Biological Conservation 147 (2012) 3–12

Contents lists available at SciVerse ScienceDirect

Biological Conservation
journal homepage:

Determining the ecological value of landscapes beyond protected areas
K.J. Willis a,b,⇑, E.S. Jeffers a, C. Tovar a, P.R. Long a, N. Caithness c, M.G.D. Smit d, R. Hagemann d, C. Collin-Hansen d, J. Weissenberger d

Biodiversity Institute, Oxford Martin School, Department of Zoology, University of Oxford, Oxford OX1 3PS, UK Department of Biology, University of Bergen, Allégaten 41, N-5007 Bergen, Norway c Oxford e-Research Centre, University of Oxford, Oxford OX1 3QG, UK d Statoil ASA, Forusbeen 50, 4035 Stavanger, Norway

a r t i c l e

i n f o

a b s t r a c t
Whilst there are a number of mapping methods available for determining important areas for conservation within protected areas, there are few tools available for assessing the ecological value of landscapes that are ‘beyond the reserves’. A systematic tool for determining the ecological value of landscapes outside of protected areas could be relevant to any development that results in a parcel of land being transformed from its ‘natural’ state to an alternative state (e.g., industrial, agricultural). Specifically what is needed is a method to determine which landscapes beyond protected areas are important for the ecological processes that they support and the threatened and vulnerable species that they contain. This paper presents the results of a project to develop a method for mapping ecologically important landscapes beyond protected areas; a Local Ecological Footprinting Tool (LEFT). The method uses existing globally available web-based databases and models to provide an ecological score based on five key ecological features (biodiversity, vulnerability, fragmentation, connectivity and resilience) for every 300 m parcel within a given region. The end product is a map indicating ecological value across the landscape. We demonstrate the potential of this method through its application to three study regions in Canada, Algeria and the Russian Federation. The primary audience of this tool are those practitioners involved in planning the location of any landscape scale industrial/business or urban (e.g., new town) facility outside of protected areas. It provides a pre-planning tool, for use before undertaking a more costly field-based environmental impact assessment, and quickly highlights areas of high ecological value to avoid in the location of facilities. Ó 2011 Elsevier Ltd. All rights reserved.

Article history: Received 18 April 2011 Received in revised form 30 October 2011 Accepted 2 November 2011 Available online 28 January 2012 Keywords: Biodiversity valuation Connectivity Ecological footprint Fragmentation Threatened species Resilience

1. Introduction Protected areas have long been the mainstay of biodiversity conservation with 12% of land currently under some form of protection (Jenkins and Joppa, 2009) and a commitment to increase this to 15–20% by 2020 (Stokstad, 2010). There are a number of excellent tools available for mapping conservation priorities within these protected landscapes (e.g., C-Plan, Pressey et al., 2009; Marxan, Ball et al., 2009; and Zonation, Moilanen, 2007). However, users of the outputs of systematic conservation planning tools have traditionally treated land outside of the protected area network as ‘‘scorched earth’’, i.e., as providing no benefit to biodiversity conservation (Edwards et al., 2010). Landscapes beyond protected areas are increasingly being recognised as important for providing ecological and evolutionary processes essential for the long-term persistence of biodiversity
⇑ Corresponding author at: Biodiversity Institute, Oxford Martin School, Department of Zoology, University of Oxford, Oxford OX1 3PS, UK. Tel.: +44 1865 281321. E-mail address: (K.J. Willis).
0006-3207/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2011.11.001

both within and beyond protected areas (Bengtsson et al., 2003; Carroll et al., 2010; Chazdon et al., 2009; Mathur and Sinha, 2008). Key biotic and abiotic features of these landscapes include their role in the provision of corridors between reserves (e.g., through waterways, wetlands, Edwards et al., 2010) and as refuges for species given future range-shifts resulting from climate change (e.g., Carroll et al., 2010). It is also recognised that many threatened and protected species have significant populations outside of protected areas (IUCN, 2009) and that the majority of species migration routes occur beyond protected areas (Riede, 2004). Consideration is also needed of the landscape scale features that are important to maintain resilient ecosystems (Bengtsson et al., 2003; Klein et al., 2009). These biotic and abiotic features have numerous different spatial configurations across the landscape. All landscapes outside of the protected areas are not, therefore, equal in terms of ecological value. This point is particularly relevant when considering placement of business facilities, industrial operations and even the building of solar and wind farms; where can they be built that will have least ecological impact? There are currently very few tools available for mapping conservation

