- Email: [email protected]
Landscape response functions for biodiversity—assessing the impact of land-use changes at the county level
Landscape and Urban Planning 67 (2004) 157–172
Landscape response functions for biodiversity— assessing the impact of land-use changes at the county level Marc Zebisch a,∗ , Frank Wechsung b , Hartmut Kenneweg a a
Institute for Landscape and Environment Planning, Technical University of Berlin, Berlin, Germany b Potsdam Institute for Climate Impact Research, Potsdam, Germany
Abstract Assessing the impact of land-use change on biodiversity is an important task in the context of global-change scenarios. Here, conceptual considerations, descriptions of a model solution and results of a case study for the regional scale are presented. Land-use was seen as an integrative variable, which depends on natural, as well as on socioeconomic parameters. Economic processes have been externalized by using results of economy driven base-scenarios about land-use change at the county level. Tendencies from these scenarios were extracted, expanded to a set of sub-scenarios, and transformed into land-use maps by a land-use model. These land-use maps were evaluated with respect to biodiversity at the ecosystem level. The results of the evaluation of the single sub-scenarios were summarized to response functions, which describe the sensitivity of landscape attributes toward land-use changes. It is stated that biodiversity is not a generic indicator and can only be assessed after defining the context. Here, only ecosystems with low hemeroby (‘semi-natural’ ecosystems) were considered. The concept of hemeroby, which describes the degree of human disturbance on ecosystems, was used as a qualitative complement to the quantitative concept of biodiversity. Biodiversity was assessed by means of six indicators for three aspects of biodiversity: composition, structure and function. The model was applied in a case study dealing with the impact of extensification of grassland on biodiversity in the county Havelland, west of Berlin. In general, the compositional aspect of biodiversity demonstrated the clearest response. Structural diversity reacted only moderately, but a strong impact of land-use change on connectivity was indicated by an increasing proximity of semi-natural biotopes. The latter was proven by evaluating the connectivity of semi-natural grasslands from the perspective of the white stork (Ciconia ciconia). All response functions showed a high heterogeneity in spatial and functional aspect. At the landscape level, the heterogeneity was hidden behind a supposed moderate reaction of landscape. This underlines the demand for a spatially explicit realization of land-use scenarios and for the consideration of a wide range of scenarios by means of response functions. © 2003 Elsevier Science B.V. All rights reserved. Keywords: Biodiversity; Landscape diversity; Landscape response functions; Land-use model
1. Introduction 1.1. Motivation ∗
Corresponding author. Tel.: +49-30-314-73215; fax: +49-30-314-71226. E-mail address: [email protected] (M. Zebisch). 0169-2046/$20.00 © 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0169-2046(03)00036-7
For sustainable future planning it is essential to investigate possible land-use changes and the impact on ecological functions and processes at the local level.
158
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Therefore scenarios about land-use change and an assessment system for the ecological function under observation are necessary. Economy driven scenarios about land-use composition do exist on higher levels of aggregation, for example, as an output of agricultural sector models (ASMs) (Adams et al., 1996; Heckelei and Britz, 2000). But for an impact assessment of ecological functions, a scenario driven realization of landscape pattern is required (Veldkamp and Lambin, 2001). Various projects have shown the applicability of land-use models for impact assessment (Verburg et al., 1999; Weber et al., 2001). Assessment systems for selected landscape functions and potentials at the regional scale exist (e.g. Marks et al., 1992; Bastian, 1999), but the evaluation is mostly limited to the status quo and the meaning of spatial context or landscape pattern is scarcely considered explicitly. Furthermore, an evaluation of biodiversity is usually not included. Here, biodiversity of ecosystems was the selected target parameter. It is strongly related to the land-use composition and the land-use pattern (Wiens, 1976), and has become a major issue in the context of land-use change investigations. In this study, an approach is presented, which employs economy driven scenarios about land-use change for a spatially explicit realization of landscape pattern and an assessment of biodiversity at the ecosystem level under consideration of site-dependent as well as spatial-context-dependent attributes. 1.2. Modeling landscape Model assumptions have to be drawn in respect to the holistic character of landscape (Naveh, 2000; Antrop, 2000). Concerning the spatial character, elementary (site-dependent) and configurative (spatial-context-dependent) properties have to be considered. Functionally, a landscape can be seen as an open system in a steady state equilibrium. Land-use is an integrative variable in this system, as it is part of the natural as well as of the cultural sphere (Palang et al., 2000). Hence, for modeling land-use change, economic and natural variables have to be considered. However, at the landscape level considerations can be limited to the natural variables, if a two-phase modeling approach is applied. In the presented approach, economy driven scenarios, which have been provided by a spatially implicit simulation of eco-
nomic activities for agriculture and timber production at the county level (RAUMIS, Henrichsmeyer et al., 1996) were used for a spatially explicit simulation of land-use change as a function of natural variables (PAGE, Section 2.3). 2. Methodology 2.1. General model properties The model system consists of two sub-models: (1) the pattern generator (PAGE); and (2) the biodiversity assessment tool (BAT). Land-use change and impact assessment were performed on a grid basis with 50 m × 50 m cell size. Both components were realized in the ArcInfo macro language (AML) using the ARC/INFO GIS environment. Digital maps of biotopes, soil, groundwater and a digital elevation model were used as input. 2.2. Landscape response functions In land-use change investigations, usually scenarios are formulated for well defined future trajectories as unique statements about land-use composition for defined regions and a defined point of time. In our case study these type of point scenarios were provided by RAUMIS and translated into landscape pattern by PAGE. To broaden this static kind of scenarios, we wanted to investigate the impact of land-use change within a single trajectory as a function of the intensity of land-use change. For the latter, we have introduced the concept of response functions (Fig. 1). Here the RAUMIS scenarios were interpreted as possible impact trends. The scenarios were expanded to a range of sub-scenarios which represent land-use changes along such a trend in an occurrence range from weak to strong. For example, if RAUMIS considers a scenario with 30% of extensification of grassland for a county for the year 2010, we considered sub-scenarios with set-aside from 0 to 100% in steps of 10%. These sub-scenarios were transferred into spatial explicit realizations of land-use change by the land-use model PAGE. The result was a set of 11 land-use maps for each RAUMIS scenario, which served as a basis for the biodiversity assessment. The evaluation results (responses) for the single sub-scenarios were related to
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 1. Flowchart of the model concept.
159
160
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
each other and summarized to response functions. It was assumed that response function can reflect the sensitivity of landscape towards land-use change better than realizations of single scenarios. Response functions should be particularly useful for comparing areas, for determining localities with high sensitivity and for finding sensitivity thresholds in the context of biodiversity and land-use change. 2.3. Land-use model PAGE PAGE models land-use changes on the basis of external scenarios about land-use change and an initial land-use pattern. PAGE realizes a transformation from one land-use class into an other as a function of natural variables and their spatial distribution making use of rule-based as well as statistical approaches. For the realizations of single scenarios, the intensity of change has to be defaulted. In this paper we will focus on the approach which is necessary to provide the input data for the response functions analysis. Here each model run produces a set of 11 sub-scenario realizations, which represent different intensities of a given trend between 0 and 100%. PAGE works in a deterministic manner without any random procedure to guarantee comparable realizations for the impact assessment. The transformation process includes four steps: (1) the evaluation of the suitability of each grid cell for the current land-use type; (2) the detection of cells with a low suitability, which have to be changed according to the related scenario; (3) the selection of an appropriate alternative land-use type; and (4) the assignment of the new land-use type (Fig. 2). In the model context, the suitability for agriculture and forestry depends on the yield potential of the corresponding products. The calculation of the yield potential is performed on a cell-by-cell basis for each land-use category with a modified version of the evaluation scheme by Glawion (in Marks et al., 1992). In the original scheme, more than 10 single attributes (soil parameters, water supply parameters, slope and climate parameters) are evaluated in a classification scheme with five possible valuations (from low to high) and combined to one yield potential by a minimum operation (only the lowest factor counts). This scheme was modified by using evaluation functions with possible values from 1 to 99, instead of the five-level scheme, and by considering the three lowest
factors for the final evaluation instead only one. By means of these modifications, a much finer ranking was possible. In addition, the spatial meaning of field plots was considered by averaging the single evaluations for all cells within a field plot. This prevents ‘salt and pepper effects’ by ensuring that land-use change will occur for whole field plots only and not for single pixels. All raster cells were sorted and ranked according to their suitability for the land-use category under consideration. Alternatively to this evaluation procedure, the yield potential can be calculated externally by a crop growth simulation model, which was tested by using the CERES type (Jones and Kiniry, 1986) crop growth module of the eco-hydrological model SWIM (Krysanova et al., 1998) (results not discussed here). In the next step, depending on the scenario, all pixels below a certain suitability threshold are a subject of land-use change. The new land-use type can either be a ‘chosen’ land-use type, for a simple conversion selected by choice (e.g. from grassland to arable land) or a ‘predicted’ land-use type, which is simulated using a statistical model. A ‘predicted’ land-use type represents a land-use type from a set of defined alternatives, which shows the highest likelihood to appear at the given site under the given configuration of the natural variables at this site. Likelihood is calculated by a supervised maximum likelihood classification under the use of existing biotopes as training areas. One possibility to validate such a land-use model, is to simulate the present land-use pattern on the basis of natural attributes and the given proportion of land-use classes and to compare these results with the real land-use pattern. PAGE was validated successfully by simulating the present distribution of grassland and arable land within the agricultural area for various test areas with a per pixel consistency between 67 and 73%. 2.4. Assessing biodiversity As biodiversity is more ‘an end in itself’ than a measurable indicator (Noss, 1990), specifications about the understanding of biodiversity in the model context have to be given. Four aspects have to be defined: (1) the level of hierarchy; (2) the classifications scheme of elements; (3) the attributes used for evaluation; and (4) the scale under observation.
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 2. Flowchart of the land-use model PAGE.
