The pace of human-induced environmental change continues to grow and threaten biodiversity across the globe. Loss and fragmentation of habitats, hastened by climate induced changes, are restricting species to small and isolated patches of habitat, where they may face a higher risk of extinction (Crooks et al. 2017). Complementary to protecting and restoring habitats across large spatial extents, maintaining and restoring connectivity between habitat patches is an effective conservation action to facilitate animal movement, gene flow, and increase the persistence of populations and species in fragmented landscapes (Crooks and Sanjayan 2006). Connectivity also influences ecological and evolutionary processes such as local adaptation (Lopez et al. 2009) and the spread of diseases (Hess 1994; Fountain-Jones et al. 2021).

While structural connectivity refers to the physical connectedness of habitat patches, functional connectivity is the species-specific degree to which a landscape facilitates or impedes the movement of individuals, their genes or propagules (Taylor et al. 2006). A high degree of functional connectivity increases the effective amount of habitat available to a species (Villard and Metzger 2014), facilitates range shifts in response to climate change (Lenoir et al. 2020), and allows individuals to serve as mobile links within ecosystems (Jeltsch et al. 2013). Hence, maintaining connectivity, for example via corridors, may effectively decrease the rate of biodiversity loss despite habitat fragmentation through the persistence of species in metapopulations (Shtilerman and Stone 2015). Consequently, the last decade has seen a substantial growth in connectivity research (Correa Ayram et al. 2016), leading to the development of a myriad of analytical approaches (Cushman et al. 2013) and an abundance in associated computational tools for connectivity modelling.

One of the main approaches in connectivity science is built upon a set of methods that aim to quantify landscape resistance, which describes the degree of willingness, physical ability, or success of an organism to cross a particular environment (Zeller et al. 2012). The environment is represented by spatial data (usually raster layers) and different elements of the landscape are assigned values indicating movement cost for the species under study, resulting in a resistance surface. Resistance surfaces are often the foundation for connectivity analyses, and once parametrized, can be used for various applications.

Due to the large variety of methods and tools, it can be very challenging to navigate and find the appropriate method and associated tools for resistance-based connectivity research. This is especially true since not all interested scientists (e.g., ecologists, geneticists, biologists) and conservation practitioners have experience with the methods required for the numerous steps needed to estimate resistance surface such as handling and manipulating spatial data, or using the relevant software. While concepts and approaches for parameterizing resistance surfaces have been summarized before (see Spear et al., (2010, 2015); Zeller et al., (2012)), the computational tools available for preparing and working with landscape resistance surfaces have not yet been reviewed.

In order to provide a guide to the available options, we reviewed tools (programs and packages) that are commonly used in connectivity studies based on resistance surfaces. Our objectives are to provide (i) an overview of the tools involving resistance surfaces for connectivity analyses, (ii) an outline of a typical workflow in resistance-based connectivity analyses, (iii) a summary of user experiences with the most frequently used tools, and (iv) key avenues and features for development of future tools that would address the needs of the scientific and conservation community. We anticipate that this overview will help both novice and experienced researchers navigate through the maze of tools, and help developers focus on those aspects or methods that render analytical tools most useful for the research and practitioner community.

Steps involved in resistance-based connectivity analyses

Connectivity analyses based on resistance surfaces typically involve three steps (Fig. 1).

Fig. 1
figure 1

Outline of workflow in resistance-based connectivity modelling along with available tools for each step in the process. Optional steps are shaded in grey and outlined by the dotted lines. Numbers in square brackets correspond to tools described in Table 1, 2 and 3. Please refer to the tables (descriptions are color-coded to match the corresponding steps in this figure) for more attributes. Supplementary material S1 provides a detailed narrative of these three steps, files in S5 present a tutorial with sample data and a typical workflow for resistance surface modeling in the R statistical computing environment, and S2 provides users with a decision tree to guide users through the entire process of working with resistance surfaces

Step 1: Preparing data for resistance surfaces.

First, researchers have to gather and process spatial environmental data that they wish to use in their analyses. This preparation step necessitates the identification of relevant environmental variables for the study species/system, and their respective data sources. Accessing such data has been reviewed in Kwok (2018). Since resistance surfaces are usually created from multiple data layers, it is often necessary to manipulate individual layers, so that they all have the same coordinate reference system, spatial extent and spatial resolution. In addition to the spatial resolution (e.g., pixel size), the thematic resolution (e.g., number of habitat classes) can greatly affect connectivity studies (Cushman and Landguth 2010; Zeller et al. 2017). Similarly, when trying to assess impacts of environmental change on connectivity, the temporal resolution of the data should capture the process of interest (e.g., seasonal dynamics or landscape before and after infrastructure development), so that changes across time are adequately represented by the data.

