

5 YR 4/6

50 cm

47·3

37·0

15·8

L

0·72

5 YR 4/6

60 cm

46·0

37·2

16·8

L

0·64

7.5 YR 4/6

04 Rhodic Luvisol

05 Rhodic Luvisol

06 Andic Cambisol 10 cm

67·6

20·9

11·5

SL

20·5

7.5 YR 3/4

40 cm

57·7

24·2

18·1

SL

11·8

7.5 YR 4/6

a

Soil classes after USDA (1993): L loam, CL clay loam, C clay, SL sandy loam; Soil colour under dry conditions after Munsell Soil Color Charts.

59

(6) where is the surface tension of water, N/m, assumed to be 0.072 at 25°C; is the contact angle between water and pore wall, assumed = 0°, cos(0) = 1; ρ is the density of water, kg/m³; and g is the gravitational acceleration, 9.8 m/s². Therefore, a higher suction or tension corresponds to an infiltration where the bigger pore sizes in the soil matrix are successively excluded from water transmission. The water tensions of -2 cm and -4 cm equal to pore sizes of 1.47 mm and 0.73 mm, respectively. In general, we assumed that soil pores are represented to be cylindrical, vertical tubes. Generally, the pressure 0 cm water column and two suctions have been measured at 40 locations.

SPATIAL MODELLING OF SOIL HYDROLOGICAL CHARACTERISTICS To get a spatially continuous data set of infiltration values and soil texture we utilized a stochastic modelling approach. In this study, a boosted regression tree approach (BRT) (Friedman, 2002) is applied using dependent and independent variables. BRTs employ a learning algorithm to identify a model that best fits the relationship between the predictor variables (an attribute set of, in this case, environmental variables) and the response variables, here the soil texture and infiltrability. The dependent variable is the soil texture measured for 61 locations within the study area, analysed with finger test (according to USDA (1993)). As independent variables the delineated landcover from multispectral WorldView-2 data as well as four terrain indices have been selected: (i) land cover classes, (ii) topographic wetness index (TWI) (after Beven & Kirkby, 1979), (iii) vertical distance to channel network, (iv) channel network base level (both after Olaya & Conrad, 2009), and (v) transport capacity (TCI) (Moore & alii, 1991). The indices ii to v are based on the derived hydrological correct DEM (see Moore & alii, 1993). We utilized the TreeNet model (Salford Systems) also known as stochastic gradient boosting (Elith & Leathwick, 2009). The models’ predictive performance is assessed by constructing the receiver operating characteristics (ROC) curves for each response variable, both for training and test data (Fielding & Bell, 1997). In a ROC curve the sensitivity

is plotted over the false positive rate (1-Specificity) for all possible cut-off points (Swets, 1988). The quality of a ROC curve is quantified by the measurement of the parameter area under the ROC curve (AUC; Hanley & McNeil, 1982). The AUC is shown to be independent of prevalence (Manel & alii, 2001) and is considered a highly effective measure for the performance of ordinal score models. A perfect discrimination between positives and negatives has a ROC plot that passes through the upper left corner (100% sensitivity, 100% specificity), so that the AUC is equal to 1 (Reineking & Schröder, 2006). According to (Hosmer & alii, 2013) AUC values exceeding 0.7/0.8/0.9 indicate acceptable/excellent/outstanding predictions.

AUTOMATIC SURFACE RUNOFF DETECTORS Surface runoff was measured with automatic surface runoff detectors (SRD) that measure surface runoff height and duration. SRDs were distributed along the slope system to account for the runoff generation dynamics on different slope segments and thus, topographic positions. The SRD devices are placed on the surface whilst the logger unit is buried in the soil (fig. 4). When runoff height grows the electric contacts are closed sequentially transmitting a signal that is then reported to the logger storage unit. In this experimental setup, we placed logger units according to Figure 4b right side with equal spacing of contacts. The distances between the electric contacts are 5 mm, reaching up to 3 cm above ground. Hence, the device is able to measure a water column of 3cm and has a sensitivity of 5mm. The advantage of our study area and of semiarid environments in general is the sparse vegetation cover during dry periods, and hence the disturbance of runoff due to plants is minimized. Moreover, in most of the cases precipitation is moderate to low, so runoff is normally not accumulating deeper than the contact zone of the SRDs.

RUNOFF CALCULATION To calculate a simple water balance model, the relevant input parameters such as i) the precipitation input, ii) the evapotranspiration, iii) the infiltration rate, and iv) the resulting runoff hast to be measured (tab. 3). The measured infiltration values are further attached to the regionalized model of the surface texture. This cre-

b

FIG. 4 - a) Close up of the sensor head of the Surface Runoff Detector. b) Installed device on the Manyara Ranch Conservancy test slope with the drainage coming from the right, shown with the blue arrow. (Photos: G. Quénéhervé).

60

TABLE 3 - Rainfall events, potential evapotranspiration, and calculated runoff characteristics. Event Day of Rainfall Rainfall Average number year duration depth rainfall [min] [mm] intensity [mm/h]

PETTurc [mm/ event]a

Rainfall Infiltration, depth (P) - tension h02 PET [mm] [mm] 26.85

Qh0 = P - PET - I [mm]b

Infiltration, Qh2 = P Infiltration, tension h2 - PET - I tension [mm] [mm] h2max [mm]

22.87

20.59

2.28

Qh2max = P - PET I [mm]

1

283

80

54.61

40.96

4.89

49.72

14.19

35.53

2

296

260

21.46

4.95

7.16

14.30

~

~

0.11

3

232 + 233

310

10.38

2.01

-

10.38

~

~

~

4

335

150

27.99

11.20

8.10

19.89

~

~

5.70

5

340

280

16.49

3.53

-

16.49

~

~

2.30

a

PET: - indicates a night event, evapotranspiration is set to 0; bQ: ~ no runoff, all P is infiltrated into the soil column

ates a raster model with infiltration values as attributes. A single-cell raster based water balances model (Equation 7) was then utilized for a weighted flow algorithm based on a deterministic infinity flow approach following (Tarboton, 1997), the GIS modelling was done with SAGA GIS (Conrad, 2006). For the sake of simplicity the runoff calculation do not take into account surface roughness parameters. R = P−I where

is water balance and

(7) is precipitation.

RESULTS

PRECIPITATION ANALYSIS For this study we focus on the period October to December 2010, where 5 rainfall events occurred, which produced substantial runoff that was captured by the surface runoff detectors. Table 4 shows the mean daily meteorologi-

cal data for each day of the recorded events and Table 5 the soil hydraulic conditions. DGPS AND REMOTE SENSING DATA PROCESSING For the evaluation of DGPS measurements, a regular 1 m point grid was computed for the study area (ca. 440,000 points). For each point of the produced digital elevation model (the centroid of the 1 x 1 m grid) the distance to the closest DGPS measurement was calculated. The average distance to a reference point is 5.43 m with a standard deviation of 4.7 m and a maximum distance of 32.65 m. 48.8% of all interpolated points are within a maximum distance of 4 m to a DGPS measurement point, whereas only 1.2 % within a distance greater than 20 m. Including spectral and textural attributes, we discriminated five land cover classes with an overall accuracy of 88.98% (Kappa 0.86), validated with ground reference information. In contrast to the land use class “Grassland” the

TABLE 4 - Mean daily climatic dataa for the recorded events in 2010.

Day of year

Tmax [°C]

Tmin [°C]

mean T [°C]

mean RH [%]

meanhPa

PETTurc [mm/d]

283

34.0

17.6

24.8

37.65

893.01

219.89

296

29.6

18.6

23.2

38.74

892.39

198.35

232

31.2

18.6

23.4

43.99

892.73

187.64

233

32.6

18.7

24.7

44.64

891.90

199.9

335

32.1

18.1

23.4

41.23

890.98

194.47

340

31.9

19.6

23.1

42.87

889.74

186.72

a

T temperature; RH relative humidity; hPa hectopascal; PET potential evapotranspiration

61

TABLE 5 - Measured tensions and according infiltration rates with hood infiltrometer.

62

Mess_ID

Tension h0

Infiltration h0 [mm/h]

Tension h1

Infiltration h1 [mm/h]

Tension h2

Infiltration h2 [mm/h]

Air entrytension

1

0

20.52

-2.0

13.99

none

none

-5.8

2

0

29.23

-3.4

21.14

-5.5

16.79

-9.5

4

0

14.93

-2.5

11.82

none

none

-7.1

5

0

34.83

-3.0

33.58

none

none

-5.2

8

0

31.00

-1.2

30.87

-3.1

29.63

-15.1

10

0

19.28

-4.0

13.68

-8.0

9.33

-15.1

11

0

15.55

-4.0

9.33

-9.0

7.46

-22.0

13

0

32.34

-1.5

26.12

-2.4

19.90

-7.8

14

0

21.77

-3.4

18.66

none

none

-7.1

15

0

84.58

-4.0

31.72

none

none

-6.9

16

0

32.96

-4.0

25.50

-8.0

19.90

-10.7

17

0

24.25

-3.5

20.52

-6.0

15.55

-12.4

19

0

38.56

-2.7

29.85

-2.8

29.02

-7.5

20

0

43.53

-4.0

18.35

-6.0

13.68

-9.3

22

0

13.33

-2.5

8.71

-4.8

6.66

-8.5

23

0

26.65

none

none

none

none

-11.0

23

0

36.07

none

none

none

none

-11.0

24

0

14.66

-6.3

9.95

-9.2

2.69

-13.0

25

0

14.61

-4.0

7.00

-7.5

4.35

-8.8

26

0

114.01

-4.0

62.19

-7.2

38.35

-8.6

27

0

74.63

-3.0

43.53

-5.1

33.17

-18.0

28

0

202.11

-3.9

128.26

-7.0

90.17

-9.5

29

0

15.55

-4.4

9.33

-9.5

7.46

-13.1

30

0

18.66

-4.4

11.19

-7.0

5.60

-7.6

31

0

32.34

-6.0

34.83

-12.0

23.01

-17.4

32

0

33.58

-3.0

22.39

-5.4

18.66

-9.6

33

0

97.95

-4.0

60.63

-5.5

23.63

-7.6

34

0

69.44

-2.2

49.13

-4.6

42.91

-9.7

35

0

98.46

-4.0

66.85

-6.0

65.30

-9.0

36

0

31.09

-4.4

18.66

none

none

-7.5

37

0

71.52

-1.4

59.08

-2.2

49.13

-3.0

38

0

39.18

-3.5

32.34

-7.0

26.85

-13.3

39

0

18.66

-4.4

15.55

-6.7

12.83

-8.7

40

0

11.11

-4.6

7.15

none

none

-14.1

41

0

37.31

-2.0

41.04

-5.0

31.72

-10.3

42

0

55.19

-3.0

23.32

-6.1

13.33

-9.1

43

0

15.55

-4.4

9.33

-8.8

6.53

-13.8

44

0

37.31

-2.7

20.52

-5.3

13.60

-8.6

45

0

29.85

-3.2

24.88

-7.1

16.17

-7.7

FIG. 5 - a) Prediction of soil texture classes (USDA taxonomy) according to the boosted regression tree based classification model. b) Spatial distribution of water excess calculated for a precipitation event of 23 mm (~event number 1), based on the predicted soil texture class.

classes “Degraded Grassland” and “Barren Land” show a high percentage of bare soil. The later ones are highly affected by dry and wet seasons concerning their vegetation share and are in result highly affected by overgrazing and erosional processes. The highest confusion between the classes turned out to occur between “Grassland” and “Degraded Grassland” (fig. 3b). REGIONALIZATION OF SOIL SURFACE TEXTURE All DEM-based independent variables show a high contribution to the best fitted classification tree model. Only the land cover classification is found to be of no significance to soil texture distribution. The modelling based on the remaining four terrain indices shows a very good model performance. The modelled classes ‘sandy loam’ and ‘sandy clay loam’ with an AUC value of more than 0.8 show an excellent prediction output. The class ‘sandy clay’ with an AUC value of 0.93 indicates to have an outstanding prediction performance. The according regionalization based on the TreeNet model is shown in Figure 5a. The results show a sequence of textures from sandy clay loam on the eroded cambisols in the upper slope positions to sandy clay textures at mid slope positions. Toe slopes are characterized by sandy loam textures and vertic features. In Figure 6 the relation of texture and infiltration is shown for the 0 tension measurements in boxplots a) to d). As expected there is a close relation between texture and infiltration rates at 0 tension. The higher tensions h2 are measured in relation to the specific air entry value of the soil’s matric potential see boxplots e) to h). Figure 7 shows the relation between texture and reached tensions. As expected the highest tensions are measured in the soils with highest clay content and they are lowest in the soils with dominant sand fractions. The median tension infiltration value of the respective textures classes were then attributed to the soil texture map for the runoff modelling.

FIG. 6 - Infiltration rate (I) of 0 tension, in mm/h, for all locations a), and for the different soil texture classes b) – d). Accordingly infiltration rate of tension h2 for all locations e), and for the different soil texture classes f) – h).

63

be the other way round, with a high micropore flux (Luvisols) lower infiltration values at higher tension values are expected, whereas soils with macropores (Vertisols) show a higher infiltration. The water balances were subsequently utilized as weighting surface in a weighted flow accumulation algorithm. This model is based on the Multiple Flow Direction algorithm (after Seibert & McGlynn, 2007) and counts the number of cells draining into a given cell. Figure 8 shows the produced runoff volumes for the infiltration scenario shown in Figure 5b. SURFACE RUNOFF MEASUREMENT VIA SRD

FIG. 7 - The relation between predicted soil texture class and reached tensions, in a) the tensions measured at h2, and in b) tensions at air entry value.

WATER BALANCE MODELLING The water balance was calculated after Equation 7 for the test area on a 1m x 1m pixel basis. The calculation was performed for the surface infiltration at 0 tension and highest measured tensions (h2). Figure 5b shows the spatial distribution of the infiltration excess for the highest measured tension. Sandy loam soils (in dark blue) show a fast response to surface sealing due to their texture partitioning. They are therefore prone to a fast surface runoff development. The Luvisols (sandy loam and loam fraction) show a high water flux and only small runoff is calculated (in light blue) or even none (yellow). Within the soil column, the picture will

For each of the 5 rainfall events (>5 mm rainfall within 1 hour) occurring from October to December 2010, we compared the data to the recorded discharge height in the SRDs. All 5 events have been recorded in each of the available SRDs. Generally, the SRDs characterize the actual surface-runoff amounts on hillslope scale in a very high temporal resolution (5 min intervals). Consequently, SRDs can be used to roughly validate water balance calculations and related runoff processes. For the single SRD locations we estimated the upslope contributing area with SAGA-GIS software using a deterministic infinity flow direction algorithm following Tarboton (1997). Measured values of surface runoff are shown in Table 6. DISCUSSION In this study we focus on the spatial distribution of infiltration rates. As shown, infiltration in semiarid savannah dry lands of northern Tanzania is mainly depending on spe-

FIG. 8 - Runoff simulation, in blue are depicted the water accumulation areas with an input of 1.000 cells up to maximum.

