Using individual-based movement models to assess inter-patch connectivity for large carnivores in fragmented landscapes

Biological Conservation 167 (2013) 298–309

Contents lists available at ScienceDirect

Biological Conservation journal homepage: www.elsevier.com/locate/biocon

Using individual-based movement models to assess inter-patch connectivity for large carnivores in fragmented landscapes Rajapandian Kanagaraj a,b,⇑, Thorsten Wiegand a, Stephanie Kramer-Schadt a,c, Surendra P. Goyal b a

UFZ, Helmholtz Centre for Environmental Research – UFZ, Department of Ecological Modelling, Permoserstr. 15, D-04318 Leipzig, Germany Wildlife Institute of India, Post Box 18, Chandrabani, Dehradun 248001, India c Leibniz Institute for Zoo and Wildlife Research, Alfred-Kowalke-Strasse 17, D-10315 Berlin, Germany b

a r t i c l e

i n f o

Article history: Received 20 March 2013 Received in revised form 15 August 2013 Accepted 19 August 2013

Keywords: Carnivore conservation Fragmented landscapes Inter-patch connectivity Movement model Panthera tigris Spatially explicit individual-based model Terai Arc Landscape Tiger

a b s t r a c t Most rare and endangered large carnivores such as tiger (Panthera tigris) exist in human-dominated landscapes as small, fragmented and isolated populations across their range. Connectivity between the remaining populations in the habitat fragments is essential for their long-term persistence and focus of management initiatives. We describe an individual-based, spatially explicit model of tiger movement behavior based on previously developed habitat models to (i) quantify inter-patch connectivity among major (protected) habitat patches in the Terai Arc Landscape of India and Nepal and (ii) investigate the effect of potential management initiatives, e.g. restoring corridors, on enhancing connectivity among fragmented protected habitats. Connectivity was not solely a function of distance between patches, but an outcome of the interplay between movement behavior and landscape composition, with asymmetric connectivity explained by canalizing or diffusing effects of the landscape, and depending on the landscape context of the starting patch. Patch connectivity was mostly determined by autocorrelation in tiger movement, the daily movement capacity, landscape structure, and the amount of matrix habitat. Several habitat patches were likely to be island-like and already effectively isolated. However, simulating scenarios of corridor restoration showed that most habitat patches in India and between India and Nepal could recover connectivity, which may mitigate negative genetic consequences of small population size and effective isolation on tiger populations in this landscape. Combining habitat models with individualbased models is a powerful and robust approach that could be widely applied to delineate dispersal corridors of large carnivores and quantify patch connectivity even if data are scarce. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction Studies on habitat connectivity have become a central issue in conservation biology and are of vital importance to the conservation of threatened species world-wide especially in fragmented landscapes (Crooks and Sanjayan, 2006; Revilla and Wiegand, 2008; Simberloff, 1988). Landscape or structural connectivity is defined as ‘‘the degree to which the landscape facilitates or impedes movement among resource patches’’ (Taylor et al., 1993), however ‘‘without any requisite reference to the movement of organisms or processes across the landscape’’ (Crooks and Sanjayan, 2006). Depending on the spatial scale and the management question, con-

Abbreviation: TAL, Terai Arc Landscape.

⇑ Corresponding author at: UFZ, Helmholtz Centre for Environmental Research – UFZ, Department of Ecological Modelling, Permoserstr. 15, D-04318 Leipzig, Germany. Tel.: +49 341 235 1717; fax: +49 341 235 1473. E-mail addresses: [email protected] (R. Kanagaraj), thorsten. [email protected] (T. Wiegand), [email protected] (S. Kramer-Schadt), goyalsp@ wii.gov.in (S.P. Goyal). 0006-3207/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biocon.2013.08.030

nectivity may be assessed with regard to the entire landscape as typically done in landscape ecology (e.g., Tischendorf and Fahrig, 2000), or with regard to specific patches (i.e., ‘‘inter-patch connectivity’’) in metapopulation studies (e.g., Moilanen and Hanski, 2001). However, movement or dispersal success and, therefore, functional connectivity depends on both, the spatial structure of the landscape and the behavior of the dispersing species in response to landscape heterogeneity (Revilla et al., 2004; Kramer-Schadt et al., 2011). An assessment of dispersal success is especially complicated in intensively used landscapes due to movement barriers imposed by humans (Graf et al., 2007; Kramer-Schadt et al., 2004). Additionally, field studies on dispersal are very time consuming and expensive, especially for large carnivores because of high tracking-costs of individual animals. As a result, our current understanding on movement behavior of such species is limited and alternative approaches are required to complement the assessment of connectivity (Graf et al., 2007; Revilla et al., 2004; Zollner and Lima, 1999).

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

One approach to estimate inter-patch connectivity is to use models. Depending on the landscape structure, the scientific question and the organism of interest, several approximations to this complex problem have been proposed. For example, the incidence metapopulation model (Hanski, 1994; Moilanen and Nieminen, 2002) describes connectivity between two patches as a declining function of distance between the patches without taking into account details of landscape structure. Similar simplifying assumptions are made in graph-based landscape connectivity indices (e.g., Keitt et al., 1997; O’Brien et al., 2006; Pascual-Hortal and Saura, 2006; Urban and Keitt, 2001; but see Saura and Rubio, 2010). In contrast, least-cost path analysis explicitly considers the impact of landscape structure to find the optimal movement path between two patches that minimizes a given cost criterion (e.g., Adriaensen et al., 2003; Gonzales and Gergel, 2007; Klar et al., 2012; Nikolakaki, 2004; Wikramanayake et al., 2004). Friction values that represent the resistance to movement through different landscape elements (i.e., the cost) implicitly represent behavioral decisions regarding movement through particular landscape features (Schadt et al., 2002). However, this method cannot directly include dispersal behavior and is only able to assess structural connectivity and therefore often lacks biological realism (Calabrese and Fagan, 2004; Crooks and Sanjayan, 2006). Although least-cost path analysis can identify potential corridors, additional information on the movement behavior and dispersal ability of the species is required to assess if the identified corridors provide indeed functional connectivity, and if the animals may actually find them. Behavior and the landscape context of the start patch become especially important in complex landscapes comprising for example narrow passages of dispersal habitat and dead ends. In this case asymmetrical inter-patch connectivity is likely to occur (Ferreras, 2001; Gustafson and Gardner, 1996; Revilla et al., 2004; Schippers et al., 1996) because the landscape structure surrounding the start patch can have both canalizing and diffusing effects on movement. Thus, assessment of functional connectivity that considers the movement capacity and the behavioral response of the target species to the physical landscape structure (i.e. spatial information about habitats or landscape elements) (Crooks and Sanjayan, 2006) is required for planning conservation efforts in complex fragmented landscapes. Individual-based spatially explicit simulation models (Dunning et al., 1995; Grimm and Railsback, 2005; Revilla and Wiegand, 2008; Wiegand et al., 2004b) overcome the limitations of landscape connectivity indices and cost-path analysis. They simulate dispersal explicitly and behavioral movement rules describe how organisms interact with landscape structure; this type of models is therefore especially suitable for evaluation of dispersal success and connectivity between specific habitat patches in situations where details of landscape structure and behavior matter (Kramer-Schadt et al., 2011; Nathan et al., 2008; Schick et al., 2008; Tracey, 2006). This type of model has been successfully used in several studies on animals and birds (e.g., Iberian lynx (Lynx pardinus; Revilla et al., 2004; Revilla and Wiegand, 2008), Eurasian lynx (Lynx lynx; KramerSchadt et al., 2004), capercaillie (Tetrao urogallus; Graf et al., 2007), red-cockaded woodpecker (Picoides borealis; Bruggeman et al., 2010) and tortoise (Testudo graeca; Anadón et al., 2012) to analyze dispersal behavior and/or estimate connectivity between habitat patches. Large carnivores are particularly vulnerable to extinction in fragmented landscapes because of their low population density, wide ranges, low fecundity, and direct persecution by humans (Dinerstein et al., 2007; Noss et al., 1996). A typical example is the fragmented populations of tiger (Panthera tigris) that exist in the Terai Arc Landscape (TAL), which consists of twelve pro-