Step 2: Constructing and optimizing resistance surfaces.

Second, researchers have to construct the actual resistance surface by translating the information of the environmental data layers into meaningful resistance values that reflect the study species and research question of interest. This parameterization can be done via expert opinion and literature reviews, or based on empirical data collected for the focal species. There are several ways of parameterizing resistance surfaces with different data types, such as occurrence, movement, or genetic data (reviewed in Spear et al. 2010, 2015; Zeller et al. 2012; Wade et al. 2015), or a combination of different data types (e.g., Zeller et al. 2017; Meyer et al. 2020).

Resistance values can be estimated by conducting expert-opinion surveys, reviewing literature, empirical data obtained for the species, or any combination thereof. Whenever possible, it is recommended to not rely entirely on expert-opinion as the only source of information (Clevenger et al. 2002; Shirk et al. 2010; Zeller et al. 2012). When empirical data is available, it can be used to directly estimate resistance values, for example by quantifying the effect of the environment on movement probabilities using point, step- or path-selection functions from telemetry data (Zeller et al. 2012; Osipova et al. 2019; Goicolea et al. 2021). Methods for these selection functions are implemented in various R packages, including amt (Signer et al. 2019) and adehabitatLT (Calenge 2006). Another popular approach for constructing resistance surfaces from empirical data is to convert habitat suitability values into resistance values. Habitat suitability models constructed from point data (i.e., presence-absence or presence only) via resource selection functions (RSFs e.g., in R package ResourceSelection, Lele et al. 2019) or species distribution models (e.g., in R package maxent, Phillips 2021) need to be transformed into resistance values, with the general idea that lower suitability will lead to greater resistance to movement and/or gene flow. A simple linear inversion of suitability values into resistance is usually not ideal, as organisms are often able and willing to traverse sub-optimal habitats (e.g., Mateo-Sánchez et al. 2015; Keeley et al. 2017). Negative exponential relationships between habitat suitability and resistance, which can be modelled using different conversions may therefore be more accurate (see e.g., Trainor et al. 2013; Keeley et al. 2016). Thus, while habitat suitability models are an appealing way to obtain resistance values, they may not always be optimal for capturing the processes of interest in connectivity analyses (Scharf et al. 2018). However, even non-linear conversions may not adequately represent resistance to dispersal and gene flow, because organisms dispersing through the landscape can behave very differently compared to their usual within-home range behavior (Elliot et al. 2014). Therefore, data reflecting long-distance movements, exploratory behavior or actual dispersal events are often more suitable for estimating resistance, compared to data reflecting movements within home ranges (e.g., Gastón et al. 2016; Dondina et al. 2022). Genetic data is often used as a measure of functional connectivity as it represents successful reproduction post-dispersal processes. The classical use of genetic data in resistance surface modelling is through a landscape genetics framework (Manel et al. 2003), wherein estimates of gene flow are evaluated against measures of effective geographic distance under alternative resistance surfaces to find the best estimates of resistance.

Once a resistance surface is created, it is recommended to optimize the surface so that it truly represents the ecological and biological needs of the species. Empirical data can be used to both construct and optimize resistance layers. In optimization, resistance values are parameterized in a way so that they best match connectivity patterns derived from empirical data. For example, a typical approach in landscape genetics is to compare estimates of genetic connectivity (e.g., genetic distances) against estimates of functional connectivity (e.g., least-cost paths or circuit-theoretic resistances) derived from multiple resistance surfaces representing different research hypotheses. The best resistance surface is then identified as the one that leads to the highest statistical fit between the genetic and functional connectivity estimates. Optimization can be constrained when researchers explore a relatively limited parameter space, for example by assessing how well a selected set of various resistance surface realizations and their combinations match empirical data, or unconstrained where a much wider parameter space and can searched. These steps can be accomplished by different computational algorithms (e.g., Peterman 2018; Peterman and Pope 2021). Even when such optimization processes are not possible, it is highly recommended to conduct sensitivity analyses to account for the myriad sources of uncertainty in the modeling process and the propagation of errors (Zeller et al. 2012).