64

TABLE 6 - Measured discharge from surface runoff detectors.. ID

Catchment area [m²]

A

278.62

B C

Rainfall runoff event [mm] 1

RI [mm/h]

107.5

1.34

608.08

22

569.68

202.5

Rainfall runoff RI event [mm] [mm/h] 2

117.5

1.47

0.28

240

2.53

300

Rainfall event 3

runoff RI Rainfall runoff RI Rainfall runoff RI [mm] [mm/h] event [mm] [mm/h] event [mm] [mm/h] 25

0.31

3.00

85

3.75

95

4

203

2.54

1.06

321.5

1.19

298.5

5

108.5

1.36

4.02

171

2.14

3.73

162

2.03

a

RI: average runoff intensity

cific soil physical characteristics and climatic conditions. The carbonatic volcanic ash and tuff/tephra deposits of the Makuyuni area are characterized by Andic Cambisols and Rhodic Luvisols with low activity clays developed on the upper slopes and flat summit areas. These soils show an intensive carbonate leaching (decalcification) and accumulation of carbonates (concretions) in the deeper subsoil or further downslope. On the slope ridges the soils are often eroded. The material is washed along the surface and deposition dominates in the toe slope situations. The structure of the soils is changing from matrix dominated soils on the slope ridges and mid-slope positions, towards vertic soils showing a distinct secondary macropore structures due to active layer clays (typically consisting of smectite and illite clays). The soil catena itself documents the intense sediment transport with surface runoff. Infiltration at different textures vary significantly. Especially in micropore dominated soils the infiltration is mainly triggered by the tension values or in other words is triggered by different humidity conditions of the uppermost soil horizon before rainfall starts. Generally, infiltration is decreasing for highest measured tensions by 18.74 %. For the sake of simplicity, we kept the evapotranspiration constant and do not consider soil water storage to calculate the water balance and hence, overland flow generation. The GIS-based simulation of runoff depending on pixel water balances shows that for a detailed hillslope scale analysis a high-resolution digital elevation model is crucial. This is because all further analyses such as topographic indices, e.g. runoff direction and accumulation, are based on the accuracy of the height estimations. Particularly, the creation of a hydrological correct DEM is essential for further hydrological modeling. Therefore, sophisticated statistical methods provide accurate spatial pattern of soil surface texture. Finally, we calculated the different runoff volumes in mm/h and, make them comparable, converting also the rainfall events to average rainfall intensity in mm/h. We show that for 0 tension only the highest event of 40.6 mm/h (rainfall intensity) produced runoff whereas all other events do not show surface runoff. If tension increase the runoff is getting higher for the highest rainfall event (event 1) but still no runoff is produced for the other events (events 2-5). Only when the highest measured tension is utilized the runoff even for the other events can be modelled as shown in Table 3. For all events runoff is produced and detected in the distributed SRDs. This means that the infiltration values for 0 and h2 tension used in the modelling are too high for the events 2-5. Only if we use the highest tension (h2max) realis-

tic surface runoff values are simulated. This means that the contributing area for the three SRDs is mainly characterized by homogeneous soils dominated by matrix pore triggered infiltration with small pore sized at sandy loam textures. Tension for this texture (sandy loam) can reach very high suctions characterized by a high air entry value. Hence, the measured highest tensions may be even higher if the measurement is performed close to the air entry point. Consequently, the infiltration becomes even lower. CONCLUSIONS The study presents a simple low budget approach to assess soil infiltration and surface runoff values. We show that infiltration processes in semiarid dryland conditions are mainly depending on the soil physical characteristics such as soil water tension and the moisture conditions of the topsoil before the precipitation event starts. Saturated conductivities often used to calculate the surface runoff yield too high infiltration values to explain observed surface runoff. In this study we measured the shallow laminar runoff at the surface using SRDs. Moreover, infiltrations were measured at different tensions simulating various soil humidity conditions. We illustrated with a simple water balance approach, that infiltrations measured at 0 tension produce too high infiltration values in order to generate surface runoff. Especially in micropore dominated homogeneous soils water tension triggered infiltration rates play a crucial role. Thus, on the slope system studied in this investigation surface runoff was only produced if the highest tension infiltration rates were used corresponding to “drier” soil conditions. Consequently, it is very important what kind of soil types are present in a catchment and to what kind of infiltration scheme they are belonging to. Especially in homogeneous soils with low activity clays notable surface runoff can be produced if the soil is very dry when precipitation starts. This example from a hillslope scale water balance model demonstrates the importance of integrated field work measurements (especially the incorporation of surface runoff detectors) and computer-based simulations in hydrological research. The automatic surface runoff detectors offer a helpful insights into actual overland flow dynamics and, especially, for the quantification of those. This approach can provide useful additional information on water balance modelling. Primarily the verification of single parameters of the water balance equation can further be validated and evaluated. 65

REFERENCES ASHTECH (2010) - ProMark 3 RTK. http://www.spectraprecision. com/products/gnss-surveying/promark3-rtk-2655.kjsp, Nov 4, 2014. BEVEN K.J. & KIRKBY M.J. (1979) - A physically based, variable contributing area model of basin hydrology. Bulletin of the International Association of Scientific Hydrology, 24, 43-69. BLASCHKE T. (2010) - Object based image analysis for remote sensing. ISPRS Journal of Photogrammetry and Remote Sensing, 65, 2-16. BORSELLI L., TORRI D., POESEN J. & SANCHIS P.S. (2001) - Effects of water quality on infiltration, runoff and interrill erosion processes during simulated rainfall. Earth Surface Processes and Landforms, 26, 329-342. BRACKEN L.J. & CROKE J. (2007) - The concept of hydrological connectivity and its contribution to understanding runoff-dominated geomorphic systems. Hydrological Processes, 21, 1749–1763. CONRAD O. (2006) - SAGA - Program structure and current state of implementation. In: Böhner, J. & MacCloy, K.R. (Eds.) SAGA-analysis and modelling applications. Goltze, Göttingen, 39-52. CRITCHLEY W. & SIEGERT K. (1991) - Water Harvesting. A Manual for the Design and Construction of Water Harvesting Schemes for Plant Production, Rome, Italy, 157 pp. ELITH J. & LEATHWICK J.R. (2009) - Species Distribution Models: Ecological Explanation and Prediction Across Space and Time. Annual Review of Ecology, Evolution, and Systematics, 40, 677-697. FAO (2014) - World reference base for soil resources. International soil classification system for naming soils and creating legends for soil maps. FAO, Natural Resources Management and Environment Department. FIELDING A.H. & BELL J.F. (1997) - A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24, 38-49. FRIEDMAN J.H. (2002) - Stochastic gradient boosting. Nonlinear Methods and Data Mining. Computational Statistics & Data Analysis, 38, 367-378. HANLEY J.A. & MCNEIL B.J. (1982) - The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology, 143, 29-36. HOSMER D.W., LEMESHOW S. & STURDIVANT R.X. (2013) - Applied logistic regression, 3rd ed. Wiley, Hoboken N.J., 1 online resource (XVI, 500 pp.) HUTCHINSON M.F. (1989) - A new procedure for gridding elevation and stream line data with automatic removal of spurious pits. Journal of Hydrology, 106, 211-232. HUXMAN T.E., SNYDER K.A., TISSUE D., LEFFLER A.J., OGLE K., POCKW.T., SANDQUIST D.R., POTTS D.L. & SCHWINNING S. (2004) – Precipitation pulses and carbon fluxes in semiarid and arid ecosystems. Oecologia, 141, 254-268.

MAN

JENSEN M.E. & BURMAN R.D. & ALLEN R.G. (1990) - Evapotranspiration and irrigation water requirements. A manual. American Society of Civil Engineers, New York, N.Y., xxviii, 332 pp. KUTÍLEK M. & NIELSEN D.R. (1994) - Soil hydrology. Catena Verlag, Cremlingen-Destedt, 370 S. pp. LAUENROTH W.K. & BRADFORD J.B. (2006) - Ecohydrology and the Partitioning AET Between Transpiration and Evaporation in a Semiarid Steppe. Ecosystems, 9, 756-767. MANEL S., WILLIAMS H.C. & ORMEROD S.J. (2001) - Evaluating presence–absence models in ecology: the need to account for prevalence. Journal of Applied Ecology, 38, 921-931. MOORE I.D., GESSLER P.E., NIELSEN G.A. & PETERSON G.A. (1993) Soil Attribute Prediction using Terrain Analysis. Transactions of the American Society of Agricultural Engineers, 57, 443-452. MOORE I.D., GRAYSON R.B. & LADSON A.R. (1991) - Digital terrain modelling. A review of hydrological, geomorphological, and biological applications. Hydrological Processes, 5, 3-30. MUALEM Y. (1976) - A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research, 12, 513-522.

66

MUALEM Y. & ASSOULINE, S. (1996) - Soil Sealing, Infiltration and Runoff. In: Issar A. & Resnick S. (Eds.) Runoff, Infiltration and Subsurface Flow of Water in Arid and Semi-Arid Regions. Springer Netherlands, 131-181. NICO G., RUTIGLIANO P., BENEDETTO C. & VESPE F. (2005) - Terrain modelling by kinematical GPS survey. Natural Hazards and Earth System Science, 5, 293-299. OLAYA V. & CONRAD C. (2009) - Geomorphometry in SAGA. In: Hengl, T. & Reuter HI (Eds.) Geomorphometry. Concepts, Software, Applications. Elsevier, Amsterdam, 293-308. PHILIP J.R. (1957) - The Theory of Infiltration: 1. The Infiltration Equation and its solution. Soil Science, 83, 345-358. PHILIP J.R. (1975) - Stability Analysis of Infiltration. Soil Science Society of America Journal, 39, 1042-1049. REINEKING B. & SCHRÖDER B. (2006) - Constrain to perform: Regularization of habitat models. Ecological Modelling, 193, 675-690. RENGER M., STOFFREGEN H., KLOCKE J., FACKLAM M., WESSOLEK G., ROTH C.H. & PLAGGE, R. (1999) - Ein autoregressives Verfahren zur Bestimmung der gesättigten und ungesättigten hydraulischen Leitfähigkeit. Journal of Plant Nutrition and Soil Science, 162, 123-130. RITTER D.F., KOCHEL R.C. & MILLER J.R. (2011) - Process Geomorphology, 5. Ed. Waveland Press, Long Grove, Ill., 652 pp. SALA O.E. & LAUENROTH W.K. (1982) - Small rainfall events: An ecological role in semiarid regions. Oecologia, 53, 301-304. SAUER T.J. & LOGSDON S.D. (2002) - Hydraulic and Physical Properties of Stony Soils in a Small Watershed. Soil Science Society of America Journal, 66, 1947–1956. SCHWÄRZEL K. & PUNZEL J. (2007) - Hood Infiltrometer A New Type of Tension Infiltrometer. Soil Science Society of America Journal, 71, 14381447. SCHWINNING S. & SALA O. (2004) - Hierarchy of responses to resource pulses in arid and semi-arid ecosystems. Oecologia, 141, 211-220. SEIBERT J. & MCGLYNN B.L. (2007) - A new triangular multiple flow direction algorithm for computing upslope areas from gridded digital elevation models. Water Resources Research, 43, W04501. SPRINGER U. & KLEE J. (1954) - Prüfung der Leistungsfähigkeit von einigen wichtigeren Verfahren zur Bestimmung des Kohlenstoffs mittels Chromschwefelsäure sowie Vorschlag einer neuen Schnellmethode. Zeitschrift für Pflanzenernährung, Düngung, Bodenkunde, 64, 1-26. SWETS J.A. (1988) - Measuring the accuracy of diagnostic systems. Science, 240, 1285-1293. TARBOTON D.G. (1997) - A new method for the determination of flow directions and upslope areas in grid digital elevation models. Water Resources Research, 33, 309-319. UMWELT-GERÄTE-TECHNIK, 2005 - Operating instructions for Hood Infiltrometer. http://www.ictinternational.com.au/ugt/brochures/hoodinfiltrometer.pdf, Nov 4, 2014. USDA (1993) - Soil survey manual. United States Department of Agriculture (USDA). Soil Survey Division, Washington, D.C., 437 pp. VAN GENUCHTEN, M.T. (1980) - A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils1. Soil Sci. Soc. Am. J., 44, 892-898. WAINWRIGHT J. & BRACKEN L.J. (2011) - Runoff generation, overland flow and erosion on hillslopes. In: Thomas, D. (Ed.) Arid Zone Geomorphology. Process, Form and Change in Drylands. Wiley-Blackwell, Oxford, 237-267. WALLACE J.S., JACKSON N.A. & ONG C.K. (1999) - Modelling soil evaporation in an agroforestry system in Kenya. Agricultural and Forest Meteorology, 94, 189-202. WOODING R.A. (1968) - Steady Infiltration from a Shallow Circular Pond. Water Resources Research, 4, 1259-1273. ZEMA A.D., BOMBINO G., DENISI P., LICCIARDELLO F. & ZIMBONE S.M. (2012) - Prediction of Surface Runoff and Soil Erosion at Watershed Scale: Analysis of the AnnAGNPS Model in Different Environmental Conditions. In: Godone D. (Ed.), Research on Soil Erosion. InTech, 3-31. (ms received 15 December 2014; accepted 15 May 2015)

Geogr. Fis. Dinam. Quat. 38 (2015), 67-78, 13 figg., 5 tab.

DOI 10.4461/GFDQ.2015.38.07

PETR SKLENÁř(*), ANDREA KUčEROVÁ(**), JANA MACKOVÁ(***) & PETR MACEK(****)

TEMPORAL VARIATION OF CLIMATE IN THE HIGH-ELEVATION PÁRAMO OF ANTISANA, ECUADOR

ABSTRACT: SKLENÁř P., KUčEROVÁ A., MACKOVÁ J. & MACEK P., Temporal variation of climate in the high-elevation páramo of Antisana, Ecuador. (IT ISSN 0391-9838, 2015). We monitored the climate in high-elevation páramo of Antisana (Ecuador) to analyze its diurnal and annual variation. We established two climatic stations on the western (leeward) side of the mountain at 4280 m and 4600 m, and two stations on the north-eastern (windward) side at 4120 m and 4430 m. We recorded air temperature at 100 cm above ground, relative air humidity, and global solar radiation in hourly intervals from July 2007 to December 2010. Moreover, we recorded precipitation at the two lower stations. The western side received 1098 mm of rainfall per year with two maxima in April–June and October–November. In contrast, the north-eastern side received 2694 mm with a single maximum in June. The air was almost permanently saturated with moisture, both during the day and the year, on the north-eastern side with mean relative air humidity of 98%. On the western side, the air humidity showed distinct daily and seasonal variation; it dropped to 80% at noon and had two annual minima in January and September. The western side of Antisana received 35% more solar radiation than the north-eastern side. Furthermore, the mostly cloudy weather on the north-eastern side tended to eliminate the mid-day maximum of radiation. The stations at higher elevations received 11% more

———————— (*), Department of Botany, Charles University, Benátská 2, 128 01 Prague, Czech Republic (**), Institute of Botany, Czech Academy of Sciences, Dukelská 135, 379 82 T ebo , Czech Republic (***), Institute of Soil Biology, Biology centre, Czech Academy of Sciences, Na Sádkách 7, 370 05 eské Bud jovice, Czech Republic (****), Faculty of Science, University of South Bohemia, Branišovská 31, 370 05 eské Bud jovice, Czech Republic The study was supported by the Grant Agency of the Academy of Sciences, Czech Republic (IAA601110702), by the long-term research development project no. RVO 67985939, and partly also by the Ministry of Education of the Czech Republic (MŠMT 0021620828). PM was additionally supported by CzechPolar grant LM2010009. An anonymous reviewer is acknowledged for insightful comments on the earlier draft of the manuscript. Katya Romoleroux and Hugo Navarrete (PUCE, Quito) are thanked for providing research facilities in Ecuador, Luis Maisincho (INAMHI, Quito) is thanked for providing access to the Antizana15 climatic data for comparison, and Ministerio del Ambiente and señor José Delgado are acknowledged for research and entry permits, respectively. Keith Edwards kindly provided linguistic revision.

