Forecasting species distributions: coping with uncertainties in data
Forecasts of species distributions have so far been most assessed using
correlative species distribution models. These models are based on the
observed current relationships between bioclimatic variables (yearly or
monthly normals of temperatures and precipitation, soil quality, etc).
These models, however, cannot tell whether observed correlations are
causal or merely due to historical contingencies. Thus, extrapolation
to future conditions should be undertaken with caution.
These models are widely used because they only require readily available data on species occurrences and on environmental variables. They are highly sensitive to the quality of the input data; yet current occurrence data is not univocal, even for well-known species in well-sampled areas. Prior to conducting further assessments on the future of biodiversity, we thus urge for the development of accurate databases on species distributions, for example using participatory science programs.
These models are widely used because they only require readily available data on species occurrences and on environmental variables. They are highly sensitive to the quality of the input data; yet current occurrence data is not univocal, even for well-known species in well-sampled areas. Prior to conducting further assessments on the future of biodiversity, we thus urge for the development of accurate databases on species distributions, for example using participatory science programs.