Regardless of how the construction and optimization are done, the final output of this second step is usually a single resistance surface that is based on all environmental variables that were identified as important in the study system for the species of interest, and parameterized to reflect the impact of the environment on the process of interest (e.g., dispersal, movement, or gene flow).

Step 3: Using resistance surfaces.

Finally, the resistance surface obtained from Step 2 can be used for a variety of purposes in landscape connectivity research and planning. For some studies, it is sufficient to optimize resistance surfaces, as this optimization in itself can tell us which environmental variables impact functional connectivity and how strongly they do so. However, resistance surfaces can also be used to delineate corridors, detect barriers, map pinch-points, prioritize linkages, compute connectivity metrics, and identify climate-resilient connectivity networks. Hence, it is this last step that makes connectivity analyses based on resistance surfaces particularly valuable for applied ecology and conservation. Many of these tools require not only a resistance surface or measures derived from it, but also additional user input, such as the spatial location, size and quality of source and target habitat patches.

More details on these three steps are provided in supplementary material S1, accompanied by a decision tree to guide users through the process of resistance surface construction in S2 and on the conservation corridor website, and a tutorial in S5 containing sample data and a typical workflow for resistance surface modeling in the R statistical computing environment.


Identification of the tools

We used our own experience to collate analytical tools that are currently being actively maintained and for which we could find functional links, i.e., that can still be downloaded and used on contemporary versions of platforms (for example, tools that require ESRI’s ArcView and have not been updated were not included in our assessment). We supplemented our experience by searching for published papers about landscape connectivity, scanned through the Corridor Conservation website (, and included suggestions offered by respondents to our online survey (see Sect. 2.2 below). We only included tools that involve a resistance surface, either as an input or an output. To help users select tools according to the resources available to them and their analytical requirements, we describe each tool based on seven characteristics: (1) the type of tool (i.e., a stand-alone program, an R package, or a GIS extension/plugin), (2) its main purpose, (3) whether it operates in a graphical user interface (GUI) and/or command line (CL), (4) the input data format required, (5) the compatibility with different operating system (OS), (6) the year of last update as of March 30, 2021, and (6) particular functions or abilities that make the tool suitable for either of the three steps (Table1).

Table 1 Summary of tools used in data preparation

Survey on user experience

To assess awareness of the tools, experience, and characteristics of tools important to the connectivity community, we conducted an online survey from June 01- July 19, 2020. The survey was conducted through the Qualtrics platform and consisted of 4 main sections. In Sect. 1, we asked respondents for their background (education and institutional affiliation), their research profile (taxonomic groups, ecosystems and geographic region of research), and self-assessment of their expertise with connectivity analysis. In Sect. 2, we provided a list of 39 tools and asked respondents to select those that they are familiar with. There was also the possibility to add additional tools not already included in the list provided. In Sect. 3, we asked respondents to rate all the tools that they had used according to 5 criteria—(1) ease of data formatting, (2) speed of analysis, (3) stability, (4) ease of customizing the tool, and (5) availability of help. Finally, in Sect. 4, we asked users to rank criteria they considered important when selecting tools to conduct connectivity analyses using resistance surfaces. In January 2021, we followed up the survey and asked participants to (1) identify existing methods or approaches for connectivity research that should be implemented in software tools, and (2) comment on future methodological and conceptual improvements that they envisage for connectivity research based on resistance surfaces.

Data analysis

We used a 3-point scale to rate respondents’ experience with the tools (poor, OK, super) or selection criteria (mandatory, important but not mandatory, not a concern). For Sect. 3 (rating user experience for different criteria), we only analyzed tools that received more than five evaluations. We calculated a score for each of the five criteria (data formatting, speed, stability, customization, help) by multiplying the percentage of responses that rated a tool as super with 1, neutral (OK) with 0.5, and poor with -1 and created a cumulative score by adding across all the criteria. We categorized tools according to the step it is useful for (preparing data, constructing resistance surfaces, and using resistance surfaces as given in Fig. 1), and ranked tools within each step by the cumulative score.