299

tected areas and covers ca 78,000 km2 area in the Himalayan foothills in India and Nepal (Dinerstein et al., 2006). The TAL is one of the top priority landscapes for tiger conservation (Sanderson et al., 2006) that was once continuous across the Himalayan foothills but is now highly fragmented and most of the remaining large, intact habitats are located within protected areas (Wikramanayake et al., 2004). As tigers cannot sustain viable populations in small habitat fragments (Johnsingh and Negi, 1998; Sanderson et al., 2006) a conservation project was initiated in the TAL by the World Wildlife Fund that implemented the concept of metapopulation management to restore, reconnect, and manage wildlife corridors to link 11 important protected areas that harbor wild tigers (Dinerstein et al., 2007; Smith et al., 1998; Wikramanayake et al., 2004). Consequently, potential connectivity among habitat patches was assessed based on a least-cost pathway model (Wikramanayake et al., 2004). However, least-cost analyses cannot assess functional connectivity, and hence this study could not establish a quantitative measure of potential corridors (links) that is an important property of effective conservation methods (Jordán, 2003). Here, we provide the next step required for corridor assessment in the TAL using a dynamic individual-based simulation model that incorporates behavioral details of movement within real landscapes. More specifically, we present a simple spatially explicit and individual-based dispersal model to (i) quantify the inter-patch connectivity among the major (protected) habitat patches in this heterogeneous landscape and (ii) investigate the effect of potential management initiatives, by restoring corridors, on enhancing connectivity among fragmented protected habitats. Previous studies stressed the importance of these corridors for maintaining landscape-level connectivity, but also highlighted the uncertainty surrounding successful usage of these corridors by tigers (Johnsingh et al., 2004; Wikramanayake et al., 2004). This exercise was motivated by two purposes: to assess the consequences of our uncertainty about the movement and habitat use of tigers for predicting patch connectivity and to test the effectiveness of potential landscape restoration measures by providing undisturbed corridors for tiger. To overcome the problem of uncertainty arising from scarce data in parameterizing the dispersal model, which is common in endangered species (Kramer-Schadt et al., 2007; Wiegand et al., 2003, 2004b), we conducted exhaustive sensitivity analyses. Finally, we discuss our results in respect of tiger management in the TAL.

2. Materials and methods 2.1. The habitat map We used probabilistic habitat suitability (HS) maps with a cell size of 500 m  500 m derived for tiger in the TAL by logistic regression and ecological niche factor analysis as described in Kanagaraj et al. (2011); their Fig. 3) (see also Appendix A and Table A1). We divided the TAL into four functional habitat types: breeding habitat, dispersal habitat, matrix and barrier (e.g., Kramer-Schadt et al., 2004; Revilla et al., 2004; Revilla and Wiegand, 2008). In our model, the movement decisions of tigers depended directly on these four categories (see Section 2.3). The four habitat types were defined by three threshold values 0.9, 0.5 and 0.01 dividing the probability-of-use given by the logistic regression equation into four classes (Appendix A). Because the predicted probability of occurrences in our habitat map (hereafter ‘landscape map’ l) was an almost binary function with either a high (>0.9) or a low (<0.25) probability of tiger occurrence (Fig. 3 in Kanagaraj et al., 2011), we only changed the central threshold of 0.5 in our original

300

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

landscape map (l = I; Fig. A1a) and created two additional landscape maps (l = II, III, Figs. A1b and A1c) that enabled us to assess the effect of changes in available matrix and dispersal habitat on tiger dispersal. We repeated all analyses for these additional maps, which used the following cut-off values: (0.9, 0.75, 0.01; landscape map II; Fig. A1b) and (0.9, 0.25, 0.01; landscape map III; Fig. A1c). Note that we used a low threshold (0.01) for matrix and included with this selection degraded habitats that may only occasionally be used for dispersal and have therefore a crucial role for connectivity (Revilla et al., 2004). As a consequence, dispersal habitat and matrix showed a much lower habitat suitability score in our landscapes than breeding habitat. This recognizes that corridors are not necessarily composed of habitats suitable for territories of resident tigers. We derived an additional habitat quality map (hq; IV, Fig. A2) which describes human disturbances (Table A1). We used this map to determine stochastic mortality during dispersal (see ‘‘Mortality during dispersal’’ in Appendix C), which, together with landscape structure and movement behavior, is crucially determining functional connectivity. Twelve source and target patches for the connectivity analysis were defined (Fig. 1), based on the distribution of breeding habitats and their protected area status in the TAL (Johnsingh et al., 2004; Wikramanayake et al., 2004). Here, we studied the connectivity among 10 important target patches (Fig. 1): Rajaji National Park (NP) west (2), Chilla range of Rajaji NP east (3), Corbett Tiger Reserve (TR) (4), Pilibhit Forest Division (FD) (5), Suklaphanta Wildlife Reserve (WLR) (6), Basanta forest block I (7), Dudhwa NP (8), Basanta forest block II (9), Katerniaghat Wildlife Sanctuary (WLS) (11) and Bardia NP (12).

2.2. Dispersal corridor restoration scenarios 2.2.1. Scenario 1: corridor creation based on natural vegetation Dispersing tigers avoid cultivated areas, but may use degraded forest habitats that suffer from high human disturbances (Smith, 1993). However, clear evidence for tiger usage of corridors composed of degraded forests is lacking (Johnsingh et al., 2004; Smith et al., 2003; Wikramanayake et al., 2004). To assess the effect of restoration of degraded forest habitats to dispersal habitat we manipulated the landscape structure of five important potential corridors in the model. We selected for this purpose corridors that would allow maintenance of a metapopulation structure between India and Nepal (Johnsingh et al., 2004; Wikramanayake et al., 2004). To implement landscape restoration in the model, we consulted the satellite images, which were used to develop the habitat suitability map (Kanagaraj et al., 2011), and classified all cells having natural vegetation located between the two target patches as dispersal habitat. This landscape alteration created corridors that were composed of narrow forests (<2 km width; inset in Fig. 1, Fig. B1a–e). In our habitat suitability map, these habitats were largely identified as either matrix or barrier, except in case of corridor II (Chilla – Corbett TR; Fig. B1b) which comprises dispersal habitat.

2.2.2. Scenario 2: increased width of corridors Wider corridors may have the highest potential to facilitate dispersal between suitable habitat patches and could be more effective than narrow continuous conduits that connect two patches (Wikramanayake et al., 2004). We therefore increased the width of the corridors, created in scenario 1, by including a 500 m buffer

Fig. 1. Study area, spatial structure of tiger habitat, source and target patches, and corridor areas used for measuring inter-patch connectivity. The white background represents barriers, light-gray areas are matrix and medium-gray areas represent dispersal habitat. The patches studied here are selected from available suitable patches marked with dark-gray shading: 2 (Rajaji NP west), 3 (Chilla range of Rajaji NP east), 4 (Corbet TR), 5 (Pilibhit FD), 6 (Suklaphanta WLR), 7 (Basanta forest block I), 8 (Dudhwa NP), 9 (Basanta forest block II), 10 (Kishanpur WLS), 11 (Katerniaghat WLS), 12 (Bardia NP). The corridor blocks show the corridors for which we analyzed the effect of landscape restorations (i.e., corridor restoration scenarios 1 and 2). The inset shows an example for landscape restorations to create a corridor between physically nonconnected suitable patches: Chilla – Rajaji National Park west (patches 2 and 3) corridor (I). In the inset, the medium-gray areas represent dispersal habitat as well as the created corridor, with ‘‘restored’’ barrier or matrix habitat (having semi natural vegetation) into dispersal habitat (scenario 1); the black border line around the corridor of restored dispersal habitat represents the 500 m buffer area used in scenario 2 (additional to the manipulation of semi natural vegetation in scenario 1) to reach the status of dispersal habitat.