html). 1994). has minimal perennial vegetation cover and low species richness and endemism. these plans are based on detailed site surveys carried out by a team of consultants. The Algeria site is located within the Saharan Desert eco-region which is hyper-arid. Oases are also present in the landscape. 2. taxonomic uniqueness (Vane Wright et al. rarity (Smith and Theberge.g. The focus here was to determine what freely available spatial data from the internet.html). Additionally. It then describes results from testing this methodology for three case study regions in Canada. How to measure ecological value is complex. At the landscape scale probably the most common measures adopted to date are indices of species richness.. Edwards et al. This paper describes the multi-criteria metrics and databases that we analysed and the algorithms developed to create a methodology to assess the ecological value of parcels of land across a landscape for any location in the world.5 km. which is also designated as a World Heritage Site and UNESCO Biosphere Reserve (UNEP-WCMC. Despite the extreme conditions. These criteria represent both the current ecological properties of . All these studies agree. In addition. 72°260 58. and (3) be at a scale that is relevant to the extent of most development concessions. 2009). models and algorithms to produce mapped output at a spatial scale relevant to most landscape scale planning decisions (<0. rather the EIA often gives an indication of necessary mitigation measures to minimise the damage that will be caused by the development. 2010).org/) which facilitate access to accurate and up-to-date biodiversity information but this again focuses on key biodiversity areas and legally protected areas and produces mapped output at a spatial scale ($20 km) that is often too coarse for most landscape planning decisions. several studies that have attempted to include additional factors in conservation planning that prioritise according to the ‘dynamic’ features of the landscape including wetlands. Many studies indicate that a more fragmented patch of landscape. Underlying assumptions associated with this method include: (1) that it should be based on multiple valuation factors.4 K. Whilst these assessments yield an estimate of the species present in an area proposed for development. which is a lowland tundra situated upon a thick layer of permafrost (http://www. 90 avian and 100 reptilian species known to reside within the Saharan Desert eco-region (http://www. these measures are based on observed biological patterns of conservation ‘assets’ or ‘actors’ but lack information on the ecological processes that support ecosystem functioning (Bennett et al. Current land use is dominated by nomadic pastoralism. Willis et al.html). connectivity and resilience. Methods 3. 111°270 13. ecoregions/item1847. 2010) overlays the study area and extends to the south for 300 km. combined with well-established models and algorithms could be used to provide a baseline of ecological information for any site in the world. The options to relocate a facility due to impact on ecological processes at this point in the decision making process are also extremely limited. however. 2010).html). Currently the mainstays of ecological assessment by businesses for landscapes beyond protected areas are site Biodiversity Action Plans and Environmental Impact Assessments (EIA) (Slootweg.2400 N. The nearest protected area to the study area is the Tassilin’Ajjer National Park. they do not consider the overall footprint of the development upon important ecological processes on the landscape.businessandbiodiversity. Beyond the oases. and/or require a high level of data knowledge and input. NatureServe www. vulnerability. this park is about 300 km to the southeast of the study area. 2009).org/action_site_bap. which is 40 km to the west of the study area (UNEP-WCMC. 2007) in order to secure both the arenas and the actors (Beier and Brost. 2°280 44. one in central Algeria (27°190 33. There is also often a consideration of the contiguous nature of the habitat supporting the species. (2) make use of free spatial data available for almost any location worldwide. Algeria and Russia. These provide the framework by which companies endeavour to minimise the impacts of their activities on ecosystems and habitats and ensure that planned development activities are in line with regional and national Biodiversity Action Plans (http:// www.. fragmentation. However. Typically. 2010). large mammal migration paths and/or methods to incorporate areas that appear to be more resilient to environmental perturbations because of the combination of biotic/abiotic features that they contain (Klein et al. would use existing globally available databases.200 E).html).worldwildlife.worldwildlife. Current land uses in the eco-region include forestry and oil and gas Crow The Yamal’skiy marine protected area (UNEPWCMC.5 km). Current land use is dominated by nomadic pastoralism..759400 E) and one on the Yamal Peninsula in Russia (72°180 47. which is characterised by continuous mid-boreal mixed coniferous and deciduous forest across a variable topography with discontinuous permafrost cover (http://www. 1986). / Biological Conservation 147 (2012) 3–12 priorities outside of reserves.. the sparse vegetation is highly adapted to the hyper-arid climate via the persistence of the seed bank that allows for masting during high rainfall science/ecoregions/item1847. Those that are available tend to be region or country-specific (e. These areas were selected to represent a wide variety of eco-regions at different geographical locations and varying data There are 70 mammalian. that effective planning for ecological processes requires a multi-criteria assessment approach that incorporates all of these important measures of ecological integrity (Regan et al.519400 N.J. There are many different metrics to consider at varying spatial and temporal scales. namely: biodiversity.ibatforbusiness. we set out to devise a method that would have a simple user input.. 2010). it contains a relatively large number of species that are adapted to this high stress environment (http://www. five criteria were determined as being of primary importance to the ecological valuation of a landscape.1600 N. 1991) and irreplaceability (Pressey et al. 2010). Identification of data sources and models to determine ecological criteria Through a series of workshops with practitioners and a detailed literature review. ideally <0.. Canada (55°490 5. 3. flyways. The aim of this project therefore was to devise a simple and quick. Recently..naturserve. where perennial vegetation and agriculture is present. There is one protected area. 2009. The Alberta site is located within the mid-continental Canadian forest eco-region. The Yamal Peninsula Russia study area is located within the Yamal-Gydan Tundra ecoregion. yet effective method for determination of important ecological features and processes on landscapes beyond protected areas. for example. Study areas Three study areas were chosen to determine an ecological valuation of the landscape: one in Alberta. They also tend to occur once a decision has been made regarding where to develop. there exist tools such as iBat for business (iBatforbusiness. has less ecological value than a large undisturbed patch (for recent review see Tjørve.319400 W).org https://www. Vegetation cover in the northern part of the peninsula is dominated by mosses and lichens with a few shrubs and grasses that support larger herds of reindeer during the summer months.worldwildlife.

