Process, correlation and parameter fitting in species distribution models: a response to Kriticos et al
Abstract
In a recent article (Dormann et al., 2012, Journal of Biogeography, 39, 2119–2131), we compared different approaches to species distribution modelling and depicted modelling approaches along an axis from purely ‘correlative’ to ‘forward process‐based’ models. In their correspondence, Kriticos et al. (2013, Journal of Biogeography, doi:10.1111/j.1365‐2699.2012.02791.x) challenge this view, claiming that our continuum representation neglects differences among models and does not consider the ability of fitted process‐based models to combine the advantages of both process‐based and correlative modelling approaches. Here we clarify that the continuum view resulted from recognition of the manifold differences between models. We also reinforce the point that the current trend towards combining different modelling approaches may lead not only to the desired combination of the advantages but also to the accumulation of the disadvantages of those approaches. This point has not been made sufficiently clear previously.
What the paper shows and why it matters
(AI-generated)
Species distribution models (SDMs) estimate where a species can occur. Correlative models relate recorded occurrences directly to environmental variables; process-based models let the distribution emerge from implemented processes such as dispersal or demography. Dormann et al. (2012) arranged the approaches on a continuum from purely correlative to “forward” process-based models. Kriticos et al. (2013) objected that this hides marked differences among models and misrepresents “fitted” process-based models, some of whose parameters are estimated from distribution data (Hartig et al. 2012) and which, they argue, combine the strengths of both approaches. The problem the paper takes on is whether fitted process-based models and hybrid models (correlative models with added processes) gain both sets of advantages without taking over the disadvantages.
The paper argues that species distribution models which mix fitting to distribution data with the representation of ecological processes may not simply add up the strengths of both approaches, but inherit some weaknesses of the correlative side and may create new ones, so a continuum describes them better than separate categories. A process-based model fitted to distribution data shares some assumptions, data requirements, validation issues and sources of uncertainty with a correlative model. Fitting is often done by necessity rather than design, because such models have many parameters, some of them hard to measure beforehand. Where model structure and independent ecological knowledge do not sufficiently constrain the outcome, fitting may give an illusion of predictive power, and may bring drawbacks in transferability and extrapolation similar to those of correlative models. Combination can also add errors: a dispersal kernel placed on a suitability map already fitted to observed distributions may, the response argues, count dispersal limitation twice, and can add resolution-dependent errors (Bocedi et al. 2012). A continuum also accommodates novel approaches, whereas fixed categories produce borderline cases.
The paper does not reject combined models. It holds that tuning is inevitable in many cases but that its consequences should be taken into account when models are validated, and restates that no method is intrinsically superior to another. The reply clarifies that the continuum was meant to acknowledge the variety of models, and it closes by inviting continued open discussion of their strengths and weaknesses.
References
- Dormann et al. (2012). Correlation and process in species distribution models: bridging a dichotomy. Journal of Biogeography 39, 2119–2131. doi:10.1111/j.1365-2699.2011.02659.x
- Kriticos et al. (2013). Essential elements of discourse for advancing the modelling of species' current and potential distributions. Journal of Biogeography 40, 608–611. doi:10.1111/j.1365-2699.2012.02791.x
- Hartig et al. (2012). Connecting dynamic vegetation models to data – an inverse perspective. Journal of Biogeography 39, 2240–2252. doi:10.1111/j.1365-2699.2012.02745.x
- Bocedi et al. (2012). Projecting species’ range expansion dynamics: sources of systematic biases when scaling up patterns and processes. Methods in Ecology and Evolution 3, 1008–1018. doi:10.1111/j.2041-210x.2012.00235.x