New article: Optimising long-term monitoring projects for species distribution modelling. How atlas data may help
In this work, recently published in Ecography, we provide a method based on modern atlas data to guide the establishment of monitoring projects that will deliver useful data to build accurate distribution models. Long-term monitoring data are used to evaluate population trends, but they may also be integrated in spatial models to produce detailed species distribution maps at large spatial scales. However, sampling design has a strong impact on the predictive accuracy of distribution models and this issue is to be addressed when using this type of data.
We examined how the number of sampling sites influences the predictive accuracy of the models and we determined the minimum number of sites required to generate reliable spatial models as a direct output of monitoring projects. To do this, we calibrated large-scale, fine-resolution distribution models for 20 bird species using data collected during the ‘Breeding Bird Atlas of Wallonia’ (BBAW) project. Modern atlas projects like BBAW provide data that are analogous to monitoring data. We manipulated the amount of calibration data to represent a range of sampling coverage and we used independent evaluation data to calculate modeling performance parameters.
The modeling performance was sensitive to particularly small sample sizes and reached an asymptotic level beyond a fairly large sampling coverage. This asymptotic level varied considerably among species depending mainly on their prevalence. Our results suggest that a sampling coverage of 4-5% of the study area is sufficient for most of the species. This innovative analytical framework will guide the design of long-term monitoring projects suited to document and regularly update species distribution maps.Aizpurua O., Paquet J.-Y., Brotons L. & Titeux N. (2014) Optimising long-term monitoring projects for species distribution modelling: how atlas data may help. Ecography, early view, doi: 10.1111/ecog.00749.