Multiple Randomized Shortest Paths for ArcGIS Pro

Least-cost paths have received widespread use in fields that aim to understand how people, animals, and particles might move across landscapes. In ecology least-cost paths have been used to examine connectivity of individuals, propagules, and genes as well as serve as the basis for many corridor-building applications for the purpose of conserving habitat. Adriaensen et al. (2003) introduced the concept into ecology as an alternative to Euclidean distance and provided examples from the Belgian landscape. Using a heterogeneous raster-based landscape in which digital maps (rasters) are assigned cost distances (also known as resistance) the least-cost path uses Dijkstra’s algorithm to identify the shortest path in terms of cumulative cost/resistance. The algorithm identifies a single least-cost path that is one cell wide. In an ecological context this assumes that the animal/propagule has sufficient knowledge of the landscape to identify and follow that “best” path. Pinto and Keitt (2009) identified that in many cases organisms don’t have perfect knowledge of their environment and that multiple realizations of the shortest path may be necessary to account for variability in movement. Their approach was to develop stochastic, rather than static, realizations of the least-cost path, which they implemented in Java software LORACS (no longer available) (Pinto et al. 2012). McRae et al. (2007) introduced the idea of using circuit theory is used to model dispersal behavior. Unlike a single least-cost path circuit theory models the dispersal of many organisms/electrons resulting in multiple paths across the landscape. Movement is based on random walk theory and is proportional to the resistance/cost surface. Circuit theory, unlike a least-cost path, doesn’t assume that an organism would have perfect knowledge of its landscape. Circuit theory has received widespread use in ecology (Dickson et al. 2019) and has been used to understand gene flow across landscapes, model animal movements, and develop conservation corridors. However, outputs from circuit theory provide the user with little control and can sometimes be difficult to translate into corridors. The randomized shortest-path was introduced by Saerens et al. (2009) and has been implemented in the R package gdistance (Van Etten 2020). The randomized shortest-path approach bears many similarities to the multiple shortest paths approach of Pinto and Keitt (2009), of which the ability to control the level of randomization is among the most important features. In the gdistance package this is done by controlling the theta parameter. The Multiple Shortest Paths Toolbox for ArcMap is built on these ideas while providing similar functionality in an ArcGIS environment. This version of the Multiple (Randomized) Shortest Paths Tool for ArcGIS Pro was built and tested using version 3.4.0 and is based on the original toolbox designed and built for ArcMap (version 10.7.1). The toolbox contains six models and no python scripts. It requires the Spatial Analyst extension. The toolbox can be executed as a tool (GUI) by double clicking on the model or run in ModelBuilder in edit mode. The tools are ordered in a logical progression running from 1 to 4 and steps 2 and 4 involve nested inner models. A key improvement over the ArcMap version of the tool is the new Create Matrix tool (step 3) which calculates the mean, maximum, minimum, and standard deviation of cost-distance and path length for all pairs of points. The Create Matrix tool can be used for applications that require a pairwise matrix, like landscape genetics.