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

of dispersal habitat on either side of a corridor (see inset in Fig. 1, Fig. B1a–e left column). 2.3. Dispersal model We developed a spatially explicit and individual-based dispersal model based on simple behavioral rules that operated on an intraday time scale, where the response of the animal to the landscape matrix took place. We based these rules on published information on tiger dispersal in the TAL (e.g. Ahearn et al., 2001; Smith, 1993) and implemented them similarly to the models of other carnivores (Kramer-Schadt et al., 2004, 2011; Wiegand et al., 2004a; Revilla et al., 2004; Revilla and Wiegand, 2008). The movement during a given day consisted of a series of steps from the focal cell to one of its adjoining (eight) neighbor cells. The actual movement decision for one step, i.e. displacement to one of the eight neighboring cells, was stochastic, but depended on the habitat type of the eight neighboring cells and on the degree of autocorrelation in the movement, with a hierarchy of habitat quality over movement direction. We also applied a stochastic mortality at each movement step that depended on a habitat quality map based on human disturbances (Table A1 and Fig. A2). For details of the dispersal submodel see Appendix C. The dispersal simulation model was implemented using a software application written in Borland Delphi 5 Programming Language. 2.3.1. Model output For a given landscape map, model parameterization and start patch (i.e., patches 3, 4, 6, 7, 9, and 12 in Fig. 1), we released one tiger and simulated its dispersal for one year (i.e., 365 days), until it died or left the landscape. This simulation was repeated 5000 times to assess functional patch connectivity. For describing a single dispersal event, we recorded a number of variables. First, we recorded all patches the tiger passed during dispersal. These data are the basis for calculating the connectivity values. Next, we counted the number of times each cell was visited by the dispersing tiger, yielding a spatial probability of matrix use and dispersal habitat. We also recorded the day the tiger left the source patch, the distribution of daily Euclidean distances moved, and the habitat use, i.e. the frequency it used dispersal habitat and matrix. If the tiger survived, we recorded the real distance moved dt, i.e. the total number of steps, the Euclidian distance dE between first and last location, and the straightness of its path being dt/dE. We used the variables described above to calculate, for a given model parameter set and landscape map, several model predictions. We calculated the probability to survive dispersal and conducted a global sensitivity analysis (see Section 2.3.2) to explore the response of surviving dispersal to variation in model parameters. We also calculated the connectivity of the source patch to all other target patches being the proportion of cases where a tiger reached a target patch, and the 5%, 50%, 95%, and 100% percentile of the distribution of the daily Euclidean distance moved, the total Euclidean distance moved, the straightness of the movement path, and the day the tiger left the patch of origin. 2.3.2. Sensitivity analyses Based on a ‘‘standard parameterization’’ we varied each parameter over its entire range to explore the response in functional connectivity to variation in single parameters (Table 1) using linear regression (Fig. 2). As sensitivity coefficient we used the slope of the linear relationship between the 0–1 standardized parameter values and the non-standardized connectivity value (Wiegand et al., 2004a). In case that the connectivity showed a threshold response, we started the regression by the last zero-connectivity value (Fig. 2D and J). With this local sensitivity analysis we assessed the relative importance of the different parameters and the re-

301

sponse in connectivity to changes in the parameters, but we did not consider interactions among parameters. To consider interactions among parameters, we conducted an extensive global sensitivity analysis and explored the full parameter space of the model for those parameters that turned out to be the most important parameters in the local analysis (Wiegand et al., 2004b). We explored four values (minimum and maximum value of ranges shown in Table 1, and two intermediate values) for each of those n parameters and simulated all possible 4n combinations within the three landscape maps (l = I, II and III, Fig. A1). Because of the factorial design of the simulations, we used analysis of variance (ANOVA) to analyze the connectivity values resulted from the global sensitivity analysis, considering first-order and second-order effects.

3. Results 3.1. Single simulation runs Example simulation runs showed that landscape structure and autocorrelation in movement had a strong influence on single dispersal events (Fig. 3, top and central row). In this example, the dispersing tigers did not always find their way through the relatively narrow region of the corridor II of dispersal habitat that connects the Rajaji NP with Corbet TR (Fig. 3A, D, and G). In many cases, they were ‘‘trapped’’ in patches which have no physical connection to the target patch (e.g., Fig. 3G). Consequently, the density maps that describe the probability that a simulated tiger reaches a given cell showed a very steep decline in the narrow region of the corridor (Fig. 3C, F, and I). Based on 5000 replicate simulations in the example simulation with parameter PC = 0.9 (Fig. 3I), we calculated exemplarily the values of the different variables that characterized tiger movement (Fig. D1). The distances between tiger locations at subsequent days were below 10 km (Fig. D1a) and tigers almost exclusively used dispersal habitat (Fig. D1b). During the one year dispersal event, the maximum distance from the release point in most cases was about 100 km, but in a few cases distances of 150 km were observed (Fig. D1c), which agrees with the maximum distance observed in the wild (Sunquist et al., 1999). For this parameterization, dispersing tigers survived dispersal in 60% of all cases (Fig. D1e) and reached patches 2 and 4 (Fig. D1g). Note that the distances between subsequent days were often low because tigers got trapped in dead ends that had no connection with the target patch (e.g., Fig. 3G).

3.2. Sensitivity analyses The response in connectivity to changes in the parameter values was mostly linear (Fig. 2) and in few cases linear with a threshold (Fig. 2D and J; and Table E1). The parameter determining autocorrelation in movement PC (e.g., PC = 0 random movement, PC = 1 strong directionality) was the most sensitive parameter with increase in parameter value increased the connectivity, followed by the maximum number of daily movement steps smax and the probability of stepping into matrix Pmatrix (e.g., Pmatrix = 0 matrix avoidance, Pmatrix = 1 random use of matrix and movement habitat; Figs. 2 and 4). The model was little sensitive to parameters influencing mortality during dispersal (b and surv) and to the parameter x (exponent of power function describing daily movement distance distribution). Connectivity values were often not symmetric (Fig. 4). For patches 3 and 4, this is an effect of landscape matrix configuration, since tigers starting in patch 3 can move almost only in one direction (i.e., southeast) to reach patch 4, whereas tigers

302

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

Table 1 Parameter ranges of the dispersal submodel (formulas see Appendix C). The column ‘‘range explored’’ refers to the local sensitivity analysis. For each parameter, we explored 21 values. The value of the standard parameterization is given in bold. Symbol

Range explored

ANOVA levels and values

Exponent of power function

x

1, 1.2, ., 2, ., 5

Maximum number of movement steps during one day

smax

5, 6, ., 10, ., 25

4 (1, 2.33, 3.67, 5) 4 (5, 11, 18, 25)

Probability of stepping into matrix Probability of keeping the previous direction Annual survival probability in optimal habitat Increase in mortality with decreasing habitat quality

Pmatrix

0, 0.05, ., 0.3, ., 1 0, 0.05, ., 0.5, ., 1 0.7, 0.715, ., 0.9, ., 1 0.001, 0.00195, ., 0.005, ., 0.02

Connectivity 6-7

Connectivity 3-4

Parameter

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

PC surv b

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

4 (0, 0.33, 0.67, 1) 4 (0, 0.3, 0.6, 0.9) – 4 (0.001, 0.0073, 0.0136, 0.02)

