Fundamental and Applied Limnology Archiv für Hydrobiologie Vol. 174/1: 63–73, February 2009 © E. Schweizerbart’sche Verlagsbuchhandlung 2009
Relation between floodplain land use and river hydromorphology on different spatial scales – a case study from two lower-mountain catchments in Germany Jochem Kail*, Sonja C. Jähnig and Daniel Hering1 University of Duisburg-Essen, Department of Applied Zoology / Hydrobiology With 2 figures and 4 tables Abstract: The influence of land use on the hydromorphological state of streams has rarely been investigated and most of the studies focused on catchment land use. Moreover, contrasting results were reported. The objective of our study was to investigate the relation between local hydromorphology and land use on different spatial scales, to identify spatial scales of special importance, and to test, if it is possible to predict hydromorphology using land use data. We differentiated between two lateral spatial scales (buffer and floodplain) and three longitudinal scales (site, reach, catchment). The results indicate that the hydromorphological state of streams is significantly related to the land use on all spatial scales investigated. Differences are small, but there is some evidence that land use on the floodplain and on the reach scale is of special importance. Considering different spatial scales simultaneously distinctly increases model predictability. But even the variance of the hydromorphological data explained by these statistical models (20–41 %) is too low to use land use as a predictor for specific channel characteristics. Land use data are better suited to predict the overall hydromorphological state of the study streams. Moreover, it is possible to derive statistically significant relations between single land use categories and single hydromorphological parameters. Key words: LAWA-Vor-Ort, spatio-temporal scale, Lahn, Eder.
Introduction Many studies showed that land use has a profound effect on processes, which act on a catchment scale by altering vegetation cover, e.g., surface-runoff and discharge (e.g., Knox 1987), erosion and sediment supply (e.g., Liebault et al. 2005). Therefore, most studies addressing the relation of land use, hydromorphology, and biota aggregate land use data on large spatial scales (e.g., Feld 2004, Townsend et al. 2004, Lautenschläger & Kiel 2005). However, land use may also impact river hydromorphology and biota on smaller spatial scales, be-
cause most modifications of the channel (straightening, bank- and bed-revetment) aim at ensuring floodplain land uses. Therefore, the intensity of human impact may depend on the economic value of floodplain land use. Based on these considerations, the present nonnatural state of rivers in intensively used landscapes can be viewed as an equilibrium state, which results from two forces: First, the natural processes, which are controlled by topography, geology, and climate. If these processes are not disturbed by man, hydromorphology reaches an equilibrium state in a time scale, which differs between river sizes and types. Second, the human impact, which intends to prevent hazards
Authors’ address: 1
University of Duisburg-Essen, Department of Applied Zoology / Hydrobiology, D-45117 Essen, Germany.
* Corresponding author: e-mail:
[email protected]
DOI: 10.1127/1863-9135/2009/0174-0063
1863-9135/09/00174-0063 $ 2.75 © 2009 E. Schweizerbart’sche Verlagsbuchhandlung, D-70176 Stuttgart
64
J. Kail et al.
and to ensure the land uses on the adjacent floodplain. If the natural hydromorphological state or the natural processes restrict these land uses, man either alters the natural state / processes or changes land use. From these considerations the questions arise if land use on spatial scales smaller than catchments is related to river hydromorphology and if there is a spatial scale of special importance. Few studies investigated the relation between land use and hydromorphology even on larger spatial scales with conflicting results. Allan et al. (1997) found that the hydromorphological state is strongly correlated to agricultural land use of the entire catchment upstream and correlations are progressively weaker at smaller spatial scales down to the stream reach. In contrast, Richards et al. (1996) reported only minor differences in the relation of hydromorphology and land use at two different spatial scales (whole catchment and buffer along the stream upstream of the reach investigated). Vondracek et al. (2005) observed no differences in the relative importance of the land use on different spatial scales in explaining channel characteristics. In most of these studies land use data with a relatively coarse spatial resolution have been used. Richards et al. (1996) sup-
posed that the coarse resolution of their land use data (61 m) may have weakened the ability to analyse finescale variation in land use and that land use data with a finer spatial resolution are required to detect relations between land use and hydromorphology at small spatial scales like buffers along stream reaches. The objective of this study was to investigate if relations between land use and hydromorphology can be found at spatial scales smaller than catchments, using land use data with a high spatial resolution and if there is a spatial scale of special importance. We further investigated if the correlation between land use and hydromorphology is strong enough to assess the hydromorphological state with land use data.