Tools identified for resistance-based analysis

We reviewed 43 tools in total, of which 10 are useful for data preparation (Table 1), 27 allow to construct and/or optimize resistance surfaces (Table 2), and 30 are suitable for connectivity analysis based on resistance surface (Table 3). All the tools we reviewed have a manual with the possibility to get help either through online forums or though the developers directly. These tools are available as R packages (n = 16), as stand-alone software (n = 16), or as GIS extensions (ArcGIS, n = 13, QGIS, n = 5), a few tools being available across more than one platform (n = 4). In addition to the tools listed in Fig. 1, we also found a few tools that were not included, either because they are no longer maintained, unavailable online (e.g., FunCon, Peer et al. 2011 and FunConn Theobald et al. 2006), no longer operational because of outdated platforms, (e.g., PATHMATRIX, Ray 2005), or are not tools per se but customized published scripts (e.g., Graves et al. 2014). The number of tools has increased sharply since 2005, when PATHMATRIX was the only tool available, to around 38 tools in early 2021, of which ~ 20 tools were added since 2014 (Fig. S3).

Table 2 Summary of tools used in constructing resistance surfaces
Table 3 Summary of tools for using resistance surfaces

Features of analytical tools

Of the ten tools we identified for the preparation step only three (export to Circuitscape, BioDispersal, and Corridor Designer Toolbox) are specifically designed for connectivity analyses. Since this preparation step involves a lot of basic spatial data manipulation, several R packages (e.g., rgdal, raster, SIMRIV), other tools (SDM and Guidos Toolbox), and GIS software (e.g., ArcGIS and QGIS) are helpful in this step.

Specialized tools for resistance-surface based connectivity analyses feature prominently in the steps for constructing, optimizing, and using resistance surfaces. Some of the tools listed in Step 2 (constructing and optimizing tools) are perhaps more popular for their application in using resistance surfaces (e.g., Linkage Mapper, Circuitscape, GFlow). We categorized these tools in Step 2 because they have functions to create outputs that can be used in the iterative optimization step where resistance surfaces are repeatedly processed through the same analytical steps several times before the final resistance surface is generated.

There are a number of tools that can use the final optimized resistance surface in Step 3. These tools cover a wide-range of applications, ranging from individual based modelling of functional connectivity (e.g., RangeShifter, UNICOR), mapping barriers to movement (e.g., HexSim, Linkage Mapper), to visualization of graph metrics (e.g., CONEFOR Sensinode, Graphab).

Several tools can serve more than one Step. Tables 1, 2 and 3 provide a summary of some properties and identify key (non-exhaustive) functions of how tools can be used in the different steps.

Survey profile

We received a total of 148 responses to the online survey, of which 120 people completed the entire survey, 134 people completed up to Sect. 2 on familiarity with tools, and 123 completed up to Sect. 3 on rating tools they have used. A majority of survey respondents were researchers (professors or postdocs; 69%) and students (16%), and affiliated with universities (52%), non-governmental organizations (16%), research or government organizations (13% and 12% respectively) (Fig. S4.1, S4.2). A majority (65%) have PhD, 27% have a Master’s degree and 7% have a Bachelor's degree as their highest degree (Fig. S4.3). About 68% of the respondents identified themselves as being experts or having good knowledge of the topic (24 and 44%, respectively) while ~ 32%, self-identified as beginners in the field of connectivity analysis (Fig. S4.4). Six percent of respondents were conducting connectivity research at global scale, 32% at a continental scale (North America − 36%, Europe − 21%, Africa − 13%, South America and Asia − 12% each, Australia − 7%), and 62% within 40 countries (US, 22%; India, 16%; Canada, 10%). (Fig. S4.5).

We found a strong bias of survey participants who worked on connectivity of vertebrates (69%) in terrestrial ecosystems (73%) (Fig. S4.6). Within the terrestrial ecosystem, 34% worked on mammals, 16% on birds, between 10 and 11% on amphibians, reptiles and insects, and about 15% worked on plants. Within the freshwater ecosystem amphibians (29%) and fish (19%) were the most frequently studied taxa, and within the marine ecosystem, fish (41%) were the most studied taxonomic group. Less than 5% respondents in each ecosystem also worked on abiotic or structural connectivity. Almost half the respondents worked on single taxa (49%), while 20% worked on two or more taxa, and 13% worked across multiple ecosystems, e.g., terrestrial and freshwater. Participants mostly used both ArcGIS and R (~ 50%) and QGIS (30%) to handle spatial data (Fig. S4.7).

