Climate change can alter the availability of suitable habitat and, thus, the fate and evolution of species. Climate does not lend itself to experimental manipulation over continental scales but natural variability in climate histories among continents provide the opportunity for comparative studies of responses of globally distributed taxonomic groups to climate change. Here we provide a test of the continental-scale variability in the response to climate change of ducks from the genus Anas over the last 21,000 years. We developed distribution models of the breeding ranges of 33 Anas species (including representatives from all inhabited continents) and used them to hindcast duck distributions to the Last Glacial Maximum (LGM ~21k years BP). We used the software MAXENT and presence-only data from multiple public and private sources (including GBIF and ORNIS) paired with temperature and precipitation data (WorldClim and PMIP2). In an effort to correct for a strong North American bias in the available data on duck breeding localities, we (1) geographically sub-sampled the presence data, and (2) adjusted the concentration of pseudo-absences (environmental background data) to sampling effort. We validated our models through (a) held-out data, and (b) comparisons to published ranges (NatureServe, GROMS). The models’ power to predict geographic ranges of sister species on different continents provided an additional indication of their biological validity.

Results/Conclusions

AUC values from held-out data were extremely high (extending over 0.9), which may, in part, reflect autocorrelation and highly uneven geographic distribution of presence localities. Adjusting the distribution of pseudo-absence data to sampling effort considerably improved model fit as evaluated against expert ranges. Predictions of sister species were mostly within known ranges (reflecting consistencies in climatic requirements), but did not fill them, possibly reflecting non-analogue combinations of climate. The area of suitable breeding habitat for duck species in the Northern Hemisphere generally increased since the LGM due to glacial retreat. In South America, models predict a decline in breeding area since the LGM resulting from rising sea level. Using changes in suitable areas as a proxy for changes in population sizes, our results agree with preliminary genetic models, which show parallel population trends. Different climate histories among continents provide a valuable natural experiment into the variability of biological responses to global climate phenomena within globally distributed taxonomic groups.