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

Observed values in the field

We set smax to the maximum of 12.5 km. Maximum distance moved by an adult female in one day was 11.3 km in Bangladesh Sunderbans (Barlow et al., 2011). A subadult male traveled 150 km in total from Chitwan to the Trijuga-Koshi-Tappu in eastern Nepal (Sunquist et al., 1999) We covered the entire range of this parameter for the sensitivity analysis We covered the entire range of this parameter in the sensitivity analysis See Fig. C2 for range of mortality functions covered by the sensitivity analysis See Fig. C2 for range of mortality functions covered by the sensitivity analysis

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

smax

x

Pmatrix

Pc

surv

b

(A)

(B)

(C)

(D)

(E)

(F)

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

smax

x

(G)

(H)

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

Pmatrix

(I)

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0

0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03

0 0.2 0.4 0.6 0.8 1

0.02 0.01 0

0 0.2 0.4 0.6 0.8 1

Pc

surv

b

(J)

(K)

(L)

Fig. 2. Local sensitivity analysis. Examples for the connectivity between patch 3 and 4 (top) and 6 and 7 (bottom). Black dots are for landscape map I and open circles for landscape map II (see Section 2.1 in the text for more details about landscape maps). The gray lines show the linear regression. We used the slope (x-axis normalized) as index of sensitivity of a given connectivity to the parameter. See Fig. 1 for the locations of patch pairs and Table 1 and legend of Fig. 4 for details about the parameters mentioned in the x-axis.).

starting at patch 4 can move in both directions (i.e., southeast and northwest; Figs. 1 and 4). The full design for the global sensitivity analysis yielded 43,008 runs for the real landscape (Table E2) and 61,440 runs for the two corridor restoration scenarios (Table E3). The global sensitivity analysis showed that the parameter PC, the number of maximal movement steps per day (smax), landscape type (l), an interaction between PC and smax, and an interaction between PC and landscape (l) were important for patch pairs explaining between 17% and 64% of the total sum of squares (Table E2). For the corridor rehabilitation scenarios, the parameter PC was the dominating parameter, accounting for more than 50% of the sum of squares in most corridors (Table E3). A notable exception was corridor IV, where the probability of stepping into matrix Pmatrix was dominant for scenario 1 (corridor restoration). Survival during dispersal was most sensitive to the parameter b describing the increase in the daily

mortality risk with decreasing habitat quality (Table E4), followed by the parameters smax, PC and Pmatrix with the effect of increase in the value of parameters decreased the survival. Although mortality is primarily driven by the parameter b, connectivity was most sensitive to the second most important parameters (smax and Pc). This shows that considering tiger mortality during dispersal in the corridors may be important for connectivity, although not being of primary importance. 3.3. Connectivity values The global sensitivity analyses revealed that almost all uncertainty in the connectivity was controlled by the parameter PC with strong directionality in the parameter increased connectivity and three alternative landscapes in the unmodified landscapes with the former controlling the connectivity in the modified landscapes

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

Pc = 0.3

Pc = 0.5

min

303

Pc = 0.9

max

Fig. 3. Examples of single dispersal events and the resulting probability of use in landscape map I (see Section 2.1. in the text for more details about landscape map). The dispersing tiger started at patch 3 (Chilla range of Rajaji NP east), the circles in C, F, and I indicate the release point. Top row: examples of single dispersal events where tigers got trapped in dead ends. Middle row: examples for longer-distance dispersal events. Bottom row: the average probability of use after 5000 simulated dispersal events [blue: lowest density (at least used once), red: maximal observed density]. Model parameters were taken from the standard parameterization (Table 1), except smax = 25 and PC = 0.3 (low autocorrelation in movement; left column), PC = 0.5 (intermediate autocorrelation in movement; middle column), and PC = 0.9 (high autocorrelation in movement; right column).

under scenarios 1 and 2. As already observed in the local sensitivity analysis, the response of connectivity to the parameter PC is characterized by a threshold behavior (Fig. 2D and J). Autocorrelation in movement determined the maximum distance moved, and patches could only become connected if a tiger was at least occasionally able to cover the distance the two patches were apart. However, once the movement is directed enough to reach the target patch, connectivity increased monotonously with parameter PC. In most cases, the threshold value was <0.5 (Table E1), which coincides with the value in PC that the probability to return becomes low (Fig. C1b). With PC 6 0.3 the movement path is rather curvy and undirected (see Fig. 3). For straighter movement (i.e., PC = 0.6 and 0.9), connectivity in non-manipulated landscapes was low (Fig. 5a). We however find that landscape management, which restored natural vegetation to reach the status of dispersal habitat (i.e., scenario 1) was sufficient for all corridors to produce positive connectivity values (Fig. 5b). In all cases with PC = 0.6 and 0.9, connectivity was larger than 0.05 for both directions, meaning that at least one of every 20 dispersing tigers may reach the target patch (Fig. 5b). The 0.5 km buffer around the restored dispersal habitat (scenario 2) showed positive change in connectivity for corridors I, IV and V with substantially enhanced the connectivity observed for corridor IV. We observed no positive change for corridors II and III.

ment (Revilla et al., 2004; Tracey, 2006; Vuilleumier and Metzger, 2006). However, lack of data on actual dispersal events of the individuals makes it difficult to directly assess functional connectivity between patches. Even if a corridor provides a physical connection of patches (i.e., structural connectivity) it is not clear if the animals may actually cross (i.e., functional connectivity) because this depends in general in a complex way on the behavioral response of individuals to landscape elements (e.g., corridor position, width, length and distance between habitat patches), their movement capacity (Downes et al., 1997; Perault and Lomolino, 2000) and survival. Clearly, assessment of functional connectivity requires more complex models that explicitly incorporate movement behavior. We linked for this purpose an individual-based dispersal model with habitat suitability models (Kramer-Schadt et al., 2004, 2011) that define the landscape in which individuals move and conducted extensive sensitivity analyses to assess parameter uncertainty. Our study is encouraging in that it shows that this approach can make use of sparse observations of species dispersal events, produces robust estimates of patch connectivity within complex landscapes, and allows for an assessment of the effect of potential habitat restoration on connectivity. This is good news and suggests wide applicability of our approach for a variety of species with similar management problems.

4.1. Modeling issues 4. Discussion Assessment of inter-patch connectivity is currently one of the major challenges in conservation biology and landscape manage-

In our modeling approach, we constructed a relatively simple individual-based dispersal model based on published data on behavior of dispersing tiger (e.g. Smith, 1993) and other carnivores

304

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

Sensitivity index

(A) connectivity 3-4 and 4-3

0.2

0.8

0.15

0.6

0.1

0.4 0.05

0.2

0

0 -0.2

(C) connectivity 6-9 and 9-6

(B) connectivity 6-7, 7-6 and 9-12

1

s max x P matrix Pc surv b

(D) connectivity 7-9 and 9-7

-0.05

(E) connectivity 6-12 and 7-12

0.05

0.05

0.05

0.04

0.04

0.04

0.03

0.03

0.03

0.02

0.02

0.02

0.01

0.01

0.01

0

0

0

-0.01

-0.01

-0.01

s max x P matrix Pc surv b

s max x P matrix Pc surv b

Parameter

Parameter

s max x P matrix Pc surv b

s max x P matrix Pc surv b

Parameter

Fig. 4. Local sensitivity analysis. Shown are the average values of the sensitivity indices (averaged over the three landscapes (I, II and III)) for the different patch combinations indicated in the figure head, and the error bars give half of the range of each index taken over the three landscapes. Bars in each parameter in each figure represent the patch combinations and presented in the same order as do path pairs mentioned in each figure head. The two map insets show the locations of patch pairs (see Fig. 1 for more details) for which averaged values of the sensitivity indices shown here. smax: maximum number of movement steps during one day, x: exponent of power function describing daily movement distances, Pmatrix: probability of stepping into matrix, Pc: probability of keeping the previous direction, surv: annual survival probability in optimal habitat, and b: increase in mortality with decreasing habitat quality.