161
162
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
(1) Biodiversity can be observed on various hierarchical levels, among which three are very common in ecological studies: ecosystems, species and genes (Convention on Biological Diversity, Rio, 1993). For this study biodiversity at the ecosystem level was selected as a target parameter for the evaluation, which also can be understood as landscape diversity (Wiens, 1995). Various studies indicate the importance of ecosystems in biodiversity studies. Ecosystem destruction and fragmentation due to land-use change is considered to be one of the major causes for species extinction and biological impoverishment (Hansson et al., 1995; Leemans, 1999; White et al., 1997). (2) Every computation of diversity strongly depends on the classification scheme of the elements. We chose the concept of biotope classes, to assess biodiversity in the context of land-use change. This concept comprises intensive land-use systems as well as semi-natural systems and reflects underlying ecological properties. Furthermore we assumed that biodiversity has not only a quantitative but also a qualitative aspect. In the context of species diversity, for example, we might express our valuation by counting the diversity of red-list species only. In our approach we selected the concept of hemeroby Sukopp (1972) as a qualitative criteria for the diversity of ecosystems. Hemeroby is defined as a measure for the intensity of human disturbance of ecosystems (Table 1) (Blume and Sukopp, 1976). We assumed that this approach is especially appropriate in the context of land-use investigations, since disturbance is strongly related to land use and land-use changes. The suitability of this approach, as a qualitative comple-
ment to the quantitative concept of biodiversity, has already been demonstrated in assessment studies for cultural landscapes (Wrbka et al., 1999). In our approach hemeroby was used for a pre-selection of biotopes. Only biotopes with a hemeroby of ‘metahemerob’ or better (‘semi-natural biotopes’ according to Dierschke, 1984) were considered in the further assessment of biodiversity. (3) According to Noss (1990), biodiversity is configured by three attributes of ecosystems: (1) composition; (2) structure; and (3) function. Composition comprises aspects of identity and variety (e.g. diversity of ecosystems), structure is the physical organization or pattern of a system (e.g. patchiness), function involves ecological processes (e.g. disturbances) (Noss, 1990). For each of these aspects, two indicators were calculated (Table 2). (4) Biodiversity assessment depends on the extent of the area under consideration (Duelli, 1997). For enabling comparative studies across different areas, we calculated all area-sensitive indicators for circular shaped ‘moving windows’ with defined diameter (1 km) and assigned the indicator value to the central pixel of the ‘moving window’. Three scales of aggregation were investigated: the local level (pixel-wise), the municipality level and the district level. The BAT comprises two steps: (1) a hemeroby evaluation for the object selection; and (2) a computation of the six selected indicators, which characterize the three aspects of biodiversity (Fig. 3). (1) First, a ‘local hemeroby’ is determined site-dependent as a function of the biotope type only. Than the ‘local hemeroby’ is modified to a spatial-
Table 1 Hemeroby levels according to Blume and Sukopp (1976) Hemeroby value
Hemeroby level
Example
Processes
1 2
Ahemerob Oligohemerob
Bogs, tundra Forest with species typical for the site
3
Mesohemerob
4 5
␣-Euhemerob -Euhemerob
Forest with species atypical for the site, extensive grassland, heather Intensive grassland, extensive arable land Arable land
6 7
Polyhemerob Metahemerob
City green, pits Streets, buildings
No disturbance Extensive wood cutting, minor changes in matter circles Wood cutting, extensive grazing, rare and small doses of fertilizer Use of fertilizers and biocides, melioration Plowing, planting, major changes in matter circle, heavy use of fertilizers and biocides Strong changes in biocenoses Sealed surface, biocenoses destroyed
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
163
Table 2 Indicators for the three aspects of biodiversity Indicator
Abbreviation
Unit
Aspect of biodiversity
Percentage of land covered by semi-natural biotopes Mean richness of semi-natural biotopes in a circle of 1 km diameter Mean number of patches of semi-natural biotopes in a circle of 1 km diameter Proximity of semi-natural biotopes, represented by the area weighted mean proximity index Mean corrected hemeroby of all biotopes in the total area Percentage of semi-natural biotopes affected by external disturbance
P SNLAND M RICH 1 km M NP 1 km AMN PROX
% % Count –
Composition
M HEM C PDISTURB
– %
Function
context-dependent ‘corrected hemeroby’ considering that disturbances from streets, settlements and intensive land-use system will increases the degree of human influence and herby the hemeroby. Disturbance impact on biotopes is calculated in a cost-distance analysis. Conclusions about the strength and area of reach of disturbances are taken from Trombulak and Frissell (2000), Forman and Deblinger (2000), and Kappler (1997). It is assumed that 100% disturbance impact will raise the biotope-dependent hemeroby of the biotope type by one level. (2) Six indicators are describing the three aspects of biodiversity, composition, structure and function (Table 2). Two indicators represent composition: (1) the percentage of land covered by semi-natural ecosystems (P SNLAND); and (2) the mean richness of semi-natural ecosystem (M RICH 1 km). P SNLAND represents the most direct and simple to interpret indicator: The more land is covered by semi-natural
Structure
biotopes, the higher will be the amount of habitats for indigenous species, the potential for self-regulation, etc. Richness (M RICH 1 km) equals the diversity of semi-natural biotopes in an observation area in relation to the maximum number of semi-natural biotope classes in the whole study are (in our case a maximum of 12 semi-natural biotope classes). A richness of 50%, for example, expresses that 50% of all semi-natural biotope types, which can be found in the whole study area, can also be found in the observation area. Richness is calculated for circular shaped ‘moving widows’ and assigned to the central pixel of the moving window. Thereby, a pixel-wise evaluation is possible. For the case study, 1 km was used as an observation diameter for the ‘moving window’. Richness for larger areas is the average of the richness of each pixel within this area. Richness stands for the basic aspect of biodiversity which is the diversity of elements. Both indicators (P SNLAND, M RICH 1 km) reflect the classical understanding of biodiversity, which focuses mainly on the elements of a system.
Fig. 3. Flowchart of the biodiversity assessment.
164
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Structure is represented by the number of patches in a moving window (M NP 1 km) and the area weighted mean proximity index (AMN PROX). The number of patches (M NP 1 km) is calculated with the same technique like the calculation of richness: the number of patches is determined for a circular shaped moving window and assigned to the central pixel. AMN PROX (Gustafson and Parker, 1992), was calculated using Fragstats 3.0 (McGarigal, 2001). The proximity index quantifies the distances to adjacent patches of the same class in relation to their size in a defined search radius. The index distinguishes sparse distributions of small biotope patches from configurations where biotopes form a complex cluster of larger patches (McGarigal, 2001). The search radius for proximity evaluation was set to 1 km according to the size of the moving window for the diversity observation. Since landscapes are highly characterized by their heterogeneous structure, the structural aspect of biodiversity is a very essential supplement to the compositional aspect. Within the structural aspect the two indicators reflect a more or less antagonistic character. A high M NP 1 km reflects a landscape with a very ‘patchy’ structure (small, fragmented patches). This has positive effects, like a high structural diversity and the occurrence of edge-related biotopes (hedgerows, forest edges and other transition biotopes) which often show a very high species diversity. On the other hand, small patches are often not suitable as habitats for species with a high demand for large and connected areas. In contrast, AMN PROX targets the connectivity of semi-natural biotopes which is high in landscapes with only some large patches. Function is represented by the mean corrected hemeroby (M HEM C) and by the percentage of semi-natural biotopes affected by external disturbances (PDISTURB). PDISTURB is an outcome of the cost-distance analysis of disturbance sources. All semi-natural biotopes, which show more than 25% of disturbance impact are classified as ‘affected by external disturbance’. PDISTURB is calculated as the proportion of affected habitats within all semi-natural habitats in the area under observation. This functional approach considers the fact, that diversity is not only a function of the elements and the structure of a system, but also of the processes within a system. Disturbance from roads, for example, might significantly alter the diversity of birds, even if all other factors
(composition and structure) are identical (Forman and Deblinger, 2000). Besides AMN PROX, all calculations are based on per pixel calculations, which can be summarized for every area of interest. Supplementary to all indices mentioned above, an approach to measure connectivity from the species perspective was tested (see Section 4.5). In the context of the case study, a simple habitat model for the white stork (Ciconia ciconia) was developed. The model is based on descriptions of suitable biotope classes, threshold distances for connectivity and minimal areas for patch complexes. Descriptions were collected from different sources (e.g. Bastian, 1999; Flade, 1994). Connectivity was measured by calculating the percentage of area covered by suitable habitats for the white stork (P H STORK) within the group of all potentially suitable biotope types (humid and fresh grasslands). Additionally, the proportion of the largest habitat complex (largest patch index) within the group of all humid and fresh grasslands (LPI H STORK) was calculated. 3. Case study: extensification of grassland in Havelland county 3.1. Extensification of grassland The objective of this case study was the evaluation of biodiversity response-functions for the impact of grassland extensification and their functional parameterization. We have assumed that extensification of grassland will increase in Havelland with increasing liberation of the European market, due to low margins for agricultural goods, low productive sites and increasing subsidies for ecological measures. Eleven sub-scenarios represented increasing extensification of intensive grassland from 0 to 100%. It was assumed that through extensification, a transformation process from intensive grassland to semi-natural grassland types is possible in the long run. Three types of semi-natural grassland were considered: (1) humid grassland; (2) fresh grassland; and (3) arid grassland. Besides the fact that biodiversity was investigated at the ecosystem level, which targets practically an evaluation of landscape diversity, results may also be interesting for reflections about biodiversity on species levels, since various studies indicate that
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
165
This map is the result of a state-wide 1:25,000 mapping of biotopes from CIR aerial photographs, which includes more than 100 biotope types. We have aggregated the biotope types to 25 biotope classes well suited to represent land-use change and biodiversity attributes. A map of soil quality in a 1:200,000 scale with seven classes (MUNR), a map of groundwater distance originally in a 200 m × 200 m grid resolution (MUNR) and a digital elevation model (DEM) with 50 m × 50 m resolution (Brandenburger Landesamt für Vermessung) served as input maps for the pattern generation. All maps were rasterized to the 50 m × 50 m resolution of the DEM to obtain a contingent data base. Some tests to improve groundwater data by means of the DEM have shown a significant sensitivity of the results to data quality. Hence, a consistent and proved data basis is crucial for land-use modeling.