gbif.e. GDM predictions of compositional dissimilarity for every possible pair of sites within the study area were made in the statistical software package R using the Table 1 Factors to be included in a local ecological footprint tool for assessing ecological integrity across a landscape and the available data source. plants. (2008) Hijmans et al. It is therefore necessary to model predictive diversity across the landscape using a combination of point species occurrences and environmental variables to predict diversity on landscape (Hirzel et al. In order to determine a measure of species occurrence in the study area. fragmentation. global data were available at every 90 m sized pixel (i. Biodiversity data were obtained for a larger area (i.. 2002). Riede et al. we selected vegetation cover at 300 m pixel size resolution (GLOBCOVER.a annual mean the study area is divided into 300 m sized pixels and the other data layers are superimposed upon this in order to calculate values for each of the ecological indicators.. habitat integrity. corridors).) at 10 for inclusion of each group in the modelling in order to be able to robustly parameterise a GDM for that group. precipitation seasonality. birds.. birds. one picture element in a raster or remotely sensed image) for elevation (topography) to 10 km sized pixels for soil characteristics.. We found that there are very few regions in the world that have appropriate densities of GBIF records to allow this approach. 2006). Output from this model highlights areas that are more different to their neighbours than other areas within the landscape in terms of their species assemblage and thus provides a measure of heterogeneity per unit area. etc.g. Factor Biodiversity Vulnerability Fragmentation Connectivity Resilience a b c d e f g h Definition Compositional turnover with respect to environmental covariates The number of threatened species present The size of the vegetation patch River. However. Ferrier and Guisan. Table 2 Environmental covariates used in generalised dissimilarity modelling per biological group. total annual precipitation. temperature seasonality.e. mammals and reptiles Plants a b c Predictor variables Distance to water bodies. 2002). therefore. Species richness is then estimated as the sum of species for which habitat suitability exceeds a given threshold. Since GDMs evaluate assemblages of species. temperature seasonality. soils. 3. http://data. Given how important a knowledge of the vegetation cover is for calculation of many of the variables (biodiversity. hydrosheds) were obtained from other global databases (see Table 2). 2004) as contained in the study area and to remove any duplicate records of species occurrence in one location. however. and it was these data which were used to generate predicted compositional dissimilarity for our study area. The biodiversity data were filtered to retain only those occurrences recorded in locations within the same eco-regions (Olson et al. (2005). spatially continuous geographic data. Biological group Amphibians. biodiversity.2.b total annual precipitation. led by MEDIAS-France. we used the data contained in the Global Biodiversity Information Facility database (GBIF. % nitrogen in soil. Lehner and Döll (2004) Lehner et al. The number of migratory species present The ability to sustain high rates of net primary productivity in areas of low precipitation Data source GBIFa. / Biological Conservation 147 (2012) 3–12 5 the landscape (e. IUCN...gbif. It also contains fossil records but these were removed from the analyses. To determine the finest spatial resolution possible for determining an ecological value of a landscape we examined the range of readily available. Ferrier et al. 2002). These data were then used in a Generalised Dissimilarity Model to determine the compositional turnover in the study area which is the rate of change in species composition with respect to the environmental variables (Ferrier et al.g. wetlands. (2009). Ó ESA GlobCover Project. threatened species) and the key features important for supporting ecosystem functions (e. resilience). precipitation seasonality Annual mean temperature. BIOCLIM. The GBIF Data Portal currently provides access to more than 300 million records of species occurrence worldwide. 700 km  700 km) around our study areas for constructing the GDM. Harmonized World Soil Database. 2009.K. Zhao et (2001).g. BIOCLIMd IUCN Red Liste GLOBCOVERf Global Lakes and Wetlands Databaseb.. we set a minimum number of species per biological group (e.. This method predicts habitat suitability for individual species based on occurrence records and environmental covariates. Data and models (described below) were then assimilated to calculate these criteria and an algorithm developed to sum the criteria into an overall dimensionless indicator of ecological value to enable a mapped output (see Table 1). lake and wetland features. Ideally we would use species richness as well as compositional turnover to estimate diversity across a landscape using a approach such as ecological niche factor analysis (Hirzel et al. Data for the abiotic variables in study area (climate. there is no minimum requirement for the number of species occurrences. Ó ESA GlobCover Project. Biodiversity For most regions in the world there will rarely be enough detailed species data to obtain a clear picture of the biodiversity. to estimate species richness using such models requires more than around 30 records per species (Zaniewski et al. led by MEDIAS-France) as the base layer for the LEFT. 2002. connectivity (migration routes.c soil water holding capacity Global Lakes and Wetlands Database and HYDROSHEDS. HYDROSHEDSc and Global Registry of Migratory Speciesg MODIS/TERRA NPP Yearly L4 Global 1 km SIN Grid V004h and BIOCLIM Global Biodiversity Information Facility (http://www. . and resilience.. Willis et al. 2002.

Folke et al.groms. 3. we retained all globally threatened species to maintain the breadth of species potentially present in our study areas. lakes.. then applied the following rules for assessing the resilience value of each pixel: a value of 1 (i. however. the lowest value) for resilience. 1973. we summed the value of each factor together to provide an overall indication of ecological value per pixel according to the formula: Summary ecological value = Biodiversity + Vulnerability + Fragmentation + Connectivity + Resilience. all other pixels were given a value of zero (i. Two complimentary factors that represent migration processes across the landscape were included namely (i) migration routes as identified in the Global Register of Migratory Species GROMS ( and (ii) waterways and wetlands. Hijmans et al. reptiles and amphibians) were obtained and overlain in order to count the number of globally threatened terrestrial vertebrates potentially present in each pixel. / Biological Conservation 147 (2012) 3–12 GDM package (Ferrier et al. 2004). 3. we used the IUCN Red List of Threatened Species (IUCN. However. The 2010 IUCN Red List of Threatened Species has assessments for $56. we used total NPP in the year 2005 since this was the year in which the GLOBCOVER data was created. 2005) to identify patterns across space in the level of productivity of each vegetation type given spatial variations in rainfall.). .000 species globally. The software FRAGSTATS (McGarigal et al.. similar vegetation cover were assumed to have a higher ecological value. 3. The biodiversity. the actual maximum value is dependent on the availability of the data for each study site. we reclassified the GLOBCOVER vegetation categories into the following groups: closed forest. All bare areas.J.g. Fragmentation There is a vast ecological literature demonstrating the ecological importance of habitat integrity and ‘intactness’ and the impact of fragmentation on biodiversity (see Fahrig (2003) for review). The output was transferred back into a GIS platform for visualisation..7. 3. 2008) and the Global Lakes and Wetlands database GLWD3 wetland classification (Lehner and Döll. We used the GROMS maps in the same way as the IUCN polygons: we calculated the number of migratory species ranges which intersect each pixel and those pixels that had the highest number of migration routes occurring across them were deemed to be of higher ecological value. soil erosion) should also be calculated.e.. bare areas.e. 2007) and the results were imported into MATLAB where a distance weighted average (using the inverse square distance measure) of all pair-wise combinations was calculated per site. Pixels which were situated within a riverine corridor or wetland were given a higher ecological score. Vulnerability In order to assess vulnerability of an area in terms of the potential loss of important species if the area was damaged. This was achieved by setting the minimum value to zero and dividing the individual values per pixel by the maximum value of each measure. Values of annualized net primary productivity (NPP. Connectivity Connectivity across a landscape either through riverine corridors and/or other migratory routes is essential to any ecologically functioning landscape.5 was assigned to pixels that were in the upper quartile of productivity and the second lowest quartile of rainfall. connectivity (rivers. Areas characterised by human activities. the higher its functionality (greater diversity. the highest resilience) was assigned to pixels that were in both the upper quartile of NPP and the lower quartile of rainfall. more pollinators. non-natural and bare areas were automatically set a value of zero for patch size.6. EN. Ideally a measure of resilience to environmental disturbances (e. we retained the highest value of compositional dissimilarity per pixel in order to report the highest potential value of compositional dissimilarity which could occur in that pixel. Resilience Ecological resilience is the capacity of a system to undergo disturbance and maintain its functions and controls (Holling. 2001). mammals. Summary ecological value In order to ascribe a dimensionless ecological value to each 300 m pixel that incorporates all five measures described above. Where multiple biological groups were modelled.6 K.. a value of 0.5. A key feature to try and retain on any landscape. In general the greater the patch size.4. We therefore include a calculation to prioritize pixels that support migratory processes. Larger areas of continuous. To ensure that NPP data were consistent with the vegetation cover data. is habitat integrity. Prior to running FRAGSTATS. 3. We calculated quartiles for the precipitation and NPP data per vegetation type within the study area.. After standardisation. The maximum ecological value for any pixel was therefore six given that there are two complementary measures of connectivity. 2009). (2009) that focuses on the ability of a parcel of land to maintain productivity – a fundamental aspect of ecosystem functioning – despite relatively high water stress. GROMS currently contains a list of 2880 migratory vertebrate species in digital format and digital maps for 545 of these species. Pixels with more globally threatened species potentially present were treated here as having a higher ecological value. Zhao et al. we standardised the values of each factor to an index between zero and one. To determine the pixels that supported river corridors and/or wetlands (and the pixels immediately adjacent to these) we used data available in the HYDROSHEDS 15-arc-second river network database (hydrosheds. 2009) per vegetation type (determined by GLOBCOVER) were overlaid with data of the typical precipitation of the driest quarter (WORLDCLIM bio17.3. Willis et al. therefore. Range polygons of globally threatened (CR. We calculated patch size according to similar vegetation type based on the classes in the GLOBCOVER dataset. as yet no global datasets are yet available that can be harnessed to produce a relative measure of resilience to such environmental disturbances across space. lakes and agricultural areas were assigned a value of zero. VU. forest/shrub/grass mosaic. lakes and wetlands) and resilience measures were already in binary integer values and thus required no conversion.. Resilience is therefore an important ecological feature of any landscape and areas that can maintain resilience despite climate/environmental disturbance are of high ecological significance. greater complexity in food webs etc.. The resulting index highlights the pixels containing the most resilient patches of vegetation based on the ability to retain productivity despite low rainfall. 2004). water/snow/ice. of which about 28.wwf) (Lehner et al. To determine climatic resilience we adopted the approach used by Klein et al.iucnredlist. Pixels demonstrating higher levels of compositional dissimilarity were treated as having a higher ecological value. open forest.000 have spatial data (http:// www. It therefore provides a large number of polygons of migratory ranges and distributions of species that migrate across national boundaries (Riede and Kunz. While it is possible to filter or weight species by their threat category. 2002) was used to define a patch using the rule of eight pixels (orthogonal and diagonal adjacency) and to estimate the natural logarithm of the area of each patch (ha). or NT) terrestrial vertebrate within each study area (birds..