Methods Study area The study area is located in the federal states of Hesse and Northrhine-Westphalia (Germany) and comprises the upper part of the river Lahn and Eder catchment (Fig. 1). This lower mountainous area (ecoregion 9 according to Illies (1978), modified after Briem (2003)) is dominated by shale (77 %) and new red sandstone (13 %). Population density is low (~175 p/km2)
Fig. 1. Location of the study catchment within Germany (left) and streams in the Eder and Lahn sub-catchments for which land use and hydromorphological data were available (right). Ecoregions are bordered by dotted lines (according to Illies 1978, modified according to Briem 2003): I = lowland, II = lower mountain area, III = alpine region.
Land use and river hydromorphology compared to the mean value for the whole of Germany (~230 p/ km2 ) and land use of the floodplain is dominated by forest, pasture, and grassland. Area of the Eder and Lahn sub-catchments is 1173 km2 and 719 km2, respectively (major tributary of the Lahn with catchment size 985 km2 not considered in this study). Mean flow at the downstream end of the sub-catchments is approximately 10 m3 s–1 and 20 m3 s–1, respectively. Downstream of the sub-catchments, both rivers are important waterways heavily impacted by bank revetments and sluices and hence, the influence of other land uses on hydromorphology cannot be investigated. The stream classification used for the hydromorphological survey mentioned below is mainly based on stream size and valley shape: (A) Small confined headwater streams in V-notched valleys: small, steep, and straight channels downstream of the stream sources with step-pool sequences. (B) Small headwater streams in alluvial deposits: small, sinuous channels with rifflepool sequences, dominated by cobbles in the shale region and dominated by sand in the sandstone region. (C) Medium-sized streams in alluvial deposits and (D) large streams in alluvial deposits essentially correspond to the stream types 5 and 9, respectively described by Pottgiesser & Sommerhäuser (2004) (a short description in English is available on the web http://www. wasserblick.net/servlet/is/24739/?lang=de).
Hydromorphological data The study is based on a large hydromorphological data set that has been compiled from regional authorities in NorthrhineWestphalia and Hesse. Since the mid-1990’s, hydromorphological surveys have been conducted in the two federal states. Slightly different methods have been applied in the surveys performed by the two federal states, but they do essentially correspond to the field survey method of the “Länderarbeitsgemeinschaft Wasser” (LAWA) briefly described by Raven et al. (2002) and Kail & Hering (2005). A total of 25 parameters were mapped for 100 m channel segments and compared to a reference condition, which is defined as the “potential natural state” of the stream (the condition that would result from the present natural setting without further human intrusion, comparable to the definition of the potential natural vegetation by Tüxen 1956). For the statistical analysis, only those parameters were used, which were mapped on an ordinal scale with at least five classes and which show a gradient ranging from “natural” to “heavily degraded” (17 out of the 25 parameters, Table 2). Data on these 17 parameters are available for about 13,600 sections (data sets) in the study catchment.
ports, railway stations, railway constructions), (5) urban area, low-density (e.g. parks, sports grounds, camping sites), (6) urban area, high-density (e.g. private housing, industry, churches, hospitals); and (7) “other”. Line: (1) dirt roads, (2) traffic routes (roads, railway lines). The land use category “other” includes areas, where land use could not be determined on the aerial photographs and areas without vegetation, which in many cases are ruderal areas. The ruderal areas in the floodplain of stream type D are generally located along main roads and railway lines, which are too small to be mapped as an areal land use category (traffic infrastructure). Therefore, in general, the land use category “other” represents traffic infrastructure in stream type D. The units are: area of the buffer or floodplain segment covered by the areal land use categories (%) and length of traffic routes and dirt roads related to the area of the buffer or floodplain segment for the linear land use categories (m m–2).
Spatial scales It is hypothesized that land use adjacent to the channel is most strongly impacting hydromorphology. Therefore, we distinguished between the land use within a buffer strip (two times bankfull channel width at both sides of the stream) and the land use of the floodplain. The floodplain was demarcated using topographic, soil, and geological maps. The buffer and the floodplain area were cut at the start- and endpoints of the 100 m stream sections of the hydromorphological survey. Besides these two lateral spatial scales, we distinguished three longitudinal spatial scales: site, reach, and catchment. At the site scale, only the land use within the buffer or floodplain of the 100 m stream section was considered. At the reach scale, the land use within some 100 m stream sections up- and downstream of the site was considered additionally. The number of considered adjacent sections depends on stream size (Table 1). At the catchment scale, the land use in the buffer or floodplain in the whole catchment upstream from the site was considered.