4. Results 4.1. Land-use change
Fig. 4. Study area of Havelland.
extensification has also a significant effect on species diversity (e.g. Mander et al., 1999). 3.2. Study area The study area covers about 1700 km2 in a typical lowland landscape west of Berlin (Fig. 4). The area is characterized by riparian floodplains with boggy or loamy soils covered with grassland, plains with arable land or deciduous forest on medium to poor soils and pine forests on sandy hills. A 20.5% of the total land are classified as grassland under intensive use. Semi-natural grassland types cover about 3% of the area. 3.3. Data sets The base map for land use and ecosystem representation was a biotope map (Brandenburger Ministerium für Umwelt, Naturschutz und Raumordnung, MUNR).
The initial situation showed only small patches of semi-natural grasslands. Humid and fresh grassland could be found along the river Havel in the western part and along some smaller water bodies (Fig. 5). After the first 30% of extensification, transformation appeared mainly in an area in the north-west of the study area. This area is characterized by very low groundwater distances and loamy or boggy soils. Here intensive grassland was mostly substituted by humid grasslands. Subsequently, intensive grassland along rivers and in wet lowland areas was substituted by semi-natural grassland of humid or fresh type. Intensive grassland on fertile soils with moderate groundwater level was the last to be transformed. Here predominately fresh grassland appeared. Additional arid grassland was rare and emerged mostly in the neighborhood of coniferous forests. To illustrate the effects of grassland extensification on landscape composition and pattern, the proportions of grassland types and two simple landscape metrics were plotted as functions of the intensity of extensification (Fig. 6). While the linear decline of the proportion of intensive grassland was the given model assumption, the shape of the graphs for the semi-natural types is a model result
166
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 5. Land-use change sequence for the scenario: extensification of grassland.
(Fig. 6a). It shows the early emergence of humid grasslands, the later emergence of fresh grassland and the minor emergence of arid grasslands. From the relation of mean patch size (MPS) (Fig. 6b) and the number of patches for each type (NP) (Fig. 6c), conclusions about fragmentation and compactness can be drawn. Humid grassland shows no increase in patch number but a strong increase in patch size. This indicates a high level of compactness. In contrary, fresh grassland shows an increasing number of patches until a maximum at 30% of extensification with no increase of MPS, followed by a decrease of the number of patches and a rapid increase in the MPS. This indicates the introduction of small, fragmented patches at the earlier levels of extensification, followed by a ‘clumping process’ of patches after a ‘percolation threshold’ (Turner et al., 2001) at 30% extensification.
4.2. Biodiversity at the pixel level Fig. 7 displays the richness for an extensification level of 50% at the pixel level. A high level of spatial heterogeneity is observable. Areas with high richness in the west (along river Havel) can clearly be distinguished from areas with low richness in the eastern part. The range reaches from 0 to 91% richness. 4.3. Biodiversity at the municipality level The first level of aggregation is the municipality level. The difference map for evaluating the gain in percentage of area covered by semi-natural biotopes (P SNLAND) between the initial situation and 100% extensification (Fig. 8), shows a high spatial heterogeneity. Some areas profited by extensification with a gain of more than 50%, others took no profit from
Fig. 6. Change of land-use composition and land-use pattern within the grassland area: (a) percentage of total grassland covered by single grassland types; (b) MPS for semi-natural grassland types; (c) total number of patches for semi-natural grassland types.
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
167
Fig. 7. Pixel-based evaluation of biotope richness.
extensification. Additionally, the response functions for each municipality can be drawn (Fig. 9). Fig. 9a displays P SNLAND curves for a subset of municipalities with significant gains in P SNLAND. The diversity of the shapes of the graphs represents clearly the functional heterogeneity of landscape response between different municipalities. While some municipalities realized gains in P SNLAND already at 20% extensification, others only benefited, if more than 80% of the grassland was extensified. A lot of the graphs show a threshold behavior. Two municipalities
with quite a similar shape of the graph and a high gain in semi-natural area were selected for further evaluation. For these municipalities, the total set of indicators is presented in Table 3. Some selected indicators were plotted as response functions (Fig. 9b and c). Indicators, which represent the compositional aspect of biodiversity showed the clearest response. Although both municipalities comprised a similar percentage of semi-natural biotopes (P SNLAND) after 100% extensification, richness (M RICH 1 km) and structural diversity (M NP 1 km) in Witzke were
Fig. 8. Difference map for the percentage of semi-natural biotopes at the municipality level.
168
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 9. Response functions at the municipality level: (a) percentage of area covered by semi-natural biotopes for a subset of municipalities with a significant gain in P SNLAND; (b) percentage of area covered by semi-natural biotopes for two selected municipalities; (c) mean richness of semi-natural biotopes in a circle with 1 km diameter for two selected municipalities.
about twice as high as in Kienberg. This indicates heterogeneities in the correlation of the single indicators. Very remarkable was the decrease of the proportion of disturbance (PDISTURB) among the semi-natural biotopes. Especially in Kienberg, the rate dropped from 57% down to only 12%. This can be explained as an effect of increasing patch size, which can buffer disturbances much more effectively. 4.4. Biodiversity at the landscape level At the highest level of aggregation indices were calculated for the total district. Response functions on the landscape level displayed only moderate sensitivity (Fig. 10, Table 4). Again, the compositional
indicators showed the clearest response (Fig. 10a and b). An extraordinary response function was displayed by the proximity (AMN PROXIM). Proximity, as a structural measure, raised by more than seven times and showed a nonlinear behavior (Fig. 10c). The high rise at about 30% of extensification can be explained by the beginning of a ‘clumping’ process of semi-natural patches when enough new patches have been introduced (see Section 4.1). After that, proximity decreased due to the introduction of new, smaller, patches in areas, where only little percentage of the land was constituted by semi-natural ecosystems. These new patches were isolated until adjacent intensive grassland also was extensified. While the other indicators followed more or less linearly the increasing
Table 3 Indicators for biodiversity at the municipality level Indicator
Extensification of grassland (%)
Total increase
Increase (%)
Municipality
0
50
100
P SNLAND
21.0 6.0
31.0 17.0
68.0 59.0
47.0 53.0
223.8 883.3
Witzke Kienberg
M RICH 1 km
31.0 14.3
36.4 18.8
37.5 23.3
6.5 9.0
21.0 62.8
Witzke Kienberg
M NP 1 km
7.0 3.1
8.3 3.7
8.1 4.3
1.0 1.2
14.9 39.8
Witzke Kienberg
M HEM C
4.0 4.3
3.9 4.2
3.5 3.8
−0.5 −0.6
−12.4 −12.8
Witzke Kienberg
PDISTURB
19.0 56.7
18.1 27.9
8.1 12.0
−11.0 −44.6
−57.5 −78.8
Witzke Kienberg
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
169
Fig. 10. Response functions at the landscape level: (a) percentage of area covered by semi-natural biotopes in the county Havelland; (b) mean richness of semi-natural biotopes in a circle with 1 km diameter in the county Havelland; (c) area weighted mean proximity index for the class of semi-natural biotopes in the county Havelland.
Table 4 Indicators for biodiversity at the landscape level Indicator
P SNLAND M RICH 1 km M NP 1 km AMN PROX M HEM C PDISTURB
Extensification of grassland (%)
Total increase
0
50
100
31.3 17.1 5.2 689.6 4.1 14.8
39.2 26.0 5.6 3388.9 4.0 14.3
48.0 27.8 5.8 5972.3 4.0 13.3
16.7 10.8 0.6 5282.7 −0.2 −1.5
Increase (%)
53.5 62.9 11.6 766.0 −4.4 −9.8
Fig. 11. Connectivity within grassland area from a functional point of view: (a) area weighted mean proximity index for semi-natural grassland types; (b) percentage of area within humid and fresh grassland suitable as stork habitat (P H STROK) and the percentage of area within humid and fresh grassland, which is covered by the largest habitat complex (LPI H STORK).
170
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
percentage of semi-natural land, the proximity has shown to be sensitive to the spatial heterogeneities, which can be observed in the landscape pattern. The two functional parameters showed a relatively weak response (Table 4). 4.5. Habitat approach to measure connectivity within the grassland area Within the grassland area the proximity index of all semi-natural grassland types was calculated (Fig. 11a). For humid grassland, proximity increased rapidly and stayed at the high level. Proximity for fresh grassland responded very similar to the proximity of semi-natural biotopes on the landscape level (Fig. 10c). The results for the proximity index indicated an increase of connectivity among semi-natural grassland types and an increase of the network character of semi-natural biotopes as a whole. But this measure is difficult to compare, even in a relative approach. To measure connectivity from a species point of view, we identified the percentage of area suitable for the white stork within the area of humid and fresh biotope types (P H STORK) (Fig. 11b). This ‘functional’ connectivity increased from 69 to 86% after 100% extensification. Hence, 86% of all fresh and wet grassland pixels were potentially suitable as stork habitats. This equals a total gain in area suitable for the stork from 4046 to 37,303 ha. LPI H STORK (the percentage of area covered by the largest habitat complex for the stork) increased rapidly at 50% extensification which indicates the clumping of all stork habitats to one big complex at this level of extensification.
