The transfer of modern organic carbon by ... - Wiley Online Library

11 downloads 0 Views 2MB Size Report
Aug 9, 2012 - impact on above- and below-ground carbon pools and its potential for ... sites, mass wasting could be either a net source or sink of carbon.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117, G03016, doi:10.1029/2011JG001838, 2012

The transfer of modern organic carbon by landslide activity in tropical montane ecosystems Carlos E. Ramos Scharrón,1,2 Edwin J. Castellanos,3 and Carla Restrepo1 Received 17 August 2011; revised 4 May 2012; accepted 23 June 2012; published 9 August 2012.

[1] Geomorphic processes play an important role in the transfer and storage of carbon within steep mountainous terrain. Among these, mass wasting stands out because of its impact on above- and below-ground carbon pools and its potential for releasing or sequestering carbon. A combined remote-sensing and GIS modeling approach was used to quantify the amount and spatial redistribution of modern organic carbon mobilized by mass wasting activity in a tropical mountain setting. The study focused on a population of hundreds of shallow, translational landslides triggered by Hurricane Mitch (1998) on seven watersheds draining the southern flank of the Sierra de Las Minas mountain range (SLM) in central-eastern Guatemala. Results illustrate that mass wasting contributed to the transfer of 43  104 MgC, or 3%, of the pre-event C in above-ground vegetation and soils for an equivalent carbon flux rate of 0.08–0.33 MgC ha 1 y 1, depending on whether we consider Hurricane Mitch to be a landslide-triggering event with a 20-year or an 80-year recurrence interval. While 30% of this carbon was delivered to hillslopes or first-order streams with a presumed high potential for long-term sequestration, the remaining 70% was delivered to higher-order streams with unknown carbon retention capabilities. Therefore, the ultimate fate of the carbon released by landsliding is very uncertain, but depending on the proportion sequestered by colluvial deposits, the recurrence interval of landslide-triggering events, and the rate of ecosystem recuperation at the landslide failure sites, mass wasting could be either a net source or sink of carbon. In a simulated setting based on the SLM study results in which all carbon transferred by landslides from all tropical mountains of the globe is released to the atmosphere, it would represent an amount equivalent to 1%–11% of the global carbon currently being released by the burning of fossil fuels. Meanwhile, in a projected scenario where a significant proportion of the carbon transferred by landslides is retained within sedimentary deposits, sequestration rates would equal 2%–19% of the residual land sink. Citation: Ramos Scharrón, C. E., E. J. Castellanos, and C. Restrepo (2012), The transfer of modern organic carbon by landslide activity in tropical montane ecosystems, J. Geophys. Res., 117, G03016, doi:10.1029/2011JG001838.

1. Introduction [2] Observed linkages between increases in atmospheric carbon (C) concentration and global climate change have stimulated innovative research that has significantly improved our conceptual and quantitative understanding of the Earth’s C cycle [Solomon et al., 2007]. Although fossil fuel burning and land-use change have captured most of the attention due to their direct and first-order control on atmospheric C levels 1 Department of Biology, University of Puerto Rico–Río Piedras, San Juan, Puerto Rico, USA. 2 Department of Geography and the Environment, University of Texas at Austin, Austin, Texas, USA. 3 Centro de Estudios Ambientales, Universidad del Valle de Guatemala, Guatemala City, Guatemala.

Corresponding author: C. E. Ramos Scharrón, Department of Geography and the Environment, University of Texas at Austin, Mailcode 3100, Austin, TX 78712, USA. ([email protected]) ©2012. American Geophysical Union. All Rights Reserved. 0148-0227/12/2011JG001838

[Houghton et al., 1998; Solomon et al., 2007], human-altered hydrologic and geomorphic processes are increasingly being examined due to their indirect, and potentially central role on C dynamics [Berhe et al., 2007; Schulze et al., 2000]. In particular, accelerated surface erosion by sheetflow and rilling on cultivated lands coupled with sediment retention in a variety of storage compartments has been proposed to play a substantial role in C budgets [Downing et al., 2008; Harden et al., 1999; Liu et al., 2003; Smith et al., 2005; Stallard, 1998; Van Oost et al., 2007]. In addition to sheetflow and rilling, other geomorphic processes such as gullying [Gomez et al., 2003], soil creep [Yoo et al., 2005], and mass wasting [Hilton et al., 2008a; Stark et al., 1999] may influence the C cycle both in natural and human-modified environments. Among these, mass wasting stands out due to its noticeable impact on aboveand below-ground C pools in mountainous terrain, and consequent potential influence on organic C fluxes at multiple scales (Figure 1). [3] The contribution of mass wasting to the C cycle is typically assessed using two independent but complimentary

G03016

1 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Figure 1. Conceptual model illustrating main transfers of organic C associated with landsliding: pools are illustrated with circles and fluxes with arrows. (a) C is stored in above- (ABG) and below-ground biomass (BGB), dead organic matter (litter (LL) and coarse woody debris (CWD)), and soil organic matter (SOM). Transfers among these pools at small scales are mediated primarily by biotic processes (primary production, litterfall, and decomposition (black continuous arrows) and respiration (black dashed arrows)). At increasingly larger scales geomorphic processes such as landsliding (green arrows) contribute to lateral transfers of carbon, including storage in colluvium [modified after Schulze et al., 2000]. (b) At even larger scales, the carbon in the biomass (B) and SOM mobilized by landsliding (green arrow) and fluvial processes (blue arrows) enters colluvial and alluvial deposits along the hillslope-fluvial interface. The degree of coupling between hillslope and fluvial processes mediated by mass wasting will likely determine the residence time of C mobilized by landsliding within watersheds. All arrows in Figure 1b as in Figure 1a. approaches (Figure 1). The first, centers on the hillslope system and quantifies rates of biomass and C removal by mass wasting [Hilton et al., 2011; Restrepo et al., 2003; Stark et al., 1999] as well as rates of biomass and C accumulation in recently formed landslides [Pandey and Singh, 1985; Reddy and Singh, 1993; Turk and Graham, 2009; Walker and Shiels, 2008; Wilcke et al., 2003; Zarin and Johnson, 1995a, 1995b] (Figure 1a and Figure 1b, green arrows). This volume of work has shown that mass wasting can mobilize significant amounts of C (0.1–0.9 MgC ha 1 y 1) and contribute to C recovery (0.3–1.5 MgC ha 1 y 1) on landslide scars during ecosystem development. The second approach has focused on the fluvial system, particularly on watershed-scale measurements of particulate (POC) and dissolved (DOC) organic C delivered by rivers draining mountain ranges strongly influenced by mass wasting [Goldsmith et al., 2008; Gomez et al., 2003; Hilton et al., 2008a, 2008b; Leithold et al., 2006] (Figure 1b, blue arrows). Studies based on POC and DOC analyses not only have corroborated early observations about the disproportionate contribution of small mountainous rivers to global C fluxes [Lyons et al., 2002] but also, by establishing the age and general provenance of C, have provided insight into the role of mass wasting activity in the delivery of organic C derived from standing vegetation and soil (modern C) as well as from sedimentary rocks (fossil C) [Gomez et al., 2003; Hilton et al., 2008a, 2008b; Leithold et al., 2006, 2005] (Figure 1b, blue arrows). This approach has yielded fluxes of modern organic C associated to mass wasting in the range of 0.1–1.1 MgC ha 1 y 1 [Gomez et al., 2003; Hilton et al., 2008a, 2008b]. Altogether this work not only has highlighted the importance, albeit little understood role, of mass wasting in the C cycle but also has uncovered the difficulties in quantifying its net effect due to complex

linkages between hillslope and fluvial processes, including C burial, and their variability at multiple spatiotemporal scales. [4] Mass wasting may induce changes beyond the boundaries of the initial zone of hillslope failure. Individual landslides may simultaneously cause both scour and deposition of organic C-rich material, such as large woody debris (LWD) and soil, in different parts of the landscape [Benda and Dunne, 1997a; Benda et al., 2005]. The potential for further remobilization or long-term storage of the scoured material will be influenced by a variety of colluvial and fluvial processes, and the extent to which these processes are coupled [Lin et al., 2008; Harvey, 2001; Korup, 2005; Schwab et al., 2008; West et al., 2011]. For example, different rates of C decomposition and transport are naturally expected once the material moves from an unchannelled hillslope where colluvial processes dominate to a stream segment with larger contributing areas where fluvial processes control the movement and storage of material [Benda and Dunne, 1997b; Dietrich et al., 1995]. Therefore, redistribution from the upper, steeper portions of a hillslope to its lower, and more gently sloping base, including low-order headwater streams can also represent important changes in its potential for further displacement [but see Densmore and Hovius, 2000; Hovius et al., 2011]. The end result is that while in some cases export to the watershed outlet may occur within a matter of days following its release by mass wasting [e.g., Hovius et al., 2000], material contained within landslide deposits may remain locally stored from centennial to millennial time scales [e.g., Benda et al., 2005; Kelsey, 1982; Lancaster and Casebeer, 2007; Pearce and Watson, 1986] and thus represent a considerable C sink (Figure 1b). [5] Mass wasting varies widely in space and time, and this not only translates into equally variable C fluxes but also into difficulties for its quantification. For example, sediment

2 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Figure 2. Map of Guatemala showing the locations of (top) the Sierra de Las Minas, including the Polochic and Motagua rivers; (middle) the seven studied watersheds on the Motagua side of the Sierra de Las Minas, with the gray features representing the landslides triggered by Hurricane Mitch in 1998 and digitized as part of this work. (bottom) A closeup of the Rio Hondo watershed showing main rivers and landslide deposits, with the red circles illustrating the amounts of C that came to rest on individual deposits. The names and main characteristics of the watersheds listed in the table are given in east-to-west order. fluxes associated with mass wasting can vary greatly among watersheds [Hovius et al., 1997], whereas the production of scree [Hales and Roering, 2005] and the number of landslide-initiation points can vary with elevation [Coe et al., 2004]. Likewise, the transport and storage of sediment delivered by mass wasting varies with the three-dimensional structure of stream networks [Benda et al., 2004]. Ecosystem processes and C densities are also variable in mountainous terrain due to the existence of sharp environmental gradients. Therefore we can expect that the production, redistribution, and storage of C by mass wasting to be equally heterogeneous throughout the landscape (Figure 1a) [e. g., Bruijnzeel and Veneklass, 1998; Hilton et al., 2011]. As a consequence of this, conclusions drawn from a particular study are likely to be influenced by the spatial and temporal extent of the observations. [6] Although previous studies have highlighted the potential contribution of mass wasting to C fluxes via removal of modern organic C from hillslopes [Restrepo et al., 2003; Stark et al., 1999] or consequential transport by streams [e.g., Gomez et al., 2003; Hilton et al., 2008a,

2008b], a fundamental question regarding the fate of this C remains unaddressed. Here we develop a multiscale approach to address the following questions. First, how much modern organic C is mobilized by mass wasting activity in a tropical mountain setting during a singular landslide-triggering event? Second, how is this C redistributed by landsliding in these complex landscapes? In developing this multiscale approach we built a GIS procedure that explicitly recognizes the landscape positions where C transferred by mass wasting originates and where it comes to rest. The linkages between the hillslope and fluvial systems are used as a first approximation to understand the potential for burial or further transport. At the same time, this approach opened the possibility to examine C budgets associated to mass wasting at spatial scales ranging from individual hillslopes and stream segments to the watershed (3960–20,700 ha) and the entire study region encompassed by all watersheds (656 km2). We focus on the Sierra de Las Minas region (SLM) of eastern Guatemala and a population of shallow landslides triggered by rainfall associated to Hurricane Mitch in 1998 [Bucknam et al., 2001; Restrepo