solar radiation than the stations at lower elevations and experienced a more distinct seasonal variation with the maximum during August–September. Mean annual air temperature varied within 2K at all stations, which contrasted with the mean daily oscillation of 8–10K on the western side and 5K on the north-eastern side. Night frosts were frequent on the western side whereas high humidity and cloudiness on the north-eastern side reduced the number of frost nights. Frosts were rather mild and of short duration; the minimum temperature recorded was –6.1°C and most frost periods lasted less than four hours. Freezing temperatures were most frequent during periods of reduced humidity. Temperature lapse rates calculated for the 300 m elevational gradient were 0.44 K/100 m and 0.55 K/100 m for the western and north-eastern sides, respectively. Potential evapotranspiration values suggested that water was in surplus year-round on the north-eastern side of Antisana, but its availability was limited during drier periods on the western side. KEY WORDS: Equatorial Andes, Mountain climate, Seasonality, Superpáramo, Tropical alpine environment. RESUMEN: SKLENÁř P., KUčEROVÁ A., MACKOVÁ J. & MACEK P., Variación temporal del clima en el páramo de Antisana, Ecuador. (IT ISSN 0391-9838, 2015). La variación diaria y anual del clima se estudió en el páramo de Antisana (Ecuador) en dos estaciones climáticas localizadas en el lado occidental de la montaña a alturas de 4280 m s.n.m. y 4600 m s.n.m. y en dos estaciones localizadas en el lado nor-oriental a alturas de 4120 msnm y 4430 msnm. La temperatura del aire a 100 cm por encima del suelo, la humedad relativa del aire y la radiación global de sol se registraron de manera continua cada hora desde julio 2007 hasta diciembre 2010. La precipitación se registró en las dos estaciones de alturas bajas. El promedio anual de precipitaciones fue 1098 mm en el lado occidental con épocas mas húmedas en abril–junio y octubre–noviembre. Por el contrario, en el lado nor-oriental el promedio anual fue 2694 mm con junio cómo el mes más húmedo. En el lado nor-oriental la humedad relativa del aire permaneicó muy alta casi todo el tiempo con un valor promedio de 98%. En el lado occidental la humedad relativa mostró un ritmo diario y annual; la humedad bajó hasta 80% en el mediodía y mostró dos minimos anuales en enero a septiembre. El lado occidental recibió un 35% más de radiación solar que el lado nor-oriental. Las dos estaciones de alturas altas recibierón un 11% más de radiación que las dos estaciones de alturas bajas y mostrarón una variación anual distinta con máximos en agosto–septiembre. La variabilidad interanual de la temperatura del aire no alcanzó 2K en ninguna de las estaciones, por el contrario, el promedio de las oscilaciones diarias fue 8–10K en el lado occidental y 5K en el lado nor-oriental. En el lado occidental se registraron frecuentemente temperaturas nocturnas bajo 0°C mientras que en el lado nor-oriental la frecuencia de días con heladas fue reducida por la alta

67

PALABRAS CLAVES: Ambiente alpino tropical, Andes ecuatoriales, Clima de montaña, Estacionalidad, Superpáramo.

established microclimatic stations on the slopes of the Antisana volcano in Ecuador to monitor variation of the highelevation páramo climate. In this paper, we describe the patterns and quantify the temporal variation of air temperature, solar radiation, air humidity, precipitation, and potential evapotranspiration using data from three years of observations. In particular, we examined how much the daily and seasonal variation in temperatures are mirrored by variation of other climatic variables.

INTRODUCTION

METHODS

The warm and humid climate of the lowland tropics along with the seasonal climate with warm summers and cold winters of the temperate–to–arctic zones are the norm for vast geographic areas. In contrast, the tropical alpine climate is one of the most specific climates on Earth, which is encountered only on the highest equatorial mountains (Sarmiento, 1986; Barry, 2008). General features of the tropical alpine climate are well-known – it is most characteristically described by a pronounced daily temperature oscillation along with frequent night frosts throughout the year (Troll, 1968; Sarmiento, 1986; Rundel, 1994). Precipitation, in contrast to temperature, is distributed more unevenly both during the year due to annual patterns of atmospheric circulation and between years due to e.g., ENSO events (Bendix & Lauer, 1992; Vuille & alii, 2000a, b). Precipitation occurs mainly in the form of rainfall, although the highest elevations experience snow and hail (Sarmiento, 1986; Bendix & Rafiqpoor, 2001; Jomelli & alii, 2009). Knowledge about the climate of the tropical high Andes, i.e., páramo, is often derived from short-term climatic records spanning several days or weeks (e.g., Diemer, 1996; Sømme & alii, 1996; Sklená , 1999). Climatic measurements lasting longer than one year are uncommon and especially the highest elevations are insufficiently covered (Vuille & alii, 2008). Long-term climatic observations are available from the mid-elevations of the páramo belt (Azócar & Monasterio, 1980; Anonymous, 1978–1990; Jørgensen & Ulloa, 1994) and sometimes data are gathered from the vicinity of glaciers (e.g., Wagnon & alii, 2009). The most complex climatic data representative of the entire páramo belt were provided by the ECOANDES project in the Central Cordillera of Colombia (Javellas & Thouret, 1995; van der Hammen & alii, 1995; Witte, 1995). In tropical alpine environments, the daily temperature variation greatly surpasses the seasonal variation and so the climate is often paraphrased as “summer every day, winter every night” (Hedberg 1964). In contrast to this common knowledge about the temperature course, patterns of daily versus seasonal variation of other climatic variables have been addressed less frequently (e.g., Azócar & Monasterio, 1980). Yet the annual climatic variation is important in determining e.g., phenology of tropical alpine biota, although causal mechanisms usually remain poorly understood (Kudo & Suzuki, 2004; Fagua & Gonzalez, 2007). Moreover, longterm climatic measurements from the upper páramo belt of the humid equatorial Andes are virtually lacking. We thus

STUDY SITE

humedad y nubosidad. En general, las noches con heladas fueron moderadas (la temperatura mínima registrada fue –6.1°C) y la duración de temperaturas bajo 0°C fue corta (menos de cuatro horas). Temperaturas bajo 0°C ocurrierón más frecuentemente en periodos con la humedad reducida. El gradiente térmico altitudinal fue 0.44 K/100 m en el lado occidental y 0.55 K/100 m en el lado nor-oriental. Según los valores de evapotranspiración potencial, en el lado nor-oriental siempre hubo exceso de agua mientras que en el lado occidental hubo escasez de agua durante los periodos secos.

68

The study was carried out on the slopes of Antisana (5704 m; 0°30’S 78°10’W), which is a mountain located in the eastern cordillera of Ecuador. Glaciers crown the top of Antisana and reach as far down as 4800 m and 4600 m on the western and eastern sides, respectively. However, older moraines indicate earlier glaciers about 1200 m lower (Hastenrath, 1981; Clapperton & alii, 1997; Rabatel & alii, 2013). Antisana is considered an extinct volcano, although the latest activity dated to the beginning of the 19th century (Sauer, 1971; Hall, 1977). A high-elevation plateau (between 4000 – 4200 m) is present on the south-western side of the mountain (fig. 1) and is mostly covered by grass páramo vegetation. Antisana hosts a diverse páramo flora and distinct types of vegetation (Sklená & Lægaard, 2003; Sklená & alii, 2008). On the eastern (windward) side, sclerophyllous shrubs (Loricaria, Diplostephium) share dominance with cushion plants (Azorella, Plantago, and Werneria), and locally also with bamboo and tussock grasses (Neurolepis, Calamagrostis, Festuca) at lower elevations. At higher elevations, prostrate subshrubs (Disterigma, Pernettya) along with herbs and small grasses (e.g., Lachemilla, Oritrophium,

FIG. 1 - Topography map of the Antisana volcano with location of the four climatic stations; contour lines are in 100 m intervals; WL means western lower site, WH means western higher site, EL means north-eastern lower site, and EH means north-eastern higher site.

TABLE 1 - Geographic location, instrumentation, and period of operation of the four microclimatic stations on the slopes of the Antisana volcano.

Locations

Western lower

Western higher

North-eastern lower

North-eastern higher

Elevation

4280 m

4600 m

4120 m

4430 m

Latitude

00°29’09’’

00°28’55’’

00°27’06’’

00°27’37’’

Longitude

78°11’52’’

78°10’02’’

78°07’32’’

78°07’45’’

Position

Period of operation

Period of operation

Period of operation

Period of operation

Precipitation, 8’’ Raingauge 372C, MetOne, Oregon, USA + MicroLog datalogger, EMS Brno

100 cm

Jul 2007–Dec 2010 NA

14 Aug–30 Sep 2007, 3 Nov 2007–29 Feb 2008, 16 Jul–21 Oct 2008, 18 Mar–9 Dec 2010

NA

Air temperature, Pt100 sensor + Minikin datalogger, EMS Brno

100 cm

Jul 2007–Dec 2010

Jul 2007–Dec 2010

Jul 2007–Jan 2010

Jul 2007–Jun 2008

Global radiation, sensor EMS 11 + Minikin dataloger, EMS Brno

100 cm

Jul 2007–Nov 2010, missing Aug–Oct 2009

Jul 2007–Nov 2010

Jul 2007–Nov 2010, missing Jul 2008

Jul 2007– Nov 2010, missing Jun–Jul 2008

Relative air humidity, sensor EMS 33 + Minikin datalogger, EMS Brno

100 cm

Jul 2007–Dec 2010

Jul 2007–Dec 2010

Jul 2007–Jan 2010

Jul 2007–Jun 2008

Instrument

Agrostis) are common. On the western (leeward) side, tussocks of Calamagrostis grasses are abundant at lower elevations along with sclerophyllous shrubs (Chuquiraga), rosulate herbs (Valeriana, Werneria), and prostrate subshrubs (Baccharis, Lupinus). This vegetation gives way to patchy upper superpáramo vegetation of small herbs and grasses (Black, 1982; Sklená , 2000). Cushion-bogs occur along streams and in terrain depressions. The region was declared an ecological reserve, Reserva Ecológica Antisana, in 1993 in order to protect vast reaches of montane forest and páramo. In spite of that, the páramo on the western side of the mountain is heavily grazed by cattle and sheep which commonly results in severe degradation of the soil and vegetation (Grubb, 1970; Black, 1982; Sklená , 2000). Data acquisition and treatment Four microclimatic stations were established on two opposite sides of the Antisana volcano in July 2007 (fig. 1; tab. 1) and operated until December 2010. Two stations were located on the western side at 4280 m and 4600 m, and two stations were located on the north-eastern side at 4120 m and 4430 m. The roughly 200 m shift in the elevational position of the stations between the western and eastern sides reflected the climatic asymmetry of the equatorial Andes (Troll, 1968; Lauer, 1979), so that the stations on the opposite sides thus occurred in corresponding vegetation types (Sklená , 2000; Sklená & Lægaard, 2003). The two lower sites corresponded to the transition between the grass (western side) and bamboo (north-eastern side) páramo vegetation to lower su-

perpáramo vegetation and were characterized by closed (or almost so) plant cover. The two higher sites were located in the sparsely vegetated upper superpáramo (e.g., Sklená , 2006). All stations were established in open areas of the páramo, i.e., they were unobstructed by any tall vegetation, rocks, etc. The stations were equipped with a set of dataloggers with built-in sensors to record incident global radiation (Rg), air temperature (T100), and relative air humidity (RH) (tab. 1). The Rg, T100, and RH sensors and loggers were screened by white radiation shields and were positioned on aluminum poles at 100 cm height above ground; due to technical limitations, the temperature sensors were not actively ventilated. The sensors and dataloggers were supplied by EMS Brno, Czech Republic (www.emsbrno.cz). The stations at lower elevations were equipped with automated Model 372 precipitation gauges with funnel diameter of 20.3 cm designed to measure rain and snow. The gauges, supplied by MetOne, Wisconsin, USA (www.metone.com), were placed on aluminum frames at 100 cm height above ground. Measurements were taken from 26 July 2007 to 9 December 2010. The recording interval was set at 1 hour in all dataloggers except the rain gauges. Records were not complete for all variables due to damage of the sensors and/or dataloggers (tab. 1). Whereas the rain gauge recorded continuously during the entire period on the western side, the instrument on the north-eastern side operated only periodically, which yielded four common periods of rainfall measurements (14 August–30 September 2007, 3 November 2007–29 February 2008, 16 July–21 October 2008, 18 March 2009–9 December 2010). 69

FIG. 2 - The seasonal course of precipitation in the páramo of Antisana measured on the western (black) and north-eastern (white) sides of the volcano during 2007–2010; mean annual sum of rainfall equals 1098 mm and 2694 mm, respectively.

FIG. 3 - Frequency distribution of 24-hours rainfall sums on the western (black) and north-eastern (white) sides of the Antisana volcano during four overlapping measurement periods between August 2007 and December 2010 (see text for details). The distribution median equals 1.6 mm (western side) and 5.4 mm (north-eastern side), and the number of rainfall days are 639 (western side) and 781 (north-eastern side).

Potential evapotranspiration was calculated from Rg, RH and T100 using the Penman-Monteith method (Allen & alii, 1998). The calculations followed the Equation 1:

mean annual sum of 1098 mm (due to incomplete records in December 2010, sums for the years were calculated for the continuous period December 2007–November 2010). The precipitation varied from 1050 mm in 2008 to 781 mm in 2009, and 1465 mm in 2010. The north-eastern side of the mountain had a unimodal distribution of rainfall with the maximum in June and the minimum in October–December (fig. 2) with a mean annual sum of 2694 mm. Although the records were incomplete in the first half of the measuring period, this estimate seems to be unbiased since there was 2626 mm of precipitation over the 12-month continuous record of December 2009–November 2010. Overall, 2562 mm of precipitation fell on the western side and 6594 mm on the north-eastern side, during the period of common measurements, which totaled 28 months. Individual time periods had the respective values of 64 mm and 341 mm (14 August–30 September 2007), 248 mm and 804 mm (3 November 2007–29 February 2008), and 247 mm and 916 mm (16 July–21 October 2008), and 2002 mm and 4533 mm (18 March 2009–9 December 2010) for the western and northeastern sides, respectively. Precipitation was distributed more evenly during the year on the north-eastern side of the Antisana volcano, varying by 33% (the value is a coefficient of variation calculated from monthly sums, n = 26), in contrast to 57% variation on the western side (n = 40). There were 639 days with rain on the western side compared to 781 rainy days on the north-eastern side, meaning that about 28% and 12% of days were without measurable precipitation, respectively. Although the absolute 24-hour maxima did not differ remarkably between the two sides, i.e., being 49.4 mm on the western and 63.2 mm on the north-eastern side (both values were recorded in the first half of 2010), rain fell in distinctly smaller amounts on the western side (fig. 3). Whereas more than 42% of rainy days yielded less than 1 mm of rain on

Eto = 0,408 D (Rn - G) + g (900 / T+ 273) U2 (ea - ed) +

Eq. 1.

(1 + 0,34 U2)

in which Eto = evapotranspiration [mm .day-1], Rn = net radiation [MJ m-2day-1], G = soil heat flux [ MJ m-2day-1], T = mean daily temperature [°C], U2 = wind speed at 2 m height [ms-1], ea - ed = the water vapour saturation deficit, and D = the slope of the saturation vapor pressure [kPa °C-1], while 900 is a conversion factor. Eto describes the potential evapotranspiration from a water-saturated lawn. The soil heat flux was considered close to zero in a daily budget and was thus omitted. The equation includes an empirical dependence on wind speed. Since wind measurements could not be taken due to limited instrumentation, we used the value of 2 ms-1 for the whole data set as an approximation of low wind speed, following the FAO recommendation. The daily and annual variation in climatic variables was compared by means of the coefficient of variation, which was calculated from hourly and monthly means or sums (where appropriate), respectively. RESULTS

PRECIPITATION Rainfall was distributed bimodally on the western side of the Antisana volcano with the main maximum in April–June and a secondary maximum in October–November with a 70

FIG. 4 - Daily course of solar radiation at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and north-eastern higher (triangles and dashed curve) climatic stations; values are means of hourly radiation.