Table 1. Main characteristics of the stream types investigated. Number of sections, for which land use and hydromorphological data are available, range of bankfull channel width, length of the reach scale, and land use in the floodplain are given for the four stream types investigated. *length of the reach scale results from the 100 m site section and a specific number of 100 m sections up- and downstream of the site section, the number of sections considered depends on stream size (bankfull channel width).
stream type
Land use data We used “ATKIS®-Basis-DLM” (Official Topographical Cartographical Information System) land use data, which were mapped by regional authorities on topographical maps with a scale 1:5,000 (www.atkis.de, www.lverma.nrw.de). The data consist of areal (polygons) and line data (e.g., roads, streams) with a total of 192 different land use categories being distinguished. For multivariate statistical analysis and to ease interpretation of the results, we aggregated the areal and line land use categories to more general categories of increasing economic value and land use pressure “exerted” on the stream sections: Areal: (1) woody vegetation (e.g. forest, gallery of trees), (2) extensive agriculture (e.g. pasture, grassland), (3) intensive agriculture (e.g. cropland), (4) traffic infrastructure (e.g. air-
65
A n sample sites bankfull channel width (m) reach scale (m)* land-use on floodplain (%area) urban high density urban low density traffic infrastructure intensive agriculture extensive agriculture woody vegetation other
C
D
1604 8646 10:1) to “very deeply entrenched” (< 3:1)
culverts
5
“no culverts” to “culvert > 50 % of section length”
artificial impoundments
5
“no impoundment” to “no apparent flow at mean flow”
bed-fixation
5
bank-revetment
7
“no bed fixation” to “concrete > 50 % of section length”
cross-section form
7
“natural cross-section” to “rectangular cross-section”
woody riparian vegetation
5
“natural forest” to „none”
non-woody riparian vegetation
5
“natural vegetation” to “none”
Combining the lateral and longitudinal scales, six different spatial scales can be distinguished, for which the %area covered by the different land use categories was calculated. In the following, they are referred to as: site buffer (sb), site floodplain (sf), reach buffer (rb), reach floodplain (rf), catchment buffer (cb), and catchment floodplain (cf).
Data analysis The relation of land use and hydromorphology may be confounded by natural parameters, such as geology or valley slope, because they are also related to both, land use and hydromorphology. Furthermore, the pressure of natural processes like flooding or lateral channel migration on the land use depends on stream size, slope and bank material, i.e. it is stream-type specific. For these reasons, it is crucial to partial out the effects of the natural parameters. This could be done by using these natural controls as co-variables in multivariate statistics. Alternatively, natural controls can be considered through a stream type specific analysis; stream types “subsume” natural controls, since they are defined by the natural setting.
Land use data were used as independent variables and hydromorphological data as dependent variables. For each of the six spatial scales, the data on the percentage-area covered by the different land use categories were arcsin-transformed (Podani 2000). Data on hydromorphology were log-transformed, centred and standardized to fit the normal distribution and because the data were measured on different scales (different number of ordinal classes). We used ordination techniques to investigate the relation between the land use and hydromorphological data. Detrended correspondence analysis (DCA) was performed for the hydromorphological data to determine the length of the gradient in the data sets and to choose an appropriate ordination technique. The length of the gradient is 3 for all data sets. Therefore, RDA was used instead of CCA according to Jongman et al. (1995). We used a Monte Carlo test with 499 permutations to investigate, if the land use variables have a significant conditional effect when they are included in the statistical model (p < 0.05). Because the sample size for the four stream types is comparably large, virtually all land use categories significantly contribute to the variance of the hydromorphological data-set explained by the land use data. To identify the most important
67
Land use and river hydromorphology land use categories, we only considered variables, for which the additional variance explained at the time they are included in the statistical model (lambda-A) is ≥ 0.03.