Users experience of tools used in resistance-based connectivity analysis

Of the 43 tools included in our final review, 33 tools were also part of the survey. Respondents who considered themselves to be experts in connectivity analyses were, on average, familiar with significantly more tools than beginners (Student's t-test, p < 0.05, df = 76, t = −3.3 and −2.8 for experts and beginners, respectively). Experts had heard of a median of eight and used five tools in relation to beginners, who had heard of a median of four and used one tool. Few tools are very popular as a majority of respondents had either used it or at least heard of it (Fig. S4.8). In particular, most users have at least heard of Circuitscape (83%), R package raster (71%), ArcMap extension for evaluating corridors (66%), Linkage Mapper (61%), R packages rgeos/sp/rgdal/sf (59%), R package gdistance (54%), Conefor (53%), and Circuitscape.jl (51%). In contrast, a majority of people (> 90%) were not aware of MultyLink, Makurhini and MatrixGreen. There was no correlation between the number of years since a particular tool became available and the percentage of people who had heard of it (Pearson’s r = 0.23, p > 0.5) or used it (Pearson’s r = 0.20, p > 0.5). Circuitscape.jl was an exception. The tool was released in 2019 and published in 2021 (Hall et al. 2021) but already received 26 evaluations (Fig S4.10), underscoring the popularity of circuit theory based tools in connectivity research.

Based on the sample size of tool ratings, the most widely used connectivity tools are Circuitscape (n = 77), R packages that specifically handle spatial data such as raster (n = 76), rgeos/sp/rgdal/sf (n = 68), Linkage Mapper (n = 41), and gdistance (n = 37). Only 22 tools received evaluations from more than five participants (Fig. 2) and were therefore included in rankings.

Fig. 2
figure 2

Spider-plots of the top 5 ranked tools for each step in resistance-based connectivity analyses. The five criteria stability (St), customization (C), data format (F), getting help (H), and speed (S) are arranged in a radial axis with lowest ranks in the center and highest ranks towards the periphery. Tools with a larger and uniform web rank high across all criteria, whereas smaller or non-uniform webs indicate a poor ranking overall, or in the criteria represented by that axis respectively


Out of the 10 tools that allow to prepare the data for further analysis (Table 1), seven were included in the survey and five tools received more than five responses. R packages rgeos/sp/rgdal/sf and raster were ranked the highest and equally across all the criteria. Guidos toolbox, a stand-alone tool, ranked third, but ranked poorly on data formatting and customization. Among ArcGIS toolboxes, SDM toolbox ranks evenly across all criteria whereas Corridor Designer Toolbox rank high on getting help and data formatting, but rank low on the other criteria.


Of the 27 tools that we identified for constructing and optimizing resistance surfaces, 20 were included in the survey and 13 tools had more than five responses. Raster package was again ranked high. Of the tools that are specifically developed for this step, gdistance, GFlow, Linkage Mapper and ResistanceGA ranked in the top five tools. GFlow, a tool specifically developed to model circuit theory-based analysis at large extents in a High Performance Computing environment was ranked high in all criteria except getting help and data format requirements. Linkage Mapper and ResistanceGA have good help and easy formatting requirements, but these tools often crash, leading to poor ranking in stability. Evaluations of the other tools used in constructing and optimizing resistance surfaces are presented in Fig. S4.9.


Among the 30 tools that use resistance surfaces, 22 were included in the survey, and 14 tools had more than five responses. gdistance ranks high across most criteria. GFlow and Linkage mapper and SDM toolbox also appear in the top five ranked tools in this step. The R package Grainscape ranked well in all criteria except for speed and stability whereas SDM toolbox Evaluations of the other tools that use resistance surfaces are presented in Fig. S4.10.

Users expectations in the future

Of the nine criteria we provided, consideration of the cost of tools, availability of user manuals and sample datasets, and the OS platform were critical factors in deciding which tools users would select for their analyses (Fig. 3a). In addition, good availability of help, ability to customize tools, and previous use as evidenced in reports and peer reviewed publications for example, were considered important in making this decision. In contrast, availability of a GUI, ability to run analyses on a cloud or server, and training workshops were not of much concern. We received an abundance of other criteria that are important to users when deciding which tool to use for their analyses (Fig. S4.11, Sect. 4 in S4). Some common themes that emerged are that details about the assumptions and caveats should be clearly stated, and that tools should be grounded in ecology and life-history of species. Users need high computational ability (i.e., the tool should be able to handle large high-resolution datasets at a high speed), compatibility with other processes in the workflow (i.e., not just limited to one step), and would prefer tools that run on commonly used open-source programming languages such as R or python. Participants also prefer tools that use input files in common data formats (e.g., *.tif, *.csv), and produce outputs that can be interpreted by stakeholders and decision-makers (e.g., visualization, maps).