FIG. 5 - Seasonal course of mean daily solar radiation at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and north-eastern higher (triangles and dashed curve) climatic stations.

the western side, such small amounts occurred only in about 16% of rainy days on the north-eastern side. As a result, the median of the daily rainfall sums for the two sides was 1.6 mm and 5.4 mm, respectively.

ondary peak in solar radiation was observed in December– January at both western sites. Such seasonal variation was essentially lacking on the north-eastern side being between 7.7–7.8% for the two sites, although slightly higher values occurred at the higher site during July–September.

SOLAR RADIATION The daily mean value of solar radiation on the western side of the Antisana volcano was 346 ± 123.9 Wm-2 at the lower site and 387 ± 138.1 Wm-2 at the higher site, while respective values for the north-eastern side were 218 ± 101.9 Wm-2 and 243 ± 93.1 Wm-2. Along with the major difference in the radiation values between the opposite mountain sides and a minor difference due to elevation, there was a distinct variation in temporal patterns among the sites. A mid-day peak of incident solar radiation can be seen on the western side, although the maximum seems to be shifted to 11:00 at the higher site in contrast to the lower site. On the north-eastern side, however, no distinct daily maximum was seen with almost constant values between 10:00 and 13:00 (fig. 4). Furthermore, solar radiation peaked in July–September on the western higher site with a similar, albeit much less distinct, increase at the lower site (fig. 5). As a result, the seasonal variation of incoming radiation was 11.6% at the higher site whereas it was only 7.5% at the lower site. A sec-

AIR TEMPERATURE The mean daily maximum surpassed 9°C at the western lower site while the mean night minimum dropped below zero yielding a mean daily oscillation of about 10K. In contrast, the mean daily range was about 5K on the northeastern side (tab. 2). These differences in the daily range of air temperature between the two sides of Antisana were consistently present throughout the entire year, i.e., the western sites always experienced greater temperature oscillations. Air temperature peaked at noon on the western side, but the temperatures were rather constant and without a distinct maximum on the north-eastern side (fig. 6). The day-time climate was thus cooler on the north-eastern side as the summed hour-degrees (7:00–18:00) yielded values of 58.2 and 37.4 at the lower and higher sites, respectively, compared to 74.6 and 48.3 at the respective western sites. On the other hand, the climate on the north-eastern side was milder during the night compared to the western side, providing re-

TABLE 2 - Means and extreme values of air temperature measured at 100 cm above ground.

Station

Number of records

Absolute minimum (°C)

Mean daily minimum (°C)

Mean (°C)

Mean daily maximum (°C)

Absolute maximum (°C)

Western 4280 m

29785

-6.1

-0.5

3.7

9.6

15.0

Western 4600 m

29537

-4.1

-0.7

2.3

6.9

13.0

North-eastern 4120 m

29124

-3.1

1.6

3.8

6.6

14.3

North-eastern 4430 m

9176

-2.5

0.2

2.1

4.9

10.5

71

FIG. 6 - Daily course of mean air temperatures at 100 cm above ground at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and northeastern higher (triangles and dashed curve) climatic stations.

spectively 33.7 hour-degrees and 14.0 hour-degrees for the north-eastern and western lower sites, and 14.3 hour-degrees and 7.1 hour-degrees for the north-eastern and western higher sites. In general, night frosts were mild at all sites and particularly on the north-eastern side of the mountain where the air temperature never dropped below –4°C (fig. 7); night air temperatures below –4°C occurred at a frequency of 1.3% at the western lower site and of 0.2% at the western higher site. The lowest air temperature of –6.1°C measured at 100 cm height was recorded at the western lower site. Whereas the western side of the Antisana volcano was more variable on a daily basis, the variation during the year tended to be greater on the north-eastern side. Monthly mean air temperatures varied between 1.5–1.8K at the northeastern sites, which corresponded to a relative yearly variation of 15–23%, whereas the variation was between 1–1.3K (~11–12%) at the western sites (fig. 7). Mean daily maximum temperatures varied between 2.1–3.2K (~13–14%) on the north-eastern side in contrast to 1.6–2K (~7%) on the western side, and similar patterns were seen for the absolute

FIG. 7 - Seasonal variation of (from the bottom in each graph) the absolute minimum (diamonds, dashed line), mean minimum (diamonds, full line), mean (squares, full line), mean maximum (triangles, full line), and the absolute maximum (triangles, dashed line) air temperatures measured during the month at 100 cm above ground at the western lower (above left), western higher (above right), north-eastern lower (below left), and north-eastern higher (below right) climatic stations.

72

FIG. 8 - Relative frequency of freezing period duration at 100 cm above ground during the night (19:00–6:00) expressed as the number of consecutive measurement periods with temperatures below zero (i.e., one period ~ 1 hour) at the western lower (black columns, solid outline; n = 16125 records), western higher (black columns, dashed outline; n = 16006), north-eastern lower (white columns, solid outline; n = 14775), and northeastern higher (white columns, dashed outline; n = 9203) climatic stations.

FIG. 9 - Number of days per month with air temperatures at 100 cm above ground falling below zero at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and north-eastern higher (triangles and dashed curve) climatic stations.

daily maxima and absolute night minima. Only the seasonal mean night minimum temperature tended to vary more at the western sites (1.4–2.1K corresponding to ~51–56% of relative variation) than at the north-eastern sites (1.2–1.7K, ~32–52%). No day with permanent frost, i.e., with temperatures below zero during the entire 24 hours, was recorded at any site. The longest freezing period of 16 hours was observed twice at the western higher site. Except for the north-eastern lower site, freezing periods lasting the whole night, i.e., 12 consecutive hours, were occasionally observed (fig. 8). But in most cases, i.e., more than 75% of days on the western side and more than 85% of days on the north-eastern side, continuous freezing temperatures lasted less than four hours. Mean daily air temperature below zero was observed on less than 1% of days at the two higher sites, and there was a single such day at the western lower site. The occurrence of frost days had a distinct seasonal pattern with the maximum frequency during July–October at all sites (fig. 9). However, the number of frost days varied significantly among the sites. Whereas there were 29 frost days at the western higher site in September, the north-eastern lower site experienced only 6 such days. Although the number of freezing days at the

north-eastern higher site significantly increased during September, the frosts were nevertheless mild and temperatures only rarely dropped below –1°C. RELATIVE AIR HUMIDITY The air was almost permanently saturated by water vapor on the north-eastern side of the Antisana volcano (tab. 3; figs. 10 and 11). As a result, mean air humidity was about 98% at both sites and varied by less than 1% during the day, although exceptionally values dropped as low as 20%. Essentially the same pattern was seen during the entire year as the seasonal variation in RH was only 0.6–0.8% for these two sites. In contrast, the RH fluctuated more on the western side both on a daily (4.6–8.2%) and seasonal (1.6–2.9%) basis. The air humidity still remained relatively high for most of the time with mean values between 90–92% (tab. 3). But since RH was largely determined by air temperature (correlation coefficients between RH and temperature, r = 0.733 and r = 0.512 for the two western sites), air humidity declined from the morning towards noon and rose in the late afternoon (fig. 10). Whereas air saturation by water vapor remained close to 100% at the

TABLE 3 - Means and extreme values of relative air humidity measured at 100 cm above ground.

Station

Number of records

Absolute minimum (%)

Mean daily minimum (%)

Mean (%)

Mean daily maximum (%)

Absolute maximum (%)

Western 4280 m

29231

17.9

69.8

91.8

99.9

100

Western 4600 m

29517

10.0

69.8

89.6

99.7

100

North-eastern 4120 m

22169

14.9

93.6

98.6

99.9

100

North-eastern 4430 m

9176

17.1

89.8

97.9

100.0

100

73

FIG. 10 - Daily course of relative air humidity at 100 cm above ground at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and northeastern higher (triangles and dashed curve) climatic stations.

FIG. 11 - Seasonal course of mean daily relative air humidity at 100 cm above ground at the western lower (squares and solid curve), western higher (squares and dashed curve), north-eastern lower (triangles and solid curve), and north-eastern higher (triangles and dashed curve) climatic stations.

lower site during the entire night, RH culminated after the sunset at ~96% at the western higher site and then declined gradually to 90% in the morning. As regards the seasonal variation, mean daily RH peaked at ~94% at both western sites in April but dropped in September to 89% and to 84% at the lower and higher sites respectively.

exceeding evaporative demands during the entire year. The seasonal variation in Eto was 16.9% at the western lower site and 11.6% at the north-eastern lower site.

POTENTIAL EVAPOTRANSPIRATION Differences in solar radiation, air temperature, and relative air humidity between the opposite sides of the Antisana volcano resulted in the potential evapotranspiration (Eto) at the north-eastern sites being only 70% of that estimated for the western sites (tab. 4). On the other hand, due to minor differences in the incoming radiation and relative air humidity between the lower and higher sites, there was only a minor difference in Eto due to elevation. Cumulative values of Eto and precipitation for the one-year period of December 2009–November 2010 (for which a complete dataset was available) were, respectively, 732 mm and 1455 mm for the western lower site. A modest water balance deficit developed there during December–March due to lower precipitation and slightly higher Eto (fig. 12). The cumulative values of Eto and precipitation were, respectively, 515 mm and 2626 mm for the north-eastern side with precipitation TABLE 4 - Means and extreme values of potential evapotranspiration Et0.

Station Western 4280 m

Minimum Mean Maximum (mm day -1) (mm day -1) (mm day -1) 0.8

2.0

4.0

Western 4600 m

0.6

2.1

4.0

North-eastern 4120 m

0.6

1.4

4.2

North-eastern 4430 m

0.5

1.4

3.3

74

DISCUSSION Three years of observations is too short a period to provide a complete picture of temporal climate variation in the equatorial high Andes (Francou & alii, 2004). Nevertheless, the annual and inter-annual variation patterns documented by our measurements on the western side of the Antisana volcano are fully consistent with those observed in the vicinity of the Antisana glacier on the same mountain side during 2007–2010 (e.g., Rabatel & alii, 2013), which supports the validity of our data. During our measurements, the ENSO index was negative except for May-2009–May-2010 when it turned positive; in fact, one of the strongest El Niño/ La Niña oscillations over the past few decades occurred between 2009 and 2011 (e.g. Boening & alii, 2012). The ENSO is imprinted in the inter-annual variation of all measured variables and particularly so on the western side of the mountain (fig. 13). Relative air humidity and precipitation tend to correlate negatively whereas global radiation tends to correlate positively to ENSO. These correlations became stronger with the onset of the 2009 El Niño event. However, since less cloud cover during an El Niño event not only reduces precipitation in the Andes but it also increases air temperature (Francou & alii, 2004; Vuille & alii, 2008), the negative correlation of air temperature with ENSO in the first part of the measurements turned strongly positive since mid-2009 on both sides of the mountain. Our observations thus covered the variation between “standard” years and a strong El Niño/La Niña event which allows for describing major climatic trends that occur in the high-elevation páramo of Antisana. Heavy convective clouds from the Amazon basin are lifted by the Andean cordillera and bring plentiful rainfall

to the eastern slopes of the Ecuadorian Andes (Bendix & Lauer, 1992; Vuille & alii, 2000a; Bendix & Rafiqpoor, 2001; Favier & alii, 2004; Laraque & alii, 2007). The precipitation increases from 1434 mm in the montane forest (Papallacta, 3160 m; Bendix & Rafiqpoor, 2001) to 2694 mm on the north-eastern side of the Antisana volcano (4100 m). Such an amount ranks this páramo among the most humid in Ecuador (Jorgensen & Ulloa, 1994) and by the temporal rainfall pattern makes it similar to the eastern slope of the Sumapaz páramo in the Colombian eastern cordillera (Las Dantas, 3996 m, annual sum of 5445 mm with a distinct maximum in June–July; Rangel & Arellano, 2008). The clouds flow around the cone of the volcano, which often leaves its western slope without rain or even sunlight. This leeward side of the mountain thus receives 2–4 times less rain than the opposite, windward side. While a significant portion of precipitation in the eastern cordillera originates from humidity that is advected from the Quito basin (Bendix & Rafiqpoor, 2001; Lauer & alii, 2001), the western (leeward) side of the Antisana volcano has two rainy periods (see also Villacís, 2008) corresponding to the general precipitation pattern of the Ecuadorian inter-Andean valley. This is similar to the western side of Sumapaz (Laguna Chisaca, 3800 m, 1245 mm) and páramo Los Nevados in the Colombian central cordillera (precipitation sums between 1200–1500 mm) (Bendix & Lauer, 1992; van der Hammen & alii, 1995; Bendix & Rafiqpoor, 2001; Laraque & alii, 2007; Rangel & Arellano, 2008; IAEA, 2009). Moreover, contrary to the eastern side, rainfall declines with elevation from 1607 mm observed at the climatic station Pinantura (3250 m) to about 1100 mm measured in this study (4280 m) and 977 mm observed at 4700 m (Villacís, 2008). Nevertheless, precipitation tends to increase again in the vicinity of the glacier (4860 m) as suggested by reports of annual sums above 1000 mm (Favier & alii, 2004, 2008; Maisincho & Carceres, 2007; Wagnon & alii, 2009). Such contrasting precipitation patterns determine the spatial climatic variation observed on the Antisana volcano. Solar radiation input is almost 1.6 times greater on the western (leeward) side of the mountain and peaks in July–September, i.e., during the period of reduced cloudiness and rainfall (Favier & alii, 2004; Wagnon & alii, 2009). There is a seasonal decline of relative air humidity to mean values below 90% during the driest month (September), as well as a distinct mid-day minimum, which is consistent with patterns observed in the Colombian (central cordillera) and Venezuelan páramos, although the RH values are lower there (Azócar & Monasterio, 1980; Witte, 1995). The nightly decline of RH at the western higher site is similar to the Venezuelan Páramo de Mucubají (Azócar & Monasterio, 1980) and may be due to the down-flow of cold and relatively dry air (RH ~78%) from the Antisana glacier (Wagnon & alii, 2009). Evaporative demand is relatively high on the western side and a modest deficit in water supply may develop temporarily during periods of reduced rainfall in December–March, similar to the seasonal Páramo de Mucubají, although the water deficit during the driest month is much higher there (Azócar & Monasterio, 1980). The vigorous advective regime on the eastern (windward) side of

FIG. 12 - Monthly sums of potential evapotranspiration ET0 (above) and difference between precipitation and ET0 (below) for the western (black columns) and north-eastern (white columns) sides during December 2009– November 2010.

Antisana brings atmospheric moisture which eliminates any daily or seasonal pattern in relative air humidity, consistent with the climate at Papallacta (Bendix & Rafiqpoor, 2001). Heavy clouds, along with high relative air humidity and frequent fogs, produce very low evapotranspiration values, which are comparable to the humid grass páramo of southern Ecuador and to Sumapaz in Colombia (Buytaert & alii, 2007; Rangel & Arellano, 2008). The lack of thermal seasonality is the most obvious feature of the tropical alpine and montane climates (e.g., Hedberg, 1964; Sarmiento, 1986; Rundel, 1994; Bendix & Rafiqpoor, 2001). This is confirmed by our measurements in the high-elevation páramo of Antisana although we also document distinct variation in this general pattern. For instance, the diurnal variation of air temperature is more than seven times greater than the annual variation on the western side of the mountain (tab. 5). Due to permanently high humidity on the opposite north-eastern side, however, daily temperature oscillation is much more limited, which reduces the ratio of daily/annual variation to 2–2.4, i.e., values which are below the common range reported for 75

FIG. 13 - Temporal variation of measured climatic variables and their correlation to the ENSO index. Each variable has plotted the running means of deviations of monthly means (air temperature, relative air humidity, global radiation) or monthly sums (precipitation) from the grand mean (e.g., air temperature and relative air humidity in Tables 2, 3), consistently with the ENSO index provided by NOAA (http://www.cpc.noaa.gov/products/ analysis_monitoring/ensostuff/ensoyears.shtml; accessed at March 2015); solid black – WL, dotted black – WH, solid gray – EL, dotted gray – EH, dashed double line – ENSO index.

tropical alpine environments (Rundel, 1994). The absolute seasonal minimum, along with the mean daily minimum, exhibits the greatest relative variation during the year, consistent with observations from the Venezuelan páramos (Sarmiento, 1986). There is no consistent difference between the two mountain sides in terms of daily vs. seasonal relative variation of solar radiation (tab. 5). This is mainly due to the fact that the occurrence of clouds on the eastern side of the Antisana

TABLE 5 - Ratio of daily to monthly variation of climatic variables at the four climatic stations.