5. Discussion First of all, the results of the investigations have shown a high level of heterogeneity of landscape responses in spatial and functional aspect. Heterogeneity was visible in an irregular spatial distribution of the evaluation results for the three aspects of biodiversity, in an irregular behavior of landscape responses of single spatial units through the various levels of extensification and in an irregular correlation between the indicators for three aspects of biodiversity. Heterogeneity increased when scaling down. At the
landscape level, heterogeneity was ‘hidden’ behind a supposed ‘moderate’ landscape response. The impact of land-use change on biodiversity strongly depended on the intensity of change, the configuration of the land-use pattern and the spatial distribution of the natural variables. Among the three aspects of biodiversity (composition, structure, function), the compositional aspect showed the clearest reaction. Particularly the percentage of area covered by semi-natural biotopes increased considerably at all levels of scale. Compositional diversity, represented by the richness of semi-natural ecosystems, increased less significantly, but showed remarkable gain in areas with an initial low percentage of semi-natural ecosystems. Furthermore, a high spatial variation of richness was observable. The structural attributes showed contradictory responses. While structural diversity expressed by the patchiness has shown a relatively weaker response, proximity was very sensitive. Structural diversity increased as a result of a partitioning process of large patches of intensive grassland into smaller patches of semi-natural grassland types. At this point, data resolution plays an important role. Semi-natural grasslands can show a very ‘patchy’ pattern at small scales in reality, e.g. due to heterogeneities in groundwater distances. This can only partly be reflected by the groundwater database with its 200 m × 200 m resolution. Thus, the interpretation of the structural diversity should not be overestimated. The structural aspect ‘proximity’ is very difficult to evaluate, since it has no defined unit and since it is not clear, if proximity response is really proportional to the impact. Results cannot be compared for an objective evaluation. Nonetheless, the results indicate an important role of isolation and connectivity within the response of landscape to land-use change. This aspect should be evaluated in an impact assessment, but must be seen under defined functional aspects (see the example of connectivity from a stork’s point of view). The functional aspect of biodiversity was assessed by the mean hemeroby and the proportion of disturbed biotopes. Hemeroby itself has shown a very weak response, which can be explained by the fact that hemeroby is a classified parameter, not a generic measure. This makes hemeroby less suitable for building mean calculations. In future studies a more appropriate approach should be introduced which
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
might, for example, include a ranking system instead of arithmetical calculations. Hemeroby has shown to be suitable for the qualitative pre-selection of the objects under observation. The percentage of disturbed biotopes showed a moderate response on the landscape level, but very significant results at the municipality level. The moving window approach for calculating ecosystem richness and structural diversity was very promising, since it enables a comparison across areas and scales. For future investigations, the diameter of the observation circle should be varied, to consider biodiversity aspects at various spatial scales. This might also contribute to further understanding of scale dependencies of biodiversity. In contrast to the island biology related approach of the pure patch-matrix concept (Forman and Godron, 1986), the moving window approach takes into account also transition phenomena between ecosystems, like considered in the ecotone approach (Delcourt and Delcourt, 1992). 6. Conclusion The presented case study has demonstrated the sensitivity of biodiversity attributes at the ecosystem level towards land-use change at the regional scale. The most sensitive attributes were the compositional ones, but also the structural and the functional attributes, particularly the connectivity, were significantly affected. Response functions of biodiversity attributes have shown to be particularly useful to detect thresholds, functional irregularities (one attribute changes more than another) and spatial heterogeneity in the impact of land-use shifts on biodiversity. This might be beneficial for determining sensitive regions and for assessing chances and threads for biodiversity at the ecosystem level under supposed changes in land-use composition. Acknowledgements This project is financed by the German Volkswagen foundation. References Adams, D., Alig, R., Callaway, J., McCarl, B., Winnett, S., 1996. The forest and agricultural sector optimization model (FASOM):
171
model structure and policy simulations. USDA Forest Service Research Paper PNW-RP-495. Portland, Oregon, p. 60. Antrop, M., 2000. Background concepts for integrated landscape analysis. Agric. Ecosyst. Environ. 77, 17–28. Bastian, O. (Ed.), 1999. Analyse und ökologische Bewertung der Landschaft. Springer, Berlin. Blume, H.-P., Sukopp, H., 1976. Ökologische Bedeutung anthropogener Bodenveränderung. Schriftenreihe für Vegetationskunde 10, 75–89. Delcourt, P.A., Delcourt, H.R., 1992. Ecotone dynamics in space and time. In: Hansen, A.J., di Castri, F. (Eds.), Landscape Boundaries: Consequences for Biotic Diversity and Ecological Flows. Springer, New York. Dierschke, H., 1984. Natürlichkeitsgrade von Pflanzengesellschaften unter besonderer Berücksichtigung der Vegetation Mitteleuropas. Phytoconeologia 12, 173–184. Duelli, P., 1997. Biodiversity evaluation in agricultural landscapes: an approach at two different scales. Agric. Ecosyst. Environ. 62, 81–91. Flade, M., 1994. Die Brutvogelgemeinschaften Mittel und Norddeutschlands. IHW-Verlag, Eching. Forman, R.T.T., Godron, M., 1986. Landscape Ecology. Wiley, New York. Forman, R.T.T., Deblinger, R.D., 2000. The ecological road-effect zone of a Massachusetts (USA) suburban highway. Conserv. Biol. 14, 36–46. Gustafson, E.J., Parker, G.R., 1992. Relationships between landcover proportion and indices of landscape spatial pattern. Landscape Ecol. 7, 101–110. Hansson, L., Fahrig, L., Merriam, G. (Eds.), 1995. Mosaic Landscapes and Ecological Processes. Chapman & Hall, London. Heckelei, T., Britz, W., 2000. Concept and explorative application of an EU-wide, regional agricultural sector model (CAPRI-project). In: Proceedings of the Paper Presented at the 65th EAAE-Seminar in Bonn. Henrichsmeyer, W., Cypris, C., Löhe, W., Meudt, M., 1996. Entwicklung des gesamtdeutschen Agrarsektormodells RAUMIS96 am Lehrstuhl für Volkswirtschaftslehre, Agrarpolitik und Landwirtschaftliches Informationswesen der Universität Bonn. Agrarwirtschaft 4–5, 213–215. Jones, C.A., Kiniry, J.R. (Eds.), 1986. CERES MAIZE. A Simulation Model of Maize Growth and Development. Texas A&M University Press, College Station, TX. Kappler, O., 1997. Some methods for determination and the evaluation of non fragmented and minimal disturbed landscape areas. In: Proceedings of the Paper Presented at the 10th European Colloquium on Theoretical and Quantitative Geography in Rostock, Germany, 6–10 September 1997. Krysanova, V., Mueller-Wohlfeil, D.I., Becker, A., 1998. Development and test of a spatially distributed hydrological/ water quality model for mesoscale watersheds. Ecol. Modeling 106, 261–289. Leemans, R., 1999. Modeling for species and habitats: new opportunities for problem solving. Sci. Total Environ. 240, 51– 73.
172
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Mander, Ü., Mikk, M., Külvik, M., 1999. Ecological and low intensity agriculture as contributors to landscape and biological diversity. Landscape Urban Planning 46, 169–177. Marks, R., Müller, M.J., Leser, H., Klink, H.-J. (Eds.), 1992. Anleitung zur Bewertung des Leistungsvermögens des Landschaftshaushaltes, vol. 229. Selbstverlag des Zentralausschusses für deutsche Landeskunde, Trier. McGarigal, K., 2001. User manual FRAGSTATS 3.0. Naveh, Z., 2000. What is holistic landscape ecology? A conceptual introduction. Landscape Urban Planning 50, 7–26. Noss, R.F., 1990. Indicators of monitoring biodiversity: a hierarchical approach. Conserv. Biol. 4, 355–363. Palang, H., Mander, U., Naveh, Z., 2000. Holistic landscape ecology in action. Landscape Urban Planning 50, 1–6. Sukopp, H., 1972. Wandel von Flora und Vegetation in Mitteleuropa unter dem Einfluß des Menschen. Berichte über die Landwirtschaft 50, 112–139. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conserv. Biol. 14, 18–30. Turner, M.G., Gardener, R.H., O’Neil, R.V., 2001. Landscape Ecology in Theory and Practice. Springer, Berlin. Veldkamp, A., Lambin, E.F., 2001. Predicting land-use change. Agric. Ecosyst. Environ. 85, 1–6. Verburg, P.H., de Koning, G.H.J., Kok, K., Veldkamp, A., Bouma, J., 1999. A spatial explicit allocation procedure for modeling the pattern of land use change upon actual land use. Ecol. Modeling 116, 45–61. Weber, A., Fohrer, N., Möller, D., 2001. Long-term land-use changes in a mesoscale watershed due to socio-economic factors—effects on landscape structures and functions. Ecol. Modeling 140, 125–140. White, D., Minotti, P.G., Barczak, M.J., Sifneos, J.C., Freemark, K.E., Santelmann, M.V., Steinitz, C.F., Kiester, A.R., Preston, E.M., 1997. Assessing risk of biodiversity from future landscape change. Conserv. Biol. 11, 349–360. Wiens, J.A., 1976. Population responses to patchy environments. Annu. Rev. Ecol. Syst. 7, 81–120. Wiens, J.A., 1995. Landscape mosaics and ecological theory. In: Hansson, L., Fahrig, L., Merriam, G. (Eds.), Mosaic Landscapes and Ecological Processes. Chapman & Hall, London, pp. 1–26.
Wrbka, T., Szerencsits, E., Moser, D., Reiter, K., 1999. Biodiversity patterns in cultivated landscapes: experiences and first results from a nationwide Austrian survey. In: Maudsley, M., Marshall, J. (Eds.), Heterogeneity in Landscape Ecology. Proceedings of the 1999 Annual IALE (UK) Conference, Bristol. Marc Zebisch has a Master’s degree in geoecology (University of Potsdam, Germany, 1999). Currently he is a PhD student at the Institute for Landscape and Environment Planning, Technical University of Berlin (TUB), Germany. He specializes in land-use investigation by means of remote sensing and GIS techniques with a particular interest in spatial explicit modeling of processes in the landscape. He was a co-worker in projects about landscape dynamics in Germany and Mongolia. The investigations presented in this article will be subject of his PhD thesis. His work is embedded in the project mesoscale simulation study assessing the consequences of global change (MESSAGE), which is a co-project of the TUB, the Potsdam Institute of Climate Impact Research and the Society for Agricultural Politics and Sociology (FAA, Bonn). Frank Wechsung is a Senior Research Scientist at the Potsdam Institute of Climate Impact Research, Department of Global Change and Natural Systems (Germany). His principal fields of interest are modeling climate impact on crops at regional and global scale and the response of crops to CO2 enrichment. He holds a PhD in agriculture and operation research. Currently he is working as the coordinator of the MESSAGE project and other projects at PIK. Hartmut Kenneweg is the Managing Director of the Institute of Landscape and Environment Planning at the Technical University of Berlin (TUB), Germany. He took his PhD in forestry at the University of Freiburg. From 1980 to 1985 he has been Professor for forest inventory and forest planning at the University of Göttingen. Since 1985 he has been Professor for Landscape Planning at the TUB. He has been a member in several advisory boards and steering committees (e.g. 1988–1997: ‘regional applications of satellite remote sensing’, DARA; 1990–1993: ‘large area operational experiment for forest damage monitoring in Europe’, UNEP/ECE). He has special interests in remote sensing of forestry and land-use dynamics.