3 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Table 1. Ecosystem Types Found Within the Study Area in the Sierra de Las Minas Arranged According to Elevation Except Pastures and Perennial Crops Ecosystem Type

Total Area (ha)

Percent Area

Carbon Density (Mg ha 1) (Mean  SD)

Tropical deciduous broad-leaved xerophytic lowland forest (1–07) Tropical evergreen mixed lower montane forest (1–27) Tropical evergreen mixed upper montane forest (1–30) Tropical evergreen broad-leaved altimontane forest (1–32) Tropical evergreen mixed altimontane forest (1–33) Tropical deciduous broad-leaved xerophytic lowland shrubland (2–04) Tropical deciduous broad-leaved lowland shrubland (2–05) Tropical evergreen mixed shrubland (2–07) Pastures and shrublands in degraded mountains (3–04) Perennial crops – Cultivated mixed forest (4–02) Perennial crops – Shaded coffee, cacao, and/or cardamom plantations (4–04) Perennial crops – Cultivated open forest with pastures or shrublands in understory (4–05)

130 5300 16500 5100 6700 700 4700 15900 650 2300 840 69

0.2% 8.1% 25% 7.8% 10% 1.1% 7.1% 24% 1.0% 3.5% 1.3% 11%

16  0e,h 135f 211  62b,c,h 549  90h 80  59c 24c,i 18  11d 87  79d,h 3.8  2.3a,g 32  13.4h 39  15.0d,h 3.8  2.3a,g

a

Acosta et al. [2001]. Castellanos et al. [2007]. Castellanos and Flores [2006]. d Castellanos unpublished data. e Clark et al. [2001]. f De Jong et al. [1999]. g Etchevers et al. [2001]. h Fundación Defensores de La Naturaleza [2002]. i Jaramillo et al. [2003]. b c

and Alvarez, 2006]. The study of a population of landslides in the SLM can help elucidate the role of extreme climatic events on the C cycle in this mountain range, and more broadly speaking in other high-relief areas of Central America. At the same time, the SLM typifies a continental montane setting in a region renown worldwide for its susceptibility to landsliding due to its rugged topography, frequent seismic activity, and intense precipitation [Bommer and Rodriguez, 2002; Franco et al., 2009; Lodolo et al., 2009; Nadim et al., 2006; Restrepo and Alvarez, 2006; Rodríguez, 2007].

2. Methods 2.1. Study Area [7] The SLM mountain range located in central-eastern Guatemala is one of several sub-parallel high ranges comprising the northern sierras of Central America that extend from southern Mexico to Guatemala, Honduras, and northern Nicaragua into the Caribbean coast (Figure 2) [Marshall, 2007; Weyl, 1980]. Stretching in a roughly NW-SE direction for approximately 130 km, the SLM lies within the active plate deformation zone of the Polochic and the Motagua faults at the boundary between the North American and Caribbean tectonic plates. Most of the exposed rocks in the SLM are of Paleozoic age and represent a complex mixture of metamorphic rocks of the Chuacús group, that include gneisses and migmatites (San Agustín Formation), phyllites, schists, and amphibolites (Jones Formation), and marble (Santa Rosa Formation) [Bosc, 1971; Newcomb, 1978]. Younger rocks from the Cretaceous-Tertiary include limestone cobbles and serpentine pebbles (Subinal Formation), fossiliferous mudstone (Chanchán Formation), and tuffs (Guastatoya Formation). [8] The SLM exhibits variable climatic conditions explained by a SE to NW increase in maximum elevation and position relative to the direction of the trade winds. The

northern flank of the SLM is humid, whereas the southern is mesic to dry, and these conditions are modified within each flank by changes in elevation. Not surprisingly, a wide range of total annual rainfall (500–4000 mm) and mean monthly temperatures (5 –30 C) have been reported for the SLM [Dix, 1997; IPGH, 1976]. Within the context of our study this climatic variability is important because, in combination with the diverse geological substrates, it translates into equally diverse topography, soils, and vegetation, and in turn C content of vegetation and soils. Existing 1:250,000 land cover and soils maps of Guatemala identify 12 ecosystem types (Table 1) [TECSULT Foresterie, Inc., 2000] and 7 soil series grouped in 3 soil orders (Table 2) [Ministerio de Agricultura, Ganadería y Alimentación (MAGA), 2000; Simmons et al., 1959] within our study area. Roughly 80,000 people depend on the numerous streams originating in the SLM [Fundación Defensores de la Naturaleza, 2003]. Recognition of the importance of the SLM for the region’s ecologic integrity and economic development led to the creation of the 242,000 ha Sierra de Las Minas Biosphere Reserve in 1990 [Fundación Defensores de la Naturaleza, 2003]. [9] Our study focuses on seven watersheds draining an area of 65,700 ha in the southern flank of the SLM. The Table 2. Soil Types Mapped in the Study Area With Carbon Densities for Tropical Soils to a Depth of 1 m

Order

Series

Total Area Coverage (ha)

Inceptisols

Marajuma Suelos de Valles Zarzal Chol Tamahu Civija Shola

24400 5000 620 22000 10900 2100 710

Entisols Ultisols

4 of 18

a

Eswaran et al. [1993].

Percent Area Coverage

Carbon Densitya (Mg ha 1)

37% 8% 1% 33% 17% 3% 1%

130 130 130 60 60 90 90

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

average area and slopes of these watersheds are 9400  5700 ha (mean  sd) and 25  3 , respectively, with elevations ranging between 140 m and 2900 m (values derived from a 29-m resolution DEM; Figure 2). The seven watersheds get progressively smaller from west to east, and are drained by streams of increasingly smaller orders (Figure 2). The largest and westernmost watershed, Río Teculután, has an area of 20,700 ha and is drained by a fifth-order stream network. The smallest and easternmost watershed is Río Mayuelas with roughly 3960 ha and, like Río Los Achiotes and Río Mayuelas, a third-order river stream network. The Río Pasabién, Río Hondo, Río Jones, and Río Santiago watersheds are intermediate in size and are drained by fourth-order stream networks (Figure 2). 2.2. Hurricane Mitch [10] Hurricane Mitch was one of the strongest and most devastating tropical storms impacting Central America over the past 250 years [Lott et al., 1999]. On 22 October 1998, four days after being declared a tropical depression, Hurricane Mitch reached a Category 5 status on the Saffir-Simpson Hurricane scale. The center of the storm moved westward through the Caribbean Sea toward the coast of Honduras and reached the Guatemalan border by 31 October. Although wind speeds rapidly diminished once Hurricane Mitch reached the Central American landmass, its slow translational speed resulted in extreme precipitations between 27 October and 1 November 1998. Rainfall totals in Guatemala associated with Hurricane Mitch amounted to approximately an entire year’s worth of normal precipitation according to some reports [Bucknam et al., 2001]. Within our study region, the 1998 cumulative October precipitations almost tripled average monthly values (267 mm versus 75 mm at the Pasabién station located at 250 m.a.s.l. and 438 mm versus 132 mm at the San Lorenzo station at 1780 m.a.s.l.). In a third station west of our study watersheds the October 1998 cumulative precipitations more than doubled a 10-year average (705 mm versus 208 mm at the Albores station; http://www.insivumeh. gob.gt/principal/alertas.htm). [11] Return intervals for precipitations of the magnitude observed during Mitch have been difficult to establish. Using historical hurricane data, Lopez [1999] concluded that in Guatemala “meteorological events capable of producing major disasters occur every twenty years.” On the other hand, Guerra-Noriega [2010], using precipitation data, concluded that the return times of maximum precipitations accumulated in 24 h and 5 consecutive days such as those observed during Mitch ranged between 30 and 80 years, respectively. In El Salvador the return time of precipitations such as those that accumulated in a period of 1–2 days during Mitch was estimated at 25 years, whereas in Nicaragua the return times for the unusual October 1998 cumulative precipitations varied between 35 and 500 years depending on location [Friedel, 2008; INETER, 1998].

3. GIS Modeling Approach [12] Our modeling approach accounts for the spatial distribution of above-ground biomass and soil C pools, and its redistribution by mass wasting activity at the local (i.e., hillslope and stream segment), watershed, and regional scales. It is based on a set of GIS routines that integrate spatial

G03016

data and C stock inventories into a single geo-database framework (Figure 3). The pre landslide-triggering event routines quantify C pools based on the combination of vegetation and soil maps with C density values obtained from literature reviews. The post-event routines estimate the total amount and spatial redistribution of C caused by landslide activity based on material transfer linkages among landslide scars and debris flow scours with deposits. We first describe the spatial and non-spatial databases, and then the main routines. Overall, this is a simplified implementation of the conceptual model outlined in Figure 1. 3.1. Databases 3.1.1. Hydrologic Feature Geo-Database [13] ArcHydro was used in combination with a 29-m resolution DEM obtained from Guatemala’s Ministerio de Agricultura, Ganadería y Alimentación (MAGA) to generate a linked geo-database that included a stream network (‘StreamOrder’), subcatchment (‘SubCatchment’), and watershed (‘Watershed’) features [Maidment, 2002]. These were projected into WGS 1984, UTM Zone 15N. The ‘StreamOrder’ feature class contains stream segments defined on the basis of a 0.5 km2 source area threshold; the resulting stream network had a density of 0.9 km km 2. Stream segments are identified with a unique code and classified according to stream order following Strahler’s method [Strahler, 1964]. Polygons defining ‘SubCatchment’ features also received a unique code and stream order value corresponding to the stream segment located directly downslope; average subcatchment area was 93  72 ha. Last, the ‘Watershed’ feature class contains polygons that represent the seven study watersheds. 3.1.2. Carbon Database [14] Existing 1:250000 vegetation [TECSULT Foresterie, Inc., 2000] and soil [MAGA, 2000; Simmons et al., 1959] maps were combined with a C density database to generate vegetation and soil C geo-databases for the study area. The vegetation map was generated by TECSULT Foresterie, Inc. [2000] from the visual interpretation of 30-m resolution Landsat TM images (1996–1999) and manual delineation of vegetation units. We slightly modified this map using IRS 5-m resolution images taken in 1999 and 2000 to resolve some observed discrepancies, particularly in the distribution of shrublands at high elevations. The soil map was generated in 1959 by Simmons et al. [1959] and digitized in 2000 by MAGA. Every ecosystem type and soil series was linked with C density values in our compiled database. The latter was assembled through the review of published and unpublished sources reporting stand-level above-ground biomass or C densities of vegetation (Table 1) and soil (Table 2). The strength of this approach lies in the fact that we were able to estimate C pools for a poorly documented region, namely the SLM. Nevertheless, we acknowledge that the combination of data from multiple sources may introduce an unknown degree of variability associated with the delineation of vegetation types and the estimation of C densities. Due to the paucity of root biomass C-content data, we excluded it from our analyses. For the purposes of our calculations, any reference to C exclusively denotes the modern organic fraction contained in the above-ground biomass and soil organic matter pools and excludes all fossil sources of C contained within rocks.