Air temperature

Solar radiation

Relative air humidity

Western 4280 m

7.1

7

5.2

Western 4600 m

7.3

4.8

1.6

North-eastern 4120 m

2

5.8

1.7

North-eastern 4430 m

2.4

6.4

1.1

76

volcano reduces not only the daily radiation input but also its seasonal variation. Air humidity is permanently high on the north-eastern side, so there is virtually neither daily nor annual RH variation. At the western lower site, however, the daily variation of RH is more than five times greater than the annual variation (tab. 5). Since air humidity at the western higher site varies both diurnally and annually, the daily variation relative to the seasonal one is comparable to the northeastern mountain side. Night frost may occur anytime during the year in the tropical high mountains, but particularly during periods of reduced rainfall (Azócar & Monasterio, 1980; Sarmiento, 1986). This pattern is confirmed in the Antisana páramo as the occurrence of freezing temperatures has a distinct seasonal pattern and correlates to the dry months. The minimum temperature (–6.1°C) observed during the three years, as well as the rare occurrence of days with mean air temperature below zero, and no day with permanent frost, are all consistent with the environment of the Colombian superpáramo Los Nevados (4500 m, minimum temperature of –8°C) (Witte, 1995; Javellas & Thouret, 1995). Most freezing periods lasted much less than four hours in the páramo

of Antisana, and only rarely continued over the whole night. Thus, although night freezing temperatures are a common and year-round phenomenon in the equatorial páramo, they are mild and of a short duration, at least in the humid regions. Moreover, the nights are becoming warmer since the frequency of the coldest temperatures is decreasing (Vuille & alii, 2008). In such thermal conditions, freezing avoidance by, e.g., insulation or transient supercooling, appears to be a convenient strategy for páramo plants and insects to cope with subzero-temperature stress (Beck, 1994; Sømme & alii, 1996; Sklená & alii, 2010, 2012). Since day time temperatures become cooler with increasing elevation, whereas nights remain comparably cold on the western side of the Antisana volcano (fig. 7; compare also calculated summed hour-degrees for day and night), the temperature balance is largely driven by radiation during the day rather than heat loss during the night. The mean annual air temperature thus declines from 3.7°C to 2.3°C between the western sites giving a temperature lapse rate of 0.44 K/100 m. This lapse rate (0.46 K/100 m) stays consistent in the upper reaches of the páramo belt since the mean annual temperature is 1.1°C in the vicinity of the glacier (4860 m) on the western side of the mountain (Wagnon & alii, 2009). The north-eastern side of the Antisana volcano is different as both the day and night mean temperatures decrease with elevation. The mean annual air temperature declines from 3.8°C to 2.1°C giving a lapse rate of 0.55 K/100 m. The lapse rates estimated for the high-elevation páramo belt are quite consistent with the values of 0.43 K and 0.50 K per 100 m provided by Cañadas Cruz (1983; cited in Jørgensen & Ulloa, 1994) for the western and eastern slopes of the Ecuadorian Andes, respectively. On the other hand, our values are lower than the lapse rate of 0.59 K/100 m reported for the 2000–4500 m elevational gradient in the Colombian Los Nevados (Javellas & Thouret, 1995). Due to its location virtually at the Equator, the annual variation of mean air temperature is less than 2K in the páramo belt, which is consistent with observations from the Antisana glacier foreland (Villacís, 2008). In this regard, the thermal conditions are indeed aseasonal with important inter-annual variation due to the ENSO events (Francou & alii, 2004; Rabatel & alii, 2013). However, distinct seasonal variation exists with regards to precipitation, to which variations in solar radiation, frequency of frosts, temperature extremes, relative air humidity, and evapotranspiration are correlated. The climatic variation is more pronounced on the western side of the mountain, which thus experiences greater extremes than the north-eastern mountain side. Reduced cloud frequency on the western side, along with lower precipitation, results in higher radiation, higher air temperatures, and lower relative air humidity during the day. Water availability to páramo plants may be occasionally limited during drier periods, especially in poorly developed soils or rocky outcrops (e.g., Pfitsch, 1994). In contrast, high rainfall, in combination with low air temperatures and high air humidity, reduces potential evapotranspiration, thus there is always surplus of water on the eastern side of the Antisana volcano. Therefore, the eastern side of the mountain has very high water yield connected with a steadily

high water discharge. Such climatic variation between the opposite sides of the Antisana volcano is reflected by the distinct species composition and community structure of the páramo vegetation (Sklená & Lægaard, 2003) as well as by variation in the 13C stable isotope signal in plant tissues (Macková & alii, unpublished data).

REFERENCES ANONYMOUS. (1978–1990) Anuario Meteorológico. Instituto Nacional de Meteorología e Hidrología. Ministerio de energia y Minas, Quito. ALLEN R.G., PEREIRA L.S., RAES D. & SMITH M. (1998) - Crop evapotranspiration - Guidelines for computing crop water requirements - FAO Irrigation and drainage paper 56, FAO - Food and Agriculture Organization of the United Nations, Rome. http://www.fao.org/docrep/X0490E/ x0490e00.htm#Contents AZÓCAR A. & MONASTERIO M. (1980) - Caracterización ecológica del clima en el Páramo de Mucubají. In: Monasterio, M. (Ed.): “Estudios ecológicos en los páramos andinos”. Editorial de la Universidad de los Andes, Mérida, 207–223. BARRY R.G. (2008) - Mountain Weather and Climate. 3rd Ed. Cambridge. BECK E. (1994) - Cold tolerance in tropical alpine plants. In: Rundel P.W., Smith A.P. & Meinzer F.C. (Eds.), “Tropical Alpine Environments. Plant Form and Function”. Cambridge, 77–110. BENDIX J. & LAUER W. (1992) - Die Niederschlagsjahreszeiten in Ecuador und ihre klimadynamische Interpretation. Erdkunde, 46, 118–134. BENDIX J. & RAFIQPOOR M.D. (2001) - Studies on the thermal conditions of soils at the upper tree line in the páramo of Papallacta (Eastern Cordillera of Ecuador). Erdkunde, 55, 257–276. BLACK J.M. (1982) - Los páramos del Antisana. Revista Geográfica, 17, 25–52. BOENING C., WILLIS J.K., LANDERER F.W., NEREM S.R. & FASULLO J. (2012) - The 2011 La Niña: So strong, the oceans fell. Geophysical Research Letters, 39, L19602, doi:10.1029/2012GL053055. BUYTAERT W., INIGUEZ V. & DE BIÉVRE B. (2007) - The effects of afforestation and cultivation on water yield in the andean páramo. Forest Ecology and Management, 251, 22–30. CAÑADAS CRUZ L. (1983) - El mapa bioclimático y ecológico del Ecuador. Quito. CLAPPERTON C.M., Hall M., Mothes P., Hole M.J., Still J.W., Helmens K.F., Kuhry P., & Gemmell A.M.D. (1997) - A Younger Dryas icecap in the equatorial Andes. Quaternary Research, 47, 13–28. DIEMER M. (1996) - Microclimatic convergence of high-elevation tropical páramo and temperate-zone alpine environments. Journal of Vegetation Science, 7, 821–830. FAGUA J.C. & GONZALEZ V. H. (2007) - Growth rates, reproductive phenology, and pollination ecology of Espeletia grandiflora (Asteraceae), a giant Andean caulescent rosette. Plant Biology, 9, 127–135. FAVIER V., WAGNON P., CHARAZIN J.P., MAISINCHO L. & COUDRAIN A. (2004) - One-year measurements of surface heat budget on the ablation zone of Antizana Glacier 15, Ecuadorian Andes. Journal of Geophysical Research, 109, D18105, doi:10.1029/2003JD004359. FAVIER V., COUDRAIN A., CADIER E., FRANCOU B., AYABACA E., MAISINCHO L., PRADEIRO E., VILLACIS M. & WAGNON P. (2008) - Evidence of groundwater flow on Antizana ice-covered volcano, Ecuador. Hydrological Sciences Journal, 53, 278–291. FRANCOU B., VUILLE M., FAVIER V. & CÁCERES B. (2004) - New evidence for an ENSO impact on low-latitude glaciers: Antizana 15, Andes of Ecuador, 0°28’S. Journal of Geophysical Research 109: D18106, doi:10.1029/2003JD004484

77

GRUBB P.J. (1970) - The impact of man on the Páramo of Cerro Antisana. Journal of Applied Ecology, 7, 7-8.

SKLENÁř P. (1999) - Nodding capitula in superpáramo Asteraceae: An adaptation to unpredictable environment. Biotropica, 31, 394-402.

HALL M. (1977) - El Volcanismo en el Ecuador. Quito.

SKLENÁř P. (2000) - Vegetation Ecology and Phytogeography of Ecuadorian Superpáramos. Ph.D. dissertation, Charles University, Prague.

HASTENRATH S. (1981) - The Glaciation of the Ecuadorian Andes. Rotterdam. HEDBERG O. (1964) - Features of Afroalpine plant ecology. Acta Phytogeographica Suecica, 49, 1-144. IAEA (2009) - Atlas of isotope hydrology - the Americas. International Atomic Energy Agency, Vienna. JAVELLAS R. & THOURET J.C. (1995) - Estudio de las temperaturas de las estaciones meteorologicas del transecto Parque Los Nevados (Cordillera Central, Colombia). In: van der Hammen T. & Dos Santos A.G. (Eds.), “La Cordillera Central Colombiana. Transecto Parque Los Nevados”. Studies on Tropical Andean Ecosystems 4. Berlin, 241-278. JOMELLI V., FAVIER V., RABATEL A., BRUNSTEIN D., HOFFMANN G. & FRANCOU B. (2009) - Fluctuations of glaciers in the tropical Andes over the last millennium and palaeoclimatic implications: A review. Palaeogeography, Palaeoclimatology, Palaeoecology, 281, 269-282 JØRGENSEN P.M. & ULLOA C.U. (1994) - Seed plants of the high Andes of Ecuador - a checklist. AAU Reports, 34, 1-443. KUDO G. & SUZUKI S. (2004) - Flowering phenology of tropical-alpine dwarf trees on Mount Kinabalu, Borneo. Journal of Tropical Ecology, 20, 563571. LARAQUE A., RONCHAIL J., COCHONNEAU G., POMBOSA R. & GUYOT J.L. (2007) - Heterogeneous distribution of rainfall and discharge regimes in the Ecuadorian Amazon Basin. Journal of Hydrometeorology, 8, 13641381. LAUER W. (1979) - Die hypsometrische Asymmetrie der Páramo Höhenstufe in den nördlichen Anden. Innsbrucker Geografische Studien, 5, 115-130. LAUER W., RAFIQPOOR M.D. & THIESEN I. (2001) - Physiogeohrapgie, Vegetation und Syntaxonomie der Flora des Páramo de Papallacta (Ostkordillere Ecuador). Erdwischenschaftliche Forschung 39. Stuttgart. MAISINCHO L. & CACERES B. (2007) - Glaciares del Ecuador: Antisana y Carihuayrazo. IRD – INAMHI – EMAAP-Quito, Quito. NOAA (2015) - http://www.cpc.noaa.gov/products/analysis_monitoring/ ensostuff/ensoyears.shtml. PFITSCH W.A. (1994) - Morphological and physiological radiation in páramo Draba. In: Rundel P.W., Smith A.P. & Meinzer F.C. (Eds.), “Tropical Alpine Environments. Plant Form and Function”. Cambridge, 151165. RABATEL A. & alii (2013) - Current state of glaciers in the tropical Andes: a multi-century perspective on glacier evolution and climate change. The Cryosphere, 7, 81-102. RANGEL J.O. & ARELLANO H. (2008) - El clima en el área del transecto Sumapaz (cordillera Oriental). In: van der Hammen T. (Ed.), “La Cordillera Oriental Colombiana. Transecto Sumapaz”. Studies on Tropical Andean Ecosystems 7. Berlin, 143-184. RUNDEL P.W. (1994) - Tropical alpine climates. In: RUNDEL P. W., SMITH A. P. & MEINZER F. C. (Eds.), “Tropical Alpine Environments. Plant Form and Function”. Cambridge, 21-44. SARMIENTO G. (1986) - Ecological features of climates in high tropical mountains. In: Vuilleumier F. & Monasterio M. (Eds.), “High Altitude Tropical Biogeography”. Oxford, 11-45. SAUER W. (1971) - Geologie von Ecuador. Berlin.

78

SKLENÁř P. (2006) - Searching for altitudinal zonation: species distribution and vegetation composition in the superpáramo of Volcán Iliniza, Ecuador. Plant Ecology, 184, 337-350. SKLENÁř P., BENDIX J. & BALSLEV H. (2008) - Cloud frequency correlates to plant species composition in the high Andes of Ecuador. Basic and Applied Ecology, 9, 504-513. SKLENÁř P., KUčEROVÁ A., MACEK P. & MACKOVÁ J. (2010) - Does plant height determine the freezing resistance in the páramo plants? Austral Ecology, 35, 929-934. SKLENÁř P., KUčEROVÁ A., MACEK P. & MACKOVÁ J. (2012) - The frost resistance mechanism in neotropical alpine plants is related to geographic origin. New Zealand Journal of Botany, 50, 391-400. SKLENÁř P. & LÆGAARD S. (2003) - Rain-shadow in the high Andes of Ecuador evidenced by páramo vegetation. Arctic Antarctic and Alpine Research, 35, 8-17. SØMME L., DAVIDSON R.L. & ONORE G. (1996) - Adaptation of insects at high altitudes of Chimborazo, Ecuador. European Journal of Entomology, 93, 313-318. TROLL C. (1968) - The cordilleras of the tropical Americas. Aspects of climatic, phytogeographical and agrarian ecology. In: Troll C. (Ed.), “Geoecology of the tropical mountainous regions of the tropical Americas”. Colloquium Geographicum 9. Bonn, 15-56. VAN DER HAMMEN T., WITTE H.J.L. & VAN REENEN G.B.A. (1995) - Aspectos ecoclimáticos del area del transecto Parque Los Nevados. In: VAN DER HAMMEN T. & DOS SANTOS A.G. (Eds.), “La Cordillera Central Colombiana. Transecto Parque Los Nevados”. Studies on Tropical Andean Ecosystems 4. Berlin, 329-364. VILLACÍS M. (2008) - Ressouces en eau glaciaire dans les Andes d’Equateur en relation avec les variations du climat: Le cas du volcan Antisana. Ph.D. Thesis, Université Montpellier II. VUILLE M., BRADLEY R.S. & KEIMIG F. (2000a) - Climate variability in the Andes of Ecuador and it relation to tropical Pacific and Atlantic sea surface temperature anomalies. Journal of Climate, 13, 2520-2535. VUILLE M., BRADLEY R.S. & KEIMIG F. (2000b) - Interannual climate variability in the Central Andes and its relation to tropical Pacific and Atlantic forcing. Journal of Geophysical Research, 105, 12447-12460. VUILLE M., FRANCOU B., WAGNON P., JUEN I., KASER G., MARK B.G. & BRADLEY R.S. (2008) - Climate change and tropical Andean glaciers: Part, present and future. Earth-Science Reviews, 89, 79-96. WAGNON P., LAFAYSSE M., LEJEUNE Y., MAISINCHO L., ROJAS M. & CHAZARIN J.P. (2009) - Understanding and modeling the physical processes that govern the melting of snow cover in a tropical mountain environment in Ecuador. Journal of Geophysical Research, 114, 1-14. WITTE H.J.L. (1995) - Seasonal and altitudinal distribution of precipitation, temperature and humidity in the Parque Los Nevados transect (Central Cordillera, Colombia). In: van der Hammen T. & Dos Santos A.G. (Eds.), “La Cordillera Central Colombiana. Transecto Parque Los Nevados”. Studies on Tropical Andean Ecosystems 4. Berlin, 279-328. (Ms. received 30 August 2014; accepted 1 February 2015)

Geogr. Fis. Dinam. Quat. 38 (2015), 79-87, 4 figg., 2 tab.