Fig. 3
figure 3

Important criteria in the development and usage of connectivity tools. Importance of criteria provided in the online survey (A). Our suggestions to users to improve their experience with tools (B), and developers to improve the usage and future development of tools (C)

We received a total of 20 responses to our two follow-up questions regarding existing methods that should be implemented in the future. Participants did not identify any existing approach for resistance analysis that is not currently implemented in an analytical tool, but they stressed the need to computationally optimize existing methods so that large data sets can be handled in a time-efficient manner. They also recommended improved visualizations, flexibility to deal with a diversity of datasets, and tools with interactive interfaces to facilitate usage by practitioners. In response to our question about future conceptual improvements, participants called for more explicit consideration of the uncertainties involved in resistance surface parameterization, e.g., Rayfield et al. (2011). Automated parameter optimization methods, e.g., ResistanceGA (Peterman 2018) or radish (Peterman and Pope 2021) are seen as an important ways to reduce some of these uncertainties while ensuring that such methods have reasonable computational demands. Another suggestion was to include dynamic connectivity to address changing spatio-temporal variability in connectivity. Some of the other recommendations are to increase biological realism by considering individual behavioral variability and demographic effects, and developing methods that can simultaneously estimate and predict different components of connectivity. For example, it would be useful to obtain a more differentiated picture of when and where connectivity will likely occur, of what quality it will be (e.g., range shift to new habitats vs. dispersal among existing habitats), and whether it will be of sufficient strength to affect patterns and processes of interest (e.g., connectivity effects on survival or population growth). Based on our review, and responses to Sect. 4 of the survey and follow up questions, we summarize key recommendations for users and developers in Fig. 3B and 3C.


Connectivity science is a rapidly growing field (Correa Ayram et al. 2016), accompanied by a proliferation of tools to perform such analysis (Fig. S3). Given that landscape connectivity is now part of global (CBD 2020) and regional (European Commission 2020) targets, it can be expected that the number of new tools will continue to rise, making the selection of tools even more challenging, especially for beginners in the field. Through this paper, we summarize available tools, tabulate the key characteristics, and synthesize experiences from the research and conservation community. In order to facilitate an easy entry point for non-experts, we provide a detailed outline of the steps involved (S1), a decision tree that can help users select tools (S2) and a tutorial in R with script and sample data that performs several of the steps (S5). Our snapshot of user experiences with these tools should help non-experts to shortlist tools, and developers to improve the criteria that received low scores in the survey. We emphasize that our goal is not to rank one tool versus the other, but rather to provide an overview of their utilities on the basis of our own experience and of other researchers who have used them.


One potential limitation of our study is that the research focus of all authors and a majority of the survey respondents is heavily biased towards academics working in terrestrial ecosystems and vertebrates. Therefore, it is quite possible that we have not covered tools applicable to the aquatic realm of connectivity research as thoroughly. However, we shared the survey on multiple list-serves across all geographic regions, taxonomies and ecosystems, and believe that this bias towards connectivity in terrestrial ecosystems is a true reflection of this field. Correa Ayram et al. (2016) also reported terrestrial landscapes being represented in 88% of published connectivity studies, and a bias towards vertebrates (mammals > birds > reptiles). Despite this extensive effort to receive responses across all sectors, we acknowledge a bias towards academic respondents, who may be more familiar with coding, and therefore could have influenced some of the results, for example the lack of preference for tools with GUI. However, the participants represent a combination of experts and beginners (Fig. S4.4) and were therefore likely to present a balanced perspective of their experiences and what they look for when choosing tools for their projects.

There are several interesting tools available for connectivity research that we did not include in this study, because we focused only on approaches and tools that explicitly involve resistance surfaces. For example, Condatis (Wallis and Hodgson 2015) is a decision support tool for habitat creation and restoration which models flows (of individuals or their genes) through a habitat network using habitat area, dispersal kernels (based on Euclidean distances) and emigration rates (Hodgson et al., 2012, 2016). The approach is based on circuit theory, similar to the well-known Circuitscape software, but is not based on a resistance surface. Similarly, LConnect (Mestre and Silva 2021) uses vector data to derive landscape connectivity metrics and assess connectivity, but not based on resistance surfaces. ResDisMapper (Tang et al. 2020) is an interesting R package that uses genotype data and geographic coordinates to generate resistance surfaces, but does not use any environmental data. Additional tools also exist for including wind speed and direction in connectivity analysis e.g., R package rWind, (Fernández-López et al. 2020), and for assessing flows and resulting connectivity in stream networks e.g., FIPEX (Oldford 2020) with Dendritic Connectivity Index DCI (Cote et al. 2009) accessible at and R package smnet (Rushworth 2020). The fact that we did not include these tools in our overview does not overlook their merit and applicability for connectivity research and conservation.