5 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Figure 3. Flowchart of GIS model highlighting routines used to calculate pre- and post-event C pools, including fluxes associated with mass wasting. Raster data represented with dashed ovals, non-spatial data with continuous diamonds, and vector data with continuous ovals (original input data) and rectangles (vector data with new attributes resulting from the union or join of input data). 3.1.3. Landslide Database [15] Rainfall associated with Hurricane Mitch triggered 11500 landslides within a 10000 km2 area centered on the SLM [Bucknam et al., 2001; Coe et al., 2004]. Even though maps of landslide initiation points and scars based on the interpretation of aerial photographs (1:40000) and photogrammetric restitution on 1:50000 maps existed [Bucknam et al., 2001], new maps were generated for this study. The reason for this is that crucial to our work was the need to identify spatial linkages among initial landslide scars, debris scours, landslide deposits, and the stream network. We screen digitized these features from panchromatic, IRS 5-m resolution images and created three feature classes, namely ‘Scar’, ‘Scour’, and ‘Deposit’. All of the digitized scar features had a typical landslide- or scour-like morphology with a general elongated shape parallel to slope, and only those showing a high contrast with the surrounding vegetation were digitized to avoid confusing landslides triggered by Hurricane Mitch with those from any previous events. Features digitized within the ‘Scars’ feature class only included the initial failure zone and therefore, upslope areas immediately above features identified as scars always had to be clear of any other slope failure disturbance. A lip or a narrowing of the polygon feature typically marked the lowermost point of the scar. Given the resolution of the IRS images and our self-imposed requirement to define a scar by its morphological features and alignment relative to hillslope topography, the smallest landslide scar that we were able to

map had an area of 0.01 ha (4 pixels). When denudation was continuous for more than approximately 50 m downslope from the lip of a scar then a debris scour was digitized in the ‘Scour’ feature class. Debris scours were digitized on both hillslopes and along streams and its downstream edge was defined where no denudation was visible for more than approximately 50 m. The downslope extent of any scar or debris scour with no further visible transport by scouring were recorded as deposits and mapped as points in the ‘Deposit’ feature class. Deposits were mapped as points because the limited resolution of the satellite images did not allow an accurate mapping of their areal extent and depth, which precluded any attempt to estimate the volumetric distribution of the material each had accumulated. The ‘Scar’ and ‘Scour’ feature classes both contain a featurelinking attribute (TRANSF_TO) that allowed the C redistribution routines to trace the downslope transfer of material from scars and debris scours to deposits. It is important to note that even when a deposit is said to lie within the boundaries of a stream, our method is unable to differentiate the portion of C coming to rest directly within the active channel from that which has settled on floodplains or terraces and out of the immediate reach of the active stream. 3.2. GIS Carbon Budgeting Routines [16] The GIS-based C budget routines were developed using the Model Builder tool available within ArcGIS. These routines take advantage of the linked geo-database

6 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

structure to quantify the size of pre- and post-event vegetation and soil C pools as a function of elevation, and geomorphic setting, namely distinguishing among hillslopes and channels of various stream orders. The first set of routines calculates the total C contained within individual watersheds prior to the landslide-triggering event. The second set quantifies the amount of C mobilized by landslide activity. The last set of routines transfers the C removed by landslides downstream to their associated deposits. 3.2.1. Pre-Event C [17] The sequence of routines begins by generating a reclassed DEM (‘DEM Reclass’) that combines elevation values into polygons each representing a 50 m elevation range or elevation band (‘Elevation’). Subsequent routines generate a new feature class (‘EleStrHill’) that combines attributes from the ‘Elevation’ and ‘Sub Catchment’ feature classes in combination with a buffered version of the ‘StreamOrder’ feature class based on a buffer distance that is a simple multiple of stream order (Figure 3). Stream buffers to define the active channel zone were created to aid in representing the degree of coupling between landslide deposits and the stream network. Based on field observations of Mitch’s impact on streams, we assigned a 30 m total buffer width to first-order channels and increased this amount by 30 m for each subsequent stream order. The choice of a constant buffer width increment between consecutive stream orders is not expected to have any effect on our calculations given that the choice of the downstream location of a deposit was explicitly chosen to belong to a hillslope or stream according to its visual expression on the panchromatic images. [18] The ‘Vegetation’ and ‘Soil’ feature classes are joined with the ‘Vegetation C density’ and ‘Soil C density’ tables, respectively, and with the ‘EleStrHill’ feature class to generate ‘Total C’ which contains the distribution of pre-event vegetation and soil C by elevation, and distinguishes among hillslopes and streams of various orders (Figure 3). Polygons contained within stream buffers are assigned a total C of zero, as they are meant to represent the active channel and not forested surfaces with soil profiles. Unfortunately, literature reviews revealed no empirical data that could be used to estimate the in-stream C content for active channels such as the ones found in SLM. 3.2.2. Post-Event C Removal [19] Two sets of routines, Post-Event Area Analysis and Post-Event Total Carbon Removal, are used to estimate the total C mobilized and to track its transfer to its resting position, respectively. The Post-Event Area Analysis generates a new table (‘ScarScourArea’) summarizing scar size and total denuded area statistics. [20] The Post-Event Total Carbon Removal routine calculates the amount of C contained within the boundaries of the scar and debris scour features. Carbon removed by scouring along previously defined stream buffer zones was assigned a value of zero because of the lack of in-channel C content data for SLM. This routine creates a new feature (‘ScarCarbon’) that combines attributes of elevation, stream order, vegetation and soil C content, and topographically defined linkages (TRANSFER_TO). The results of these routines are stored in ‘ScarCarbon’ and in two tables that summarize the distribution of C removed with regards to elevation and subcatchment stream order.

G03016

3.2.3. Post-Event C Deposition [21] Our GIS Post-event C deposition approach is based on several assumptions: (1) all vegetation and soil C within a scar or scour feature is removed as a result of landsliding, (2) soil C is independent of variations in soil depth, (3) C losses by mineralization while the material is in transit or within a depositional environment are negligible, and (4) all C transferred by landsliding remains within the watershed at the end of the landslide triggering event. The first two assumptions are not that critical given that the satellite images and field observations confirm that most of the vegetation was removed from scars and that most landslides removed material down to the soil–rock interface. In addition, these two assumptions are supported by the literature in that it agrees that most organic C is contained within the uppermost first meter of the soil [Eswaran et al., 1993; Lal, 2005] and that shallow landslides remove the soil down to saprolite [Wilcke et al., 2003]. [22] The assumption of negligible atmospheric losses due to mineralization is not likely to be a major source of error, given that our primary goal is to quantify the amount of C being transferred by landslide activity, a process lasting over a relatively short duration. Nevertheless we acknowledge that C losses are plausible as shown by numerous studies documenting C fluxes associated with overland flow in agriculturally controlled landscapes [Berhe et al., 2007; Jacinthe and Lal, 2001]. The fourth assumption is unrealistic, and would have been important if the main goal was related to the fluvial transport and net export of C. This is because depending on the setting, significant amounts of C may be exported through watershed outlets as the result of the combined effects of mass wasting and fluvial transport during an event such as Hurricane Mitch. Instead, our main goal was to estimate the amount of C delivered by mass wasting to hillslopes and streams. Therefore, the assumption is retained here in accordance to our overall goal and with the limitations posed by our remote sensing methods that did not allow a clear differentiation of the portion of C delivered to hillslopes, active channels, or out-ofchannel deposits along the fluvial network. [23] The Post-Event Transfer to Deposit Calculation routine transfers all C within scar and scour features to their associated deposits and saves them into the ‘DepositsFinal’ feature class (Figure 3). The Post-Event Final Carbon Redistribution routine combines all attributes in ‘Total C’ with those from the ‘DepositFinal’ feature class. The “Summary Statistics” function is used to create two final tables that organize total C from vegetation and soils contained in deposits by elevation, and differentiating between streams of different order and unchannelled hillslopes.

4. Results 4.1. Pre-Event Carbon Distribution [24] Total C stored in the study area before hurricane Mitch was estimated at 15  106 MgC. Approximately 60% (9  106 MgC) of this C was in the above-ground biomass and the remaining 40% (6  106 MgC) in the soil. When expressed on a per-unit-area basis, the total C stored in the study area was equivalent to 230 MgC ha 1 (Table 3). [25] Carbon storage varied greatly whether examined on a watershed basis, as a function of elevation or based on subcatchment order. At a watershed scale, the general trend was for total C and C density to decrease in an eastward direction

7 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Table 3. Pre- and Post-Event C Pools and Fluxes Associated With Landslides Triggered by Hurricane Mitch in 1998 Teculután

Pasabién

Hondo

Jones

Santiago

Pre-Event 6.8 1.9 1.8 1.8 1.1 Total C (106 MgC) 217 87 114 110 78 Vegetation C density (MgC ha 1) 112 90 70 84 76 Soil C density (MgC ha 1) 329 176 184 194 155 Total C density (MgC ha 1) Post-Event Scars and debris scours 13.0 [2%] 2.0 [1%] 5.0 [2%] 14.0 [8%] 5.0 [5%] Total C (104 MgC) and %C [] mobilized by landslides 1 6.2 1.6 4.5 15.2 7.2 C mobilized per unit watershed area (MgC ha ) Deposits 1.6 [12%] 0.2 [14%] 0.9 [19%] 1.0 [7%] 1.1 [21%] Total C (104 MgC) and %C retained in hillslopes [] 4 3.7 [29%] 0.1 [3%] 0.9 [29%] 0.8 [6%] 0.2 [3%] Total C (10 MgC) and %C reaching 1st-order channels [] 7.5 [59%] 1.4 [83%] 2.8 [62%] 12.2 [93%] 3.8 [76%] Total C (104 MgC) and %C reaching 2nd–5th-order channels []

(Table 3). Teculután, the westernmost watershed, stored the greatest amount of C (6.8  106 MgC or 45% of total), whereas Mayuelas, at the eastern end of the study area, stored the least (0.7  106 MgC or 5%). When normalized by area, C density was greatest (329 MgC ha 1) in Teculután and least (155 MgC ha 1) in Santiago (Table 3). These patterns are influenced by changes in the areal extent of ecosystem types and their respective C densities, which in turn reflect the existing variation in drainage area and elevation range across the study area (Figure 2; Table 3). [26] The size of the pre-event regional C pools varied greatly with elevation (Figures 5a, 5e). Although total C exhibited a small peak at elevations ranging between 150 and 300 m (71  104 MgC or 5% of total; Figure 5a), it was between 1450 and 2300 m where the largest proportion of C was found (755  104 MgC or 50% of total; Figure 5a). Median total C across all seven watersheds exhibited a similar pattern, yet the variability increased above 1675 m (Figure 5e). When C is normalized by area, a relatively steady increase in C is observed with elevation (Figure 5a). Several non-mutually exclusive factors may explain the above trends. First, C content of soil and vegetation types, as well as their areal extent,

Los Achiotes Mayuelas

All

0.9 115 101 216

0.7 95 92 188

15.0 130 99 229

3.0 [3%]

2.0 [3%]

43.0 [3%]

6.3

5.6

6.6

0.5 [16%]

0.8 [36%]

6.0 [14%]

0.5 [19%]

0.7 [31%]

6.8 [16%]

1.8 [65%]