DOI 10.4461/GFDQ.2015.38.08

CLAUDIO SMIRAGLIA (*), ROBERTO SERGIO AZZONI (*), CARLO D’AGATA (*), DAVIDE MARAGNO (*), DAVIDE FUGAZZA (*) & GUGLIELMINA ADELE DIOLAIUTI (*)

THE EVOLUTION OF THE ITALIAN GLACIERS FROM THE PREVIOUS DATA BASE TO THE NEW ITALIAN INVENTORY. PRELIMINARY CONSIDERATIONS AND RESULTS ABSTRACT: SMIRAGLIA C., AZZONI R.S., D’AGATA C., MARAGNO D., FUD.&. DIOLAIUTI G.A., The evolution of the Italian glaciers from the previous data base to the New Italian Inventory. Preliminary considerations and results. (IT ISSN 0391-9838, 2015). A glacier inventory is a fundamental tool for describing and managing the Alpine glacierized environment and evaluating the impacts of the ongoing climate change. After the 1959-1962 Italian glacier inventory published by the Italian Glaciological Committee (CGI) in cooperation with the National Research Council (CNR), only regional glacier lists have been developed in Italy, thus giving partial pictures of the evolution of the Italian glaciers. In this work, we summarized the main results from the New Italian Glacier Inventory, a national glacier atlas recently completed and based on the analysis of high resolution color orthophotos which were acquired in the time frame 2005-2011. In the New Italian Glacier Inventory 903 glaciers are described, covering a total area of 369.90 km2 ± 2%. The largest part of the glacier coverage is located in the Aosta Valley Autonomous Region (36.15 % of the total), followed by the Lombardy Region (23.71 %) and the Autonomous Province of Bolzano (23.01 %). The highest number of glaciers was found in Lombardy (230), then in the Autonomous Province of Bolzano (212), in the Aosta Valley Autonomous Region (192), and in the Autonomous Province of Trento (115). About 84 % of the census is composed by glaciers minor than 0.5 GAZZA

———————— (*) Dipartimento di Scienze della Terra “Ardito Desio”, Università degli Studi di Milano Corresponding author: [email protected] The New Italian Glacier Inventory project has been developed in the framework of a cooperation among the Università degli Studi di Milano, Sanpellegrino Spa brand-Levissima and the Association EvK2CNR. This project also took advantage from the precious cooperation of the Comitato Glaciologico Italiano and of several regional and local partners. These latter are both public administrations and private associations of glaciologists, trekkers, climbers and mountain lovers. This work was also performed in the framework of the PRIN project 2010/2011 (2010AYKTAB_006), local leader C. Smiraglia. The project was also accredited and recognized by the World Glacier Monitoring Service, the organization who developed and is managing the World Glacier Inventory. The New Italian Glacier Inventory is an open access data base. The digital copy of the inventory (reporting tables, diagrams and maps) is available on line at http://users.unimi.it/glaciol. The glacier outlines are visible at the open access web-GIS SHARE GEO network ((http://geonetwork.evk2cnr.org/) developed by the EvK2CNR Association which is periodically checked and updated.

km2 covering only the 21% of the total area. Glaciers wider than 1 km2 are 9.4 % of the whole number, but they cover 67.8 % of the total area. In the widest size class (>10 km2), only three glaciers are found. Only 25 glaciers (2.8 % of the census) were classified as “valley glacier”, while the largest part (57.3%) was classified as “mountain glacier” and “glacieret” (40%), thus underlining that the Italian glaciers are spread into several small ice bodies with few larger glaciers. A first comparison between the total area reported in the New Italian Glacier Inventory and the value reported in the CGI –CNR Inventory (1959-1962) suggests an overall reduction of the glacier coverage of about 30% (from 526.88 km2 in the Sixties to 369.90 km2 in the present time). A second comparison was performed with the WGI (World Glacier Inventory) dataset which in the Eighties listed 1381 Italian glaciers covering a total area of 608.56 km2. This comparison suggests a loss of 478 glaciers and an area reduction of 238.66 km2 (-39 %). KEY WORDS: Glacier Inventory, Orthophotos, WebGIS, Alpine glaciers, Italian Alps RIASSUNTO: SMIRAGLIA C., AZZONI R.S., D’AGATA C., MARAGNO D., FUGAZZA D.&. DIOLAIUTI G.A., L’evoluzione del glacialismo italiano dalle banche dati del secolo scorso al nuovo Catasto dei Ghiacciai Italiani. Risultati e considerazioni preliminari. (IT ISSN 0391-9838, 2015). Un catasto glaciale è uno strumento fondamentale per descrivere e gestire l’ambiente glacializzato di alta quota alpino e per valutare gli effetti del Cambiamento Climatico in atto. Dopo il Catasto dei Ghiacciai Italiani prodotto tra il 1959 e il 1962 a cura del Comitato Glaciologico Italiano (CGI) in collaborazione con il Consiglio Nazionale delle Ricerche (CNR), in Italia sono stati pubblicati solo inventari a scala regionale, utili ad una descrizione parziale del glacialismo nazionale e pertanto non completamente esaustivi di tendenze ed evoluzioni generali. Vengono qui presentati in sintesi i risultati conseguiti dal progetto che ha portato alla compilazione del Nuovo Catasto dei Ghiacciai Italiani, un atlante glaciale nazionale prodotto sulla base dell’analisi di ortofoto a colori ad alta definizione acquisite nel periodo 2005-2011. Nel Nuovo Catasto dei Ghiacciai Italiani vengono descritti 903 ghiacciai estesi complessivamente su un’area di 369.90 km2 ± 2%. La maggior parte della superficie glaciale è localizzata nella Regione Autonoma Valle d’Aosta (36.15 % del totale), seguita dalla Regione Lombardia (23.71 %) e dalla Provincia Autonoma di Bolzano (23.01 %). Il maggior numero di apparati glaciali è risultato in territorio lombardo (230), nella Provincia Autonoma di Bolzano (212), nella Regione Autonoma Valle d’Aosta (192), e nella Provincia Autonoma di Trento (115). Circa l’ 84 % dei ghiacciai (per numero) è rappresentato da apparati di dimensioni minori a 0.5 km2 che insieme coprono il 21% della superficie totale. Ghiacciai più estesi di 1 km2 sono solo il 9.4 % del numero complessivo, ma coprono il 67.8 % dell’area glaciale nazionale.

79

Nella classe dimensionale maggiore (>10 km2) sono risultati tre apparati glaciali. Solo 25 ghiacciai (2.8 % del numero totale) sono stati classificati come “vallivi” e la maggior parte del campione (57.3%) è stato classificato come “montano” mentre il 40% è risultato di tipo “glacionevato”, a sottolineare come la risorsa glaciale italiana sia frammentata in un gran numero di piccoli apparati. Un primo confronto tra i dati riportati nel Nuovo Catasto dei Ghiacciai Italiani e quanto pubblicato nel Catasto CGI-CNR (1959-1962) indica una riduzione della superficie glaciale nazionale del 30% (dai 526.88 km2 degli anni ’60 del secolo scorso agli attuali 369.90 km2). Un secondo confronto effettuato considerando il Catasto Glaciale Mondiale (World Glacier Inventory o WGI), pubblicato alla fine degli anni ’80 e che descrive 1381 ghiacciai estesi su un’area di 608,56 km2, suggerisce la perdita di 478 apparati ed una contrazione areale di 238.66 km2 (-39 %). TERMINI CHIAVE: Catasto Glaciale, Ortofoto, WebGIS, Ghiacciai alpini, Alpi Italiane

INTRODUCTION In the recent past glaciers have begun melting at rates that cannot be explained only by natural climate variability (Dyurgerov & Meier, 2000). Glacier shrinkage is particularly severe upon the Alps, and it is likely driven by the important changes occurring in mid-tropospheric conditions, such as the widely acknowledged rapid increase in temperature during the last few decades (IPCC, 2001; 2007; 2013). In the Alps, air warming was estimated to be more than double of the World average over the last half a century (Böhm & alii, 2001), with a significant Summer warming since Seventies (Casty & alii, 2005). Between 1850 and 1980, glaciers in the European Alps lost approximately 1/3rd of their area and ca. 50% of their mass, and since 1980 another 20 to 30% of the ice has melted (EEA, 2004). Among possible methods to analyse the ongoing evolution of cryosphere, collection and analysis of glacier inventories (e.g. glacier area and geometry features) can be used to investigate mountain glaciation in a changing climate (Paul & alii, 2004a), and potential scenarios on the regional Alpine scale (Zemp & alii, 2006). In fact, glacier geometry changes are key variables with respect to strategies for early detection of enhanced greenhouse effects on climate (Kuhn, 1980; Hoelzle & alii, 2003). Glacier inventories should be carried out at intervals compatible with the characteristic response time of mountain glaciers (a few decades or less in the case of small glaciers), and the currently observed glacier down-wasting calls for frequent updates of inventories (Paul & alii, 2007; 2011; Pfeffer & alii, 2014). Also the Italian glaciers are experiencing over the last decades a strong area reduction which is comparable with magnitude and rates of other Alpine glacierized sectors (Diolaiuti & alii, 2012a; 2012b, Paul & alii, 2004a; 2011; Zemp & alii, 2008). Here we summarize data and information reported in the New Italian Glacier Inventory, a recent glacier data base listing all the Italian glaciers and describing their present size, area and geometry. This new record also permits fruitful comparisons with previous national glacier data base, thus giving information on magnitude and rates of glacier shrinkage in the Italian Alps. 80

PREVIOUS ITALIAN GLACIER INVENTORIES Italy has a long and valued tradition in developing glacier inventories. After the first glacier data base realized by Carlo Porro in 1925, the most important work dealing into this context and also being a novelty for that time was the Italian Glacier Inventory, developed by the Italian Glaciological Committee (CGI) in cooperation with the National Research Council (CNR) in the time frame 19591962. This data base, developed by analyzing maps and through field surveys, reported 838 glaciers (considering both actual glaciers and glacierets) which covered a total area of 526.88 km2 (CGI-CNR, 1959; 1961a; 1961b; 1962). At the end of the Seventies (XX century) the CGI was part of the international team leading the World Glacier Inventory (WGI), which was published later as a synthesis (Haeberli & alii, 1989; Serandrei-Barbero & Zanon, 1993) and with full details at the dedicated web page hosted by the WGI web site (www.wgms.ch/). The WGI listed 1381 Italian glaciers covering a total area of 608.56 km2 (Belloni & alii, 1985). The Italian data inserted in the WGI derived from aerial photos, in several cases affected by a non negligible snow coverage. At the end of the Eighties the CGI was charged by the Environment Ministry of the Italian Government of developing a new updated glacier inventory. In the obtained data base the Italian glaciers resulted covering an area of about 480 km2, indicating an actual decrease with respect to both the CGI inventory and the WGI (Ajassa & alii, 1994; 1997). This was the last Italian general inventory, no other studies to develop a national glacier data base were performed. Afterwards, only local (mainly regional) inventories were published (among the others: Desio, 1967; Zanon, 1990; Servizio Glaciologico Lombardo, 1992; Comitato Glaciologico Trentino, 1994; Federici & Pappalardo, 1995, 2009; Citterio & alii, 2007; Bonardi & alii, 2012; Diolaiuti & alii, 2012a; 2012b; Secchieri, 2012; Cerutti, 2013). Then the project named “The New Italian Glacier Inventory” was ideated and developed to fill this scientific gap and in a quite short time frame it has produced an actual updated data base describing the whole Italian glaciation (Smiraglia & Diolaiuti, Eds., 2015). APPLIED METHODS AND SOURCE OF DATA The main source of information for data collecting (in particular glacier area data) are high resolution orthophotos. In fact, to mark glacier boundaries and to calculate glacier areas, recent color orthophotos have been analyzed, which were made kindly available for this research project by regional and local administrations. The orthophotos are derived from high resolution aerial photos featuring low or absent cloud coverage and mainly acquired at the end of the summer when glaciers show the minimum snow mantle and then their limits result clearer and better detectable. The analyzed orthophotos have been surveyed in 2005 (Valle D’Aosta, RAVA Flight); 2007 (Lombardia, digital color orthophoto BLOM-CGR S.p.A.-IIT2000/ VERS.2007); 2008 (Provincia Autonoma di Bolzano Alto

Adige, PAB Flight); 2009 (Veneto: LIDAR survey performed by Regione Veneto-ARPAV Centro Valanghe di Arabba); 2009-2011 (Regione Piemonte, ICE Flight); 2011 (Provincia Autonoma di Trento, PAT Flight). The orthophotos are purchasable products, with a planimetric resolution specified by 1 pixel (pixel size = 0.5 m). The planimetric accuracy stated by the manufacturers is ±1 m. In few cases satellite imagines have been used to improve the glacier mapping (Valle d’Aosta, 2009 SPOT imagines featuring a resolution of 6 m) and field and literature data as well (Friuli -Venezia Giulia and Abruzzo). The color orthophotos have been used as base layer in a GIS (Geographic Information System) environment to detect and map glacier boundaries. These latter permitted the calculation of glacier areas. The area data together with other crucial information (e.g.: glacier name, identification code, coordinates, catchments, etc.) were inserted in a data base which constitutes the New Italian Glacier Inventory. A cross check of the obtained glacier data was performed considering already existent regional or local inventories, recent published maps and cartography and performing dedicated field surveys. The final validation of all the New Italian Glacier Inventory data was performed by a team of experts selected from the Italian Glaciological Committee and /or from technical personnel of local administrations (Smiraglia & Diolaiuti, Eds., 2015). To assess the potential error affecting data inserted in the new inventory the approach introduced by Vögtle & Schilling (1999) has been followed. This method, largely applied in the recent past to evaluate the glacier area error in some Italian regions (Diolaiuti & alii, 2012a, Diolaiuti & alii, 2012b), is based on the calculation for each mapped glacier of the surface area buffer. The final precision of the whole glacier coverage was determined by taking the root of the squared sum of all the buffer areas. Thanks to the

high quality and resolution of the orthophotos and to the accurate manual mapping, the obtained glacier area data featured an error minor than ± 2% of the actual value. Exceptions occur in the case of supraglacial debris presence (i.e.: debris covered glaciers, Kirkbride, 2011; Smiraglia & Diolaiuti, 2011). These conditions are becoming more frequent on the last years (Diolaiuti & Smiraglia, 2010) and make more difficult and uncertain to detect and map glacier outlines. In such conditions the mapped glacier area can be underestimated up to 10% of the actual value. To reduce the error, whenever debris occurs at the glacier surface, we also considered particular glacier morphological features (ice cliffs and ice pinnacles, bediérès and epiglacial streams, supraglacial lakes and water ponds, glacier moulins and surface roughness, dark areas due to higher water content or areas showing clear changes of elevation, meltwater streams originating from heavily debris covered areas) indicating the occurrence of buried ice (Paul & alii, 2004b; 2009). From the orthophotos analysis we also derived information on glacier aspect and type (following the recommendations listed by Paul & alii, 2010; Cogley & alii, 2011; Pfeffer & alii, 2014); these data were inserted in the New Italian Glacier inventory, as well (Smiraglia & Diolaiuti, Eds., 2015). RESULTS THE ACTUAL PICTURE OF THE ITALIAN GLACIATION On the Italian side of the Alps about 1/5th of the whole Alpine glaciation is located with a total glacierized area of 369.90 km2 ± 2%, a non-negligible value if compared to the Alps as a whole (2050 km2, Paul & alii, 2011). The total number of Italian glaciers results 903 (also consider-