Assessment of the tools

A major challenge for connectivity research lies in analyzing data across large spatial extents, at fine spatial scales, and with methods that are computationally demanding. It is important to note that in principle, the efficiency of all tools can be improved through additional steps, for example by running ArcGIS tools in batch mode, by stringing steps together in sequential pipelines through ArcGIS model builder and in R, or by running tools in parallel through R. Several tools are much faster and stable if executed through the command line version.

From the survey responses, we can safely conclude that there is no such thing as ‘the perfect tool’. Tools that rank the highest in all five criteria e.g., raster, and rgeos, sp, rgdal and sf, are generic tools that mainly help in preparing data for resistance-based analyses. Several tools were rated more heterogeneously across the five criteria, including ResistanceGA, which is specifically designed for optimizing resistance surfaces, and different tools that use resistance surfaces (Linkage Mapper, Circuitscape and Circuitscape.jl). They are either too slow, crash frequently, demand a lot of data formatting, are not customizable, or have poor access to help. Although we could not find any clear patterns, we observed trade-offs between speed and stability, and between stability and access to a help forum (Figs. S4.9, S4.10).

While the perfect tool may not exist, users almost always have a choice of several tools that perform similar tasks. This redundancy of tools, dependency on OS platforms and modelling options (e.g., through a GUI or programming languages like R and Python), and heterogeneity of performance across the different criteria provides users the option to decide what trade-offs they are willing and able to make. For example, some users may not have access to a server or high computing facility, so they may choose a tool that is relatively fast but needs some complicated data formatting requirements. Other users working with large landscapes may be willing to sacrifice speed for stability. Much of this decision making depends on the research question, the resources available (some software require paid licenses, or computers with high computational capacity), the extent and resolution of the input data, and the computational skills of the user.

The two most important characteristics that users care about are the associated costs, for example to purchase a license for the software that can implement the tool, and the access to a user manual with detailed instructions and example projects with sample data. While seemingly obvious, this is critical information. It is important to note here that all tools we found are available free of cost, but several toolboxes that run on ESRI ArcGIS or other proprietary software require users to buy a license with advanced extensions. There has been a huge push towards making customizable GIS software available without the need to invest in expensive proprietary software (Steiniger and Hay 2009) that allows collaborators to share data, code, and results seamlessly (Palomino et al. 2017). Although there are other programming languages used by ecologists, R appears to be very popular in the field of ecology because it improves open science, reproducibility of analyses and captures workflows when scripts and codes are included and shared (Lai et al. 2019).

Once the barrier of access (through costs of a GIS software) is removed, it is exceedingly difficult for beginners to start a project if they do not have a well-documented user guide, or no way to resolve questions through a peer group or forum. Here, we would like to emphasize that just providing a manual does not imply that the information is easy to follow. If functions and workflows are not explicitly defined, manuals can lead to large gaps of understanding and thus, limited usefulness of a tool. Most of the information is (usually) available, but often hidden in the manual. This may appear to be trivial and intuitive information for developers. However, during this review, we, experienced researchers in connectivity analyses, spent a significant amount of time sieving through user manuals to extract such basic information.

Recommendations for users

As already mentioned, we recommend users to combine expert-opinion with literature reviews and empirical data wherever possible. When habitat suitability models are used to generate resistance surfaces, we recommend testing a variety of relationships between habitat suitability and resistance (e.g., linear, negative exponential) with different conversions as mentioned earlier. Movement and genetic movement data are ‘gold standards’ as they capture functional connectivity, but in fact, these two data-types in fact capture different underlying ecological processes and temporal-scales. We reiterate here that the optimization and sensitivity analyses step is essential to overcome several of these issues.

For practical reasons, we recommend the minimum resistance values to be at least 1, as some of the tools that use resistance surfaces cannot handle smaller resistance values. Note that in the case of combining several resistance surfaces (for example when assessing multi-species connectivity), before multiple layers are combined through addition or averaging, their resistance values are often re-scaled to have the same range, and sometimes also standardized (e.g., z-score) to have the same distribution, (e.g., Row et al. 2017). This ensures that no single layer outweighs other layers in their contribution to final resistance values. However, one must be cautious when comparing resistance surfaces among different species, because the resistance values are species-specific meaning that the same landscape can lead to very different minimum and maximum resistance values for, e.g., a bird vs. an amphibian. As a result, the maximum value of a resistance surface will vary between different species.