Results Lateral scales buffer and floodplain Land use at the floodplain scale is more strongly impacting hydromorphology than land use at the buffer scale. The share of variance of the hydromorphological data explained by the nine land use categories is significantly larger at the floodplain scales compared to the corresponding buffer scales (Wilcoxon Matched Pairs Test, p < 0.01, n = 12). Only for stream type C at the catchment scale, land use in the stream buffer explained hydromorphology better. Therefore, we reject the hypothesis that the land use near the channel is more influential on the hydromorphological state of the stream compared to land use far away from the stream section. However, differences between the two lateral scales are small (Table 3). The differences between the floodplain and buffer scale range from 0.1 to 2.4 percentage points with a median value of 1.6 for stream types A-C. Only in larger streams (stream type D), the influence of the floodplain scale is markedly higher. This is especially true for the site and reach scale, where the share of variance explained at the floodplain scale is 1.75 and 1.33 times the share of variance explained at the buffer scale. Longitudinal scales site, reach, and catchment For three out of four stream types (A–C) land use at the reach scale is most strongly impacting hydromorphology. However, differences between the three longitudinal scales are small (Table 3). This is especially true for the stream types, which represent headwater streams (stream type A and B, maximum difference for six spatial scales 2.5 and 1.6 percentage points, respectively). For stream type D, which represents larger streams with a bankfull channel width > 10 m, the land use on the catchment scale has a distinctly greater influence on local hydromorphology compared to the site and reach scales (Table 3). The share of variance of the hydromorphological data explained by the floodplain land use data is about 1.6 and 1.3 the variance explained by the site and reach scales, respectively. Moreover, the share of variance explained on the catchment floodplain scale is the largest compared to all other spatial scales and for all stream types.
Table 3. Share of variance (%) of the hydromorphological data explained by the land use data at the six different spatial scales and for the four stream types. The total variance explained in a RDA by all nine land use categories is given and the highest value for each stream type is highlighted in bold (virtually all land use categories significantly contribute to the variance explained, p < 0.05). Moreover, the total variance explained by all six spatial scales (6*9 = 54 land use categories) and the variance explained by the land use categories, which significantly contribute to the model (p < 0.05) is given for each stream type.
stream type spatial scale site buffer site floodplain
A 14.1 15.9
B 14.7 15.6
C 13.5 15.9
D 8.8 15.4
reach buffer reach floodplain
14.4 16.0
16.0 16.3
15.9 18.3
13.9 18.5
catchment buffer catchment floodplain
13.5 15.7
15.2 15.3
18.0 17.1
22.7 23.9
all spatial scales
24.4
20.2
31.0
41.3
23
20
29
41
all spatial scales which significantly contribute to the model
The share of variance explained is markedly higher for all stream types, if land use categories of all spatial scales are considered (20–41 %), even if only categories are included, which significantly contribute to the model (p < 0.05). Land use categories of special importance In streams of stream type A and B, representing small headwater streams, sections bordered by forest are in a near-natural state and therefore, the land use category “woody vegetation” is strongly related to hydromorphology (Table 4). The variance explained by this land use category is highest at all single six spatial scales and for the complete model, where all land use categories of all six spatial scales are considered. The area covered by urban infrastructure is small, which obviously is the reason, why this land use category has a minor conditional effect. In contrast, the land use category “urban area (high density)” is an indicator for a poor hydromorphological state in streams of stream type C (Table 4), where land use pressure is higher and – compared to stream type A and B – a larger part of the buffer and floodplain area is covered by urban infrastructure (Table 1). The variance explained by this land use category is highest at all spatial scales, and woody vegetation is the only one with a substantial conditional effect (lambda-A ≥ 0.03). For stream type D, the length of dirt roads and the area covered by traffic infrastructure are indicators
68
J. Kail et al.
for a heavily altered hydromorphological state on the catchment scale (Table 4). The variance explained by the land use category “traffic infrastructure” is highest for the complete model, where all land use categories of all six spatial scales are considered. On the reach
and site scale, intensive agriculture (cropland) and the land use category “other” (mainly ruderal areas along traffic routes) are the best indicators for a heavily altered hydromorphological state.