This area is marked by a dense river network and both small and large wetlands are associated with these rivers (Supplementary data Fig. Canada The Globcover map for the study region is presented in Fig. within the study area there are predicted to be atleast 59 internationally migratory species and the number in any location ranges from 52–59 (Supplementary data Fig. 1c). shrubland and grasslands.e. Data handling and display For the study areas described above.2. Once calculated. 5. Other areas of high resilience are scattered around the study area. which is associated in space with the boundary between unconsolidated bare areas (i.. mammals (n = 18 species) amphibians (n = 10 species) and plants (n = 545 species).K.1. plants (n = 1). with large areas of continuous boreal forest and with some smaller areas of wetland along the western edge of the study area. In terms of connectivity. Alberta. In terms of habitat fragmentation. A desert region with sparse vegetation cover except for that present on intermittent wetlands (oases) in the centre of the area. connectivity (due to migratory species plus the landscape features that support migration) and resilience. data from the IUCN Red Data list of Threatened species indicates that there three different globally threatened species predicted to occur within the study Fig. 2a). Algeria The Globcover map for the study region is presented in Fig. Canada study area. the summary ecological value for each pixel was plotted on a final map layer ArcGIS 10 to display differences in value across the landscape. 3. 5. The highest values (darkest grey-scale) appear to be consistent with the river boundaries. 1. GBIF species occurrences for the Algeria site indicate that only bird species can be used in the GDM modelling (n = 27) since there were not enough data for mammals (n = 7). we manually downloaded all of the spatial data and projected it into the Universal Transverse Mercator (UTM) coordinate system. Output from GBIF indicates that in this region there are point occurrences for birds (n = 282 species). The IUCN threatened species list indicates that the study area contains two globally threatened species both of which potentially occur across the whole study area (Supplementary data Fig. sand dunes) and consolidated bare areas (Supplementary data Fig. in pixels that contain closed needle-leaved evergreen forest and lakes (Supplementary data Fig. This large patch is surrounded by much smaller patches of a mosaic of forest. Using this data in combination with the environmental co-variates (Table 2) the GDM output demonstrates that predicted biodiversity is the highest in the northwestern portion of the study area. . 1f). fragmentation. reptiles (n = 3) or amphibians (n = 0). especially in the east where there are numerous wetlands. 1e). 1d). 1a). 2. Output from the GDM modelling using the bird data combined with environmental covariates (Table 2) indicated highest beta diversity in the central and southwestern portions of the study region. This demonstrates that the highest possible ecological value for this study area is 6 (values assigned for biodiversity. / Biological Conservation 147 (2012) 3–12 7 4. vulnerability. results from FRAGSTATS reveals that there is a wide range in vegetation patch size in this study area from less than 1 ha to over 1000 ha (Supplementary data Fig.J. In terms of the vulnerability measure. Results 5. 1. The largest continuous patch consisting of closed needle-leaved deciduous and evergreen forest occupies the central region of the study area. Willis et al. The greatest number of international migratory species appears to be concentrated in the north central part of the large continuous patch of closed needle-leaved forest. We set the study area at 100 km  100 km and each 300 m pixel within this area was evaluated for its ecological value using the algorithm described above. The resilience measure indicated that the areas of highest resilience are in the north-west corner of the landscape (Supplementary data Fig. The summary ecological valuation map is presented in Fig. 1b).