Landscape response functions for biodiversity— assessing the impact of land-use changes at the county level Marc Zebisch a,∗ , Frank Wechsung b , Hartmut Kenneweg a a
Institute for Landscape and Environment Planning, Technical University of Berlin, Berlin, Germany b Potsdam Institute for Climate Impact Research, Potsdam, Germany
Abstract Assessing the impact of land-use change on biodiversity is an important task in the context of global-change scenarios. Here, conceptual considerations, descriptions of a model solution and results of a case study for the regional scale are presented. Land-use was seen as an integrative variable, which depends on natural, as well as on socioeconomic parameters. Economic processes have been externalized by using results of economy driven base-scenarios about land-use change at the county level. Tendencies from these scenarios were extracted, expanded to a set of sub-scenarios, and transformed into land-use maps by a land-use model. These land-use maps were evaluated with respect to biodiversity at the ecosystem level. The results of the evaluation of the single sub-scenarios were summarized to response functions, which describe the sensitivity of landscape attributes toward land-use changes. It is stated that biodiversity is not a generic indicator and can only be assessed after defining the context. Here, only ecosystems with low hemeroby (‘semi-natural’ ecosystems) were considered. The concept of hemeroby, which describes the degree of human disturbance on ecosystems, was used as a qualitative complement to the quantitative concept of biodiversity. Biodiversity was assessed by means of six indicators for three aspects of biodiversity: composition, structure and function. The model was applied in a case study dealing with the impact of extensification of grassland on biodiversity in the county Havelland, west of Berlin. In general, the compositional aspect of biodiversity demonstrated the clearest response. Structural diversity reacted only moderately, but a strong impact of land-use change on connectivity was indicated by an increasing proximity of semi-natural biotopes. The latter was proven by evaluating the connectivity of semi-natural grasslands from the perspective of the white stork (Ciconia ciconia). All response functions showed a high heterogeneity in spatial and functional aspect. At the landscape level, the heterogeneity was hidden behind a supposed moderate reaction of landscape. This underlines the demand for a spatially explicit realization of land-use scenarios and for the consideration of a wide range of scenarios by means of response functions. © 2003 Elsevier Science B.V. All rights reserved. Keywords: Biodiversity; Landscape diversity; Landscape response functions; Land-use model
1. Introduction 1.1. Motivation ∗
Corresponding author. Tel.: +49-30-314-73215; fax: +49-30-314-71226. E-mail address: [email protected] (M. Zebisch). 0169-2046/$20.00 © 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0169-2046(03)00036-7
For sustainable future planning it is essential to investigate possible land-use changes and the impact on ecological functions and processes at the local level.
158
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Therefore scenarios about land-use change and an assessment system for the ecological function under observation are necessary. Economy driven scenarios about land-use composition do exist on higher levels of aggregation, for example, as an output of agricultural sector models (ASMs) (Adams et al., 1996; Heckelei and Britz, 2000). But for an impact assessment of ecological functions, a scenario driven realization of landscape pattern is required (Veldkamp and Lambin, 2001). Various projects have shown the applicability of land-use models for impact assessment (Verburg et al., 1999; Weber et al., 2001). Assessment systems for selected landscape functions and potentials at the regional scale exist (e.g. Marks et al., 1992; Bastian, 1999), but the evaluation is mostly limited to the status quo and the meaning of spatial context or landscape pattern is scarcely considered explicitly. Furthermore, an evaluation of biodiversity is usually not included. Here, biodiversity of ecosystems was the selected target parameter. It is strongly related to the land-use composition and the land-use pattern (Wiens, 1976), and has become a major issue in the context of land-use change investigations. In this study, an approach is presented, which employs economy driven scenarios about land-use change for a spatially explicit realization of landscape pattern and an assessment of biodiversity at the ecosystem level under consideration of site-dependent as well as spatial-context-dependent attributes. 1.2. Modeling landscape Model assumptions have to be drawn in respect to the holistic character of landscape (Naveh, 2000; Antrop, 2000). Concerning the spatial character, elementary (site-dependent) and configurative (spatial-context-dependent) properties have to be considered. Functionally, a landscape can be seen as an open system in a steady state equilibrium. Land-use is an integrative variable in this system, as it is part of the natural as well as of the cultural sphere (Palang et al., 2000). Hence, for modeling land-use change, economic and natural variables have to be considered. However, at the landscape level considerations can be limited to the natural variables, if a two-phase modeling approach is applied. In the presented approach, economy driven scenarios, which have been provided by a spatially implicit simulation of eco-
nomic activities for agriculture and timber production at the county level (RAUMIS, Henrichsmeyer et al., 1996) were used for a spatially explicit simulation of land-use change as a function of natural variables (PAGE, Section 2.3). 2. Methodology 2.1. General model properties The model system consists of two sub-models: (1) the pattern generator (PAGE); and (2) the biodiversity assessment tool (BAT). Land-use change and impact assessment were performed on a grid basis with 50 m × 50 m cell size. Both components were realized in the ArcInfo macro language (AML) using the ARC/INFO GIS environment. Digital maps of biotopes, soil, groundwater and a digital elevation model were used as input. 2.2. Landscape response functions In land-use change investigations, usually scenarios are formulated for well defined future trajectories as unique statements about land-use composition for defined regions and a defined point of time. In our case study these type of point scenarios were provided by RAUMIS and translated into landscape pattern by PAGE. To broaden this static kind of scenarios, we wanted to investigate the impact of land-use change within a single trajectory as a function of the intensity of land-use change. For the latter, we have introduced the concept of response functions (Fig. 1). Here the RAUMIS scenarios were interpreted as possible impact trends. The scenarios were expanded to a range of sub-scenarios which represent land-use changes along such a trend in an occurrence range from weak to strong. For example, if RAUMIS considers a scenario with 30% of extensification of grassland for a county for the year 2010, we considered sub-scenarios with set-aside from 0 to 100% in steps of 10%. These sub-scenarios were transferred into spatial explicit realizations of land-use change by the land-use model PAGE. The result was a set of 11 land-use maps for each RAUMIS scenario, which served as a basis for the biodiversity assessment. The evaluation results (responses) for the single sub-scenarios were related to
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 1. Flowchart of the model concept.
159
160
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
each other and summarized to response functions. It was assumed that response function can reflect the sensitivity of landscape towards land-use change better than realizations of single scenarios. Response functions should be particularly useful for comparing areas, for determining localities with high sensitivity and for finding sensitivity thresholds in the context of biodiversity and land-use change. 2.3. Land-use model PAGE PAGE models land-use changes on the basis of external scenarios about land-use change and an initial land-use pattern. PAGE realizes a transformation from one land-use class into an other as a function of natural variables and their spatial distribution making use of rule-based as well as statistical approaches. For the realizations of single scenarios, the intensity of change has to be defaulted. In this paper we will focus on the approach which is necessary to provide the input data for the response functions analysis. Here each model run produces a set of 11 sub-scenario realizations, which represent different intensities of a given trend between 0 and 100%. PAGE works in a deterministic manner without any random procedure to guarantee comparable realizations for the impact assessment. The transformation process includes four steps: (1) the evaluation of the suitability of each grid cell for the current land-use type; (2) the detection of cells with a low suitability, which have to be changed according to the related scenario; (3) the selection of an appropriate alternative land-use type; and (4) the assignment of the new land-use type (Fig. 2). In the model context, the suitability for agriculture and forestry depends on the yield potential of the corresponding products. The calculation of the yield potential is performed on a cell-by-cell basis for each land-use category with a modified version of the evaluation scheme by Glawion (in Marks et al., 1992). In the original scheme, more than 10 single attributes (soil parameters, water supply parameters, slope and climate parameters) are evaluated in a classification scheme with five possible valuations (from low to high) and combined to one yield potential by a minimum operation (only the lowest factor counts). This scheme was modified by using evaluation functions with possible values from 1 to 99, instead of the five-level scheme, and by considering the three lowest
factors for the final evaluation instead only one. By means of these modifications, a much finer ranking was possible. In addition, the spatial meaning of field plots was considered by averaging the single evaluations for all cells within a field plot. This prevents ‘salt and pepper effects’ by ensuring that land-use change will occur for whole field plots only and not for single pixels. All raster cells were sorted and ranked according to their suitability for the land-use category under consideration. Alternatively to this evaluation procedure, the yield potential can be calculated externally by a crop growth simulation model, which was tested by using the CERES type (Jones and Kiniry, 1986) crop growth module of the eco-hydrological model SWIM (Krysanova et al., 1998) (results not discussed here). In the next step, depending on the scenario, all pixels below a certain suitability threshold are a subject of land-use change. The new land-use type can either be a ‘chosen’ land-use type, for a simple conversion selected by choice (e.g. from grassland to arable land) or a ‘predicted’ land-use type, which is simulated using a statistical model. A ‘predicted’ land-use type represents a land-use type from a set of defined alternatives, which shows the highest likelihood to appear at the given site under the given configuration of the natural variables at this site. Likelihood is calculated by a supervised maximum likelihood classification under the use of existing biotopes as training areas. One possibility to validate such a land-use model, is to simulate the present land-use pattern on the basis of natural attributes and the given proportion of land-use classes and to compare these results with the real land-use pattern. PAGE was validated successfully by simulating the present distribution of grassland and arable land within the agricultural area for various test areas with a per pixel consistency between 67 and 73%. 2.4. Assessing biodiversity As biodiversity is more ‘an end in itself’ than a measurable indicator (Noss, 1990), specifications about the understanding of biodiversity in the model context have to be given. Four aspects have to be defined: (1) the level of hierarchy; (2) the classifications scheme of elements; (3) the attributes used for evaluation; and (4) the scale under observation.
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 2. Flowchart of the land-use model PAGE.