(Revilla et al., 2004) to address our management motivated questions concerning inter-patch connectivity in the fragmented TAL. To overcome the problem of parameter uncertainty that arises for this cryptic and endangered species, we conducted extensive sensitivity analyses to compensate for the lack of field data for parameter estimation. Note that dispersal models are only one component of a spatially-explicit population model and have, therefore, usually much less parameters. This allowed for a complete sensitivity analysis involving variation of all parameters of the model simultaneously. We combined both local and global methods for assessing the sensitivity of inter-patch connectivity to model input parameters. In unmodified landscape, the global sensitivity analysis showed that inter-patch connectivity was mostly determined by strong directionality in movement (high PC values), the three different landscape maps and the combined effect of these two parameters. This finding is in agreement with a study by Zollner and Lima (1999) where nearly straight correlated random walks provided an effective search strategy in situations similar to that found in our study. More exhaustive systematic search (i.e., lower PC values) resulted in low Euclidian distance between first and last location and individuals were often trapped in dead ends. An additional reason for the overpowering effect of the parameter PC in the corridor restoration scenarios 1 and 2 (revealed by the global sensitivity analysis) is that the corridors between the target patches were treated as dispersal habitats without human disturbance (i.e., having low mortality risk). This diminishes the effect of the two parameters ‘probability of stepping into matrix’ Pmatrix and ‘increase in mortality with decreasing habitat quality’ b. Models of individual dispersal have often used a strong directionality in the movement (e.g., Letcher et al., 1998; Schippers et al., 1996; Zollner and Lima, 1999), but Revilla et al. (2004) found a relatively low de-

gree of autocorrelation in intraday movement in lynx. Our model therefore points to a need to address this general aspect of animal movement in further field studies. 4.2. Inter-patch connectivity Our model clearly indicated that connectivity is not solely a function of distance between patches. Dispersal success is in general influenced by a combination of factors that additionally includes characteristics of the matrix in a landscape (e.g., vegetation type, structure and land use; Wiens et al., 1993) and the spatial pattern and texture of a landscape (Ferreras, 2001). We showed that spatial structures of the landscape can strongly affect the movement path of an individual (Wiens, 2001) and the ability to find a corridor. In our simulations, which are based on real landscape structure, we found that the simulated tigers may become frequently trapped in dead ends of the landscape or do not find a narrow entrance to a corridor (see also Kramer-Schadt et al., 2011). This shading effect reduces the net flux into the corridor and substantially reduced inter-patch connectivity. Transformation of former grassland or forest areas to agriculture can function as effective dispersal barriers (Smith et al., 2003) and create complex fragmented landscapes with dead ends. We observed shading effects e.g. for tigers starting from the Chilla part of Rajaji NP in the Bijnor FD dead end (corridor II, Fig. 1). When simulated tigers moved from the large Corbett TR westward (corridor II) they frequently did not find the very narrow corridor entry near Kotdwar town. Similarly, tigers starting dispersal at Pilibhit FD (corridor III) did not find Suklaphanta WLR because Sarada dam acted as dead end in Pilibhit FD and diluted the narrow entry to the restored corridor. Shading because of narrow entry to the corridor also affected the connectivity between Dudhwa NP and Basanta

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

3-4

6-7

Mean patch connectivity

(a)

305

6-12

7-9

7-12

Mean patch connectivity

Mean patch connectivity

6-9

Mean patch connectivity

9-12

Pc Fig. 5. Average connectivity and 95% CI intervals for (a) non-manipulated landscapes (I, II and III) and (b) for the scenarios with corridor restoration. Average connectivity values were calculated based on all simulations of the global sensitivity analysis for (a) landscapes without corridor restoration and for (b) landscape manipulation scenario 1 (black) and scenario 2 (blue) separately, for the four values of the parameter PC that controls the autocorrelation in the movement. To visualize asymmetric connectivity, dark circles represent the animal movement in the direction pointed out in the corridor title. Inverted open triangles represent the opposite direction. For comparative purpose we also show the results of the analyses without corridor restoration (red) in corridor II. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

and movement from Bardia NP to Katerniaghat WLS (corridors 4 and 5, respectively). We found clear evidence for asymmetrical inter-patch connectivity which has been previously observed in simulation studies (e.g., Gustafson and Gardner, 1996; Revilla et al., 2004; Schippers et al., 1996) and field studies (e.g., Ferreras, 2001). Ferreras (2001) found that habitat configuration is an important factor for connectivity between fragmented Iberian lynx populations and

responsible for asymmetrical connectivity between populations. In our study, asymmetrical inter-patch connectivity arose because landscape structure could have both canalizing and diffusing effects on movement, which depended strongly on the context of the start patch. Our results outline that the details of landscape structure, such as dead ends, island patches, or matrix and its interactions with species specific behavior may matter substantially in determining

306

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

Corridor I (2-3)

Corridor II (3-4)

Corridor III (5-6)

Corridor IV (8-9)

Mean patch connectivity

Mean patch connectivity

(b)

Mean patch connectivity

Corridor V (11-12)

Pc Fig. 5 (continued)

inter-patch connectivity and that simplifying approaches may not be able to effectively capture this complexity, potentially leading to management failures in real world conservation issues. Our results thus are in concert with other recent studies on functional connectivity based on individual-based models that showed that including the behavioral ecology of the target species and the landscape structure are imperative when assessing connectivity (Gustafson and Gardner, 1996; Gardner and Gustafson, 2004; Graf et al., 2007; Kramer-Schadt et al., 2004, 2011; Pe’er et al., 2011; Revilla et al., 2004; Revilla and Wiegand, 2008; Severns et al., 2013; Tracey, 2006; Wiegand et al., 2004b) rather than considering it purely a function of distance. However, when demographic data for the target species is available, these models can be further improved by including them (e.g., Kramer-Schadt et al., 2005). For example, Carroll and Miquelle (2006) assessed the landscape connectivity using a spatially explicit population model that linked survivorship and fecundity of tiger to landscape parameters in the Russian Far East, and Revilla and Wiegand (2008) found that the demographic state of the population strongly influences dispersal and patch connectivity. Models based on least-cost path analysis only provide an indication of pathways with the lowest relative costs without explicitly considering the behavior of the dispersing animal in the model (Gonzales and Gergel, 2007; Sawyer et al., 2011). Our study provides strong evidence that animals may not find the optimal path (or even an approximately optimal path) in complex fragmented landscapes, but may become trapped in specific landscape structures. If there is a big difference between the optimal path and alternative paths connectivity will be severely overestimated by the least cost path. Overprediction of potential connectivity, i.e.

commission errors, might have severe consequences in conservation planning as this may prevent conservation actions from showing the requested results and may lead to extinction of populations despite conservation efforts (e.g., Rondinini et al., 2006). Thus, cost path may work well in simple landscapes for nearly random walks, but may fail in more complicated landscapes like in our case with narrow passages, dead ends, etc. which are, however, the ones of interest for conservation. The use of a statistical habitat suitability map based on species occurrence data has advantages over the direct use of land cover data because it statistically translates the land cover data into a probability of use which is simpler to parameterize with respect to habitat preferences of dispersers and mortality and already represents the landscape from the viewpoint of the dispersing animal. Not surprisingly, parameters associated with habitat preference and mortality resulted in large uncertainties in individual-based simulations (Lookingbill et al., 2010; Morzillo et al., 2011). Transformation of complex landscape such as that found in our study into a patch and network representation as done in graph-based analyses is also problematic because the track of the simulated individuals depended essentially on the actual landscape configuration which is difficult to translate into a patch approximation. 4.3. Management implications Most of the remaining large patches of intact habitat in the TAL are located within protected areas because the forests outside become increasingly disturbed by human activity. Although tigers occur at relatively high densities in these protected areas, these refuges are quickly becoming insular, and indications of inbreeding