Another region of high ecological value is the northern boundary of the study area. (e) connectivity from rivers. Willis et al. The summary ecological value for the Algerian study area is presented in Fig. there is little to no vegetation cover within the study area. 3d and 3e) The resilience measure (Supplementary data Fig. These layers are overlaid to determine the (g) summary value of each pixel within the Canadian study area (see also Supplementary data Fig. The resulting map indicates that the majority of the study area has low relative ecological value (Fig. 6). 2. fragmentation and both indicators of connectivity. In terms of a vulnerability calculation. It was not therefore possible to calculate a measure of betadiversity for this region. fragmentation.. a landscape that is predominantly covered by grasslands and shrublands and interspersed with small water bodies. 2b). 5. vulnerability. Results from GBIF indicate a serious data gap of species occurrences for this region with only four mammal species and eight plant species recorded (Biodiversity Data from GBIF). Results from the fragmentation calculation using FRAGSTATS was good and indicated variation in patch sizes ranging from <1 ha to 3200 ha (Supplementary data Fig. (f) vegetation resilience. As this region is part of the Saharan desert. This is due to the extreme aridity of the Saharan Desert and the resulting high variability in vegetation cover over each year resulting in no consistency between the NPP data and vegetation cover. 3f) indicated a patchy distribution of resilient vegetation across the landscape. 4) while there are small patches of higher ecological value at the sites of the intermittent wetlands (in the northeast and southwest) and around the few patches of vegetation cover (central area). where the land cover is primarily consolidated bare areas (Supplementary data Fig. The largest patch is found in the southern part of the study area. GROMs indicates that there are at least 21 migratory animals predicted to occur in the study area. there are intermittent wetlands (i. It was not possible to assess the differences in resilience of vegetation across space for this area (Supplementary data Fig. Despite the hyper-aridity of the area. including both deciduous and evergreen forest. . (c) fragmentation. Measures of ecological value for the Canadian study area including: (a) beta-diversity. / Biological Conservation 147 (2012) 3–12 Fig.. Russia The Globcover map for the Yamal peninsula study region is presented in Fig. All three occur in the northern portion of the study area.3.e. The predicted range of these species appears to be concentrated in the northwestern region of the study area. 2f). thus the only areas subject to the fragmentation assessment is the tiny patch of vegetation in the centre of the study area (Supplementary data Fig. the maximum value is five). Output from GROMS indicates that there are at least eight internationally migrating species that are expected to utilise the habitat around the Algerian study area (Supplementary data Fig.) region. 3c). therefore. lakes and wetlands. Results from this map indicate the areas of highest ecological value are along the coastline and along the river and wetland networks (Fig. In the northern part of the study area there are also numerous patches of regularly flooded vegetation and open forest. (d) connectivity due to migratory species.J. the IUCN Red data list for Threatened Species indicates that there are two globally threatened terrestrial vertebrates that are predicted to occur in the study area (Supplementary data Fig. The summary ecological value map (Fig. This area is surrounded by largely continuous desert land cover. 4 and is based on biodiversity. A measure of connectivity was also possible for this site. 6) for the Russian study area is therefore based on 5 of the possible 6 layers: vulnerability. there was little spatial difference in their migratory ranges. 1a–f. which is consistent with the highest values of connectivity and vulnerability. 2c). oases) within the study area that may provide support for the migratory bird species that are known to occur in the Tasslin’Ajjer National Park and these pixels are detected in the connectivity layer (Supplementary data Fig.e. 2e). This vegetation cover is so sparse that the maximum patch size is about 0. (Supplementary data Figs.03 ha.8 K. 2d). 5. (b) vulnerability. the maximum value of the summary ecological value for this site is 5. 3b). which is desert habitat and far away from human settlements that tend to concentrate around the wetland areas. two connectivity measures and resilience (i.

(d) connectivity due to migratory species. Measures of ecological value for the Algerian study area including: (a) beta-diversity. (f) vegetation resilience.) 6. The poorest data-region from our . but using a multi-criteria metric approach such as described here. (c) fragmentation. Discussion For all three study areas examined. 4. 3. this study demonstrates that it is possible to obtain an estimation of the spatial distribution of some important ecological features across a landscape and at a relatively fine spatial resolution (here at a 300 m pixel resolution).K. 2a–f. Willis et al. On a global scale. (b) vulnerability. / Biological Conservation 147 (2012) 3–12 9 Fig. using globally available databases and existing models and algorithms. some data gaps are inevitable.J. (e) connectivity from rivers. Fig. it would appear that even for data poor regions. lakes and wetlands. Algeria study area. there is now sufficient information available in global databases to make some estimation of ecological value. These layers are overlaid to determine the (g) summary ecological value of each pixel within the Algerian study area (see also Supplementary data Fig.

Russia study area. (e) connectivity from rivers. especially from the species distribution ranges (Palminteri et al. (b) vulnerability. (f) vegetation resilience. Measures of ecological value for the Russian study area including: (a) beta-diversity. 6. These layers are overlaid to determine and (g) summary ecological value of each pixel within the Russian study area (see also Supplementary data Fig. In addition to the data gaps however. there are also sources of uncertainty associated with these global databases which need to be acknowledged from the outset. (d) connectivity due to migratory species. information that could have relevance to sighting of industrial facilities and highlight areas of high ecological risk. 2011). / Biological Conservation 147 (2012) 3–12 Fig. 5.J. Fig. three case-studies was northern Russia – but it was still possible to obtain spatial information sufficient to map vulnerability. The difference .. 3a–f). fragmentation and connectivity across this landscape. Willis et al. The main limitation is the coarse resolution of some of the input data. lakes and wetlands.10 K. (c) fragmentation.

