Geographical patterns of congruence and incongruence between correlative species distribution models and a process-based ecophysiological growth model

AimOur aim was to map the climate dependence of tree species distributions (probability of occurrence) and forest growth (net primary productivity) by comparing the congruence and incongruence between correlative and process-based modelling approaches. LocationIberian Peninsula, south-western Europe. MethodsWe used forest inventory data for three widespread tree species (Quercus ilex, Pinus halepensis and Pinus sylvestris) to model climatic suitability with an ensemble of seven correlative species distribution models (using biomod). We then simulated forest net primary productivity (NPP) as a surrogate of forest growth for forests of each species using an ecophysiological process-based model (gotilwa+) along a gradient of climatic suitability. The spatial distribution of the growth estimates was then compared with that of the suitability estimates, and robust regression was used to classify regions in terms of model congruence. ResultsQuercus ilex and P.sylvestris both showed a positive relationship between forest NPP and climatic suitability. The main discrepancies were found in the north of the peninsula, where there was high potential forest growth but low climate suitability. Low forest-growth estimates in areas of high suitability only appeared for P.sylvestris in southern montane regions. Pinus halepensis always showed a negative relationship between estimated growth and climatic suitability. The analysis of other ecophysiological parameters (mean leaf life and leaf area index) suggests that this tree species has different physiological strategies that allow differential growth rates in areas of low suitability. Main conclusionsWe found that the relationship between estimated growth and distribution varies strongly in different areas and species. Mapping the incongruences between the predicted climatic suitability and growth allowed us to identify regions where other factors (e.g. biotic interactions) may be more significant than the physiological limits on growth. We show that new insights into species distributions can be gained from mapping the differences between correlative and process-based models.