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

depression have been reported in populations which are isolated within reserves (Smith and McDougal, 1991). Patch-level effects of habitat fragmentation on population persistence may only become manifest some decades after this process started and, possibly, after a certain habitat amount threshold value is surpassed (Fahrig, 2001). Thus, it is imperative to initiate management actions before isolation critically affects the persistence of populations. The persistence of tiger populations within protected areas can be enhanced if populations in the TAL are managed as a metapopulation (Wikramanayake et al., 2004). Our analysis showed that several of these habitat patches may be island-like and already effectively isolated. This applies for patches located between Nepal and India (corridor III–V) and also to patches on the Indian side of the landscape (corridor I: Chilla Motichur area, Dudhwa NP-Kishanpur WLS (patch 10), and Corbett TR-Pilibhit FD). A landscape management in terms of corridor restoration may be a relatively cheap management action. We found that most of the patches in India and between India and Nepal could become connected under the corridor restoration scenario 1. To our surprise, an additional 0.5 km buffer around the restored dispersal habitats (i.e., habitat restoration scenario 2) resulted into a substantially enhanced connectivity only for corridor IV, rendering this management option questionable for other corridors. We however recommend adding the 0.5 km buffer around the already available habitat in the Dudhwa–Basanta corridor IV. Our results suggest that adding this buffer should enhance the suitability of this corridor and increases the connectivity between tiger populations in India and the Nepal part of the TAL. This will pave the way for successful implementation of metapopulation management concepts in this landscape (Wikramanayake et al., 2004). Connectivity between protected areas is crucial for effective and sustainable landscape level conservation. An exercise proposing a network of protected areas connected by corridors as a conservation strategy in India (Rodgers and Panwar, 1988; Sukumar, 1991) resulted in highlighting the importance of Chilla–Motichur (corridor I) within Rajaji NP and Rajaji–Corbett (corridor II) corridors for large mammal conservation in the Rajaji and Corbett NP areas (Johnsingh, 1992; Johnsingh et al., 1990; Sunderraj et al., 1995). Despite the fact that the Chilla–Motichur corridor was identified in the early 1980s, its conservation status has constantly declined over time (Johnsingh, 1992; Johnsingh et al., 1990), subsequently resulting in considerable loss of corridor area (Nandy et al., 2007). However, our model results showed that corridor restoration based on the currently remaining vegetation may allow regaining connectivity for tiger in Rajaji NP. Our results based on the current landscape predicted relatively high connectivity scores for the Rajaji–Corbett corridor. This is encouraging and the preservation of good quality habitats in this corridor may be a success of the on-going conservation efforts in this area. However, although some part of corridor areas are known to be used by tigers (Landsdown FD; Johnsingh et al., 2004; R.K. personal observation), there is no clear field evidence showing that tiger moved indeed from one protected area to another. We therefore recommend augmenting protection efforts in this important corridor area. This would create a large block of a single functioning unit (4052 km2) for tiger in the Indian side of TAL (Johnsingh et al., 2004). The three functional trans-boundary dispersal corridors, Pilibhit-Suklaphanta, Dudhwa–Basanta and Katerniaghat-Bardia, between India and Nepal are vital for creating a single landscape level functioning unit of the entire TAL (Johnsingh et al., 2004; Wikramanayake et al., 2004). These corridors connect India’s Dudhwa NP, Katerniaghat and Kishanpur WLSs and Pilibhit FD with the Nepal’s Bardia NP and Suklaphanta WLR through the Churia foothill forests. In line with results by Wikramanayake et al. (2004) we found that restoration of corridor areas among these protected

307

areas may potentially unite subpopulations. Although different habitat blocks showed connectivity under the corridor restoration scenarios, there are no protected areas in the long forest stretch between Corbett NP in India and Sulaphanta WLR in Nepal and between Bardia NP and Sukalphanta WLR in Nepal. It is recommended to create new protected areas in these regions because further human disturbances and habitat fragmentation could potentially prevent the successful dispersal of tiger (Johnsingh et al., 2004; Wikramanayake et al., 2004). 5. Conclusion Our dispersal model assessed the permeability of dispersal habitat and estimated the connectivity between pairs of habitat patches in the TAL. The approach presented that combines use of statistical habitat suitability models with individual based dispersal models can be easily extended to other species that exist in human-dominated landscapes as small, fragmented and isolated populations across its range. The parameter settings in the individual-based simulation model can be adapted to account for differences in species’ dispersal behavior (Baguette and Van Dyck, 2007). Several techniques are available to use data on species occurrences to assess habitat selection and derive habitat suitability maps, which can then be used to represent the landscape in the simulation. Our approach requires data on species dispersal behavior for parameter estimation; a common problem associated with most cryptic and endangered species. However, extensive global sensitivity analyses (involving variation of all parameters of the model simultaneously) can be used to assess the robustness of the connectivity estimates. Dispersal is only one albeit essential element of connectivity and a next step in assessment of functional connectivity is to consider the entire population dynamics since dispersal kernels and patch connectivity are critically influenced by the demographic condition of the populations which governs production of sufficient potential dispersers (Carroll, 2006; Carroll and Miquelle, 2006; Kramer-Schadt et al., 2005, 2011; Revilla and Wiegand, 2008; Tian et al., 2011). Acknowledgements This study was financed by the Save the Tiger Fund, National Fish and Wildlife Foundation, USA (Grant: 2004-0103-002) with the collaboration of Wildlife Inst. of India (WII), Dehradun. We thank the Director and Dean, WII for their support. RK was supported by a DAAD (German Academic Exchange Service) Sandwich Model Fellowship (Code-A/05/57364). SKS was supported by a Marie Curie Individual Fellowship provided by the EU (MEIF-CT2006-039985). We would like to thank E. Wikramanayake for useful comments on an earlier version of the manuscript. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.biocon.2013. 08.030. References Adriaensen, F., Chardon, J.P., De Blust, G., Swinnen, E., Villalba, S., Gulinck, H., Matthysen, E., 2003. The application of ‘least-cost’ modelling as a functional landscape model. Landscape Urban Plan. 64, 233–247. Ahearn, S.C., Smith, J.L.D., Joshi, A.R., Ding, J., 2001. TIGMOD: an individual-based spatially explicit model for simulating tiger/human interaction in multiple use forests. Ecol. Model. 140, 81–97. Anadón, J.D., Wiegand, T., Giménez, A., 2012. Individual-based movement models reveals sex-biased effects of landscape fragmentation on animal movement. Ecosphere 3. http://dx.doi.org/10.1890/ES11-00237.1, art64.

