High resolution mapping of population change in breeding birds in Wallonia (Southern Belgium)

Patterns of fine-scale change in bird abundance across different landscapes may inform about the driving forces behind bird communities evolution. We compare km²-resolution map of breeding birds in Wallonia (Southern Belgium) at 10 years interval. The maps are based on repeated sampling transects conducted inside a km²-grid. Spatial modelling techniques were applied on these two dataset using environmental variables produced by the LifeWatch-WB Project. Variables are issued from pixel-based land cover classification of orthophoto mapping and satellite images, with a resolution of 2 meters and are available for the two periods corresponding to bird data. Others variables included in the model are climatic, topographic or soil attributes. For each bird species, spatial models built with data from the first period are projected with the value of the environmental variables for the more recent period, and viceversa. Moreover, another method is to compare models built independently on the two periods. Modelling methods mainly are of two types: Generalized Additive Model (GAM) and RandomForest. Comparison between prediction and real data at the km² level offers insight about the causes of change in bird populations.