Background
The authors write that "accurately documenting and predicting declines and shifts in species' distributions is fundamental for implementing effective conservation strategies and directing future research," as well as for projecting climate-change-induced range shifts. And they note, in this regard, that "species distribution models (SDMs) have become a powerful tool for such work." Nevertheless, they indicate that "these models are opportunistic and often violate some of their basic assumptions."

What was done
To illustrate the significance of these shortcomings, Puschendorf et al. employed "amphibian declines and extinctions linked to the fungus Batrachochytrium dendrobatidis (Bd) to examine how sampling biases in data collection can affect what we know of this disease and its effect on amphibians in the wild." Working in the Australian Wet Tropics near the northeast coast of Queensland, they first "developed a distribution model for Bd incorporating known locality records for Bd and a subset of climatic variables that should correctly characterize its distribution," after which they (1) tested the validity of the model with additional surveys, (2) recorded new Bd observations in novel environments, and (3) where required, revised the original distribution model. Then, they "investigated the difference between the original and new models, and used frog abundance and infection status data from two of the new localities to look at the susceptibility of the torrent frog Litoria nannotis to chytridiomycosis."

What was learned
The five Australian researchers found that "the original SDM underestimated the distribution of Bd," due to the fact that subsequent sampling in supposedly unsuitable drier environments led to the discovery of "abundant populations of susceptible frogs with pathogen prevalences of up to 100%." They also found that the validation surveys they conducted led to their discovery of "a new population of the frog Litoria lorica coexisting with the pathogen," which species was previously believed to have been "an extinct rain forest endemic."

What it means
Puschendorf et al. say their results "indicate that SDMs constructed using opportunistically collected data can be biased if species are not at equilibrium with their environment or because environmental gradients have not been adequately sampled," while for disease ecology, they suggest that "better estimations of pathogen distribution may lead to the discovery of new populations persisting at the edge of their range."

Clearly, one has got to know the full ranges of where various plants, animals and pathogens currently reside, before one can even begin to predict (successfully) where they will live in a CO2-enriched and possibly warmer (or colder) future world.