308

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309

Baguette, M., Van Dyck, H., 2007. Landscape connectivity and animal behavior: functional grain as a key determinant for dispersal. Landscape Ecol. 22, 1117– 1129. Barlow, A.C.D., Smith, J.L.D., Ahmad, I.U., Hossain, A.N.M., Rahman, M., Howlader, A., 2011. Female tiger Panthera tigris home range size in the Bangladesh Sundarbans: the value of this mangrove ecosystem for the species’ conservation. Oryx 45, 125–128. Bruggeman, D.J., Wiegand, T., Fernández, N., 2010. The relative effects of habitat loss and fragmentation on population genetic structure. Mol. Ecol. 19, 3679–3691. Calabrese, J.M., Fagan, F.F., 2004. A comparison-shopper’s guide to connectivity metrics. Front. Ecol. Environ. 2, 529–536. Carroll, C., 2006. Linking connectivity to viability: insights from spatially explicit population models of large carnivores. In: Crooks, K.R., Sanjayan, M. (Eds.), Connectivity Conservation. Cambridge University Press, pp. 369–389. Carroll, C., Miquelle, D.G., 2006. Spatial viability analysis of Amur tiger Panthera tigris altaica in the Russian Far East: the role of protected areas and landscape matrix in population persistence. J. Appl. Ecol. 43, 1056–1068. Crooks, K.R., Sanjayan, M., 2006. Connectivity conservation: maintaining connections for nature. In: Crooks, K.R., Sanjayan, M. (Eds.), Connectivity Conservation. Cambridge University Press, pp. 1–19. Dinerstein, E., Loucks, C., Heydlauff, A., Wikramanayake, E., Bryja, G., Forrest, J., Ginsberg, J., Klenzendorf, S., Leimgruber, P., O’Brien, T.G., Sanderson, E., Seidensticker, J., Songer, M., 2006. Setting priorities for the conservation and recovery of wild tigers: 2005–2015. A User’s Guide. WWF, WCS, Smithsonian, and NFWF-STF, Washington, D.C.–New York. Dinerstein, E., Loucks, C., Wikramanayake, E., Ginsberg, J., Sanderson, E., Seidensticker, J., Forrest, J., Bryja, G., Heydlauff, A., Klenzendorf, S., Leimgruber, P., Mills, J., O’Brien, T.G., Shrestha, M., Simons, R., Songer, M., 2007. The fate of wild tigers. Bioscience 57, 508–514. Downes, S.J., Handasyde, K.A., Elgar, M.A., 1997. The use of corridors by mammals in fragmented Australian eucalypt forests. Conserv. Biol. 11, 718–726. Dunning Jr., J.B., Stewart, D.J., Danielson, B.J., Noon, B.R., Root, T.L., Lamberson, R.H., Stevens, E.E., 1995. Spatially explicit population models: current forms and uses. Ecol. Appl. 5, 3–11. Fahrig, L., 2001. How much habitat is enough? Biol. Conserv. 100, 65–74. Ferreras, P., 2001. Landscape structure and asymmetrical inter-patch connectivity in a metapopulation of the endangered Iberian lynx. Biol. Conserv. 100, 125–136. Gardner, R.H., Gustafson, E.J., 2004. Simulating dispersal of reintroduces species within heterogeneous landscapes. Ecol. Model. 171, 339–358. Gonzales, E.K., Gergel, S.E., 2007. Testing assumptions of cost surface analysis: a tool for invasive species management. Landscape Ecol. 22, 1155–1168. Graf, R.F., Kramer-Schadt, S., Fernández, N., Grimm, V., 2007. What you see is where you go? Modeling dispersal in mountainous landscapes. Landscape Ecol. 22, 853–866. Grimm, V., Railsback, S.F., 2005. Individual-based Modeling and Ecology. Princeton University Press, Princeton, N.J. Gustafson, E.J., Gardner, R.H., 1996. The effect of landscape heterogeneity on the probability of patch colonization. Ecology 77, 94–107. Hanski, I., 1994. A practical model of metapopulation dynamics. J. Anim. Ecol. 63, 151–162. Johnsingh, A.J.T., 1992. Protected areas and elephant conservation in India. In: Proceedings of the Asian Elephant Specialist Group Meeting. Bogor, Indonesia, 20–22 May, 1992, Asian Elephant Conservation Centre, IUCN/SSC Asian Elephant Specialist, Group, pp. 137–147. Johnsingh, A.J.T., Negi, A.S., 1998. Only disturbance-free well-managed habitats can save the tiger. Cat News 28, 4. Johnsingh, A.J.T., Prasad, S.N., Goyal, S.P., 1990. Conservation status of the Chilla– Motichur corridor for elephant movement in Rajaji–Corbett national parks area, India. Biol. Conserv. 51, 125–138. Johnsingh, A.J.T., Ramesh, K., Qureshi, Q., David, A., Goyal, S.P., Rawat, G.S., Rajapandian, K., Prasad, S., 2004. Conservation status of tiger and associated species in the Terai Arc Landscape, India. RR-04/001, Wildlife Institute of India, Dehradun. Jordán, F., 2003. Quantifying landscape connectivity: key patches and key corridors. In: Tiezzi, E., Brebbia, C.A., Usó, J.-L. (Eds.), Ecosystems and Sustainable Development IV. WIT Press, Southampton, pp. 883–892. Kanagaraj, R., Wiegand, T., Kramer-Schadt, S., Goyal, S.P., Anwar, M., Harihar, A., 2011. Assessing habitat suitability for tiger in the fragmented Terai Arc Landscape of India and Nepal. Ecography 34, 970–981. Keitt, T.H., Urban, D.L., Milne, B.T., 1997. Detecting critical scales in fragmented landscapes. Conserv. Ecol. 1, 1–17, . Klar, N., Herrmann, M., Henning-Hahn, M., Pott-Dörfer, B., Hofer, H., Kramer-Schadt, S., 2012. Between ecological theory and planning practice: (Re-)Connection of forest patches for the Wildcat in Lower Saxony, Germany. Landscape Urban Plan. 105, 376–384. Kramer-Schadt, S., Revilla, E., Wiegand, T., Breitenmoser, U., 2004. Fragmented landscapes, road mortality and patch connectivity: modelling influences on the dispersal of Eurasian lynx. J. Appl. Ecol. 41, 711–723. Kramer-Schadt, S., Revilla, E., Wiegand, T., 2005. Lynx reintroductions in fragmented landscapes of Germany: projects with a future or misunderstood wildlife conservation? Biol. Conserv. 125, 169–182. Kramer-Schadt, S., Revilla, E., Wiegand, T., Grimm, V., 2007. Patterns for parameters in simulation models. Ecol. Model. 204, 553–556. Kramer-Schadt, S., Kaiser, T., Frank, K., Wiegand, T., 2011. Analyzing the effect of stepping stones on target patch colonisation in structured landscapes for Eurasian lynx. Landscape Ecol. 26, 501–513.

