The fragmentation of wildlife habitats caused by anthropogenic activities has reduced biodiversity and impaired key ecosystem functions. Wildlife corridors play an important role in linking detached habitats. The optimal design of such corridors considering spatial, ecological, and economic factors is addressed in this paper.

We present a novel graph-theoretic optimisation approach and a mixed-integer linear programming model to determine an optimal wildlife corridor connecting two given habitat patches. The model maximises the total quality of the corridor and satisfies pre-specified corridor width and length requirements under a resource constraint.

Compared to the corridor design models presented in the literature, our model is conceptually simpler, and it is computationally convenient. We applied the model to a real data set for Eldorado National Forest in California, USA, involving 1,363 irregular land parcels.

The model can be extended to design multiple corridors that connect two or more existing habitat patches.

This dataset contains a GAMS code used to test the model's computational efficiency, and another GAMS code and the data used in the empirical application for Eldorado National Forest in California, USA. 

We prepared the empirical application data using ArcGIS. Specifically, we transformed the 1,363 polygons of the Eldorado National Forest into 4,214 polylines using the Polygon-To-Line function in ArcToolbox and identified all pairs of adjacent land parcels and their common edges using the attribute table of the resulting polyline file. We identified 2,858 junction points for those polylines using the Planarize-Lines function in Advanced Editing, and we found all pairs of edges within a distance under 300 meters using the Generate-Near-Table function included in Analysis Tools in ArcToolbox. We created a personal geodatabase in ArcGIS to place these data and used the program mdb2gms.exe included in GAMS to convert them to a GAMS readable format (.inc).

We used Gurobi 9.0.2 integrated with GAMS 32.1.0 as the solver for the mixed-integer linear programming model presented in the paper.