Where and at which scales does the latitudinal diversity gradient fail?

International audience; AimThis study analyses the places and the scales at which the latitudinal diversity gradient (LDG) for vertebrate species holds or, conversely, fails. Although much empirical support for the LDG at a global scale exists, several marked departures have also been previously reported for some regions, scales and organisms. Here, we aim to test the generality of the LDG by accurately quantifying its variability among regions and scales using a novel analysing approach applied on environmental and species data. LocationGlobal; Vertebrate species (amphibians, mammals and birds). MethodsTo test the LDG, we proceeded in two successive steps. First, we collected environmental and vertebrate species data to summarize the overall environment (likely the LDG). In a second step, we compared the two datasets (environment and species biodiversity index maps) using the Comparison Map and Profile (CMP) method. It computes the correlation between datasets in moving windows, all over both maps (at one scale), and by increasing the size (scale) of the moving windows. Cross-correlations are then spatially computed through all scales in order to map and accurately quantify the relationship between environment and biodiversity on Earth. ResultsThe best environmental variable, explaining almost 70% of the spatial variation in vertebrate species richness, was the temperature seasonality composite (TSC), a good surrogate candidate for the LDG. While the TSC-diversity relationship followed an expected LDG-like pattern in the Southern Hemisphere and in the tropics, it showed significant departures from latitudinal trends in Asia and North America. This observation can be explained by the increase in animal species number with increasing northern latitudes. Main conclusionsWe showed that the LDG (i.e. through the TSC surrogate) relationship fails in many places and at all scales in average. We must then reject the mainstream hypothesis of the overall worldwide LDG pattern. We suggest that it should not be taken for granted in macroecological studies, especially over the Northern Hemisphere. The spatial distribution of vertebrate richness is both area and scale dependent, and this observation is robust to changes in spatial resolution, in data coverage (e.g. ecoregions) and in taxa.