The bioclimate modelling community has been undertaking an increasingly degree of self evaluation regarding the assumptions and methodology that underpins a large research output. The criticisms of the bioclimate modelling are diverse including a contentious relationship with ecological and evolutionary theory, which can be coarsely summarised into an assumption that the modelled is ‘niche’ spatially and temporally invariant. We investigate methods that can account for variability in the ‘realised niche’ by capturing the processes important in determining the species distributions. Our approach is facilitated by the utilisation of simple ecological models, and parameter estimation techniques derived from Maximum Likelihood and Bayesian methods.  

We explore the properties of the ecological models in a theoretical context and examine the ability of the methodology to recover the correct model from simulated data. Two broad classes of model are used, a simple effect-response model and a consumer-resource model. Within the effect-response model we investigated the effect of changing the topology of the interactions matrix, investigating the effect of hierarchical competition against an assumption of niche overlap.

Results/Conclusions

Our analysis demonstrates thresholds in the methodology’s ability to distinguish underlying processes as data becomes increasingly rare, showing that errors made are dependent on the process and no single model is likely to be universally found to be correct. The strength of interspecific interactions also affects the ability to discriminate between processes, though differentially under the different modes of interaction. The topology of the competitive interaction matrix has a reduced effect when all of the community has similar absolute environmental tolerances.

We explore a practical application of the methodology using a data set of Mediterranean trees. The degree to which the fundamental niche is exhibited varies, with more than half of the species inhabiting <80% of their potential range. We demonstrate how a traditional approach may over estimate the effect of climate change on diversity patterns.

Finally we discuss some of the computational and statistical limitations of modelling biotic interactions, providing some simple suggestions that improve scalability in the number of parameters with an increasing number of focal species that will make these methods tractable.