Table 4. Land use categories for which the additional variance explained at the time they are included in the statistical model
(conditional effect, lambda-A) is ≥ 0.03. Variance explained (%) is given in parentheses. Spatial scales: sf = site floodplain, rf = reach floodplain, cb = catchment buffer, cf = catchment floodplain. Land use categories: wood = woody vegetation, urbhd = urban areas (high density), intag = intensive agriculture (e.g., cropland), traffic = traffic infrastructure. Single spatial scale, for which variance explained by the RDA model is highest is highlighted in bold for each stream type (compare Table 3). stream type spatial scale
A
B
C
D
site buffer
wood (10)
wood (10)
wood (3)
site floodplain
wood (11)
wood (12)
reach buffer
wood (10)
wood (11)
urbhd (7) wood (4) urbhd (9) wood (4) urbhd (8) wood (5)
reach floodplain
wood (11)
wood (13)
urbhd (10) wood (5)
catchment buffer
wood (9)
wood (11)
catchment floodplain
wood (9) urbhd (3)
wood (11)
urbhd (7) wood (4) urbhd (7) wood (3)
other (9) intag (3) traffic (11) other (5) dirtroad (11) traffic (3) wood (3)
sf_wood (11) cf_urbhd (3)
rf_wood (13)
all spatial scales
Fig. 2. Biplots of land use categories and hydromorphological data of the spatial scale, for which the variance explained is highest (compare Table 3). I: stream type B, reach floodplain spatial scale; II: stream type C, reach floodplain spatial scale; III: stream type D, catchment floodplain spatial scale; biplot for stream type A is not given, because it is almost similar to the biplot of stream type B. Abbreviations used: div = diversity, ext. = extensive, indic. = indicating, int. = intensive, sub = substrate, veg. = vegetation.
rf_urbhd (10) cb_wood (6)
intag (6) other (4) other (4) intag (3)
cb_traffic (11) cf_traffic (7) cf_other (3)
Land use and river hydromorphology
Relations between individual hydromorphological parameters and land use categories In stream types A and B (small headwater streams), woody vegetation is the best indicator for a natural hydromorphological state (Fig. 2). Many of the hydromorphological parameters, which indicate a good hydromorphological state (number of riffles/steps, substrate diversity), are positively related to the land use category “woody vegetation” and negatively related to cross-section depth, which is an indicator for entrenchment / degradation. In contrast, parameters representing hydromorphological alterations (e.g.,
69
bank-revetment) are positively related to land use categories with the highest land use pressure (urban / high density, traffic routes, traffic infrastructure). Due to the large sample size, most of the hydromorphological parameters are correlated to at least one of the land use categories (Spearman Rank Correlation, p < 0.01). For example, the presence of riffles or steps (Rs = 0.48), natural channel planform (Rs = 0.46), and width diversity (Rs = 0.45) are positively related to the presence of woody vegetation. In stream type C (mid-sized alluvial streams) crosssection depth is negatively related to the land use category “woody vegetation”, but in contrast to stream
70
J. Kail et al.
types A and B, woody vegetation is not as closely related to parameters indicating a good hydromorphological state (median Rs significantly lower in stream type C, Wilcoxon-Matched Pairs Test, n = 12, p < 0.05). Land use categories with the highest land use pressure (e.g. urban / high density) are equally good indicators for the absence of natural channel features (median Rs not significantly different, Wilcoxon Matched-Pairs Test, n = 12, p = 0.31). Moreover, they are positively related to parameters representing hydromorphological alterations, e.g. the presence of bank revetment is positively related to presence of urban high density areas (Rs = 0.47). In stream type D (large alluvial streams), woody vegetation is the best indicator for a natural hydromorphological state, but not so closely related to these parameters as in stream types A and B (median Rs significantly lower in stream type D Wilcoxon Matched Pairs Test, n = 12, p < 0.01). Moreover, land use categories with the highest land use pressure (urban / high density) are positively related to bank-revetments as in stream type A–C (Rs = 0.55). In contrast to the other stream types, the strongest relation was observed between land use categories with a medium land use pressure (e.g. other / ruderal areas) and some single hydromorphological parameters, namely cross-section depth (Rs = 0.53) and substrate diversity / number of bed features (Rs = –0.7 and –0.62, respectively).
Discussion One of the major objectives of this study was to investigate, if land use at spatial scales smaller than catchments is related to hydromorphology and if there is a spatial scale of special importance. Concerning the two lateral scales investigated (buffer and floodplain) the results indicate that land use on the floodplain is more strongly impacting hydromorphology than the land use directly adjacent to the stream (buffer scale). In contrast, Vondracek et al. (2005) found that riparian vegetation on a spatial scale corresponding to the reach buffer scale of our study was the only land use form related to hydromorphology. One possible reason for these different findings is the different parameters used to describe hydromorphology. Vondracek et al. (2005) used parameters, which are mainly related to the riparian land use (e.g., % cover for fish, % nonvegetated streambank), whereas we used a wider set of parameters, which also describes channel-bed features and the general channel form (cross-section form, planform).