0.7 [33%] 30.2 [70%]

varies along the elevation gradient. Second, watersheds differ in terms of their areal extent and elevation range. At low elevations alluvial soils distributed along the main channels have some of the highest soil C densities (Suelos de Valles in Table 2) and this explains the peak in total C at lower elevations. On the other hand, increases in the areal extent of vegetation and soil types with high C densities explain the two peaks at high elevations. In our study area, hillslopes at low elevations are dominated by shrublands, whereas those at high elevations are covered by tropical evergreen mixed upper montane and tropical evergreen broad-leaved altimontane forests, two high C density ecosystem types (Table 1). [27] Pre-event C pools varied among subcatchments of different order and mirrored to some extent those observed along the elevation gradient (Figures 6a, 6f). First-order subcatchments stored more C (10  106 MgC or roughly 70% of total) than second- to fifth-order subcatchments combined (5  106 MgC). The dominating role of firstorder subcatchments is not only related to the fact that they represent approximately 65% of the entire SLM surface area but also because they contain the largest overall C density values (Figure 6a). Median total C values were highest and

Figure 4. (a) Landslide scars, and associated debris scours, triggered by Hurricane Mitch in the Sierra de Las Minas. (b) Picture of a debris flow deposit at the junction of a low-order stream with a major tributary (pictures courtesy of J. A. Coe). 8 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Table 4. Characteristics of Landslides Triggered by Hurricane Mitch in 1998 in the Seven Studied Watersheds

Number of landslide scars Average scar size (ha) Denuded area (ha) Scar density (number ha 1) Scar density (m2 ha 1)

Teculután

Pasabién

Hondo

Jones

Santiago

Los Achiotes

Mayuelas

Total

434 0.41 374 0.021 181

186 0.23 100 0.018 98

553 0.22 236 0.057 243

594 0.49 712 0.064 770

491 0.35 248 0.071 359

195 0.29 114 0.044 259

258 0.24 95 0.065 239

2711 0.34 1876 0.041 289

more variable in first-order subcatchments (1  106 MgC) and lowest and less variable in fifth-order subcatchments (0.1  106 MgC) (Figure 6f ). 4.2. Landslide Scars and Debris Scours [28] A total of 2711 landslide scars were identified and mapped from the IRS satellite images and together with their associated debris scours they denuded 1876 ha or 3% of the study area (Figure 2, Figure 4a, and Table 4). The studied watersheds, however, varied in terms of the number of landslides as well as denuded area: the largest number and density of landslides, average scar size, and total denuded area was found in Jones, whereas the lowest was found in Pasabién (e.g., 594 versus 186 landslides in Jones and Pasabién, respectively; Table 3). Although scars and debris scours were found at most elevations, the largest area of denudation was observed at 1475 m (72 ha, equivalent to 4% of the land within this elevation range or a median of 8 ha across all watersheds; Figures 5b and 5f ). Above this elevation the variability in the area denuded by landslides increased due to the influence of some large landslides (>20 ha for a given elevation) in some watersheds (Figure 5f ). While denuded area at high elevations was related to both scars and scours, most denudation at lower elevations was due to the effects of debris flows that induced scouring through long stretches of the main tributaries, as observed within the Jones watershed. When examined as a function of subcatchment order, the area denuded by landslides decreased from first- (1115 ha) to fifth- (30 ha) order subcatchments (Figures 6b and 6g). 4.3. Post-Event Carbon Distribution 4.3.1. Carbon Removal [29] Landslide scars and debris scours mobilized 43  104 MgC, or equivalently 3% of the total C stored in the study area (Table 4). Of this C, 61% originated from the aboveground vegetation while the remaining originated from the soils. Unlike the pre-event total C, there was no observable east-west trend in the total amount of C mobilized by landslides. The largest amounts of C were mobilized in Teculután (13  104 MgC) and Jones (14  104 MgC), and this was equivalent to 2% of the entire study area C pool, or 63% of the total C mobilized by landsliding (Table 3). Yet, when expressed on a per-watershed-area basis, the amounts of C mobilized were highest in Jones (15 MgC ha 1) and Santiago (7 MgC ha 1) due to the elevated density of landslides in these two watersheds (Table 3). [30] The greatest amount of total C mobilized by landsliding (17932 MgC) was observed at around 1625 m (Figure 5c). Likewise, the largest median value of total C (2130 MgC) was observed at 1575 m, but there was considerable variability across the elevation gradient (Figure 5g). When total C is categorized by the subcatchment order from which it was mobilized, first-order catchments predominate

by generating 65% (0.3  10 6 MgC) of the total mobilized C. Meanwhile the relative importance of all other subcatchments steadily decreases with increasing order (Figures 6c and 6h). 4.3.2. Carbon Deposition [31] Our data shows that 14% (6  104 MgC) of the C mobilized by landsliding came to rest on hillslopes, whereas the remaining 86% (37  104 MgC) was delivered to fluvial landscape units. The amount of C deposited on the hillslopes varied greatly among watersheds. When expressed on an absolute basis, it ranged between 0.2–1.6  104 MgC, with Teculután storing the largest amount of C on its hillslopes. When expressed on a per watershed area basis it ranged between 0.1–2 MgC ha 1, with Mayuelas having the largest amount of C deposited on hillslopes. As expected by the dominant role first-order catchments played in transferring C, about 70% (41700 MgC) of the material re-deposited on hillslopes remained within first-order catchments (Figures 6d and 6i). [32] The amount of C entering the fluvial system varied over an entire order of magnitude among the seven watersheds whether expressed in absolute (1.4–13  104 MgC) or in a per unit area (1.4–14 MgC ha 1) basis. Streams in Jones received the largest amounts of C among the studied watersheds and this may be explained by the presence of the two largest landslides triggered by hurricane Mitch in the study area, both of which delivered all of the scoured material directly into the stream network. When examined as a function of subcatchment order, 18% (6.8  104 MgC) of the mobilized C reaching the stream network was deposited in first-order segments, whereas the remaining 82% (30  104 MgC) came to rest in higher-order systems (Table 3). In particular, fourth-order stream systems were recipients of 47% (17  104 MgC) of the total C delivered to the fluvial network (Figures 6e and 6j). [33] Landslide deposits were found at a wide range of elevations extending from 175 m to 2575 m. Discriminating between hillslope and stream deposits showed that the largest amount of C deposited on hillslopes and first-order channels occurred between 1225–2375 m (Figure 5c), whereas the largest amount of C reaching streams of second order or higher occurred between 225–1525 m (Figure 5d). Carbon deposits in stream channels at low elevations contained as much as 40,000 MgC (Figure 5d). The highest median C content of deposits on hillslopes and first-order streams (341 MgC) was recorded at 1425 m, but a large variability among different watersheds was observed at that range of intermediate elevations (Figure 5h). Similarly, the highest median C content in second- and higher-order stream deposits (778 MgC) was recorded at 1275 m (Figure 5i). Carbon mobilized by landsliding was greater in first- than in higher-order subcatchments (Table 3; Figures 6c and 6h); a similar trend was seen for C reaching hillslope deposits (Figures 6d and 6i) but not for those reaching streams

9 of 18

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

G03016

Figure 5. (a–d) Total and (e–f) median variation in the size of C pools and C mobilized by landslides triggered by Hurricane Mitch in the SLM as a function of elevation. (a) Pre-event total C pool (solid circles) and C density (open circles). (b) Absolute hillslope surface area (green) and total surface area denuded (scarred or scoured) by landsliding (brown). (c) Total C mobilized from hillslopes by mass wasting (gray) and total C re-deposited on hillslopes or transferred to first-order streams (green). (d) Total C deposited in streams of second or higher order (green). (e) Box plot of pre-event C pool showing the 25th and 75th quartiles (box), median (horizontal black line), maximum and minimum values (vertical gray bars), and extreme values (gray filled circles) among the seven study watersheds. Central tendency and dispersion values are based on data representing each of the seven study watersheds. (f) Box plot of total area denuded by landslides. (g) Box plot of total C mobilized from hillslopes by landsliding. (h) Box plot of total C deposited on hillslopes and first-order stream deposits. (i) Box plot of total C delivered to deposits on second- and higher-order streams. Symbols in Figures 5f–5i as in Figure 5e. (Figures 6e and 6j). This finding suggests an SLM-wide tendency for the effective delivery of C from hillslopes to streams by landslide activity.

5. Discussion [34] The development of an approach to quantify the role of mass wasting activity in C budgets has allowed the

simultaneous quantification of C pools and fluxes before and after extensive mass wasting activity at scales ranging from the local to the regional. In the SLM study region where Hurricane Mitch triggered hundreds of landslides in 1998, mass wasting contributed to the rapid transfer of 43  104 MgC, or equivalently 3% of the estimated C contained within the above-ground biomass and soil pools. Furthermore, distinguishing between hillslopes and streams, two potential

10 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Figure 6. (a–e) Total and (f–j) median variation in the size of C pools and C mobilized by landslides triggered by Hurricane Mitch in SLM as a function of subcatchment stream order. (a) Pre-event total C pool (open green bars) and C density (solid green circles). (b) Total hillslope (open gray bars) surface area and total surface area denuded by landslides (solid brown circles). (c) Total C mobilized by mass wasting from hillslopes within subcatchments of various stream orders (gray bars). (d) Total C deposited on hillslopes (green bars). (e) Total C delivered to stream environments (olive green bars). (f) Box plot of pre-event C pool showing the 25th and 75th quartiles (box), median (horizontal black line), maximum and minimum values (vertical gray bars), and extreme values (gray solid circles). Central tendency and dispersion values are based on data representing each of the seven study watersheds. (g) Box plot of total scoured area. (h) Box plot of total C mobilized by landsliding. (i) Box plot of total C deposited on hillslopes and first-order stream deposits. (j) Box plot of total C entering second- and higher-order streams. Symbols in Figures 6g–6i as in Figure 6f. storage compartments for the C mobilized by mass wasting, allowed us to establish that approximately 14% of the C was re-deposited in unchannelled hillslopes, 16% reached firstorder streams, while the remaining 70% was transferred to higher-order streams. Although these are overall figures across the seven study watersheds, our approach also allowed us to distinguish a significant variability among them. Finding and characterizing this spatial variability is important for

two reasons. First, it provides valuable information about factors controlling the mobilization and deposition of C associated with mass wasting activity. Second, it highlights the spatial variability exhibited by ecosystem and geomorphic processes, including C sources and sinks. Ultimately, characterizing this spatial variability raises questions about feedbacks between ecological and geomorphic processes.

