Spatial uncertainty effects on a species-landscape relationship model in ecology

In this study, we explore the effects of geometrical uncertainty in an existing species-landscape relationship model in the hoverfly communities. We also investigate how geometrical uncertainties affect a more complex model including both current forest patch features and past forest features. Because of a possible time-lag in biological responses to forest changes such as fragmentation, the historical dimension is added to the first model. The proposed approach relies on three spatial sources enabling to get forest fragments at different times: historical map (~1850), aerial black and white photographs (1954) and orthorectified photographs (2010). Firstly, we analyze the effect of the spatial data production method (manual versus automatic) on models using current forest patches only. Then, we build a more complex model including past changes in forest size. As previously, the effect of production-based uncertainty was assessed by comparing the models based on forests extracted manually and automatically. We address finally the impact of positional accuracy on the historical map by using a Monte Carlo simulation approach. Global results show that responses of the statistical models are strongly affected by spatial uncertainty in inputs