it can both compensate for the data gaps in some layers and also be used to support a wider variety of land-use decisions. A.biocon.M. which are available for a variety of costs but can provide high-resolution data. Oxford.J. B. Use of land facets to plan for climate change: conserving the arenas.L. R.. Whilst they can (and in a few cases.. and Zonation.D. As a pre-planning tool to be used before any site locations are determined to highlight the most ecologically sensitive regions. As mentioned in the introduction. Conclusions Using the approach described.. ground-truthed data. while preliminary checks are routinely conducted on GBIF data.D. at doi:10. CPlan... P. much less compiled in a comprehensive way’’. I. 2009) therefore ground-truthing can also verify the classification of vegetation and whether vegetation change has occurred since the MERIS data used in the GLOBCOVER classification were acquired (i. Mackey.2011. we found that for all three case studies it was possible to obtain a measure of the ecological value . there is a urgent need to devise methodologies that can map at a landscape these measures and tools have been developed almost exclusively for use in strategic conservation planning exercises. New. of the landscapes under investigation even if for some regions it was only possible to measure a few indicators. / Biological Conservation 147 (2012) 3–12 11 between predicted distributions and actual occurrences of the species can only be properly detected by ground-truthing the data.. Moilanen.R. 2. 2009. M.1016/j. The next steps in the LEFT development are to automate the methods described herein.. To provide additional information for the Environmental Impact Assessment. Brost.e. Possingham. H. their requirement for very high quality species/environmental data often means that they are of extremely limited use in data poor regions.001. Haslem. Bennett.N. G. Marxan and relatives Software for spatial conservation optimization. since 2005).. 7. Quinn. H. migratory.P. Recently several methodologies have also been proposed for the conservation of landscape features important to maintaining ecosystem processes (i.. Supplementary data Supplementary data associated with this article can be found. However. Ecological Management & Restoration 10.. M..F. C. Furthermore. Some of the information obtained by this tool will not be routinely collected as part of the EIA and can therefore be used to provide additional sources of ecological information for a chosen development area.. The tool would therefore be run before the EIA and be used as a guide to appropriate areas for more detailed on-ground surveys. Lunt. Conservation Biology 24. 192–199. Reserves... Possingham... 2009. (Eds. Elmqvist.E.e. Such a trade-off approach is starting to be considered for some site facility location projects.). in the online version.. 2007). something that is of much higher consideration outside of protected areas. 389–396. To make full use of this approach for such regions may require accessing additional sources of remotely sensed data. To provide the first stage of assessment in the determination of areas for Biodiversity Offsetting projects (Business and Biodiversity Offsets programme (BBOP. for example. have) been used for systematic planning outside of protected areas.G. 2007)... Willis et al. J. M. L.. T. U.. T. 2010. Spatial Conservation Prioritization Quantitative Methods and Computational Tools.K.Q.11. The GLOBCOVER information has a known accuracy level of 79.... which will perform the analysis for any location in the world within a web-based tool and end-user platform. Russia..R. Bengtsson. In addition. Pressey et al. the important ecological properties and functions of landscapes.g. future work plans to include a comparative analysis of the output from the approach described herein with the output from high-resolution. Marxan. It is also widely acknowledged that before this can happen. Watson. (Klein et al.W. Oxford University Press. R. References Ball.. In: Moilanen. Wilson. 701–710. Lake. J. less attention is given to the ecological versus economic tradeoffs. P.. it is inevitable that economic/ecological trade-offs will need to be considered more often (de Groot et al. Jones. 2011) ’’a gap in market infrastructure that persists is lack of landscape-scale ecological monitoring – while site-level ecological monitoring is not uncommon..P. D. Angelstam. since the focus of such approaches is on finding the best location for conservation. To assess long-term ecological impact of a development: therefore to assess ecological value of the landscape before. Folke. Ball et al. Koehn. 2003. J...R.. J. van Haaren and Fthenakis (2011) have developed a GIS-based site selection tool for locating windturbines on the landscape such that important bird areas are avoided... D. A. As recently stated in a review on the state of Biodiversity markets (Madsen et al. However.forest-trends. 2005). Yen. T. 3. K.. In order to obtain a quantitative measure of uncertainty. The uncertainty associated with the GBIF points is related to the correct estimation of coordinates from the original source.. resilience and dynamic landscapes. this methodology is such that additional data can be incorporated if and when they are available. G. A.. 4. spatial decision support tools are now being developed to inform landscape-scale forest restoration such that multiple landscape functions (ecological and economic) are maximised (Gimona & van der Horst. during and post-development to determine ecological improvement or loss over time when combined with field-based. Ambio 32. Appendix A. Ihse. (Edwards et al. Use of this methodology to analyse the three study areas inevitably highlighted significant differences in data availability across the globe despite our use of globally available datasets. 2009. 2009)).. and often for landscapes where detailed data on species is far from complete. P. Additionally. Ecological processes: a key element in strategies for nature conservation. Cheal. The approach we have presented in this paper is a first attempt at providing a framework to do this – and by its use of a range of ecological indicators. Watts. 2010).25% (Bicheron et al. anywhere in the world. With an increasing requirement to map and value landscapes according to their ecosystem service provision. B. In particular there was a general lack of freely available remotely sensed environmental data in high latitude regions such as in the Yamal Peninsula. Lumsden. therefore. it is essential to ground-truth these species data where possible (Chapman.A. Beier. Radford. http://bbop. is provides a way of visualising the ecological value of the whole EIA study area. Moberg.C. there is an extensive literature and an excellent number of tools now available to measure ecological value for the selection of priority areas for conservation (e.. Mac Nally. The other question that needs to be addressed is how this new approach aligns with the other methodologies currently available. 2010) and evolutionary.P. Menkhorst. Robinson. Clarke.S. Nystrom. ground-truthed datasets. M. Similarly in restoration projects. I.. not the actors. Newell. We envisage that this methodology has potential for use in four stages of planning and facility location in landscapes beyond protected areas as follows: 1.. F.. O’Hara.F.. P. What this methodology has also demonstrated is that there is a vast amount of global data available on the web to provide enough information to demonstrate differences in ecological value across the landscape for even the remotest of study areas. 2009.. Emanuelsson.F. A. the data is not easily available.. we are undertaking a ground-truthing and uncertainty estimation exercise by comparing the results of our LEFT methodology with those obtained from high-resolution field datasets.