11 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

5.1. Pre-Landslide Triggering Event Carbon Pool Estimates [35] Our focus has been on mass wasting and its role on the removal and transfer of modern organic C contained in the above ground biomass and soil. Not included are C estimates from below-ground biomass, litter, alluvial or colluvial deposits along streams, and coarse woody debris (CWD); nor have we included estimates from the fossil C pool, because these have not been quantified for the SLM. Our overall C pool and C transfer estimates have taken into account the variability in C densities and spatial distribution observed among the main ecosystem types and soil orders found in the SLM (Tables 1 and 2). The results shown in Table 3 indicate that prior to Hurricane Mitch the above ground biomass held 1.5 times more C than soils (60% versus 40% of total, respectively). These percentages are within known ranges for moist tropical forests where the above ground biomass may represent between 20%–77% of the total C stock and the soil between 23%–80% [Raich et al., 2006]. In addition, the C density values in the above-ground biomass for some of the SLM ecosystems are similar and in some instances exceed those reported in the literature (Table 1; max–min values reported for world moist forests are 75–266 MgC ha 1 [Raich et al., 2006]). Included among high-C-content ecosystems of the SLM are the tropical evergreen mixed upper montane and the tropical evergreen broad-leaved altimontane forests. [36] Perhaps the greatest uncertainties in our C pool estimates originate from the soils. Due to the paucity of soil data for the SLM, the soil C pool estimates presented here were based on a pre-existing soil map that identifies three soil orders in the studied watersheds [MAGA, 2000; Simmons et al., 1959] and C density data from elsewhere (Table 2). Nevertheless, limited existing soil C content data for the SLM are within the values used here and in some instances exceed them. One study along a 1000-m elevation gradient (493–1523 m) encompassing tropical evergreen mixed lower montane and tropical evergreen mixed upper montane forests in the Santiago watershed, documents between 6 and 12 MgC ha 1 within the first 10 cm of soil (E. J. Castellanos, unpublished data, 2005). A second data set spanning a 1200-m elevation gradient (1000–2200 m) in the Pasabién watershed documents between 32 and 211 MgC ha 1 within the top 30 cm of the soil profile, the largest value found under the tropical evergreen broad-leaved altimontane forest (C. Restrepo, unpublished data, 2011). The existing variation in C density in the above-ground biomass and soil most likely reflects that of temperature, precipitation, and geology along the multiple environmental gradients found within our study region as it has been shown elsewhere [Cavelier, 1996; Jordan, 1985; Raich et al., 2006]. 5.2. Mass Wasting, Transfer and Deposition of Organic Carbon [37] Mass wasting activity associated with Hurricane Mitch contributed to the rapid removal and re-distribution of 3% of the C stock over the SLM study area. Two factors may ultimately influence transport, storage, and decomposition of this material. First, the sources of this C differ in size and composition such that tree trunks, large tree branches, and coarse woody debris (CWD) are likely to have transport potentials different from those of litter and leaves

G03016

(Figure 1) [Cadol and Wohl, 2010]. Second, the particular characteristics of the sites where this material comes to rest may influence C burial or further C transport. [38] Spatial accounting of the landslide deposits at small scales (i.e., within hillslopes) showed that 14% (6  104 MgC) of the C mobilized by mass wasting came to rest at the base of small landslide scars that did not generate debris flows (Table 3) similar to what has been shown in sediment studies elsewhere [Hansen, 1984; Pearce and Watson, 1986; Peart et al., 2005]. If the C contained in the unchannelled hillslope deposits remains protected from oxidation due to increased wetness and reduced aeration relative to hillslopes [Stallard, 1998] and becomes physically stabilized by plant growth [Guariguata, 1990; Peart et al., 2005; Velázquez and Gómez-Sal, 2008; cf. Walker and Shiels, 2008], this becomes a mechanism by which mass wasting can contribute to long-term C storage. [39] Spatial accounting of C at larger scales requires establishing the extent of coupling between landslides and the stream network to quantify the amount of material that is available for fluvial transport [Harvey, 2001; Korup, 2005; Schwab et al., 2008] or burial (Figure 1) [Benda et al., 2005; Bigelow et al., 2007; Schrott et al., 2003]. The stability of in-channel deposits is controlled by the its location relative to the active channel, and the interplay of stream power, hydraulic roughness, and resistance to erosion, all of which are highly variable in time and space. While the downstream distribution of stream power is influenced by climatic and topographical factors that control the distribution of runoff and effective channel slope, hydraulic roughness and resistance to erosion are influenced by local factors such as vegetation, flow obstructions, and substrate properties including lithology and structure [Benda et al., 2005; Goode and Wohl, 2010; Knighton, 1999]. [40] In the particular case of first-order streams they often display clast-supported deposits composed of unsorted, angular rocks of varying sizes with plant material wedged in between, that are oftentimes stabilized by plant growth. We presume that most of the C fraction transferred to first-order streams (16% or 7  104 MgC) during Hurricane Mitch likely entered long-term storage compartments due to the often described transport-limited capacity and reducing conditions of these environments [Benda et al., 2005; Hyatt and Naiman, 2001; Kelsey, 1982; Lancaster and Casebeer, 2007; Pearce and Watson, 1986]. In contrast, the C fraction reaching second and higher-order stream channels (70% or 30  104 MgC) has a more dubious fate as it may as well have entered or bypassed long-term storage deposits (Figure 4b). On the one hand, two sources of evidence exist that tend to support the effective retention of C in colluvial deposits. First, sediment cores obtained from Lake Izabal, a large lake collecting runoff generated by the northernfacing watersheds of the SLM, lack a distinct layer of organic material associated to Hurricane Mitch [Michot et al., 2002]. Second, debris deposits are abundant along low order streams and are easily recognized by the vegetation that has been established on them. Those attributed to Hurricane Mitch (Figure 4b) have small Hazelnut pines (Pinus oocarpa), whereas those associated with previous events are easily recognized by 15–25 m tall Hazelnut pines and American sweetgums (Liquidambar styraciflua), suggesting longevity and stability from fluvial erosion. On the other hand,

12 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

Table 5. Carbon Fluxes Associated With Mass Wasting Country Panamaa,c Costa Ricaa,g Papua New Guineaa,k Guatemalal Globallya,b,k Puerto Ricoh,j Hawai’ii Taiwanf Taiwanf New Zealande New Zealandd

Region Darien Pacuare watershed Sierra de Las Minas Luquillo mountains Ninole ridges LiWu watershed LiWu watershed Waitangitaona watershed Waiapoa watershed

Method

C Pools

Disturbance Disturbance Disturbance Disturbance Disturbance Land cover Land cover Land cover POC POC POC

AGC, AGC, AGC, AGC, AGC, AGC, AGC, AGC,

BGC, SOC BGC, SOC BGC, SOC SOC BGC, SOC SOC BGC CWD, SOC

C Fluxes (MgC ha 0.13 0.03–0.29 0.45–1.33 0.08–0.33 0.2–0.6 0.13 0.14–0.51 0.26 0.16–2.02 0.39 0.12

1

y 1)

C Fluxes (PgC y 1)

0.1–0.3

a

Estimates based on compiled figures for AGC (carbon in above-ground biomass), BGC (carbon in below-ground biomass), SOC: soil organic carbon, reported for tropical/subtropical regions. For Taiwan CWD (coarse woody debris) was included in the published estimates. b Based on the area of highstanding islands of southern Asia and Oceania which are reported to deliver 70% of the total suspended sediments reaching the ocean [Milliman and Syvitski, 1992]. c Garwood et al. [1979]. d Gomez et al. [2003]. e Hilton et al. [2008a]. f Hilton et al. [2008b]. g Hyman [1997]. h Ortiz-Zayas [1998]. i Restrepo et al. [2003]. j Scatena and Lugo [1995]. k Stark et al. [1999]. l This study. Estimates are based on three methods referred to as Disturbance rates: magnitude and frequency of landslide triggering events; Land cover: total area in landslides between two dates, and POC: riverine particulate organic carbon.

contrasting evidence suggests that a fraction of the C reaching second- and higher-order stream channels may have bypassed long-term storage deposits and exited the SLM watersheds. Specifically stream-scouring associated with debris flows generated during Hurricane Mitch contributed to the degradation of many high-order stream channels throughout the SLM. Also down-cutting of alluvial terraces by debris scours might have exposed and transported additional stored C (C. Restrepo and C. E. Ramos-Scharrón, personal observation, 2009). [41] In summary, the degree of protection from oxygenation granted by burial, the transportability of the differently sized and shaped C fragments released by mass wasting, the accessibility of the fluvial network to the newly established debris flow deposits, and the relative importance of depositional versus transport processes within streams largely influence the residence time of C mobilized by mass wasting within watersheds. While conditions on hillslopes and firstorder streams appear to promote storage, the fate of organic material delivered to second- and higher-order streams appears more uncertain. 5.3. Mass Wasting and C Fluxes [42] We can examine the results of our work in a broader perspective if we transform area-normalized C transfer totals associated with mass wasting activity into C fluxes. To calculate these we use return times of precipitation events such as those associated with Hurricane Mitch. Albeit with large uncertainties (see section 2 above), in Guatemala these events are estimated to occur once every 20–80 years. Therefore, the 6.6 MgC ha 1 transferred by mass wasting activity during Hurricane Mitch (Table 3) are equivalent to C fluxes in the order of 0.08–0.33 MgC ha 1 y 1 over the entire SLM study

area. Isolating the transfer of C to what might be considered landscape compartments with a high potential for long-term storage, i.e., hillslopes and first-order streams, results in C flux estimates in the order of 0.02–0.13 MgC ha 1 y 1 depending on return times (Table 3). The potential for long-term C storage may be greater if a fraction of C entering second- and higher-order streams (0.06–0.20 MgC ha 1 y 1) is stored locally, and if mass wasting activity associated with varyingmagnitude storms or with seismic activity is taken into account. [43] The C mass transfer or flux estimates associated with mass wasting activity for our study area (0.08–0.33 MgC ha 1 y1) are within ranges reported elsewhere by various methods (Table 5). On the one hand, studies that followed a hillslope-centered approach in which the area denuded by landslides is integrated with C content of vegetation and soil have yielded C flux values between 0.1 and 0.9 MgC ha 1 y 1. In addition to allowing an estimation of C fluxes at a variety of scales, this approach has the potential to develop a spatial accounting of the deposits, as illustrated by our work (Figure 1b, green arrows; Table 5). One of its drawbacks is that it is unable to fully elucidate the ultimate fate of the C mobilized by mass wasting. On the other hand, studies in which watershed processes are integrated with spatially and temporally discrete stream measurements of POC and DOC have yielded C flux values associated with mass wasting between 0.1 and 1.1 MgC ha 1 y 1 (Table 5). This C-yield based methodology has allowed an estimation of watershed scale C fluxes while also characterizing the age and potential source of C (Figure 1b, blue arrows) [Blair et al., 2004; Gomez et al., 2003; Hilton et al., 2008a, 2008b; Leithold et al., 2006; Galy and Eglinton, 2011]. This approach may provide a good approximation of C fluxes associated with

