Climate is a complex, multidimensional phenomenon. The Bioclim variables provide a statistical summary of climate in a set of static spatial variables that are suitable for bioclimatic modelling. Details of their derivation can be found on the Bioclim Registry page

Concerns have been raised about the stability (and hence utility) of some of the seasonality variables (e.g. Bio09, mean temperature of the driest quarter), particularly where the variable defining the quarter has little variance across the year. Typically, in arid regions the climate profile includes a uniform rainfall pattern, reflecting the temporal average of an erratic low-frequency process. In such cases the quarter that is defined as the driest can also vary erratically. Modellers should be cautious in the use of these variables and the interpretation of resulting models in these regions.

Concerns have also been raised about the practise of "data dredging", using all of the available variables as covariates. This practice can lead to over-fitted models, with limited ability to be applied outside the training domain. The best practice is to select covariates based upon an understanding of the factors limiting the species range. To assist in situations where this is not possible, the first five principal components have been added to the set of Bioclim variables, coded Bio36-Bio40 (Kriticos et al. 2014). Together, these five variables account for more than 90% of the variance in the first 35 Bioclim variables. However, modellers should be aware that the interpretation of models built using these variables cannot be used for climate change studies, they cannot be compared between datasets, and the interpretation of these models will necessarily be complicated by the complex nature of the variables.

Users should check their modelling package to determine which of the ASCII or ESRI grid formats will be most suitable for their purposes.