161
162
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
(1) Biodiversity can be observed on various hierarchical levels, among which three are very common in ecological studies: ecosystems, species and genes (Convention on Biological Diversity, Rio, 1993). For this study biodiversity at the ecosystem level was selected as a target parameter for the evaluation, which also can be understood as landscape diversity (Wiens, 1995). Various studies indicate the importance of ecosystems in biodiversity studies. Ecosystem destruction and fragmentation due to land-use change is considered to be one of the major causes for species extinction and biological impoverishment (Hansson et al., 1995; Leemans, 1999; White et al., 1997). (2) Every computation of diversity strongly depends on the classification scheme of the elements. We chose the concept of biotope classes, to assess biodiversity in the context of land-use change. This concept comprises intensive land-use systems as well as semi-natural systems and reflects underlying ecological properties. Furthermore we assumed that biodiversity has not only a quantitative but also a qualitative aspect. In the context of species diversity, for example, we might express our valuation by counting the diversity of red-list species only. In our approach we selected the concept of hemeroby Sukopp (1972) as a qualitative criteria for the diversity of ecosystems. Hemeroby is defined as a measure for the intensity of human disturbance of ecosystems (Table 1) (Blume and Sukopp, 1976). We assumed that this approach is especially appropriate in the context of land-use investigations, since disturbance is strongly related to land use and land-use changes. The suitability of this approach, as a qualitative comple-
ment to the quantitative concept of biodiversity, has already been demonstrated in assessment studies for cultural landscapes (Wrbka et al., 1999). In our approach hemeroby was used for a pre-selection of biotopes. Only biotopes with a hemeroby of ‘metahemerob’ or better (‘semi-natural biotopes’ according to Dierschke, 1984) were considered in the further assessment of biodiversity. (3) According to Noss (1990), biodiversity is configured by three attributes of ecosystems: (1) composition; (2) structure; and (3) function. Composition comprises aspects of identity and variety (e.g. diversity of ecosystems), structure is the physical organization or pattern of a system (e.g. patchiness), function involves ecological processes (e.g. disturbances) (Noss, 1990). For each of these aspects, two indicators were calculated (Table 2). (4) Biodiversity assessment depends on the extent of the area under consideration (Duelli, 1997). For enabling comparative studies across different areas, we calculated all area-sensitive indicators for circular shaped ‘moving windows’ with defined diameter (1 km) and assigned the indicator value to the central pixel of the ‘moving window’. Three scales of aggregation were investigated: the local level (pixel-wise), the municipality level and the district level. The BAT comprises two steps: (1) a hemeroby evaluation for the object selection; and (2) a computation of the six selected indicators, which characterize the three aspects of biodiversity (Fig. 3). (1) First, a ‘local hemeroby’ is determined site-dependent as a function of the biotope type only. Than the ‘local hemeroby’ is modified to a spatial-
Table 1 Hemeroby levels according to Blume and Sukopp (1976) Hemeroby value
Hemeroby level
Example
Processes
1 2
Ahemerob Oligohemerob
Bogs, tundra Forest with species typical for the site
3
Mesohemerob
4 5
␣-Euhemerob -Euhemerob
Forest with species atypical for the site, extensive grassland, heather Intensive grassland, extensive arable land Arable land
6 7
Polyhemerob Metahemerob
City green, pits Streets, buildings
No disturbance Extensive wood cutting, minor changes in matter circles Wood cutting, extensive grazing, rare and small doses of fertilizer Use of fertilizers and biocides, melioration Plowing, planting, major changes in matter circle, heavy use of fertilizers and biocides Strong changes in biocenoses Sealed surface, biocenoses destroyed
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
163
Table 2 Indicators for the three aspects of biodiversity Indicator
Abbreviation
Unit
Aspect of biodiversity
Percentage of land covered by semi-natural biotopes Mean richness of semi-natural biotopes in a circle of 1 km diameter Mean number of patches of semi-natural biotopes in a circle of 1 km diameter Proximity of semi-natural biotopes, represented by the area weighted mean proximity index Mean corrected hemeroby of all biotopes in the total area Percentage of semi-natural biotopes affected by external disturbance
P SNLAND M RICH 1 km M NP 1 km AMN PROX
% % Count –
Composition
M HEM C PDISTURB
– %
Function
context-dependent ‘corrected hemeroby’ considering that disturbances from streets, settlements and intensive land-use system will increases the degree of human influence and herby the hemeroby. Disturbance impact on biotopes is calculated in a cost-distance analysis. Conclusions about the strength and area of reach of disturbances are taken from Trombulak and Frissell (2000), Forman and Deblinger (2000), and Kappler (1997). It is assumed that 100% disturbance impact will raise the biotope-dependent hemeroby of the biotope type by one level. (2) Six indicators are describing the three aspects of biodiversity, composition, structure and function (Table 2). Two indicators represent composition: (1) the percentage of land covered by semi-natural ecosystems (P SNLAND); and (2) the mean richness of semi-natural ecosystem (M RICH 1 km). P SNLAND represents the most direct and simple to interpret indicator: The more land is covered by semi-natural
Structure
biotopes, the higher will be the amount of habitats for indigenous species, the potential for self-regulation, etc. Richness (M RICH 1 km) equals the diversity of semi-natural biotopes in an observation area in relation to the maximum number of semi-natural biotope classes in the whole study are (in our case a maximum of 12 semi-natural biotope classes). A richness of 50%, for example, expresses that 50% of all semi-natural biotope types, which can be found in the whole study area, can also be found in the observation area. Richness is calculated for circular shaped ‘moving widows’ and assigned to the central pixel of the moving window. Thereby, a pixel-wise evaluation is possible. For the case study, 1 km was used as an observation diameter for the ‘moving window’. Richness for larger areas is the average of the richness of each pixel within this area. Richness stands for the basic aspect of biodiversity which is the diversity of elements. Both indicators (P SNLAND, M RICH 1 km) reflect the classical understanding of biodiversity, which focuses mainly on the elements of a system.
Fig. 3. Flowchart of the biodiversity assessment.
164
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Structure is represented by the number of patches in a moving window (M NP 1 km) and the area weighted mean proximity index (AMN PROX). The number of patches (M NP 1 km) is calculated with the same technique like the calculation of richness: the number of patches is determined for a circular shaped moving window and assigned to the central pixel. AMN PROX (Gustafson and Parker, 1992), was calculated using Fragstats 3.0 (McGarigal, 2001). The proximity index quantifies the distances to adjacent patches of the same class in relation to their size in a defined search radius. The index distinguishes sparse distributions of small biotope patches from configurations where biotopes form a complex cluster of larger patches (McGarigal, 2001). The search radius for proximity evaluation was set to 1 km according to the size of the moving window for the diversity observation. Since landscapes are highly characterized by their heterogeneous structure, the structural aspect of biodiversity is a very essential supplement to the compositional aspect. Within the structural aspect the two indicators reflect a more or less antagonistic character. A high M NP 1 km reflects a landscape with a very ‘patchy’ structure (small, fragmented patches). This has positive effects, like a high structural diversity and the occurrence of edge-related biotopes (hedgerows, forest edges and other transition biotopes) which often show a very high species diversity. On the other hand, small patches are often not suitable as habitats for species with a high demand for large and connected areas. In contrast, AMN PROX targets the connectivity of semi-natural biotopes which is high in landscapes with only some large patches. Function is represented by the mean corrected hemeroby (M HEM C) and by the percentage of semi-natural biotopes affected by external disturbances (PDISTURB). PDISTURB is an outcome of the cost-distance analysis of disturbance sources. All semi-natural biotopes, which show more than 25% of disturbance impact are classified as ‘affected by external disturbance’. PDISTURB is calculated as the proportion of affected habitats within all semi-natural habitats in the area under observation. This functional approach considers the fact, that diversity is not only a function of the elements and the structure of a system, but also of the processes within a system. Disturbance from roads, for example, might significantly alter the diversity of birds, even if all other factors
(composition and structure) are identical (Forman and Deblinger, 2000). Besides AMN PROX, all calculations are based on per pixel calculations, which can be summarized for every area of interest. Supplementary to all indices mentioned above, an approach to measure connectivity from the species perspective was tested (see Section 4.5). In the context of the case study, a simple habitat model for the white stork (Ciconia ciconia) was developed. The model is based on descriptions of suitable biotope classes, threshold distances for connectivity and minimal areas for patch complexes. Descriptions were collected from different sources (e.g. Bastian, 1999; Flade, 1994). Connectivity was measured by calculating the percentage of area covered by suitable habitats for the white stork (P H STORK) within the group of all potentially suitable biotope types (humid and fresh grasslands). Additionally, the proportion of the largest habitat complex (largest patch index) within the group of all humid and fresh grasslands (LPI H STORK) was calculated. 3. Case study: extensification of grassland in Havelland county 3.1. Extensification of grassland The objective of this case study was the evaluation of biodiversity response-functions for the impact of grassland extensification and their functional parameterization. We have assumed that extensification of grassland will increase in Havelland with increasing liberation of the European market, due to low margins for agricultural goods, low productive sites and increasing subsidies for ecological measures. Eleven sub-scenarios represented increasing extensification of intensive grassland from 0 to 100%. It was assumed that through extensification, a transformation process from intensive grassland to semi-natural grassland types is possible in the long run. Three types of semi-natural grassland were considered: (1) humid grassland; (2) fresh grassland; and (3) arid grassland. Besides the fact that biodiversity was investigated at the ecosystem level, which targets practically an evaluation of landscape diversity, results may also be interesting for reflections about biodiversity on species levels, since various studies indicate that
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
165
This map is the result of a state-wide 1:25,000 mapping of biotopes from CIR aerial photographs, which includes more than 100 biotope types. We have aggregated the biotope types to 25 biotope classes well suited to represent land-use change and biodiversity attributes. A map of soil quality in a 1:200,000 scale with seven classes (MUNR), a map of groundwater distance originally in a 200 m × 200 m grid resolution (MUNR) and a digital elevation model (DEM) with 50 m × 50 m resolution (Brandenburger Landesamt für Vermessung) served as input maps for the pattern generation. All maps were rasterized to the 50 m × 50 m resolution of the DEM to obtain a contingent data base. Some tests to improve groundwater data by means of the DEM have shown a significant sensitivity of the results to data quality. Hence, a consistent and proved data basis is crucial for land-use modeling.