13 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

mass wasting in rivers where modern C bypasses potential storage compartments and where C associated with mass wasting has a clear signature as it has been proposed for several sites in the vicinity of continental margins [Blair et al., 2004; Hilton et al., 2008b; Leithold et al., 2006]. The same might not be true for most mountainous regions where C may enter numerous potential compartments before reaching the ocean. [44] The similarities among the C flux data associated with mass wasting are striking given that assumptions underlying the various methodologies are quite different. This not only constricts the range of expected values but also provides some insight into how the existing methodological approaches might complement each other (Table 5). Likewise, integrating the various approaches to quantify landsliderelated C fluxes allows the comparison with those estimated for other geomorphic processes such as surface erosion on agricultural fields (0.03–2.6 MgC ha 1 y 1) [Davidson and Ackerman, 1993; Jacinthe and Lal, 2001; Schlesinger, 1986; Van Oost et al., 2007]. 5.4. Mass Wasting: Source or Sink of C [45] A simple mass balance quantification of net gains or losses associated to landslide activity allows for the evaluation of its net effect on carbon stocks. Carbon gains to a control area such as a watershed or mountain range resulting from landsliding activity are solely associated to ecosystem recuperation on failed landslide sites (C gains on landslide scars: CgL). Losses particular to landsliding activity refer to that portion of the C transferred that is not retained by landslide deposits (C losses from deposits (ClD)) due to the direct effect of debris flow transport to watershed outlets, subsequent physical erosion and evacuation by fluvial transport, or C losses by oxidation. All other fluxes to longterm storage (C gains into deposits (CgD)) are then considered internal to the control area and represent neither a net gain nor loss to the system. Therefore, whether landslide activity results in the net gain or loss in C solely depends on the balance between CgL and ClD, whereas the ratio CgL/ClD (hereafter referred to as Cp) could aid in discriminating between conditions leading to net gains in C stocks (Cp > 1) from those leading to net losses (Cp < 1). This represents a slight reformulation of a mass balance criterion previously proposed to evaluate the effects of surface erosion on terrestrial C pools [Berhe et al., 2007]. [46] Carbon gains accrued by ecosystem recovery responding to a landslide-triggering event (CgL) can be calculated as the product of net primary productivity at these sites (NPPL) times the area denuded by mass wasting (AL). Existing landslide chronosequence studies examining changes in AGB and SOM over time [Pandey and Singh, 1985; Reddy and Singh, 1993; Restrepo et al., 2003; Zarin and Johnson, 1995b] places NPPL between 0.3 and 1.5 MgC ha 1 y 1. Typically, NPP for tropical forests range between 0.2 and 3.9 MgC ha 1 y 1, with the lower values being characteristic of primary forests and the higher values representative of secondary forests still thriving toward a mature state [Raich et al., 2006]. Using the NPPL range (0.3–1.5 MgC ha 1 y 1) and the area denuded by landslides during Hurricane Mitch in SLM (AL; 1475 ha), we obtain CgL values ranging between 442 and 2212 MgC y 1. These rates imply that it would take between 194 and 972 years

G03016

for ecosystem recovery to compensate for the entire 43  104 MgC displaced by landslide activity in SLM. [47] Meanwhile, the rate for net C transfer by landslides can be calculated by dividing the total C transferred by landslides (CT = CgD + ClD) by the estimated recurrence interval of precipitation similar to Hurricane Mitch (20–80 y). This calculation is based on the untested assumption that events of a similar rainfall magnitude will always result in a similar C transfer total, and it results in C fluxes ranging between 5375 and 21500 MgC y 1. This result implies that ecosystem recovery rates (442–2212 MgC y 1) would never be able to compensate for the net transfer of C directly associated to mass wasting within the time period in-between landslide-triggering events, but that landslide activity could still result in a net long-term C gain (Cp > 1) if CgL exceeds ClD. An important finding of this analysis it that it suggest that the net gain or loss of C resulting from landsliding is dependent upon CT, AL, NPPL, the recurrence interval of landslide-triggering events, and the proportion of C released that enters long-term storage (CgD/CT). [48] Three feasible scenarios are used here to evaluate the C sink potential (Cp) of the SLM during subsequent landslide-triggering events similar to Mitch. The first assumes that all the C released by landsliding exits the system; this is similar to scenarios described for Taiwan [Hovius et al., 2000; West et al., 2011]. The second scenario assumes that the C entering hillslopes and first-order stream deposits is physically stable and effectively protected from oxidation, while all C reaching second- and higher-order streams is exported out of the study area. The third scenario similarly assumes full retention of the portion of C entering hillslope and first-order stream deposits but in addition assumes retention of half of the C reaching second- and higher-order streams; the remaining half is exported out of the watersheds. According to Scenarios 1, 2 and 3, the relative fractions of C retained (CgD/CT) for the SLM are 0%, 30%, and 65%, respectively (Table 3). The third scenario proposes a very similar proportion of stored C to the 70% retention estimate suggested by Hilton et al. [2011]. [49] The full range of C retention values (CgD/CT; 0–1) were plotted against the resulting C sink potential ratios (Cp) for the entire range of possibilities encased by the C transfer rates by landsliding (5375–21,500 MgC y 1) and CgL rates (442–2212 MgC y 1) (Figure 7). As it might be expected, calculated values for Cp were consistently less than unity for almost the entire range of CgD/CT values if ecosystem recovery rates are slow (442 MgC y 1) and if landslide recurrence intervals are short (20 y). This implies that under these conditions, any of the three C retention scenarios described above would result in a net loss of C, and that only between 2% and 6% of ClD would be replenished by NPPL by the time the SLM would be experiencing another landslide triggering event. CgD/CT values would have to exceed 98% for any gains in C to be possible under these conditions. When ecosystem recovery rates are relatively fast (1.5 MgC ha 1 y 1) and landslide recurrence is less frequent (80 y), any retention exceeding approximately 60% of the C released by landsliding would result in a net gain of C. Although, the ecosystem would be able to recover only 41% and 59% of ClD by the next landslide-triggering event under scenarios 1 and 2 (respectively), a net gain of roughly 18% relative to the losses would be expected under the third scenario.

14 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

G03016

tropical montane regions could contribute to the global export of 0.09–0.78 PgC y 1 (average 0.27 PgC y 1). Although the ultimate fate of this C is very uncertain as it could either be released into the atmosphere or become incorporated into lowland or marine sedimentary deposits, its magnitude represents a quantity that amounts to 1%–11% of the 7.2 PgC y 1 emitted to the atmosphere by the burning of fossil fuels [Denman et al., 2007]. Under scenario 3 mass wasting would contribute to the retention of 0.06 to 0.51 PgC y 1 (average is 0.18 PgC y 1), a figure that is equivalent to 2.1%–19% of the residual land sink [Denman et al., 2007].

6. Conclusions

Figure 7. Relationship between the C sink potential ratio (Cp) for the entire SLM study area to that fraction of total C transferred by landslides that is sequestered within sedimentary deposits (CgD/CT). Cp equals the ratio of the landslide-related C gains due to ecosystem development (CgL) to C losses caused by physical exportation or exposure to oxidation (ClD). A ratio equal to unity implies neither gain nor loss of C as a result of landslide activity. Values 1 imply net loss and net gain as a consequence of mass wasting activity, respectively. The solid black line depicts ‘fast’ C recuperation rates on landslide scars and scours associated to ecosystem recovery and an 80-year recurrence interval, while the dashed black line represents the low end of C recuperation rates and a 20-year recurrence interval. 5.5. Implications [50] Results shown here imply that landslides could play an important role in the C cycle of mountainous regions at spatial scales ranging from individual watersheds to regions. At national scales mass wasting may also play an important role in countrywide C budgets. For example, in New Zealand, one of the highstanding islands in Oceania where mass wasting activity is prevalent, C export by rivers has been estimated to be equivalent to 40% of the country’s fossil fuel emissions [Preston et al., 2004; Scott et al., 2006]. [51] The mobilization of C by mass wasting activity may also represent a significant contribution to global C budgets. Earlier work integrating disturbance rates by landsliding with C content in vegetation and soil, showed that C fluxes associated with mass wasting were equivalent to 0.1– 0.3 PgC y 1 for the largest highstanding islands in Oceania whose rivers make a disproportionate contribution to global C fluxes (Table 5) [Stark et al., 1999]. We can now recalculate this figure based on the two extreme C retention scenarios discussed in the previous section (scenarios 1 and 3), C transfer rates associated with mass wasting activity calculated from Table 5, and the area covered by mountains lying between 300 and 3500 m within tropical regions around the globe (7132  103 km2) [Huddleston et al., 2003]. Under scenario 1 in which all of the C mobilized by mass wasting activity is lost, our estimate implies that mass wasting in

[52] This study quantified the effects of shallow landslide activity in the downslope transfer of modern organic C within a montane tropical environment. The study focused on the population of shallow landslides triggered by Hurricane Mitch (1998) on seven watersheds draining the southern flanks of the Sierra de Las Minas mountain range (SLM) in central Guatemala. SLM is part of the Northern Sierras of Central America, and is characterized by rugged topography, elevations ranging from 150 to 3015 m, and a wide variety of both dry and wet tropical life zones. [53] Results shown here imply that landslides may play an important role in the cycling of C within tropical mountainous regions at spatial scales ranging from the individual watershed to the regional. At the watershed level, a large variability was observed even among contiguous watersheds on the total amount of C transferred, its redistribution throughout the landscape, and in the potential for mass wasting to promote the sequestration or enhance losses of C. At the regional scale, mass wasting activity during Hurricane Mitch displaced approximately 3% of the entire C pool contained within the above ground biomass and soils. About 30% of the displaced C was delivered to hillslopes and firstorder streams, while the remaining 70% was delivered further downslope by debris flows. Landslide-related C flux rates calculated for the entire SLM region (0.08–0.33 MgC ha 1 y 1) were found to be within the range of values reported for other mountainous areas under varying climatic settings, while also being comparable to C flux rates associated to surface erosion. [54] Mass wasting has the potential to influence the cycling of C at a variety of spatial scales within mountain systems. At the watershed and regional scales the total net C losses or gains are dependent on the amount of C released by landsliding, the proportion of C released that gets sequestered in long-term sedimentary deposits, the recurrence interval of landslide-triggering events, the area denuded by landsliding, and the C recovery rates by ecosystem recuperation on denuded areas. Analyses presented here derive a key conclusion in that landslides triggered during a single event can induce either net long-term gains or net long-term losses of C. These opposing effects of mass wasting activity in C budgets could be expected even in contiguous watersheds. This conclusion should not come as a major surprise as the amount and fate of C transferred by landslides is controlled by a variety of biogeochemical factors which are in turn affected by dynamic climate and land use patterns. The significance of mass wasting activity in redistributing C

15 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

within the SLM study area, the relative importance of mountainous terrain in global C transfer rates, the limitations of current research approaches, and the inevitable dependency of mass wasting activity on land use dynamics and climate change highlights the need to make the study of landslides on the global C cycle a high-priority research topic. [55] Acknowledgments. The authors would like to thank the Ministerio de Agricultura, Ganadería y Alimentación, Consejo Nacional de Areas Protegidas, and Fundación Defensores de la Naturaleza of Guatemala for providing logistical support and sharing data. We are grateful to Oscar Nuñez, Cesar Tot, Francisco Hernandez, and Alex Guerra for their invaluable help in Guatemala, and to Niels Hovius and Fred Scatena, Rob Hilton, and an anonymous reviewer for providing thorough reviews of this article. We would also like to thank Rodrigo Sierra at the Center for Environmental Studies in Latin America at UT-Austin for providing data-processing hardware and software support. Funding for this project was provided through NSF-EPSCoR 9874782 (U.P.R.), NSF-EAR 0909271 (C.R.), and NSFDEB 0919043 (C.R.).