TABLE 1 - Area and number of the Italian glaciers sorted according to the Region or Province where they are located Region or Province

Number of glaciers

Cumulative area value (km2)

Region contribution to the total area (percentage with respect to the whole Italian coverage)

Region contribution to the total census (percentage with respect to the total number of Italian glaciers)

PIEDMONT

107

28.92

8%

12%

AOSTA VALLEY Autonomous Region

192

133.73

36%

21%

LOMBARDY

230

87.71

24%

25%

TRENTINO (Autonomous Province of Trento)

115

30.96

8%

13%

SOUTH TYROL (Autonomous Province of Bolzano)

212

85.12

23%

23%

VENETO

38

3.23

1%

4%

FRIULI-VENEZIA GIULIA

7

0.19

0%

1%

ABRUZZO

2

0.04

0%

0%

903

369.90

100%

100%

ITALY

81

FIG. 1 - Aspect frequency distribution. The percent of glacier area (black) and glacier number (black dotted line line) in 45° aspect bins are reported.

ing two small ice bodies on the Apennines which form the Calderone Glacier), a large value with respect to the Alpine census (3370 glaciers, Paul & alii, 2011); they feature an ample distribution, from the Maritime to the Julian Alps. The glacier size and type covers a wide range as well: from the largest Italian glacier, the Adamello ice plateau, to Lys and Forni, large compound basin valley glaciers, to the small mountain glaciers and glacierets. According to the regional distribution applied in the New Italian Glacier Inventory, the largest part of the glacierized area of Italy resulted to be located in the Aosta Valley Autonomous Region (36.15 % of the total), followed by the Lombardy Region (23.71 %) and the Autonomous Province of Bolzano (23.01 %). The other regions host minor values of glacier area (the minima were found in the Autonomous Region of Friuli-Venezia Giulia, 0.05 %, and in the Abruzzo Region, 0.01 %). With regards to the glacier number, the highest one is ranked in Lombardy (230), then in the Autonomous Province of Bolzano (212), in the Autonomous Region of Aosta Valley (192), in the Autonomous Province of Trento (115) and in Piedmont Region (107). A very small number of glaciers is located in Veneto Region, in the Autonomous Region of Friuli-Venezia Giulia and in Abruzzo (38, 7 and 2 respectively) (tab. 1). The mean area value featured by each glacier Region clearly indicates that in the Italian glaciers the small ice bodies dominate: the average value was found 0,41 km2 and at regional level the range of the mean values goes from 0.70 km2 (Aosta Valley) to 0.09 km2 (Veneto). With regard to the 903 inventoried Italian glaciers, the largest part of their area shows a prevalent North aspect (NW, N and NE) (fig. 1). The 61% of the glacierized area and the 54% of the glaciers feature a North, North-West and North East aspect. For a better description of the size distribution of Italian glaciers, the surfaces were sorted according to seven size classes, already used in previous regional and international Alpine inventories (i.e.: 10 km2; Paul & alii, 2004) (fig. 2). 82

The size distribution of the Italian glaciation agrees with the ones found on other sectors of the Alps and other glacierized mountain chains of the Earth (Paul & alii, 2004a, Racoviteanu & alii, 2008; Diolaiuti & alii, 2012a; 2012b) with a predominance of a large number of small ice bodies (i.e.: < 1 km2) and only few large glaciers (i.e.: > 10 km2). Applying this size classification it results that about 84 % of the total glacier number is composed by glaciers minor than 0.5 km2, but covering only the 21% of the total area. Glaciers wider than 1 km2 are about 9.4 % of the total number, but they cover 67.8 % of the total area. In the biggest size class (>10 km2) only three glaciers are found: the Forni Glacier (11.36 km2) in Lombardy Region, the Adamello Glacier (16.44 km2) in both Lombardy Region and in the Autonomous Province of Trento, and the Miage Glacier (10.47 km2) in the Autonomous Region of Aosta Valley. Altogether, these three ice bodies cover about 10.3 % of the Italian glaciation area. A similar picture of the Italian glaciation derives from the type classification. In fact, only 25 glaciers (2.8 % of the total census) were classified as “valley glacier” and the largest part of the sample was labeled as “mountain glacier” (i.e. 517 glaciers corresponding to 57.3%) and “glacieret” (i.e. 361 ice bodies corresponding to 40%), thus further underling a glacier resource spread into several small ice bodies with only few larger glaciers. Moreover the New Italian Glacier inventory also includes an updated bibliography on Italian glaciers. In fact, an updated reference list is a fundamental tool which supports any development of the scientific research, including glaciology. To accomplish this crucial issue, in the New Italian Glacier Inventory a reference list reporting all the scientific papers dealing with Italian glaciers and published in the time window 1962-2014 was included. In the first volume of the CGI Inventory (1959) the list of the scientific publications, which were focusing on the Italian glaciation and were published up to 1961, was reported. After this date some interesting publications reporting a more recent reference list were published as well, among the others the one by Pantaleo (1973) which has updated by Mortara & alii (1995). The main limits of all these works is they only

del Comitato Glaciologico Italiano”, this latter in 1978 became “Geografia Fisica e Dinamica Quaternaria”). To give a more exhaustive picture of the actual Italian Glaciation in the New Italian Glacier Inventory were listed all the papers dealing on Italian glaciers and published in the time window 1962-2014 on national and international journals, books and monographs. THE RECENT CHANGES OF ITALIAN GLACIERS

FIG. 2 - Upper panel) Area frequency distribution of the Italian glaciers (on the left, data are percentage values (%) with respect to the total coverage). Area were sorted according to 7 size classes. The labels show the area value of each size class. Lower panel) Number frequency distribution of the Italian glaciers (on the right, data are percentage values (%) with respect to the total glacier number). Number were sorted according to 7 size classes. The labels show the number of glaciers of each size class (from Smiraglia and Diolaiuti, Eds. 2015).

reported papers and monographs edited by the Italian Glaciological Committee (CGI) and then published on the CGI official journal (i.e.: up to 1977 on the “Bollettino

A first comparison between the total glacier area we reported in the New Glacier Inventory and the glacier coverage value from the CGI Inventory (1959-1962) suggests an overall reduction of the glacier extent of about 30% (from 526.88 km2 in the Sixties to 369.90 km2 in the present time). Considering the glacier regions we have introduced to analyze the Italian glaciation, it resulted an area loss ranging from the stronger reduction experienced by glaciers in Friuli, Piedmont and Veneto to the smaller decrease of Lombardy glaciers (tab. 2). The glacier number resulted increased and 68 new ice bodies were found: in fact, 835 glaciers were listed in the CGI Inventory (this latter reported 838 glaciers but three of them were located in France and then not suitable to be inserted in a national record of data), instead the new inventory described 903 ice bodies. The number increase of glaciers is manly due to both fragmentation phenomena (which are particularly frequent during a glacier retreating phase) and identification of glaciers without any previous mention in the past inventory (generally due to abundant supraglacial rock debris coverage which may have suggested in the past to consider extinct the glacier, now the high resolution source of data we analyzed permitted to detect evidence of glacier activity and then to analyze the ice body and insert its data and features in the new inventory). Then to better analyze the area changes occurred over a time frame of half a century, we compared only the glaciers listed in both the records of data (CGI Inventory and the New Italian Glacier Inventory) (fig. 3); in case of fragment-

TABLE 2 - Changes of area and number of the Italian glaciers sorted according the Regions where they are located Region

Glacier number in the New Glacier Inventory

Glacier number in the CGI Inventory

Change of glacier number

Area change (%)

PIEDMONT

107

115

-8

-48%

AOSTA VALLEY

192

204

-12

-26%

LOMBARDY

230

185

45

-24%

TRENTINO

115

91

24

-33%

SOUTH TYROL

212

206

6

-30%

VENETO

38

26

12

-43%

FRIULI

7

7

0

-50%

ABRUZZO

2

1

1

-33%

TOTAL

903

835

68

-30%

83

FIG. 3 - Area changes of a subset of Italian Glaciers (we compared the ones reported in both the records of data) in the time window 1959/62 and 2005/11. On the X axis the glacier Regions are reported, on the Y axis the area decrease (%) is indicated. The values in bold at the top of each columns are the number of glaciers compared for each glacier Region.

ed glaciers we compared the CGI area value with the one obtained by crediting the area data of all the glacier fragments listed in the New Italian Glacier Inventory. The subset of glaciers common to both the inventories is composed by 622 ice bodies. They were covering 500.96 km2 in the Sixties and 363.76 2 km in the present time thus giving a total area loss of 137.20 km2, about -27 % of the total glacier coverage in the CGI inventory. Considering the area changes occurred in each glacier region we found the stronger decreases in Piedmont and Friuli which had lost more than 40% of their previous coverage even if they contributed only to 17% of the total area change (fig. 3). Minor diminishments resulted in Lombardy and Trentino (both have experienced a decrease of ca. 24% of their previous area but together they contributed to about 34% of the total glacier loss) and South Tyrol (26% or area reduction with respect to the past regional extent equal to 21.60% of the total area change ). Aosta Valley has lost 24% of its previous area giving the largest contribution to the national glacier loss (30.50%). A second comparison was performed with the WGI dataset which in the Eighties listed 1381 Italian glaciers covering a total area of 608.56 km2. This comparison suggest a loss of 478 glaciers and an area reduction of 238.66 km2 (39 % of the total). The largest decreases area found in Friuli (73 %), Piedmont (57 %) and Trentino (56 %). DISCUSSION AND CONCLUSIONS Firstly we compare the area changes we have evaluated for the Italian glaciers with respect to the ones occurred on the second half of the XX century in other glacierized areas of Europe. As regards of glaciers in the Svalbard, Nuth & alii (2013) developed a multi-temporal digital inventory of Svalbard 84

glaciers with the most recent from the late 2000s containing 33 775 km2 of ice covering 57 % of the total land area of the archipelago. The glacierized area has decreased by an average of 80 km2 a −1 over the past 30 years, representing a reduction of 7 %. Andreassen & alii (2008) have analyzed glaciers in the Jotunheimen and Breheimen region (Norway) in the time frame 1930-2003 and the glaciers they investigated shrank since the 1930s with an overall area reduction of about 23% for 38 glaciers. As regards Iceland, this is a glaciated country. Approximately 11% of Iceland’s total area of roughly 100.000 km2 is covered by glaciers. The estimated coverage loss per year is about 0.2% overall, which amounts to 20-30 km2 becoming ice free every year (data from University of Iceland). In the European Alps, Maisch (2000) evaluated an overall decrease of 27% from the midnineteenth century to the mid-1970s, and losses were even greater in some areas. Even more striking was the recession in the Berne-Valais area during 1973-98 (Kääb & alii, 2002). The strongest area reduction is affecting small glaciers (i.e. glaciers with area < 1 km2), that cover roughly 80% of the census in the Alps and make an important contribution to water resources (Citterio & alii, 2007a; Federici & Pappalardo, 2010; Bocchiola & alii, 2010; Diolaiuti & alii , 2012a, b; D’Agata & alii, 2014). Paul & alii (2004a) estimated that 44% of the glacier area decrease during 1973-1998/1999 was charged to glaciers smaller than 1 km2, encompassing 18% of the total area in 1973. From the new Swiss Glacier Inventory SGI2010 (Fischer et alii, 2014), the total glacierized area resulted 944.3 ±24.1 km2 and the area changes are -362.6 km2 (-27.7%, or -0.75% a–1) between 1973 and 2010. Lambrecht & Kuhn (2007) estimated that the Austrian glaciation experienced an area decrease of about 17% during 1969-1998. French glaciers recent changes are analyzed by Gardent & alii (2014), who realized the first multitemporal French Glacier Inventory. They found an extent decrease by 25% between 1967–71 and 2006–09. Glaciers in the French

FIG. 4 - Glacier area changes (reported as percentage (%) of area lost per year) of a selection of European glaciers. The input data are derived from the literature cited in the text. For the Italian glaciers two bars are plotted. Italian glaciers I is the area change computed by comparing the CGI data to the values reported in the New Inventory. Italian Glaciers II is evaluated by comparing the WGI data describing Italian glaciers with the New Italian Inventory record. This second yearly area change resulted highest thus suggesting stronger retreat over the last decades. The diagram underlines that glacier size is the most important factor which modulates magnitude and rates of glacier shrinkage. Glaciers in the North of Europe are wider and are retreating slower, glaciers in the central and southern sector are smaller and are experiencing faster changes.

Alps covered 369 km2 in 1967/71 and 275 km2 in 2006/09 (fig. 4). Secondly we discuss our results to understand the meaning of our findings and their reliability. As regards the evaluation of the overall error affecting our comparisons we cannot state an actual value since no information on data accuracy were provided by the CGI surveyors; differently the New Glacier Inventory resulted featuring a mean error always minor than ± 2% of the computed area value. Then a different strategy has to be applied to obtain an estimation of the reliability of CGI Inventory data and on the potential error affecting such data base (Smiraglia & Diolaiuti, Eds., 2015). A suitable occasion to evaluate the accuracy of CGI data is offered by regional glacier database developed in the recent years by analysing aerial photographs acquired in the Fifties. Among the others, the glacier records by Diolaiuti & alii (2011) and D’Agata & alii (2014), who described two representative sub samples of Lombardy glaciers in the time frames 1954-2003 and 1954-2007 respectively, give the possibility of a comparison with CGI Inventory area data. Diolaiuti & alii (2011) estimated an area change of 28 glaciers in the Dosdè Piazzi group (Lombardy Alps) during 1954–2003 of –3.97 km2 (–51% of the area coverage in 1954). D’Agata & alii (2014) studied the recent evolution of 43 glaciers located in the Ortles-Cevedale group (Stelvio National Park, Italy) by analysing surface area changes from 1954 to 2007; these subset of glaciers chosen for the analysis are among the best known and studied of Italy, also comprising the widest Italian valley glacier (namely Forni). The analysis provided area surface changes as -19.43 km2 ± 1.2 %, approximately -40 %, from 1954 to 2007. Since these authors published detailed tables reporting all the data (stating an error minor than ± 5% of the calculated area), it

was possible for 55 ice bodies a deeper analysis by comparing the 1954 area value with the one reported in the CGI Inventory (and dating back to 1957-1958 time frame). Then for each glacier of the subset we evaluated the departure between 1954 value and CGI inventory one. The absolute values resulted ranging from a minimum of 0.02 km2 to a maximum of 5.92 km2, with an average of 0.38 km2. Adding all the absolute departure values, it resulted a misunderestimated area of 18.12 km2 and of 2.67 km2 for the OrtlesCevedale and the Dosdè-Piazzi group respectively. These values correspond to an error of about 37% and 35% of the actual area featured by the two analysed mountain groups. The largest error resulted affecting the widest Italian valley glacier (i.e. Forni, 20 km2 in the CGI Inventory and 14.08 km2in D’Agata & alii, 2014). In spite of this large overestimation experienced by the Forni glacier, the largest part of glaciers in the subset resulted affected by underestimations (from 5% to 80%). Surely this is a comparison of a small subset with respect to the whole Italian glaciation (903 glaciers in the new inventory) and the date of acquisition of the compared datasets is not perfectly overlapping (1954 vs 1957-1958), nevertheless it can give some elements for further discussion. From the one hand our results suggested that the CGI Inventory values are affected by a non-negligible error mainly due to the source of data applied, thus limiting the reliability of the area changes derived from the comparison with the New Italian Glacier Inventory (which instead is based on high resolution imagery). The error we have evaluated for the glacier subset suggests a underestimated area of the whole Italian glaciation in the CGI inventory of about 30%, thus suggesting larger area losses over the last half a century in stronger agreement with the other Alpine litera85

ture (Maisch 2000, Kaab & alii, 2002; Knoll & Kerschner 2009; Diolaiuti & alii, 2011; D’Agata & alii, 2014). From the other hand, in spite of the wide errors probably affecting the whole CGI Inventory sample, a general regression trend is found and this is a truthful environmental impact of climate change which suggests to perform further investigations on the Italian glacier resource based on high resolution imagery describing the past coverage thus allowing better estimations of glacier variations.

