Authors: Saeed Harati*, Université de Montréal, Liliana Perez, Université de Montréal, Roberto Molowny-Horas, CREAF, Spain
Topics: Spatial Analysis & Modeling, Land Use and Land Cover Change
Keywords: land cover change, validation, three-map comparison, scale, cellular automata
Session Type: Paper
Start / End Time: 2:35 PM / 4:15 PM
Room: Marshall North, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Validation of land cover change models often involves comparing simulation and observation maps. To this end, a simple approach for raster maps is pixel-by-pixel comparison. This type of map comparison rewards correct simulation even in areas of no change, and punishes small and large errors in allocation of change equally. In this study, a cellular automata (CA) model of forest insect infestation by the Mountain Pine Beetle (MPB) in the province British Columbia, was validated. To avoid the aforementioned mentioned issues of pixel-by-pixel comparison, the method of three-map comparison at multiple scales was applied. Notwithstanding the advantages of this method, results should be interpreted considering the context. Notably in the case of this study, forest insect infestation is an irreversible phenomenon where only the transition of land from non-infested to infested is possible. In this setting, simulation of no change in a non-infested zone might be credited as success for the model. Moreover, the CA models are typically built with the assumption of spatial contingency of phenomena in neighbourhoods. Therefore it may be reasonable that in the simulation of short-term spread of forest infestations of this study, multi-scale analysis finds many of the allocation errors in small distances, rather than large distances.