References Acosta, M., et al. (2001), Un método para la medición del carbono en los compartimentos subterráneos (raíces y suelo) de sistemas forestales y agrícolas en terrenos de ladera en México, paper presented at Simposio Internacional Medición y Monitoreo de la Captura de Carbono en Ecosistemas Forestales, Valdivia, Chile, 18–20 October. Benda, L., and T. Dunne (1997a), Stochastic forcing of sediment routing and storage in channel networks, Water Resour. Res., 33, 2865–2880, doi:10.1029/97WR02387. Benda, L., and T. Dunne (1997b), Stochastic forcing of sediment supply to channel networks from landsliding and debris flow, Water Resour. Res., 33(12), 2849–2863, doi:10.1029/97WR02388. Benda, L., et al. (2004), The network dynamics hypothesis: How channel networks structure riverine habitats, BioScience, 54, 413–427, doi:10.1641/ 0006-3568(2004)054[0413:TNDHHC]2.0.CO;2. Benda, L., et al. (2005), Geomorphology of steepland headwaters: The transition from hillslopes to channels, J. Am. Water Resour. Assoc., 41(4), 835–851. Berhe, A. A., et al. (2007), The significance of the erosion-induced terrestrial carbon sink, BioScience, 57, 337–346, doi:10.1641/B570408. Bigelow, P. E., et al. (2007), On debris flows, river networks, and the spatial structure of channel morphology, For. Sci. [Washington D. C.], 53, 220–238. Blair, N. E., et al. (2004), From bedrock to burial: The evolution of particulate organic carbon across coupled watershed-continental margin systems, Mar. Chem., 92, 141–156, doi:10.1016/j.marchem.2004.06.023. Bommer, J. J., and C. E. Rodriguez (2002), Earthquake-induced landslides in Central America, Eng. Geol. Amsterdam, 63(3–4), 189–220, doi:10.1016/S0013-7952(01)00081-3. Bosc, E. (1971), Geology of the San Agustín Acasaquastlan quadrangle, east central Guatemala, PhD dissertation, 131 pp., Rice Univ., Houston, Tex. Bruijnzeel, L. A., and E. J. Veneklass (1998), Climatic conditions and tropical montane forest productivity: The fog has not lifted yet, Ecology, 79, 3–9, doi:10.1890/0012-9658(1998)079[0003:CCATMF]2.0.CO;2. Bucknam, R. C., et al. (2001), Landslides triggered by hurricane Mitch in Guatemala: Inventory and discussion, Open File Rep. 01–443, U.S. Geol. Surv., Reston, Va. Cadol, D., and E. Wohl (2010), Wood retention and transport in tropical, headwater streams, La Selva Biological Station, Costa Rica, Geomorphology, 123, 61–73, doi:10.1016/j.geomorph.2010.06.015. Castellanos, E., and C. Flores (2006), Estimacion del Contenido de Carbono en Bosques del Altiplano Occidental de Guatemala, 64 pp., Cent. de Estud. Ambient., Univ. del Valle de Guatemala, Guatemala City. Castellanos, E., et al. (2007), Cuantificacion de Carbono Capturado por Bosques Comunales y Municipales de Cuatro Municipios en los Departamentos de San Marcos y Huehuetenango, 95 pp., Cent. de Estud. Ambient., Univ. del Valle de Guatemala, Guatemala City. Cavelier, J. (1996), Environmental factors and ecophysiological processes along altitudinal gradients in wet tropical mountains, in Tropical Forest Plant Ecophysiology, edited by S. S. Mulkey, A. P. Smith, and R. L. Chazdon, pp. 399–439, Springer, New York, doi:10.1007/978-1-46131163-8_14. Clark, D., et al. (2001), Net primary production in tropical forests: An evaluatuon and synthesis of existing field data, Ecol. Appl., 11(2), 371–384, doi:10.1890/1051-0761(2001)011[0371:NPPITF]2.0.CO;2.

G03016

Coe, J. A., et al. (2004), Landslide susceptibility from topography in Guatemala, in Landslides: Evaluation and Stabilization, edited by W. A. Lacerda et al., pp. 69–78, Taylor and Francis, London, U. K. Davidson, E. A., and I. L. Ackerman (1993), Changes in soil carbon inventories following cultivation of previously untilled soils, Biogeochemistry, 20, 161–193, doi:10.1007/BF00000786. De Jong, B. H. J., et al. (1999), Land-use change and carbon flux between 1970s and 1990s in central highlands of Chiapas, Mexico, Environ. Manage. N. Y., 23, 373–385, doi:10.1007/s002679900193. Denman, K. L., et al. (2007), Coupling between changes in climate system and biogeochemistry, in Climate Change 2007: The Physical Science. Contribution of Workin Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., pp. 501–587, Cambridge Univ. Press, Cambridge, U. K. Densmore, A. L., and N. Hovius (2000), Topographic fingerprints of bedrock landslides, Geology, 28(4), 371–374, doi:10.1130/0091-7613(2000) 282.0.CO;2. Dietrich, W. E., et al. (1995), A process-based model for colluvial soil depth and shallow landsliding using digital elevation data, in Scale Issues in Hydrological Modelling, edited by J. D. Kalma and M. Sivaplan, pp. 383–400, John Wiley & Sons, New York, New York, doi:10.1002/ hyp.3360090311. Dix, M. A. (1997), Sierra de Las Minas region and biosphere reserve, Guatemala, in Centres of Plant Diversity: A Guide and Strategy for Their Conservation: Vol 3, The Americas, edited by S. D. Davis et al., pp. 193–197, Int. Union for Conserv. of Nat., Cambridge, U. K. Downing, J. A., et al. (2008), Sediment organic burial in agriculturally eutrophic impoundments over the last century, Global Biogeochem. Cycles, 22, GB1018, doi:10.1029/2006GB002854. Eswaran, W., et al. (1993), Organic carbon in soils of the world, Soil Sci. Soc. Am. J., 57, 192–194, doi:10.2136/sssaj1993.03615995005700010034x. Etchevers, J., et al. (2001), Los stocks de carbono en diferentes compartimentos de la parte aérea y subterránea en sistemas forestales y agrícolas de ladera en México, paper presented at Simposio Internacional Medicion y Monitoreo de la Captura de Carbono en Ecosistemas Forestales, Valdivia, Chile, 18–20 Oct. Franco, A., et al. (2009), Seismicity and crustal structure of the PolochicMotagua Fault System area (Guatemala), Seismol. Res. Lett., 80, 977–984, doi:10.1785/gssrl.80.6.977. Friedel, M. J. (2008), Regularized joint inverse estimation of extreme rainfall amounts in ungauged coastal basins of El Salvador, Nat. Hazards, 46, 15–34, doi:10.1007/s11069-007-9179-1. Fundación Defensores de La Naturaleza (2002), Parcelas de Medicion y Monitoreo de Carbono: Campaña de Recoleccion de Datos 2002, Reserva de Biosfera Sierra de las Minas, Guatemala City. Fundación Defensores de la Naturaleza (2003), III Plan Maestro 2003– 2008, 84 pp., Reserva de Biosfera Sierra de las Minas, Guatemala City. Galy, V., and T. Eglinton (2011), Protracted storage of biospheric carbon in the Ganges-Brahmaputra basin, Nat. Geosci., 4, 843–847, doi:10.1038/ ngeo1293. Garwood, N., et al. (1979), Earthquake-caused landslides: A major disturbance to tropical forests, Science, 205, 997–999, doi:10.1126/ science.205.4410.997. Goldsmith, S. T., et al. (2008), Extreme storm events, landscape denudation, and carbon sequestration: Typhoon Mindulle, Choshui River, Taiwan, Geology, 36, 483–486, doi:10.1130/G24624A.1. Gomez, B., et al. (2003), Production, storage, and output of particulate organic carbon: Waipaoa River basin, New Zealand, Water Resour. Res., 39(6), 1161, doi:10.1029/2002WR001619. Goode, J. R., and E. Wohl (2010), Substrate controls on the longitudinal profile of bedrock channels: Implications for reach-scale roughness, J. Geophys. Res., 115, F03018, doi:10.1029/2008JF001188. Guariguata, M. R. (1990), Landslide disturbance and forest regeneration in the upper Luquillo Mountains of Puerto Rico, J. Ecol., 78(3), 814–832, doi:10.2307/2260901. Guerra-Noriega, A. (2010), Climate-related disaster risk in mountain areas: The Guatemalan Highlands at the start of the 21st century, PhD thesis, 287 pp., Univ. of Oxford, Oxford, U. K. Hales, T. C., and J. J. Roering (2005), Climate-controlled variations in scree production, Southern Alps, New Zealand, Geology, 33(9), 701–704, doi:10.1130/G21528.1. Hansen, A. (1984), Engineering geomorphology: The application of an evolutionary model of Hong Kong’s terrain, Z. Geomorphol., 51, Suppl., 39–50. Harden, J. W., et al. (1999), Dynamic replacement and loss of soil carbon on eroding cropland, Global Biogeochem. Cycles, 13, 885–901, doi:10.1029/1999GB900061. Harvey, A. M. (2001), Coupling between hillslopes and channels in upland fluvial systems: Implications for landscape sensitivity, illustrated from the

16 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

Howgill Fells, northwest England, Catena, 42, 225–250, doi:10.1016/ S0341-8162(00)00139-9. Hilton, R. G., et al. (2008a), Riverine particulate organic carbon from an active mountain belt: Importance of landslides, Global Biogeochem. Cycles, 22(1), GB1017, doi:10.1029/2006GB002905. Hilton, R. G., et al. (2008b), Tropical-cyclone-driven erosion of the terrestrial biosphere from mountains, Nat. Geosci., 1, 759–762, doi:10.1038/ ngeo333. Hilton, R. G., et al. (2011), Landslide impact on organic carbon cycling in a temperate montane forest, Earth Surf. Processes Landforms, 36, 1670–1679, doi:10.1002/esp.2191. Houghton, R. A., et al. (1998), Missing sinks, feedbacks, and understanding the role of terrestrial ecosystems in the global carbon balance, Global Biogeochem. Cycles, 12, 25–34, doi:10.1029/97GB02729. Hovius, N., et al. (1997), Sediment flux from a mountain belt derived by landslide mapping, Geology, 25, 231–234, doi:10.1130/00917613(1997)0252.3.CO;2. Hovius, N., et al. (2000), Supply and removal of sediment in a landslidedominated mountain belt: Central Range, Taiwan, J. Geol., 108(1), 73–89, doi:10.1086/314387. Hovius, N., et al. (2011), Prolonged seismically induced erosion and the mass balance of a large earthquake, Earth Surf. Processes Landforms, 304, 347–355. Huddleston, B., et al. (2003), Towards a GIS-based analysis of mountain environments and populations, Working Pap. 10, Food and Agric. Org. of the United Nations, Rome. Hyatt, T. L., and R. J. Naiman (2001), The residence time of large woody debris in the Queets Rivers, Washington, USA, Ecol. Appl., 11, 191–202, doi:10.1890/1051-0761(2001)011[0191:TRTOLW]2.0.CO;2. Hyman, G. C. (1997), Sediment transfer in the Rio Pacuare watershed and neighboring basins of Costa Rica, PhD thesis, Univ. of Tennessee, Knoxville. INETER (1998), Las Lluvias del Siglo en Nicaragua. El Huracán Mitch, las Lluvias y otros Eventos Ciclónicos y su Recurrencia en Nicaragua, Inst. Nicaraguense de Estud.Territoriales, Managua. IPGH (1976), Atlas Climatológico e Hidrológico del Istmo Centroamericano, Inst. Panam. de Geogr. e Hist., Guatemala City. Jacinthe, P. A., and R. Lal (2001), A mass balance approach to assess carbon dioxide evolution during erosional events, Land Degradation Dev., 12, 329–339, doi:10.1002/ldr.454. Jaramillo, V. J., et al. (2003), Biomass, carbon, and nitrogen pools in Mexican tropical dry forested landscapes, Ecosystems (N. Y.), 6, 609–629, doi:10.1007/s10021-002-0195-4. Jordan, C. F. (1985), Nutrient Cycling in Tropical Forest Ecosystems: Principles and Their Application in Management and Conservation, Wiley, Chichester, U. K. Kelsey, H. M. (1982), Hillslope evolution and sediment movement in a forested headwater basin, Van Dusen River, in Sediment Budgets and Routing in Forested Drainage Basins, Gen. Tech. Rep. PNW-141, edited by F. J. Swanson et al., pp. 86–96, U.S. Dep. of Agric., Portland, Oreg. Knighton, A. D. (1999), Downstream variation in stream power, Geomorphology, 29, 293–306, doi:10.1016/S0169-555X(99)00015-X. Korup, O. (2005), Geomorphic imprint of landslides on alpine river systems, southwest New Zealand, Earth Surf. Processes Landforms, 30, 783–800, doi:10.1002/esp.1171. Lal, R. (2005), Soil erosion and carbon dynamics, Soil Tillage Res., 81, 137–142, doi:10.1016/j.still.2004.09.002. Lancaster, S. T., and N. E. Casebeer (2007), Sediment storage and evacuation in headwater valleys at the transition between debris-flow and fluvial processes, Geology, 35(11), 1027–1030, doi:10.1130/G239365A.1. Leithold, E. L., et al. (2005), Sedimentation and carbon burial on the northern California continental shelf: The signatures of land-use change, Cont. Shelf Res., 25(3), 349–371, doi:10.1016/j.csr.2004.09.015. Leithold, E. L., et al. (2006), Geomorphologic controls on the age of particulate organic carbon from small mountainous and upland rivers, Global Biogeochem. Cycles, 20(3), GB3022, doi:10.1029/2005GB002677. Lin, G.-W., et al. (2008), Effects of earthquake and cyclone sequencing on landsliding and fluvial sediment transfer in a mountain catchment, Earth Surf. Processes Landforms, 33(9), 1354–1373, doi:10.1002/esp.1716. Liu, S., et al. (2003), Modeling carbon dynamics in vegetation and soil under the impact of soil erosion and deposition, Global Biogeochem. Cycles, 17(2), 1074, doi:10.1029/2002GB002010. Lodolo, E., et al. (2009), Deep structural setting of the North AmericanCaribbean plate boundary in eastern Guatemala, Geofis. Int., 48, 263–277. Lopez, F. (1999), Evaluación Hidrológia de las Crecidas Provocadas por el Paso del Huracán Mitch en Guatemala, PHI/UNESCO, Montevideo. Lott, N., et al. (1999), Mitch: The deadliest Atlantic hurricane since 1780, http://lwf.ncdc.noaa.gov/oa/reports/mitch/mitch.html, NOAA Satell. and Inf. Serv., Natl. Clim. Data Cent., Washington, D. C.