REFERENCES AJASSA R., BIANCOTTI A., BIASINI A., BRANCUCCI G., CAPUTO C., PUGLIESE F. & SALVATORE M.C. (1994) - Catasto dei ghiacciai italiani: primo confronto tra i dati 1958 e 1989. Il Quaternario, 7 (1), 497-502. AJASSA R., BIANCOTTI A., BIASINI A., BRANCUCCI G., CARTON A. & SALVATORE M.C. (1997) - Changes in the number and area of Italian Alpine glaciers between 1958 and 1989. Geografia Fisica e Dinamica Quaternaria, 20, 293-297.

COMITATO GLACIOLOGICO ITALIANO – CONSIGLIO NAZIONALE DELLE RICERCHE (1962) - Catasto dei Ghiacciai Italiani, Anno Geofisico Internazionale 1957-1958. Ghiacciai delle Tre Venezie (escluso Ortles-Cevedale) e dell’Appennino. Torino, Comitato Glaciologico Italiano, v. 4, 309 pp. COMITATO GLACIOLOGICO TRENTINO (1994) - I ghiacciai del Parco Naturale Adamello-Brenta, Nuova Stampa Rapida, Trento, 127 pp. DESIO A. (1967) - I ghiacciai dell’Ortles Cevedale. Comitato Glaciologico Italiano, Torino, 875 pp. D’AGATA C., BOCCHIOLA D., MARAGNO D., SMIRAGLIA C. & DIOLAIUTI G. (2014) - Glacier shrinkage driven by climate change in the OrtlesCevedale group (Stelvio National Park, Lombardy, Italian Alps) during half a century (1954-2007). Theoretical Applied Climatology, 116 (1-2), 169-190. DIOLAIUTI G. & SMIRAGLIA, C. (2010) - Changing glaciers in a changing climate: how vanishing geomorphosites have been driving deep changes on mountain landscape and environment. Géomorphologie: relief, processus, environnement GPRE, (2), 131-152. DIOLAIUTI G., MARAGNO D., D’AGATA C., SMIRAGLIA C. & BOCCHIOLA D. (2011) - A contribution to the knowledge of the last fifty years of Alpine glacier history: the1954-2003 area and geometry changes of Dosdè Piazzi glaciers (Lombardy-Alps, Italy). Progress in Physical Geography, 35(2), 161-182.

ANDREASSEN L. M., PAUL F., KÄÄB A. & HAUSBERG J.E. (2008) - Landsatderived glacier inventory for Jotunheimen, Norway, and deduced glacier changes since the 1930s. The Cryosphere, 2, 131-145.

DIOLAIUTI G., BOCCHIOLA D., D’AGATA C. & SMIRAGLIA C. (2012a) - Evidence of climate change impact upon glaciers’ recession within the Italian alps: the case of Lombardy glaciers. Theoretical and Applied Climatology, 109(3-4), 429-445.

BELLONI S., CATASTA G. & SMIRAGLIA C. (1985) - Problematiche e indicazioni per un nuovo catasto dei ghiacciai italiani sulla base del confronto fra il Catasto Italiano 1959-1962 e il World Glacier Inventory, Geografia Fisica e Dinamica Quaternaria, 8, 166-181.

DIOLAIUTI G., BOCCHIOLA D., VAGLIASINDI M., D’AGATA C. & SMIRAGLIA C. (2012b) - The 1975-2005 glacier changes in Aosta Valley (Italy) and the relations with climate evolution. Progress in Physical Geography, 36 (6), 764-785.

BÖHM R.I., AUER A., BRUNETTI M.B., MAUGERI M., NANNI C. & SCHÖNER B.W. (2001) - Regional temperature variability in the European Alps: 1760-1998 from homogenized instrumental time series. International Journal of Climatology, 21, 1779-1801, DOI:10.1002/joc.689.

DYURGEROV M.B. & MEIER M.F. (2000) -Twentieth century climate change: Evidence from small glaciers. PNAS, 97(4), 1406-1411. EUROPEAN ENVIRONMENT AGENCY, EEA (2004) - Impacts of Europe’s changing climate.107 pp., ISBN: 92-9167-692-6.

BONARDI L., ROVELLI E., SCOTTI R., TOFFALETTI A., URSO M. & VILLA F. (2012) - I ghiacciai della Lombardia, evoluzione ed attualità. Hoepli, Milano, 328 pp.

FEDERICI P.R. & PAPPALARDO M. (1995) - L’evoluzione recente dei ghiacciai delle Alpi Marittime. Geografia Fisica e Dinamica Quaternaria, 18, 257-269.

CASTY C., WANNER H., LUTERBACHER J., ESPER J. & BÖHM R. (2005) - Temperature and precipitation variability in the European Alps since 1500. International Journal of Climatology, 25, 1855-1880.

FEDERICI P.R. & PAPPALARDO M. (2009) - I ghiacciai delle Alpi Marittime. Vicende morfologiche, climatiche e storiche. Bollettino della Società Geografica Italiana, serie XIII, 2, 827-874.

CERUTTI A.V. (2013) - Fifty Years of glacial variations and climate evolution on the Mont Blanc Massif. Geografia Fisica e Dinamica Quaternaria, 36, 225-239.

FEDERICI P.R. & PAPPALARDO M. (2010) - Glacial retreat in the Maritime Alps. Geografis Annaler, Series A, Physical Geography, 92 (3), 353365.

CITTERIO M., DIOLAIUTI G., SMIRAGLIA C., D’AGATA C., CARNIELLI T., STELLA G. & SILETTO G. B. (2007) - The fluctuations of Italian glaciers during the last century: a contribution to knowledge about Alpine glacier changes. Geografiska Annaler, 89, A3, 164-182.

FISCHER M., HUSS M., BARBOUX C. & HOELZLE M. (2014) - The new Swiss Glacier Inventory SGI2010: relevance of using high-resolution source data in areas dominated by very small glaciers. Arctic, Antarctic, and Alpine Research, 46, 4, 933–945.

COGLEY J.G., HOCK R., RASMUSSEN L.A., ARENDT A.A., BAUDER A., BRAITHWAITE R.J., JANSSON P., KASER G., MÖLLER M., NICHOLSON L. & ZEMP M. (2011) - Glossary of Glacier Mass Balance and Related Terms, IHP-VII Technical Documents in Hydrology No. 86, IACS Contribution No. 2, Paris, UNESCO-IHP. COMITATO GLACIOLOGICO ITALIANO – CONSIGLIO NAZIONALE DELLE RICERCHE (1959) - Catasto dei Ghiacciai Italiani, Anno Geofisico Internazionale 1957-1958. Elenco generale e bibliografia dei ghiacciai italiani. Torino, Comitato Glaciologico Italiano, v. 1, 172 pp. COMITATO GLACIOLOGICO ITALIANO – CONSIGLIO NAZIONALE DELLE RICERCHE (1961a) - Catasto dei Ghiacciai Italiani, Anno Geofisico Internazionale 1957-1958. Ghiacciai del Piemonte. Torino, Comitato Glaciologico Italiano, v. 2, 324 pp. COMITATO GLACIOLOGICO ITALIANO – CONSIGLIO NAZIONALE DELLE RICERCHE (1961b) - Catasto dei Ghiacciai Italiani, Anno Geofisico Internazionale 1957-1958. Ghiacciai della Lombardia e dell’Ortles-Cevedale.Torino, Comitato Glaciologico Italiano, v. 3, 389 pp.

86

GARDENT M., RABATEL A., DEDIEU J.P. & DELINE P. (2014) - Multitemporal glacier inventory of the French Alps from the late 1960s to the late 2000s. Global and Planetary Change, 120, 24–37. HOELZLE M., HAEBERLI H., DISCHL M.& PESCHKE W. (2003) - Secular glacier mass balances derived from cumulative glacier length changes. Global Planetary Change, 36, 295-306. HAEBERLI W., BÖSCH H., SCHERLER K., ØSTREM G. & WALLÈN C.C. (eds.) (1989) - World Glacier Inventory - Status 1988, AHS-UNEP-UNESCO. KÄÄB A., PAUL F., MAISCH M., HOELZLE M. & HAEBERLI W. (2002) - The new remote sensing derived Swiss glacier inventory: II. First results. Annals of Glaciology, 34, 362-366. KIRKBRIDE M. (2011) - Debris Covered-Glaciers. Iin: Singh V.P., Singh P. & Haritashya U.K. (Eds.), “Encyclopedia of Snow, Ice and Glaciers”, Dordrecht, Springer, 190-192. KNOLL C. & KERSCHNER H. (2009) - A glacier inventory for South Tyrol, Italy, based on airborne laser scanner data. Journal of Glaciology, 50, 46-52.

KUHN M. (1980) - Climate and glaciers. IAHS, 131, 3-20. IPCC (2001) - Climate change 2001 the scientific basis. Contribution of Working Group II to the Third Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, Section 4.3.11. IPCC (2007) - Fourth Assessment Report: Climate Change 2007. Synthesis Report (AR4). WMO-UNEP. IPCC (2013) - Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change (Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley Eds.). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 1535 pp, doi:10.1017/ CBO9781107415324 LAMBRECHT A. & KUHN M. (2007) - Glacier changes in the Austrian Alps during the last three decades, derived from the new Austrian glacier inventory. Annals of Glaciology, 46, 177–184. MAISCH M. (2000) - The longterm signal of climate change in the Swiss Alps. Glacier retreat since the Little Ice Age and future ice decay scenarios. Geografia Fisica e Dinamica Quaternaria, 23, 139–52. MORTARA G., MERCALLI L., DUTTO F. & CASAGRANDE A. (1995) - Aggiornamento della “Bibliografia analitica dei ghiacciai italiani nelle pubblicazioni del C.G.I.” di Michele Pantaleo. Comitato Glaciologico Italiano, Archivio, 1, 97 pp. NUTH C., KOHLER J., KÖNIG M., VON DESCHWANDEN A., HAGEN J. O., KÄÄB A., MOHOLDT & PETTERSSON R. (2013) - Decadal changes from a multitemporal glacier inventory of Svalbard. The Cryosphere, 7, 1603–1621. PANTALEO M. (1973) - Bibliografia analitica dei ghiacciai italiani nelle pubblicazioni del C.G.I. Bollettino del Comitato Glaciologico Italiano, suppl. 21, 127 pp. PAUL F., KÄÄB A. & HAEBERLI W. (2007) - Recent glacier changes in the Alps observed from satellite: Consequences for future monitoring strategies. Global and Planetary Change, 56(1/2), 111-122. PAUL F., KÄÄB A., MAISCH M., KELLENBERGER T. & HAEBERLI, W. (2004a) - Rapid disintegration of Alpine glaciers observed with satellite data. Geophysical Research Letters, 31, L21402. PAUL F., HUGGEL C. & KÄÄB A. (2004b) - Combining satellite multispectral image data and a digital elevation model for mapping debris-covered glaciers. Remote Sensing of Environment, 89, 510–518. PAUL F., BARRY R.G., COGLEY J.G., FREY H., HAEBERLI W., OHMURA A., OMMANNEY C.S.L., RAUP B., RIVERA A. & ZEMP M. (2009) - Recommendations for the compilation of glacier inventory data from digital sources. Annals of Glaciology, 50 (53), 119–126. PAUL F., BARRY R.G., COGLEY J.G., FREY H., HAEBERLI W., OHMURA A., OMMANNEY C.S.L., RAUP B., RIVERA A. & ZEMP, M. (2010) - Recommenda-

tions for the compilation of glacier inventory data from digital sources. Annals of Glaciology, 50(53), 119-126. PAUL F., FREY H. & LE BRIS R. (2011) - A new glacier inventory for the European Alps from Landsat TM scenes of 2003: challenges and results. Annals of Glaciology, 52(59), 144-152. PFEFFER W.T., ARENDT A.A., BLISS A., BOLCH T., COGLEY J.G., GARDNER A.S., HAGEN J.O., HOCK R., KASER G., KIENHOLZ C., MILES E.S., MOHOLDT G., MÖLG N., PAUL F., RADIC´ V., RASTNER P., RAUP B.H., RICH J., SHARP M.J. & THE RANDOLPH CONSORTIUM (2014) - The Randolph Glacier Inventory: a globally complete inventory of glaciers, Journal of Glaciology, 60 (221), 537-552. RAU F., MAUZ F., VOGT S., KHALSA S. J. S. & RAUP B. (2005) - Illustrated GLIMS Glacier Classification Manual: Glacier Classification Guidance for the GLIMS Glacier Inventory. Regional Center 18 “Antarctic Peninsula”. RACOVITEANU A.E., ARNAUD Y., WILLIAMS M.W. & ORDONEZ J. (2008) - Decadal changes in glacier parameters in the Cordillera Blanca, Peru, derived from remote sensing. Journal of Glaciology, 54(186), 499-510. SECCHIERI F. (2012) - I ghiacciai delle Dolomiti. Editrice La Galaverna, Battaglia Terme, 222 pp. SERANDREI-BARBERO R. & ZANON G. (1993) - The Italian Alps. In: Williams R.S. & Ferrigno J.G. (Eds.), “Satellite Image Atlas of Glaciers of the World – Europe”. USGS Professional Paper 1386-E, Washington D.C., 37-48. SERVIZIO GLACIOLOGICO LOMBARDO (1992) - Ghiacciai in Lombardia. Bolis, Bergamo, 368 pp. SMIRAGLIA C. & DIOLAIUTI G .A. (2011) - Epiglacial Morphologies. In: Singh V.P., Singh P. & Haritashya U.K. (Eds.), “Encyclopedia of Snow, Ice and Glaciers”, Dordrecht, Springer 262-268. SMIRAGLIA C. & DIOLAIUTI G. (Eds.) (2015) - The New Italian Glacier Inventory. Ev-K2-CNR Publ., Bergamo, 400 pp. VÖGTLE T & SCHILLING K. (1999) - Digitizing Maps. In: Bähr H.P. & Vögtle T. (Eds.), “GIS for Environmental Monitoring”, Schweizerbart, Stuttgart, 201-216. ZANON G. (1990) - I Ghiacciai del Veneto. Canova Ed., Treviso. ZEMP M., PAUL F., HOELZLE M. & HAEBERLI W. (2008) - Glacier fluctuations in the European Alps, 1850–2000: an overview and a spatiotemporal analysis of available data. In: Orlove B., Wiegandt E. & Luckman B.H. (Eds), “Darkening peaks: Glacier retreat, science, and society”. University of California Press, Berkeley, CA, 152-167. ZEMP M., HAEBERLI W., HOELZLE M. & PAUL F. (2006) - Alpine glaciers to disappear within decades? Geophysical Research Letters, 33, L13504. (Ms received 1 March 2015; accepted 1 June 2015)

87