12 K.. 2004.html>. G. A.. Washington.. B. Paleobiology Database. R.. CBS fungi strains. Annual Reviews of Ecology. K. O. ZFMK Orthopteroidea collection. Terrestrialecoregions of the world: a new map of life on earth. 2010–7-25. Elmqvist. Achard. 252–264. USA. A.0. Bonn.. Norbert Kondla Collection. Shades of irreplaceability – towards a measure of the contribution of sites to a reservation goal. H. J. 1–22. Federal Agency for Nature University of New Brunswick Collection..09. DC: Forest Trends. Endo. Allnutt.L.protectedplanet. Protected Area Data were Obtained from UNEP-WCMC World Database on Protected Areas. NBM birds... Peabody Mammalogy DiGIR Service. EURISCO. Bergen (BG). 2004. Mammal Collection Catalog. Hooper Butterfly collection. / Biological Conservation 147 (2012) 3–12 Phanerogamic Botanical Collections (S). 1991. Possingham. xviii..gbif. 2007. University of Saskatchewan... M.W. Elliott. Vascular Plant Herbarium (MA). J. I. CSU Herbarium. Hedao. Holling. Edwards.... S.J. N. Vascular Plant Collection – University of Washington Herbarium (WTU). Ranera. Cambridge. Zoological Museum Amsterdam..ecosystemmarketplace. Layberry Observations. Scheffer. Ricketts.J. Peabody Invertebrate Paleontology DiGIR Service.L. Biodiversity and Conservation 3. H. MSB Mammals Specimens... Braat.E. Chessel. International Forestry Review 10. E.F. Loucks.. Oxford.. 1–23. Manion. Alberta. van Breugel. Richardson. 206– 217. de Groot.H.. E. Mackey.R. Stein. Department of Environmental Biology. Gunderson.. Valencia. A review of criteria for evaluating natural areas. EOS Transactions.. 2002. Evolution and Systematics 34. The C-Plan conservation planning system: origins. 2009.A.E.1_GEOTIFF_1km/. Expansion of the global terrestrial protected area system. M.. 403. Kirkby.S. J. Missouri Botanical Garden. 2003.. Ecological Applications 19. Mapping hotspots of multiple landscape functions: a case study on farmland afforestation in Scotland. Canada. E. Accessed through GBIF Data Portal. R. resilience and biodiversity in ecosystem management..L. Arino. R. <http:// www. Challenges in integrating the concept of ecosystem services and values in landscape planning. 2009.J.. 2027–2036. G.. Vane Wright. 2011. Watson. Cambridge University Press. Barnett. K. Gerald Hilchie Collection. N. S. S. C. Hausser. P. IUCN. Usefulness of species range polygons for predicting local primate occurrences in southeastern Peru. 604–612. M. Update: State of Biodiversity Markets. Donald F.. M. A. I.. B. Itoua. D. Global Register of Migratory Species: Database. Edmonton. 403.. L. Spatial modelling of biodiversity at the community level. (accessed 24. Andelman. <http:// www. Gollop Collection. The European Genetic Resources Search Catalogue.G.. Evolution and Systematics 4. Beyond reserves: a research agenda for conserving biodiversity in humanmodified Ttopicallandscapes. GBIF: Museum of Zoology. Morandin PhD Thesis/La Crete.. Environmental Management 10. C. Acquired for year 2005. Herbarium GZU. F. Powell. 2011. management and decision making..M. C. Germany. J. 2011. Smith. Incorporating ontogenetic dispersal. S. Looking beyond protected area networks: a paradigm shift in approach for biodiversity conservation.D. Resilience and stability of ecological systems. 2005. p. Halifax.. Saskatoon. Kunz. Ecological Modelling 157. MCZ Ornithology Collection. Verdin. Zoology (Museum of Evolution – Uppsala).. CONN GBIF data. Hijmans. S. Birds (Aves).R. Ridges. Staatliches Museum für Naturkunde Stuttgart. 260–272. Therevid PEET Project.. S. Regan.10). Yu. Lichen herbarium.. Bontemps. D. E.. FMNH Mammals Collections. In: Proceedings of the Recent Advances in Quantitative Remote Sensing Symposium. AB. Huc..E. Chapman. H.M. Canada.. M.. 2010... In: Moilanen... International Journal of Climatology 25. biodiversity declines. MartinezRamos. What to protect – systematics and the agony of choice. Y. 242–262. EUNIS. AGU 89. Using generalized dissimilarity modelling to analyse and predict patterns of beta diversity in regional biodiversity assessment. Incorporating ecological and evolutionary processes into continental-scale conservation planning. Underwood. Animal Sound Archive Berlin.L. Ross A. <http://www. ftp:// ftp.umass. A.. R. Computer Software Program Produced by the Authors at the University of Massachusetts. M. H. Amherst. 2002. 457–470. Wilson. L.C. 93–94. Lund Botanical Museum (LD).. Renewable &Sustainable Energy Reviews 15. Sinha. University of Alberta Ornithology Collection. N. FRAGSTATS: Spatial Pattern Analysis Program for Categorical Maps. Optimizing resiliency of reserve networks to climate change: multispecies conservation planning in the Pacific Northwest. 1272–1273. www. 53–61. NSW herbarium collection. A. Davis.H.07. Lehner. Invertebrates (GBIF-SE:SMNH). Tjørve... Evolution and Systematics 35. benefit functions and target-based planning: unifying reserve selection strategies. New global hydrography derived from spaceborne elevation data. Australian National Herbarium (CANB). S.C. applications and possible futures. J. D.W. R. (downloaded 21. 2005.. Project FeederWatch.G.. G. Extended statistical approaches to modelling spatial pattern in biodiversity in northeast New South Wales II. Mumby. Wikramanayake. 2008. Jarvis. Running. Herbarium (UNA). A. 437. Jardi Botanic de Valencia: VAL.... Peres. African Rodentia. Burgess....C. Biological Conservation 143. Carpenter. Jones.. Pressey. Smith. L. 2309–2338. eBird. Defourny. Ferrier.. R. Canadian Lakes Loon Survey. Brockmann.I. 2010. C. 571–579. MOD17A3. GIS Maps and Threat Analysis. Predicting species spatial distributions using presence-only data: a case study of native New Zealand ferns. NMNH Vertebrate Zoology Mammals Collections.. Herbarium WU.. 235–254. Provincial Museum of Alberta. Cushman. 