Differences between the buffer and floodplain scale are most pronounced in larger streams for several possible reasons: (1) Some variation of the land use on the buffer and floodplain scale is a prerequisite to detect differences between the relations of land use and hydromorphology on the two spatial scales. The extent of the floodplain is much larger compared to the buffer scale in larger streams. However, there are only small differences between the floodplain and buffer scale in the narrow valleys of small headwater streams (stream types A and B). Therefore, in some stream sections of the small headwater streams, land use on the buffer and floodplain scales did not markedly differ. Moreover, land use categories with a high land use pressure (urban, intensive agriculture) cover only small parts of the floodplain area of the small headwater stream types, but are more common in the floodplains of larger streams. (2) In contrast to small streams, larger streams are considered being a hazard to floodplain land uses remote from the streams and hence, have been altered by man. This supports the hypothesis that the impact of land use on hydromorphology depends on the magnitude of the pressure exerted by the stream on the land use, and therefore, is stream-type specific. Concerning the three longitudinal scales investigated (site, reach, catchment), the results indicate that land use on the reach scale is most strongly impacting hydromorphology compared to the land use on smaller (site) or larger (catchment) scales. However, differences are negligibly small for the small headwater streams (stream types A and B), in which catchment and reach area differ only slightly: For sections near the source of the stream, the upstream part of the stream (catchment scale) is not much longer compared to the fixed number of up- and downstream sections, which are part of the reach scale. As for the lateral spatial scales, some variation of the land use on the different spatial scales is a prerequisite to detect differences between the relations of land use and hydromorphology. Richards et al. (1996) and Vondracek et al. (2005) also report minor differences between scales in explaining channel characteristics. In contrast, Allan et al. (1997) found habitat quality being strongly correlated to agricultural land use of the entire catchment upstream and progressively weaker correlations at smaller spatial scales. This might be due to the fact that Allen et al (1997) did not partial out the effect of natural controls (e.g., geology, slope, discharge) before investigating the relation between land use and hydromorphology, as it has been done by Richards et al. (1996), Vondracek et al. (2005), and in this study. Many natural controls change in the downstream di-
Land use and river hydromorphology
rection, as well as hydromorphology and land use, which both are related to these natural controls. Therefore, the share of the agricultural land use of the entire catchment upstream of a reach may be an indicator for the location of the reach within the river network and hence, the hydromorphological characteristics, without being directly related to hydromorphology. In larger streams (stream type D), land use of the catchment is most strongly impacting hydromorphology compared to the other two longitudinal spatial scales. There is no obvious reason, why and how the land use far away in the upstream part of the catchment might influence local hydromorphology. Most of the parameters used in this study are more dependent on local processes like bed scour or local hydromorphological alterations (e.g., bank-revetment) than on catchment scale processes like the input of fine sediment or changes in the discharge regime (see Table 2). Buffagni et al. (2009, this issue) did use comparable hydromorphological data (modified River Habitat Survey method). Their results indicate that land use on the catchment scale is less strongly related to hydromorphology compared to the reach scale. We suspect that the high influence of the catchment scale on hydromorphology is an artefact for the following reason: most of the land use categories, which influence stream morphology in this stream type, are related to traffic (traffic infrastructure, traffic routes, dirt roads, other / ruderal areas). This seems reasonable, because traffic routes are located in many of these larger valleys and streams have correspondingly been straightened and fixed. Such water engineering works were done on a reach scale, including short sections remote from traffic routes. Possibly, the length of the up- and downstream sections summarized as “reach scale” in this study (Table 1) is too small to consider this effect; i.e. there are many sections, which were altered due to relative vicinity to a traffic route, but the traffic route is not included into the reach scale area. In contrast, on the catchment scale, information on traffic routes far away from the site is included. Therefore, we hypothesize that the traffic land use categories on the reach scale would have an even higher influence on stream morphology, if longer reaches would have been used. To test this hypothesis, a larger number of longitudinal scales should be investigated in future studies. The second objective of this study was to investigate, if the correlation between land use data and hydromorphology is strong enough to assess the hydromorphological state using land use data. Considering the six spatial scales separately, the nine land use
