The authors present 5 challenges for SDM. By confronting these models, we should be able to generate better and provide a better understanding of these models.
- Clarification of the niche concept: The authors suggest a diversion from the Hutchinson definition of the niche (n-dimensional hyper-volume of environmental variables) to more of a Grinnellian definition: the set of environmental conditions at which the birth rate of the population is greater than or equal to the death rate. They also point out the need for the clear(er) distinction between potential habitats for species (ranges/areas that the conditions are “right” for a species to inhabit) and potential geographic distributions of species (incorporates spatial factors such as dispersal)
- Improved designs for sampling data for building models: Models are sensitive to bias. Subsampling datasets are likely to remove or reduce biases. A (more expensive) alternative is to target specific, strategic additional sampling locations.
- Improved parameterization strategies: Different parameterizations of models can result in vastly different projections of species distributions. We to better understand why and when different parameterizations of the same technique provide different results.
- Improved model selection and predictor contribution: There are many different strategies that can be used to compare different models. However, finding the individual contribution of each predictor is a more difficult problem. Further testing of proposed solutions provided by the authors is needed.
- Improved model evaluation strategies: Evaluation of models can be discussed in the light of three different implementations: explanation, understanding and prediction. Typically modelers use the “wrong” evaluation tool outside of the scope of their goal or question. They present two different forms of evolution: verification (projecting using an new/independent set of data) and validation (maximizing the fit to training data).
Araújo, M. B. & Guisan, A. 2006 Five (or so) challenges for species distribution modelling. J. Biogeogr. 33, 1677–1688. (doi:10.1111/j.1365-2699.2006.01584.x)