4. Results 4.1. Land-use change
Fig. 4. Study area of Havelland.
extensification has also a significant effect on species diversity (e.g. Mander et al., 1999). 3.2. Study area The study area covers about 1700 km2 in a typical lowland landscape west of Berlin (Fig. 4). The area is characterized by riparian floodplains with boggy or loamy soils covered with grassland, plains with arable land or deciduous forest on medium to poor soils and pine forests on sandy hills. A 20.5% of the total land are classified as grassland under intensive use. Semi-natural grassland types cover about 3% of the area. 3.3. Data sets The base map for land use and ecosystem representation was a biotope map (Brandenburger Ministerium für Umwelt, Naturschutz und Raumordnung, MUNR).
The initial situation showed only small patches of semi-natural grasslands. Humid and fresh grassland could be found along the river Havel in the western part and along some smaller water bodies (Fig. 5). After the first 30% of extensification, transformation appeared mainly in an area in the north-west of the study area. This area is characterized by very low groundwater distances and loamy or boggy soils. Here intensive grassland was mostly substituted by humid grasslands. Subsequently, intensive grassland along rivers and in wet lowland areas was substituted by semi-natural grassland of humid or fresh type. Intensive grassland on fertile soils with moderate groundwater level was the last to be transformed. Here predominately fresh grassland appeared. Additional arid grassland was rare and emerged mostly in the neighborhood of coniferous forests. To illustrate the effects of grassland extensification on landscape composition and pattern, the proportions of grassland types and two simple landscape metrics were plotted as functions of the intensity of extensification (Fig. 6). While the linear decline of the proportion of intensive grassland was the given model assumption, the shape of the graphs for the semi-natural types is a model result
166
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 5. Land-use change sequence for the scenario: extensification of grassland.
(Fig. 6a). It shows the early emergence of humid grasslands, the later emergence of fresh grassland and the minor emergence of arid grasslands. From the relation of mean patch size (MPS) (Fig. 6b) and the number of patches for each type (NP) (Fig. 6c), conclusions about fragmentation and compactness can be drawn. Humid grassland shows no increase in patch number but a strong increase in patch size. This indicates a high level of compactness. In contrary, fresh grassland shows an increasing number of patches until a maximum at 30% of extensification with no increase of MPS, followed by a decrease of the number of patches and a rapid increase in the MPS. This indicates the introduction of small, fragmented patches at the earlier levels of extensification, followed by a ‘clumping process’ of patches after a ‘percolation threshold’ (Turner et al., 2001) at 30% extensification.
4.2. Biodiversity at the pixel level Fig. 7 displays the richness for an extensification level of 50% at the pixel level. A high level of spatial heterogeneity is observable. Areas with high richness in the west (along river Havel) can clearly be distinguished from areas with low richness in the eastern part. The range reaches from 0 to 91% richness. 4.3. Biodiversity at the municipality level The first level of aggregation is the municipality level. The difference map for evaluating the gain in percentage of area covered by semi-natural biotopes (P SNLAND) between the initial situation and 100% extensification (Fig. 8), shows a high spatial heterogeneity. Some areas profited by extensification with a gain of more than 50%, others took no profit from
Fig. 6. Change of land-use composition and land-use pattern within the grassland area: (a) percentage of total grassland covered by single grassland types; (b) MPS for semi-natural grassland types; (c) total number of patches for semi-natural grassland types.
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
167
Fig. 7. Pixel-based evaluation of biotope richness.
extensification. Additionally, the response functions for each municipality can be drawn (Fig. 9). Fig. 9a displays P SNLAND curves for a subset of municipalities with significant gains in P SNLAND. The diversity of the shapes of the graphs represents clearly the functional heterogeneity of landscape response between different municipalities. While some municipalities realized gains in P SNLAND already at 20% extensification, others only benefited, if more than 80% of the grassland was extensified. A lot of the graphs show a threshold behavior. Two municipalities
with quite a similar shape of the graph and a high gain in semi-natural area were selected for further evaluation. For these municipalities, the total set of indicators is presented in Table 3. Some selected indicators were plotted as response functions (Fig. 9b and c). Indicators, which represent the compositional aspect of biodiversity showed the clearest response. Although both municipalities comprised a similar percentage of semi-natural biotopes (P SNLAND) after 100% extensification, richness (M RICH 1 km) and structural diversity (M NP 1 km) in Witzke were
Fig. 8. Difference map for the percentage of semi-natural biotopes at the municipality level.
168
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Fig. 9. Response functions at the municipality level: (a) percentage of area covered by semi-natural biotopes for a subset of municipalities with a significant gain in P SNLAND; (b) percentage of area covered by semi-natural biotopes for two selected municipalities; (c) mean richness of semi-natural biotopes in a circle with 1 km diameter for two selected municipalities.
about twice as high as in Kienberg. This indicates heterogeneities in the correlation of the single indicators. Very remarkable was the decrease of the proportion of disturbance (PDISTURB) among the semi-natural biotopes. Especially in Kienberg, the rate dropped from 57% down to only 12%. This can be explained as an effect of increasing patch size, which can buffer disturbances much more effectively. 4.4. Biodiversity at the landscape level At the highest level of aggregation indices were calculated for the total district. Response functions on the landscape level displayed only moderate sensitivity (Fig. 10, Table 4). Again, the compositional
indicators showed the clearest response (Fig. 10a and b). An extraordinary response function was displayed by the proximity (AMN PROXIM). Proximity, as a structural measure, raised by more than seven times and showed a nonlinear behavior (Fig. 10c). The high rise at about 30% of extensification can be explained by the beginning of a ‘clumping’ process of semi-natural patches when enough new patches have been introduced (see Section 4.1). After that, proximity decreased due to the introduction of new, smaller, patches in areas, where only little percentage of the land was constituted by semi-natural ecosystems. These new patches were isolated until adjacent intensive grassland also was extensified. While the other indicators followed more or less linearly the increasing
Table 3 Indicators for biodiversity at the municipality level Indicator
Extensification of grassland (%)
Total increase
Increase (%)
Municipality
0
50
100
P SNLAND
21.0 6.0
31.0 17.0
68.0 59.0
47.0 53.0
223.8 883.3
Witzke Kienberg
M RICH 1 km
31.0 14.3
36.4 18.8
37.5 23.3
6.5 9.0
21.0 62.8
Witzke Kienberg
M NP 1 km
7.0 3.1
8.3 3.7
8.1 4.3
1.0 1.2
14.9 39.8
Witzke Kienberg
M HEM C
4.0 4.3
3.9 4.2
3.5 3.8
−0.5 −0.6
−12.4 −12.8
Witzke Kienberg
PDISTURB
19.0 56.7
18.1 27.9
8.1 12.0
−11.0 −44.6
−57.5 −78.8
Witzke Kienberg
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
169
Fig. 10. Response functions at the landscape level: (a) percentage of area covered by semi-natural biotopes in the county Havelland; (b) mean richness of semi-natural biotopes in a circle with 1 km diameter in the county Havelland; (c) area weighted mean proximity index for the class of semi-natural biotopes in the county Havelland.
Table 4 Indicators for biodiversity at the landscape level Indicator
P SNLAND M RICH 1 km M NP 1 km AMN PROX M HEM C PDISTURB
Extensification of grassland (%)
Total increase
0
50
100
31.3 17.1 5.2 689.6 4.1 14.8
39.2 26.0 5.6 3388.9 4.0 14.3
48.0 27.8 5.8 5972.3 4.0 13.3
16.7 10.8 0.6 5282.7 −0.2 −1.5
Increase (%)
53.5 62.9 11.6 766.0 −4.4 −9.8
Fig. 11. Connectivity within grassland area from a functional point of view: (a) area weighted mean proximity index for semi-natural grassland types; (b) percentage of area within humid and fresh grassland suitable as stork habitat (P H STROK) and the percentage of area within humid and fresh grassland, which is covered by the largest habitat complex (LPI H STORK).
170
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
percentage of semi-natural land, the proximity has shown to be sensitive to the spatial heterogeneities, which can be observed in the landscape pattern. The two functional parameters showed a relatively weak response (Table 4). 4.5. Habitat approach to measure connectivity within the grassland area Within the grassland area the proximity index of all semi-natural grassland types was calculated (Fig. 11a). For humid grassland, proximity increased rapidly and stayed at the high level. Proximity for fresh grassland responded very similar to the proximity of semi-natural biotopes on the landscape level (Fig. 10c). The results for the proximity index indicated an increase of connectivity among semi-natural grassland types and an increase of the network character of semi-natural biotopes as a whole. But this measure is difficult to compare, even in a relative approach. To measure connectivity from a species point of view, we identified the percentage of area suitable for the white stork within the area of humid and fresh biotope types (P H STORK) (Fig. 11b). This ‘functional’ connectivity increased from 69 to 86% after 100% extensification. Hence, 86% of all fresh and wet grassland pixels were potentially suitable as stork habitats. This equals a total gain in area suitable for the stork from 4046 to 37,303 ha. LPI H STORK (the percentage of area covered by the largest habitat complex for the stork) increased rapidly at 50% extensification which indicates the clumping of all stork habitats to one big complex at this level of extensification.