71
categories explain up to 16.0–23.9 % of the variance of the hydromorphological data. The share of variance explained is markedly higher (20–41 %), if land use categories of all spatial scales are considered. This is probably partly due to the fact that a large number of land use categories have a small but significant conditional effect solely because of the large sample size. But the land use categories, which have a high conditional effect do belong to different spatial scales (Table 4). This indicates that considering different spatial scales simultaneously distinctly increases model predictability and our understanding of the factors that determine the local hydromorphological state of a stream. However, even this larger share of variance explained is too low to predict the state of all the hydromorphological parameters. Moreover, compared to data from literature, land use seems to be a weak indicator for hydromorphology in our study streams. Allen et al. (1997) reported that a single land use category (agricultural land use) in the entire catchment upstream of the sites explained 75 % of the variance of the habitat index they used to describe hydromorphology (MDNR habitat index). However, as it has been mentioned above, they did not partial out the effect of natural controls and at least some of the variance of the MDNR habitat index may be explained by natural controls rather than land use. Besides, they did use a single index to describe hydromorphology, whereas we investigated the relation between land use and 17 different hydromorphological parameters. It is evident that land use data can explain the overall hydromorphological state much better than specific aspects of hydromorphology like the number of channel-bed features or the presence of bank-revetment. Applying two general indices derived from the 17 parameters the share of variance explained by the land use data is markedly higher. It is up to 27.8–31.0 % for the single spatial scales and 35–49 %, if all spatial scales, which significantly contribute to the model are considered. This is within the range of the variance of hydromorphological data explained by land use in a study by Vondracek et al. (2005). Using geology as a co-variable in a partial redundancy analysis, the remaining land use parameters explained up to 34 % of the variation in hydromorphology (Vondracek et al. 2005). Despite the influence of natural controls on hydromorphology, past historical land use changes might be the main reason for the relatively low share of variance explained by the land use data. Many of the floodplain areas, which nowadays are forested, were used as grassland, pasture or even cropland in the past and streams were altered (straightened, fixed)
72
J. Kail et al.
to maintain these land uses. According to the spatiotemporal concepts of Knighton (1984), Frissell et al. (1986) or Kern (1994), the time, which is necessary for the stream to adjust to these controls, is dependant on the spatial scale of the channel-features considered. The spatial scale of the channel-features described by the 17 hydromorphological parameters ranges from single pools and riffles to channel-planform and the time to reach a new equilibrium state after changes of natural or human induced controls is 1–1000 years for these channel features. As the relevant time scale increases, so does the potential influence of past conditions (Schumm 1991), and hence, the importance of past land use changes. Because a long time is needed for adjusting some important channel-characteristics like planform, the present hydromorphological state must be considered being the result of a long history of changing land use pressure, which the present land use does only partially reflect. Since historical land-use and land-use changes possibly still significantly influence the present hydromorphological state of streams and rivers, historical data should be additionally used in future investigations.
Conclusion The results of this study indicate that land use from the local to the catchment scale and on both, buffer and floodplain scales, are significantly related to hydromorphology. Differences between the spatial scales are small, but there is some evidence that land use on the reach and on the floodplain scale is of special importance. Considering different spatial scales simultaneously distinctly increases model predictability. However, the share of variance of the hydromorphological data explained by the land use data is too small to really predict specific aspects of hydromorphology from land use data, e.g. single parameters like the number of channel-bed features. Land use data are much better suited to explain the overall hydromorphological state of a stream section. It is possible to qualitatively describe the relation between land use and hydromorphology in detail, i.e. to derive statistically significant relations between single land use categories and specific hydromorphological parameters. The results of this study further stress the need to partial out the effect of natural controls before investigating land use / hydromorphology relations (e.g., stream type specific approach) and to use high-resolution land use data to accurately map linear land use features like roads and railway lines.
Acknowledgements This paper is a result of the EU-funded Integrated Project EuroLimpacs (6th Framework Programme; contract number: GOCECT-2003–505540).