G03016

Lyons, W. B., et al. (2002), Organic carbon fluxes to the ocean from highstanding islands, Geology, 30(5), 443–446, doi:10.1130/0091-7613(2002) 0302.0.CO;2. Maidment, D. R. (2002), ArcHydro. GIS for Water Resources, 300 pp., ESRI Press, Redlands, Calif. Marshall, J. S. (2007), Geomorphology and physiographic provinces, in Central America: Geology, Resources, and Hazards, Vol. 1, edited by J. Bundschuh and G. E. Alvarado, pp. 75–122, Taylor and Francis Group, London, U. K, doi:10.1201/9780203947043.pt2. Michot, T. C., et al. (2002), Impacts of Hurrican Mitch on water quality and sediments of Lake Izabal, Open File Rep. 03-180, 20 pp., U.S. Geol. Surv., Lafayette, La. Milliman, J. D., and J. M. Syvitski (1992), Geomorphic/tectonic control of sediment discharge to the ocean: The importance of small mountainous rivers, J. Geol., 100, 525–544, doi:10.1086/629606. Ministerio de Agricultura, Ganadería y Alimentación (MAGA) (2000), AtlasTemático de la República de Guatemala, Map 9: Mapa de Clasificación Taxonomía de Suelos, scale 1:250,000, Guatemala City. Nadim, F., et al. (2006), Global landslide and avalanche hotspots, Landslides, 3, 159–173, doi:10.1007/s10346-006-0036-1. Newcomb, W. E. (1978), Geological map of Guatemala, Rio Hondo, Hoja 2261 II G, scale 1:50,000, Inst. Geogr. Nac., Guatemala City. Ortiz-Zayas, J. R. (1998), The metabolism of the Rio Mameyes, Puerto Rico: Carbon fluxes in a tropical rain forest river, PhD thesis, Univ. of Colorado, Boulder, Colo. Pandey, A. N., and J. S. Singh (1985), Mechanism of ecosystem recovery: A case study from Kumaun, Himalaya, Reclam. Revegetation Res., 3, 271–292. Pearce, A. J., and A. J. Watson (1986), Effects of earthquake-induced landslides on sediment budget and transport over a 50-yr period, Geology, 14, 52–55, doi:10.1130/0091-7613(1986)142.0.CO;2. Peart, M. R., K. Y. Ng, and D. D. Zhang (2005), Landslides and sediment delivery to a drainage system: Some observations from Hong Kong, J. Asian Earth Sci., 25, 821–836, doi:10.1016/j.jseaes.2004.08.004. Preston, N. J., et al. (2004), Estimating national erosion-related soil carbon losses in New Zealand: An indicator of Kyoto compliance and sustainability, in Agricultural Impacts on Soil Erosion and Soil Biodiversity: Developing Indicators for Policy Analysis, edited by R. Francaviglia, pp. 301–310, Ist. Sper. per la Nutr. delle Piante, Rome. Raich, J. W., et al. (2006), Temperature influences carbon accumulation in moist tropical forests, Ecology, 87, 76–87, doi:10.1890/05-0023. Reddy, V. S., and J. S. Singh (1993), Changes in vegetation and soil during succession following landslide disturbance in the central Himalaya, J. Environ. Manage., 39, 235–250, doi:10.1006/jema.1993.1068. Restrepo, C., and N. Alvarez (2006), Landslides and their contribution to land-cover in the mountains of Mexico and Central America, Biotropica, 38, 446–457, doi:10.1111/j.1744-7429.2006.00178.x. Restrepo, C., et al. (2003), Landslides significantly alter land cover and the distribution of biomass: An example from the Ninole ridges of Hawai’i, Plant Ecol., 166, 131–143, doi:10.1023/A:1023225419111. Rodríguez, C. E. (2007), Earthquake-induced landlsides, in Central America: Geology, Resources, and Hazards, edited by J. Bundschuh and G. E. Alvarado, pp. 1217–1256, Taylor and Francis, London, U. K, doi:10.1201/9780203947043.ch40. Scatena, F. N., and A. E. Lugo (1995), Geomorphology, disturbance and soil vegetation of two subtropical wet steepland watersheds of Puerto Rico, Geomorphology, 13, 199–213, doi:10.1016/0169-555X(95)00021-V. Schlesinger, W. H. (1986), Changes in soil carbon storage and associated properties with disturbance and recovery, in The Changing Carbon Cycle: A Global Analysis, edited by J. R. Trabalka and D. E. Reichle, pp. 194–220, Springer, New York. Schrott, L., et al. (2003), Spatial distribution of sediment storage types and quantification of valley fill deposits in an alpine basin, Reintal, Bavarian Alps, Germany, Geomorphology, 55, 45–63, doi:10.1016/S0169-555X(03) 00131-4. Schulze, E.-D., et al. (2000), Climate change: Managing forests after Kyoto, Science, 289, 2058–2059, doi:10.1126/science.289.5487.2058. Schwab, M., et al. (2008), Landsliding and sediment flux in the Central Swiss Alps: A photogrammetric study of the Schimbrig landslide, Entlebuch, Geomorphology, 97(3–4), 392–406, doi:10.1016/j.geomorph.2007.08.019. Scott, D., et al. (2006), Localized erosion affects national carbon budget, Geophys. Res. Lett., 33, L01402, doi:10.1029/2005GL024644. Simmons, C. S., et al. (1959), Clasificación de Reconocimiento de los Suelos de la República de Guatemala, 1000 pp., Jose de Pineda Ibarra, Guatemala City. Smith, S. V., et al. (2005), Fates of eroded soil organic carbon: Mississippi basin case study, Ecol. Appl., 15(6), 1929–1940, doi:10.1890/05-0073. Solomon, S., et al. (Eds.) (2007), Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of

17 of 18

G03016

RAMOS SCHARRÓN ET AL.: EFFECTS OF LANDSLIDES ON CARBON BUDGETS

the Intergovernmental Panel on Climate Change, Cambridge Univ. Press, Cambridge, U. K. Stallard, R. F. (1998), Terrestrial sedimentation and the carbon cycle: Coupling weathering and erosion to carbon burial, Global Biogeochem. Cycles, 12, 231–257, doi:10.1029/98GB00741. Stark, C., et al. (1999), The pivotal role of landsliding in montane ecosystems and its quantitative significance in regional and global carbon cycles, Abstract B31A-15, presented at 1999 Fall Meeting AGU, San Francisco, Calif., 13–17 Dec. Strahler, A. N. (1964), Quantitative geomorphology of drainage basins and channel networks, in Handbook of Applied Hydrology, edited by V. T. Chow, pp. 40–74, McGraw Hill, New York. TECSULT Foresterie, Inc. (2000), Metodologia de Cartografia. Mapas de Ecosistemas Vegetales de Guatemala, Instituto Nacional de Bosques de Guatemala (INAB), World Bank, Washington, D. C. Turk, J. K., and R. C. Graham (2009), Soil carbon and nitrogen accumulation in a forested debris flow chronosequence, California, Soil Sci. Soc. Am. J., 73, 1504–1509, doi:10.2136/sssaj2008.0106. Van Oost, K., et al. (2007), The impact of agricultural soil erosion on the global carbon cycle, Science, 6, 189–207, doi:10.1126/science.1145724. Velázquez, E., and A. Gómez-Sal (2008), Landslide early succession in a neotropical dry forest, Plant Ecol., 199, 295–308, doi:10.1007/s11258008-9433-y.

G03016

Walker, L. R., and A. B. Shiels (2008), Post-disturbance erosion impacts carbon fluxes and plant succession on recent tropical landslides, Plant Soil, 313(1–2), 205–216, doi:10.1007/s11104-008-9692-3. West, A. J., et al. (2011), Mobilization and transport of coarse woody debris to the oceans triggered by an extreme tropical storm, Limnol. Oceanogr., 56(1), 77–85, doi:10.4319/lo.2011.56.1.0077. Weyl, R. (1980), Geology of Central America, Gebrüder Borntraeger, Berlin. Wilcke, W., et al. (2003), Soil properties on a chronosequence of landslides in montane rain forest, Ecuador, Catena, 53(1), 79–95, doi:10.1016/ S0341-8162(02)00196-0. Yoo, K., et al. (2005), Erosion of upland hillslope soil organic carbon: Coupling field measurements with a sediment transport model, Global Biogeochem. Cycles, 19, GB3003, doi:10.1029/2004GB002271. Zarin, D. J., and A. H. Johnson (1995a), Base saturation, nutrient cation, and organic-matter increases during early pedogenesis on landslide scars in the Luquillo experimental forest, Puerto-Rico, Geoderma, 65(3–4), 317–330, doi:10.1016/0016-7061(94)00048-F. Zarin, D. J., and A. H. Johnson (1995b), Nutrient accumulation during primary succession in a montane tropical forest, Puerto-Rico, Soil Sci. Soc. Am. J., 59(5), 1444–1452, doi:10.2136/sssaj1995.03615995005900050034x.

18 of 18