Letcher, H., Priddy, J.A., Walters, J.R., Crowder, L.B., 1998. An individual-based, spatially-explicit simulation model of the population dynamics of the endangered red-cockaded woodpecker, Picoides borealis. Biol. Conserv. 86, 1–14. Lookingbill, T.R., Gardner, R.H., Ferrari, J.R., Keller, C.E., 2010. Combining a dispersal model with network theory to assess habitat connectivity. Ecol. Appl. 20, 427– 441. Moilanen, A., Hanski, I., 2001. On the use of connectivity measures in spatial ecology. Oikos 95, 147–151. Moilanen, A., Nieminen, M., 2002. Simple connectivity measures in spatial ecology. Ecology 84, 1131–1145. Morzillo, A.T., Ferrari, J.R., Liu, J., 2011. An integration of habitat evaluation, individual based modeling, and graph theory for a potential black bear population recovery in southeastern Texas, USA. Landscape Ecol. 26, 69–81. Nandy, S., Kushwaha, S.P.S., Mukhopadhyay, S., 2007. Monitoring the Chilla– Motichur wildlife corridor using geospatial tools. J. Nat. Conserv. 15, 237–244. Nathan, R., Getz, W.M., Revilla, E., Holyoak, M., Kadmon, R., Saltz, D., Smouse, P.E., 2008. A movement ecology paradigm for unifying organismal movement research. Proc. Natl. Acad. Sci. USA 105, 19052–19059. Nikolakaki, P., 2004. A GIS site-selection process for habitat creation: estimating connectivity of habitat patches. Landscape Urban Plan. 68, 77–94. Noss, R.F., Quigley, H.B., Hornocker, M.G., Merrill, T., Paquet, P., 1996. Conservation biology and carnivore conservation. Conserv. Biol. 10, 949–963. O’Brien, D., Manseau, M., Fall, A., Fortin, M.J., 2006. Testing the importance of spatial configuration of winter habitat for woodland caribou: an application of graph theory. Biol. Conserv. 130, 70–83. Pascual-Hortal, L., Saura, S., 2006. Comparison and development of new graphbased landscape connectivity indices: towards the priorization of habitat patches and corridors for conservation. Landscape Ecol. 21, 959–967. Pe’er, G., Henle, K., Dislich, C., Frank, K., 2011. Breaking functional connectivity into components: a novel approach using an individual-based model, and first outcomes. PLoS ONE 6, e22355. Perault, D.R., Lomolino, M.V., 2000. Corridors and mammal community structure across a fragmented, old-growth forest landscape. Ecol. Monogr. 70, 401–422. Revilla, E., Wiegand, T., 2008. Individual movement behavior, matrix heterogeneity and the dynamics of spatially structured populations. Proc. Natl. Acad. Sci. USA 105, 19120–19125. Revilla, E., Wiegand, T., Palomares, F., Ferreras, P., Delibes, M., 2004. Effects of matrix heterogeneity on animal dispersal: from individual behavior to metapopulation-level parameters. Am. Nat. 164, E130–E153. Rodgers, W.A., Panwar, H.S., 1988. Planning a Wildlife Protected Area Network in India, vol. I and II. Wildlife Institute of India, Dehradun. Rondinini, C., Wilson, K.A., Boitani, L., Grantham, H., Possingham, H.P., 2006. Tradeoffs of different types of species occurrence data for use in systematic conservation planning. Ecol. Lett. 9, 1136–1145. Sanderson, E., Forrest, J., Loucks, C., Ginsberg, J., Dinnerstein, E., Seidensticker, J., Leimgruber, P., Songer, M., Heydlauff, A., O’Brien, T., Bryja, G., Klenzendorf, S., Wikramanayake, E., 2006. Setting priorities for the conservation and recovery of wild tigers: 2005–2015. The Technical Assessment. WCS, WWF, Smithsonian, and NFWF-STF, New York – Washington, D.C. Saura, S., Rubio, L., 2010. A common currency for the different ways in which patches and links can contribute to habitat availability and connectivity in the landscape. Ecography 33, 523–537. Sawyer, S.C., Epps, C.W., Brashares, J.S., 2011. Placing linkages among fragmented habitats: do least-cost models reflect how animals use landscapes? J. Appl. Ecol. 48, 668–678. Schadt, S., Knauer, F., Kaczensky, P., Revilla, E., Wiegand, T., Trepl, L., 2002. Rulebased assessment of suitable habitat and patch connectivity for the Eurasian lynx in Germany. Ecol. Appl. 12, 1469–1483. Schick, R.S., Loarie, S.R., Colchero, F., Best, B.D., Boustany, A., Conde, D.A., Halpin, P.N., Joppa, L.N., McClellan, C.M., Clark, J.S., 2008. Understanding movement data and movement processes: current and emerging directions. Ecol. Lett. 11, 1338–1350. Schippers, P., Verboom, J., Knaapen, J.P., van Apeldoorn, R.C., 1996. Dispersal and habitat connectivity in complex heterogeneous landscapes: an analysis with a GIS-based random walk model. Ecography 19, 97–106. Severns, P.M., McIntire, E.J.B., Schultz, C.B., 2013. Evaluating functional connectivity with matrix behavior uncertainty for an endangered butterfly. Landscape Ecol. 28, 559–569. Simberloff, D., 1988. The contribution of population and community biology to conservation science. Annu. Rev. Ecol. Syst. 19, 473–511. Smith, J.L.D., 1993. The role of dispersal in structuring the Chitwan tiger population. Behaviour 124, 165–195. Smith, J.L.D., McDougal, C., 1991. The contribution of variance in lifetime reproduction to effective population size in tigers. Conserv. Biol. 5, 484–490. Smith, J.L.D., Ahearn, S.C., McDougal, C., 1998. Landscape analysis of tiger distribution and habitat quality in Nepal. Conserv. Biol. 12, 1338–1346. Smith, J.L.D., Shrestha, M., Gurung, B., 2003. Tiger conservation and priority areas for ecological restoration: a landscape approach 2001–2003. The National Fish and Wildlife Foundation, USA. Sukumar, R., 1991. The Asian elephant: Ecology and management. Cambridge University Press, Cambridge, United Kingdom. Sunderraj, S.F.W., Mishra, B.K., Johnsingh, A.J.T., 1995. Elephant use of the Rajaji– Corbett corridor, northwest India. In: Daniel, J.C., Datye, H.S. (Eds.), A Week with Elephants. Oxford University Press, Bombay, India, pp. 261–269. Sunquist, M., Karanth, K.S., Sunquist, F., 1999. Ecology, behaviour and resilience of the tiger and its conservation needs. In: Seidensticker, J., Christie, S., Jackson, P.

R. Kanagaraj et al. / Biological Conservation 167 (2013) 298–309 (Eds.), 2004, Riding the Tiger: Tiger Conservation in Human Dominated Landscapes. Cambridge University Press, Cambridge, The Zoological Society of London, pp. 5–18. Taylor, P.D., Fahrig, L., Henein, K., Merriam, G., 1993. Connectivity is a vital element of landscape structure. Oikos 68, 571–573. Tian, Y., Wu, J., Smith, A.T., Wang, T., Kou, X., Ge, J., 2011. Population viability of the Siberian Tiger in a changing landscape: Going, going and gone? Ecol. Model. 222, 3166–3180. Tischendorf, L., Fahrig, L., 2000. On the usage and measurement of landscape connectivity. Oikos 90, 7–19. Tracey, J.A., 2006. Individual-based modeling as a tool for conserving connectivity. In: Crooks, K.R., Sanjayan, M. (Eds.), Connectivity Conservation. Cambridge University Press, Cambridge, United Kingdom, pp. 343–368. Urban, D., Keitt, T., 2001. Landscape connectivity: a graph theoretic perspective. Ecology 82, 1205–1218. Vuilleumier, S., Metzger, R., 2006. Animal dispersal modelling: handling landscape features and related animal choices. Ecol. Model. 190, 159– 170.

309

Wiegand, T., Jeltsch, F., Hanski, I., Grimm, V., 2003. Using pattern-oriented modeling for revealing hidden information: a key for reconciling ecological theory and application. Oikos 100, 209–222. Wiegand, T., Knauer, F., Kaczensky, P., Naves, J., 2004a. Expansion of brown bears (Ursus arctos) into the eastern Alps: a spatially explicit population model. Biodiv. Conserv. 13, 79–114. Wiegand, T., Revilla, E., Knauer, F., 2004b. Reducing uncertainty in spatially explicit population models. Biodiv. Conserv. 13, 53–78. Wiens, J.A., 2001. The landscape context of dispersal. In: Clobert J., Danchin E., Dhondt A.A., Nichols J.D. (Eds.), Dispersal: Individual, Population, and Community. Oxford University Press, pp. 96–109. Wiens, J.A., Stenseth, N.C., Van Horne, B., Ims, R.A., 1993. Ecological mechanisms and landscape ecology. Oikos 66, 369–380. Wikramanayake, E., MeKnight, M., Dinerstein, E., Joshi, A., Gurung, B., Smith, J.L.D., 2004. Designing a conservation landscape for tigers in human-dominated environments. Conserv. Biol. 18, 839–844. Zollner, P.A., Lima, S.L., 1999. Search strategies for landscape-level inter-patch movements. Ecology 80, 1019–1030.