References Allan, J. D., Erickson, D. L. & Fay, J., 1997: The influence of catchment land use on stream integrity across multiple spatial scales. – Freshwat. Biol. 37: 149–161. Briem, E., 2003: Gewässerlandschaften der Bundesrepublik Deutschland – Morphologische Merkmale der Fließgewässer und ihrer Auen. – ATV-DVWK-Arbeitsbericht GB-1, Hennef, Germany, 176 pp. Buffagni, A., Casalegno, C., & Erba. S., 2009: Hydromorphology and land use at different spatial scales: expectations for medium-sized rivers of the Western Italian Alps in a changing climate scenario. – Fundam. Appl. Limnol., Arch. Hydrobiol. 174: 7–25. Feld, C., 2004: Identification and measures of hydromorphological degradation in Central European lowland streams. – Hydrobiologia 516: 69–90. Frissell, C. A., Liss, W. J., Warren, C. E. & Hurley, M. D., 1986: A hierarchical framework for stream habitat classification: viewing streams in a watershed context. – Environ. Manage. 10: 199–214. Illies, J., 1978: Limnofauna Europaea. – Gustav Fischer Verlag, Stuttgart, Germany, 532 pp. Jongman, R. H. G., ter Braak, C. J. F. & van Tongeren, O. F. R., 1995: Data analysis in community and landscape ecology. – Cambridge University Press, Cambridge. Kail, J. & Hering, D., 2005: Using large wood to restore streams in Central Europe: potential use and likely effects. – Landscape Ecol. 20: 755–772. Kern, K., 1994: Grundlagen naturnaher Gewässergestaltung – Geomorphologische Entwicklung von Fließgewässern. – Springer, Berlin, 256 pp. Knighton, D., 1984: Fluvial forms and processes. – Edward Arnold, Baltimore, 218 pp. Knox, J. C., 1987: Historical valley floor sedimentation in the Upper Mississippi valley. – Ann. Assoc. Amer. Geogr. 77: 224–244. Lautenschläger, M. & Kiel, E., 2005: Assessing morphological degradation in running waters using Blackfly communities (Diptera, Simuliidae): Can habitat quality be predicted by land use?. – Limnologica 35: 262–273. Liebault, F., Gomez, B., Page, M., Marden, M., Peacock, D., Richard, D. & Trotter, C. M., 2005: Land-use change, sediment production and channel response in upland regions. – Riv. Res. Appl. 21: 739–756. Podani, J., 2000: Introduction to the exploration of multivariate biological data. – Backhuys Publishers, Leiden. Pottgiesser, T. & Sommerhäuser, M., 2004: Fließgewässertypologie Deutschlands: Die Gewässertypen und ihre Steckbriefe als Beitrag zur Umsetzung der EU-Wasserrahmenrichtlinie. – In: Steinberg, C., Calmano, W., Wilken, R. & Klapper, H. (eds): Handbuch Angewandte Limnologie. – Ecomed, Landsberg, Germany, pp. 3–16. Raven, P. J., Homes, N. T. H., Charrier, P., Dawson, F. H., Naura, M. & Boon, P. J., 2002: Towards a harmonized approach for hydromorpological assessment of rivers in Europe: a qualitative comparison of three survey methods. – Aquat. Conserv.: Mar. Freshwat. Ecosyst. 12: 405–424.
Land use and river hydromorphology Richards, C., Johnson, L. B. & Host, G. E., 1996: Landscapescale influences on stream habitat and biota. – Can. J. Fisher. Aquat. Sci. 53 (Suppl. 1): 295–311. Schumm, S. A., 1991: To interpret the earth: ten ways to be wrong. – Cambridge University Press, Cambridge, 133 pp. Townsend, C. R., Downes, B. J. & Peacock, K. A. C. J., 2004: Scale and the detection of land-use effects on morphology, vegetation and macroinvertebrate communities of grassland streams. – Freshwat. Biol. 49: 448–462.
Submitted: 23 January 2007; accepted: 12 September 2007.
73
Tüxen, R., 1956: Die heutige potentielle natürliche Vegetation als Gegenstand der Vegetationskartierung. – Angewandte Pflanzensoziol. 13: 5–42. Vondracek, B., Blann, K. L., Cox, C. B., Nerbonne, J. F., Mumford, K. F., Nerbonne, B. A., Sovell, L. A. & Zimmermann, J. K. H., 2005: Land use, spatial scale, and stream systems: Lessons from an agricultural region. – Environ. Manage. 36: 775–791.