5. Discussion First of all, the results of the investigations have shown a high level of heterogeneity of landscape responses in spatial and functional aspect. Heterogeneity was visible in an irregular spatial distribution of the evaluation results for the three aspects of biodiversity, in an irregular behavior of landscape responses of single spatial units through the various levels of extensification and in an irregular correlation between the indicators for three aspects of biodiversity. Heterogeneity increased when scaling down. At the
landscape level, heterogeneity was ‘hidden’ behind a supposed ‘moderate’ landscape response. The impact of land-use change on biodiversity strongly depended on the intensity of change, the configuration of the land-use pattern and the spatial distribution of the natural variables. Among the three aspects of biodiversity (composition, structure, function), the compositional aspect showed the clearest reaction. Particularly the percentage of area covered by semi-natural biotopes increased considerably at all levels of scale. Compositional diversity, represented by the richness of semi-natural ecosystems, increased less significantly, but showed remarkable gain in areas with an initial low percentage of semi-natural ecosystems. Furthermore, a high spatial variation of richness was observable. The structural attributes showed contradictory responses. While structural diversity expressed by the patchiness has shown a relatively weaker response, proximity was very sensitive. Structural diversity increased as a result of a partitioning process of large patches of intensive grassland into smaller patches of semi-natural grassland types. At this point, data resolution plays an important role. Semi-natural grasslands can show a very ‘patchy’ pattern at small scales in reality, e.g. due to heterogeneities in groundwater distances. This can only partly be reflected by the groundwater database with its 200 m × 200 m resolution. Thus, the interpretation of the structural diversity should not be overestimated. The structural aspect ‘proximity’ is very difficult to evaluate, since it has no defined unit and since it is not clear, if proximity response is really proportional to the impact. Results cannot be compared for an objective evaluation. Nonetheless, the results indicate an important role of isolation and connectivity within the response of landscape to land-use change. This aspect should be evaluated in an impact assessment, but must be seen under defined functional aspects (see the example of connectivity from a stork’s point of view). The functional aspect of biodiversity was assessed by the mean hemeroby and the proportion of disturbed biotopes. Hemeroby itself has shown a very weak response, which can be explained by the fact that hemeroby is a classified parameter, not a generic measure. This makes hemeroby less suitable for building mean calculations. In future studies a more appropriate approach should be introduced which
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
might, for example, include a ranking system instead of arithmetical calculations. Hemeroby has shown to be suitable for the qualitative pre-selection of the objects under observation. The percentage of disturbed biotopes showed a moderate response on the landscape level, but very significant results at the municipality level. The moving window approach for calculating ecosystem richness and structural diversity was very promising, since it enables a comparison across areas and scales. For future investigations, the diameter of the observation circle should be varied, to consider biodiversity aspects at various spatial scales. This might also contribute to further understanding of scale dependencies of biodiversity. In contrast to the island biology related approach of the pure patch-matrix concept (Forman and Godron, 1986), the moving window approach takes into account also transition phenomena between ecosystems, like considered in the ecotone approach (Delcourt and Delcourt, 1992). 6. Conclusion The presented case study has demonstrated the sensitivity of biodiversity attributes at the ecosystem level towards land-use change at the regional scale. The most sensitive attributes were the compositional ones, but also the structural and the functional attributes, particularly the connectivity, were significantly affected. Response functions of biodiversity attributes have shown to be particularly useful to detect thresholds, functional irregularities (one attribute changes more than another) and spatial heterogeneity in the impact of land-use shifts on biodiversity. This might be beneficial for determining sensitive regions and for assessing chances and threads for biodiversity at the ecosystem level under supposed changes in land-use composition. Acknowledgements This project is financed by the German Volkswagen foundation. References Adams, D., Alig, R., Callaway, J., McCarl, B., Winnett, S., 1996. The forest and agricultural sector optimization model (FASOM):
171
model structure and policy simulations. USDA Forest Service Research Paper PNW-RP-495. Portland, Oregon, p. 60. Antrop, M., 2000. Background concepts for integrated landscape analysis. Agric. Ecosyst. Environ. 77, 17–28. Bastian, O. (Ed.), 1999. Analyse und ökologische Bewertung der Landschaft. Springer, Berlin. Blume, H.-P., Sukopp, H., 1976. Ökologische Bedeutung anthropogener Bodenveränderung. Schriftenreihe für Vegetationskunde 10, 75–89. Delcourt, P.A., Delcourt, H.R., 1992. Ecotone dynamics in space and time. In: Hansen, A.J., di Castri, F. (Eds.), Landscape Boundaries: Consequences for Biotic Diversity and Ecological Flows. Springer, New York. Dierschke, H., 1984. Natürlichkeitsgrade von Pflanzengesellschaften unter besonderer Berücksichtigung der Vegetation Mitteleuropas. Phytoconeologia 12, 173–184. Duelli, P., 1997. Biodiversity evaluation in agricultural landscapes: an approach at two different scales. Agric. Ecosyst. Environ. 62, 81–91. Flade, M., 1994. Die Brutvogelgemeinschaften Mittel und Norddeutschlands. IHW-Verlag, Eching. Forman, R.T.T., Godron, M., 1986. Landscape Ecology. Wiley, New York. Forman, R.T.T., Deblinger, R.D., 2000. The ecological road-effect zone of a Massachusetts (USA) suburban highway. Conserv. Biol. 14, 36–46. Gustafson, E.J., Parker, G.R., 1992. Relationships between landcover proportion and indices of landscape spatial pattern. Landscape Ecol. 7, 101–110. Hansson, L., Fahrig, L., Merriam, G. (Eds.), 1995. Mosaic Landscapes and Ecological Processes. Chapman & Hall, London. Heckelei, T., Britz, W., 2000. Concept and explorative application of an EU-wide, regional agricultural sector model (CAPRI-project). In: Proceedings of the Paper Presented at the 65th EAAE-Seminar in Bonn. Henrichsmeyer, W., Cypris, C., Löhe, W., Meudt, M., 1996. Entwicklung des gesamtdeutschen Agrarsektormodells RAUMIS96 am Lehrstuhl für Volkswirtschaftslehre, Agrarpolitik und Landwirtschaftliches Informationswesen der Universität Bonn. Agrarwirtschaft 4–5, 213–215. Jones, C.A., Kiniry, J.R. (Eds.), 1986. CERES MAIZE. A Simulation Model of Maize Growth and Development. Texas A&M University Press, College Station, TX. Kappler, O., 1997. Some methods for determination and the evaluation of non fragmented and minimal disturbed landscape areas. In: Proceedings of the Paper Presented at the 10th European Colloquium on Theoretical and Quantitative Geography in Rostock, Germany, 6–10 September 1997. Krysanova, V., Mueller-Wohlfeil, D.I., Becker, A., 1998. Development and test of a spatially distributed hydrological/ water quality model for mesoscale watersheds. Ecol. Modeling 106, 261–289. Leemans, R., 1999. Modeling for species and habitats: new opportunities for problem solving. Sci. Total Environ. 240, 51– 73.
172
M. Zebisch et al. / Landscape and Urban Planning 67 (2004) 157–172
Mander, Ü., Mikk, M., Külvik, M., 1999. Ecological and low intensity agriculture as contributors to landscape and biological diversity. Landscape Urban Planning 46, 169–177. Marks, R., Müller, M.J., Leser, H., Klink, H.-J. (Eds.), 1992. Anleitung zur Bewertung des Leistungsvermögens des Landschaftshaushaltes, vol. 229. Selbstverlag des Zentralausschusses für deutsche Landeskunde, Trier. McGarigal, K., 2001. User manual FRAGSTATS 3.0. Naveh, Z., 2000. What is holistic landscape ecology? A conceptual introduction. Landscape Urban Planning 50, 7–26. Noss, R.F., 1990. Indicators of monitoring biodiversity: a hierarchical approach. Conserv. Biol. 4, 355–363. Palang, H., Mander, U., Naveh, Z., 2000. Holistic landscape ecology in action. Landscape Urban Planning 50, 1–6. Sukopp, H., 1972. Wandel von Flora und Vegetation in Mitteleuropa unter dem Einfluß des Menschen. Berichte über die Landwirtschaft 50, 112–139. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conserv. Biol. 14, 18–30. Turner, M.G., Gardener, R.H., O’Neil, R.V., 2001. Landscape Ecology in Theory and Practice. Springer, Berlin. Veldkamp, A., Lambin, E.F., 2001. Predicting land-use change. Agric. Ecosyst. Environ. 85, 1–6. Verburg, P.H., de Koning, G.H.J., Kok, K., Veldkamp, A., Bouma, J., 1999. A spatial explicit allocation procedure for modeling the pattern of land use change upon actual land use. Ecol. Modeling 116, 45–61. Weber, A., Fohrer, N., Möller, D., 2001. Long-term land-use changes in a mesoscale watershed due to socio-economic factors—effects on landscape structures and functions. Ecol. Modeling 140, 125–140. White, D., Minotti, P.G., Barczak, M.J., Sifneos, J.C., Freemark, K.E., Santelmann, M.V., Steinitz, C.F., Kiester, A.R., Preston, E.M., 1997. Assessing risk of biodiversity from future landscape change. Conserv. Biol. 11, 349–360. Wiens, J.A., 1976. Population responses to patchy environments. Annu. Rev. Ecol. Syst. 7, 81–120. Wiens, J.A., 1995. Landscape mosaics and ecological theory. In: Hansson, L., Fahrig, L., Merriam, G. (Eds.), Mosaic Landscapes and Ecological Processes. Chapman & Hall, London, pp. 1–26.
Wrbka, T., Szerencsits, E., Moser, D., Reiter, K., 1999. Biodiversity patterns in cultivated landscapes: experiences and first results from a nationwide Austrian survey. In: Maudsley, M., Marshall, J. (Eds.), Heterogeneity in Landscape Ecology. Proceedings of the 1999 Annual IALE (UK) Conference, Bristol. Marc Zebisch has a Master’s degree in geoecology (University of Potsdam, Germany, 1999). Currently he is a PhD student at the Institute for Landscape and Environment Planning, Technical University of Berlin (TUB), Germany. He specializes in land-use investigation by means of remote sensing and GIS techniques with a particular interest in spatial explicit modeling of processes in the landscape. He was a co-worker in projects about landscape dynamics in Germany and Mongolia. The investigations presented in this article will be subject of his PhD thesis. His work is embedded in the project mesoscale simulation study assessing the consequences of global change (MESSAGE), which is a co-project of the TUB, the Potsdam Institute of Climate Impact Research and the Society for Agricultural Politics and Sociology (FAA, Bonn). Frank Wechsung is a Senior Research Scientist at the Potsdam Institute of Climate Impact Research, Department of Global Change and Natural Systems (Germany). His principal fields of interest are modeling climate impact on crops at regional and global scale and the response of crops to CO2 enrichment. He holds a PhD in agriculture and operation research. Currently he is working as the coordinator of the MESSAGE project and other projects at PIK. Hartmut Kenneweg is the Managing Director of the Institute of Landscape and Environment Planning at the Technical University of Berlin (TUB), Germany. He took his PhD in forestry at the University of Freiburg. From 1980 to 1985 he has been Professor for forest inventory and forest planning at the University of Göttingen. Since 1985 he has been Professor for Landscape Planning at the TUB. He has been a member in several advisory boards and steering committees (e.g. 1988–1997: ‘regional applications of satellite remote sensing’, DARA; 1990–1993: ‘large area operational experiment for forest damage monitoring in Europe’, UNEP/ECE). He has special interests in remote sensing of forestry and land-use dynamics.