Due to the high redundancy of the tools, we recommend that users try out a few of the tools before deciding on any one. Because there are several tools with user-friendly GUIs for all three steps, we believe that users should be able to find tools that cater to their needs and training. We also recommend readers to check the Conservation Corridor website for updates on new tools linked to the decision tree.

Recommendations for developers

We suggest developers to explicitly identify the methods implemented (e.g., is the functional connectivity estimated using least-cost paths or circuit-theory), and refer to our three-step structure (Fig. 1) to systematically indicate for which steps their tool can be used, in the same way researchers refer to the “Overview, Design concepts and Details (ODD)” (Grimm et al. 2010, 2020) or the “Overview, Data, Model, Assessment and Prediction” (ODMAP) (Zurell et al. 2020) protocol to report findings on agent-based simulation and species distribution models respectively. A similar classification and description system in user manuals that state the data and computational requirements, implemented methods, and outputs on the first page of all tools would be very helpful for users to spot suitable tools.

A detailed manual with clear examples can enable researchers to navigate through the tools and learn their functions more easily, minimizing the need for additional support. This information can also be used by other developers to easily identify already existing functions, thus avoiding duplication of tools.

We believe that connectivity analyses will most likely become analytically more complex in the future, especially with the increasing availability of high-resolution remote sensing data (Kwok 2018). Parallel to this computational enhancement, we underscore the need to put greater emphasis on increasing biological realism and producing outputs that are easy to understand by conservation practitioners and policy makers. Connectivity research that is analytically complex can be presented in relatively simple, yet ecologically meaningful indices that can be used in connectivity conservation and landscape planning. For example, several of the connectivity metrics recently reviewed by Keeley et al. (2021) can incorporate effective distances (e.g., least-cost paths or circuit-theoretic resistances) estimated from resistance surfaces, yet these metrics provide numbers that are relatively easy to explain and understand for practical planning and conservation purposes. These kinds of metrics also illustrate that all connectivity analyses need not be resistance-based as the typical input for calculating the metrics are Euclidean (straight-line) distances, or binary assessments on whether patches are connected or not. Overall, we recommend that developers strive to produce tools that are not dependent on proprietary GIS platforms, and encourage them to maintain an active, searchable help forum in addition to a detailed user manual and example data.

Future areas of development identified through the survey to move from static environmental data layers to incorporate dynamic, e.g., seasonal or annual changes which can strongly impact the resistance of a landscape (Osipova et al. 2019; Fenderson et al. 2020; Zeller et al. 2020). There has been a call to incorporate anthropogenic resistance in connectivity mapping, that accounts for human behaviors that may impact the way animals use the landscape (Ghoddousi et al. 2021). Recent studies present ways to synthesize resistance surfaces derived from different underlying data-types and modelling approaches to quantify consensus on landscape permeability and therefore be useful for conservation efforts (Schoen et al. 2022). Based on our extensive overview of analytical tool that is further corroborated by the suggestions provided by survey respondents, we identified at-least three key avenues for future development of connectivity tools: (i) better incorporation and presentation of uncertainties in analyses (ii) the importance of including dynamism in connectivity models and (iii) refining and testing methods to automatically optimize resistance surfaces.


Over the past 20 years, the use of resistance surfaces has been pivotal in advancing our understanding of how landscape features affect species movement and gene flow. Such knowledge is vital for effective management of populations in space and time, but the selection of the right tool is critical to many studies and subsequent management decisions. We do not aspire to advise users exactly which tools to use when, but rather to provide a road map (Fig. 1) with a compass (Supplementary S2) to navigate through the plethora of tools that are available. Selecting a particular tool requires a trade-off between some characteristics and is highly dependent upon the research question, the resolution and size of the landscape, and the number of data points used to optimize layers.

We hope this review will help beginners get a smooth entry at resistance-based connectivity research, highlight other available options to experienced researchers, and provide developers with ideas to improve the performance and usefulness of their tools. Ultimately, the diversity of methods, algorithms, and tools should help facilitate a better understanding of drivers of connectivity in fragmented landscapes and aid in conservation of biodiversity on Earth.