715–734.N. K. Canadian Museum of Nature Bird Collection.. J.S. version 1. L.. P. Western Palearctic migrants in continental Africa. Canadian Museum of Nature Herbarium. G. Joppa. I.N. 3332–3340. (Eds. Watts.. Stokstad..A.. GLOBCOVER: A 300 m global land cover product for 2005 using ENVISAT/MERIS time series. Biodiversity and Conservation 11.. UAM Mammals Specimens. A. Berry. Herbarium of Oskarshamn (OHN).. Jenkins. Schouten.. Agriculture and Agri-Food Canada. 2715–2728. Caroll. 1255–1264.. 487–515. Illinois Natural History Survey. R.. C. Biotropica 41. K.R. Guisan. (accessed 10. Lamoreux. S.A. NMNH Invertebrate Zoology Collections.. Lichen herbarium.. B. J.305/ Improved_MOD17A3_C5.A. Copenhagen..). NMNH Botany Collections.. NSW herbarium collection.W. Possingham. Fthenakis. Bioscience 51. Bird specimens. G. J.. C.. Leroy.. Nigh. Final Report of the R&D-Projekt. V.. 891–904.. University of Amsterdam (NL) –>. Harvey. W. Biological Conservation 142. C. Despite progress. Wilson.. Hirzel.M. Widyanata. How to resolve the SLOSS debate: lessons from species-diversity models.. Science 329. 2009.. Biodiversity in Environmental Assessment : Enhancing Ecosystem Services for Human Well-being.. P.. L.. B. pp. P. University of Western Ontario Collection. Manitoba Museum of Man and Nature. 2007. Cameron.J. T. D. Provincial Museum of Alberta. Landscape Ecology 22. W..Primary species and species occurrence Olson. S.S.. F..A...J.. Freese. McGarigal. F.1. Perrin. Pressey.. Version 2009. Ferrier.. MVZ Mammals Specimens.L. M. A.. Dö MEXU/Colección de Briofitas. Riede. M.. Journal of Hydrology 296. Humphries..M. University of Guelph.G. Canadian Museum of Nature – Fish Collection (OBIS Canada). T.H.05 MODIS Annual Net Primary Productivity at 0.0083 resolution. D’Amico. J. 2166–2174. B. M.. Vancutsem. Hein. C.J. M. Arizona State University Lichen Collection. C. Walker. Annual Review Ecology.11). CGN-PGR.. Lichen Herbarium Berlin... Lund Botanical Museum (LD). 2006. American Journal of Primatology 73.. Zaniewski. Diversity and Distributions 13.. Ecological Complexity 7. Wettengel.D. United States National Plant Germplasm System Collection. Biologiezentrum Linz. University of Alberta Herpetology Collection. Very high resolution interpolated climate surfaces for global land areas. Federal Agency for Nature Conservation.. New Brunswick Museum Collection. NBM Unionoids. Carwardine. Canada.. van Haaren. Journal of Applied Ecology 43. 261–280. Australian Antarctic Division Herbarium. Neel.>. Peabody Paleoportal DiGIR Service (IP). Herold... C.. K.. 2009. Moilanen. 933–938.. Willis et al. Harvard University Herbaria. Community-level modelling. Journal of Theoretical Biology 264.. G. Regime shifts. 1994.P. Ferguson. Dinerstein.10).A. University of Alberta Mammalogy Collection. Development and validation of a global database of lakes.. NS. Peter Hall Observations. Comprehensive criteria for biodiversity evaluation in conservation planning. Biodiversity and Conservation 16.03.. University of Navarra. Spatial Conservation Prioritization: Quantitative Methods and Computational Tools. R. H. P. M.J. L. 2010. HMAP-History pf Marine Animal Populations (CoML). Klein. 2009..C.. Holling. J. R.. . 1973. 305–314. Global Register of Migratory Species – from Global to Regional Scales. R. Biological Conservation 55. C. Fahrig. Lehner. Santa Barbara Musem of Natural History. Royal Saskatchewan Museum Collection. Kandy. Bundesamt fèur Naturschutz.W. Mathur.. UAM Botany Specimens.M. Soto-Pinto.. Jarvis. Ecological-niche factor analysis: how to compute habitat-suitability maps without absence data? Ecology 83.. Senckenberg – CeDAMar Resource. The Collection of Lichenicolous Fungi at the Botanische Staatssammlung München.ntsg. C.. Bennett.F. Williams. Great Backyard Bird Count. FrogWatch Canada. Folke. A. S. 2009. Parra. p. E. The Collection of Lichenicolous Fungi at the Botanische Staatssammlung München. Alkemade. K... P. Gimona. 2004.. Theberge. T. 2009.. E. ecological processes and conservation zoning into reserve design. Komar. C. Overton. T.. Morales. M. A. Watts.... 142–153. Effects of habitat fragmentation on biodiveristy... Nova Scotia Museum of Natural History. K..R. P.. Manion. Drielsma. Strand. Nemani.. Biological Conservation 134. Ichtyologie. Arizona State University Lichen Collection. Chazdon. Johnson. 2002. J. 1965–1978. Palminteri. Slootweg. Carroll.. Zhao. Philpott.. van der Horst. O.S. Wilson. M. 2010. Oslo (O). Kura. A. Moilanen. SysTax. GIS-based wind farm site selection using spatial multi-criteria analysis (SMCA): evaluating the case for New York State. L. R. Landscape Zonation. Botany (UPS). Canadian Museum of Nature Amphibian and Reptile Collection Anura.. D. Botany Vascular Plant Collection. K. Pressey. B.. Fairchild Tropical Botanic Garden Virtual Herbarium Darwin Core format. P. 2007.. 2011. Bicheron.V.E. H. 2001.D. 2004. Annual Review of Ecology. MAL. Ene. Kassem. SABIF Resource. P. M.K.. Botanic Garden of Finnish Museum of Natural History. A. 393–404. North West Territories and Nunavut Bird Checklist. Principles and methods of data cleaning . reservoirs and wetlands. K. Real Jardin Botanico (Madrid). Atlantic Reference Centre (OBIS Canada). D. Elith. Global Change Biology 16. Dunk. 2009. Griffith. P..R. Lehmann. 557–581. R. S.. Morrison. 2010. Willemen. L. 2007. DBL Life. USU-UTC Specimen Database. 1986.... S.J. P... 2002. Ferrier. Red List of Threatened Species..umt. 2008. Bonn. Riede. Oxford University Press.B. Royal British Columbia Museum. Powell. Mammals (NRM). Phragmites of Canada. Madsen.