Birds (class Aves) follow the latitudinal gradient in species richness (more species are found closer to the tropics). However lowering the taxonomic scale to orders other patterns can emerge which can be instructive about mechanism. For example, in the order Anseriformes the pattern is reversed so richness increases towards higher and lower latitudes. This divergence from the latitudinal gradient in species richness, as well as the divergence from other marcoecological rules (e.g. Lack’s rule), calls for a study of the drivers of species richness in this order. Here we studied the spatial pattern of species richness among Anseriformes during breeding season. As in other studies of birds, we found variability in productivity (measured by NDVI) to be the most important predictor when using univariate models to predict richness. However, allowing for non-stationarity in the relationship between predictors and response by using regression trees we found day length in the breeding season to be the most important predictor followed by annual actual- or potential evapotranspiration conditioned on short or long days respectively. Predictors commonly used to explain large-scale richness patterns of birds, such as annual mean temperature and topographic heterogeneity, explained little or none of the variation in richness among Anseriformes. We discuss several possible mechanisms for this pattern. We highlight the importance of using methods allowing for non-stationarity in the process driving richness patterns, especially on a global scale where the strength of the predictors are likely change.