Study of spatial analysis methods for characterizing the green infrastructure of Louvain-la-Neuve

Details

Supervisors

Defourny, Pierre ; Radoux, Julien

Faculty

Faculté des bioingénieurs

Degree label

Master [120] : bioingénieur en sciences et technologies de l'environnement, à finalité spécialisée

Abstract

Biodiversity at the global scale is endangered by human activities, causing habitat loss and fragmentation among others. These processes apply pressure to animal populations, affecting their dispersion, reproduction, and movement between their habitats. In order to preserve these habitats, ecological networks must first be assessed, which is possible by using connectivity analysis tools to characterize the green infrastructure of landscapes. In this context, this work offers a comparison between four landscape connectivity analysis software relying on least-cost path and circuit theory algorithms: Graphab, TerrSet, Omniscape, and Circuitscape. The study is conducted on habitats surrounding Louvain-la-Neuve for which movement of small mammals is impeded by the town and its infrastructures. Small fictional landscapes are also created to extensively understand the functioning of each tool. To compare the software, input files are created so that every tool uses the same data despite diverse requirements in format and parameters. These files are obtained by precisely delineating the habitats and creating a resistance surface of the landscape through the assignment of resistance values to the WALOUS 2018 land cover classes. The computations of Circuitscape and Omniscape produce results in the form of current maps displaying the probability of movement for every cell across a landscape, while Graphab computes habitat linkages whose cost to follow is the least among all possible paths, and TerrSet models least-cost corridors of defined width. Results analysis demonstrate that TerrSet is unreliable to model least-cost corridors. Omniscape produces coherent current maps but is flawed by the lack of minimal habitat size to specify, which biases connectivity across the landscape, and is therefore not an appropriate tool for this thesis. Graphab and Circuitscape both compute consistent outputs and the comparison of their results reveals useful elements. The Graphab least-cost paths provide easily readable results but also lack details as they do not display alternative pathways. The Circuitscape current maps allow identifying all possible paths across a landscape to connect habitats but do not indicate the least costly one and demand more computation power which limits its use for large extent areas. The choice of software depends on the conducted study; if dealing with a large extent landscape and not needing an extensively detailed result, Graphab is appropriate; if the assessment of connectivity between habitats must be thorough and the volume of data is not too important, Circuitscape might be more relevant. Both tools can also be used jointly in the same study. Lastly, this thesis could be improved by analyzing larger landscapes containing more habitats, and the connectivity assessment of Louvain-la-Neuve can be enhanced by fine-tuning the resistance surface and taking better account of the landscape layout.