Least-cost modelling on irregular landscape graphs

• Published in: Landscape ecology Vol. 27; no. 7; pp. 957 - 968
• Main Author: Etherington, Thomas R.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.08.2012; Springer; Springer Nature B.V
• Subjects: Algorithms; Animal, plant and microbial ecology; Apexes; Applied ecology; Bias; Biological and medical sciences; Biomedical and Life Sciences; Connectivity analysis; Decision making; Ecology; Environmental Management; Fundamental and applied biological sciences. Psychology; General aspects; General aspects. Techniques; Graph theory; Landscape Ecology; Landscape/Regional and Urban Planning; landscapes; Life Sciences; Methods and techniques (sampling, tagging, trapping, modelling...); Modelling; Nature Conservation; Python; Research Article; Sustainable Development; Functional connectivity; Nearest-neighbour interpolation; Least-cost path; Delaunay triangulation; Triangulated irregular network; Accumulated-cost-surface; Graph theory; Cost-surface; Dijkstra’s algorithm; Python; Nearest neighbour; Costs; Landscape; Algorithm; Modeling; Ophidia; Interpolation; Vertebrata; Reptilia; Dijkstra's algorithm
• ISSN: 0921-2973; 1572-9761
• DOI: 10.1007/s10980-012-9747-y

Least-cost modelling is becoming widely used in landscape ecology to examine functional connectivity. Traditionally the least-cost modelling algorithm creates a regularly structured landscape graph for connectivity analysis by converting all the cells from a cost-surface into vertices in a landscape graph. However, use of a regular landscape graph is problematic as it: contains a great deal of redundant information that in turn increases processing times, is constructed in a deterministic manner that precludes examination of the effects of graph structure on connectivity measures, and is known to produce results with directional bias. I present, and provide Python code for, an algorithm to produce an irregular landscape graph from a cost-surface. Tests demonstrate that comparable results to those of the traditional regular landscape graph approach can be achieved, while at the same time reducing computational expense, enabling variations in graph structure to be incorporated into an analysis, and avoiding directional bias. Therefore, this approach may allow for more robust ecological decision-making when examining matters of functional connectivity using least-cost modelling.