Canadian Forest Service Publications
Cross-scale integration of knowledge for predicting species ranges: a metamodeling framework. M.V. Talluto, I. Boulangeat, A. Ameztegui, I. Aubin, D. Berteaux, A. Butler, F. Doyon, C.R. Drever, M-J. Fortin, T. Franceschini, J. Liénard, D. McKenney, K.A. Solarik, N. Strigul, W. Thuiller, D. Gravel. 2016. Global Ecology and Biogeography 25:238-249.
Year: 2016
Issued by: Great Lakes Forestry Centre
Catalog ID: 36357
Language: English
Availability: PDF (request by e-mail)
Abstract
There is great interest in modeling species ranges, and such models have particular utility in guiding conservation and management decisions. There presently exists a great diversity in modeling approaches, and individual models often only use a small subset of the total available information about a species. For example, a broad-scale correlative model might use climate variables to predict presence or absence, but ignore what is known about smaller-scale processes suchas the effect of local climate on growth, fecundity, and dispersal rates. An inevitable prob lem with the diversity of approaches is that multiple models produce differing predictions for the same organism, with no simple way to reconcile these predictions. We present a flexible framework for integrating models at multiple scales using hierarchical Bayesian methods. The resulting model produces probabilistic estimatesof species presence with uncertainty that prop agates through all models. These predictions reflect all of the information used as input for the original submodels. We illustrate the approach through two examples, with complete code and data provided assupplemental information. The framework performed well with oursimple examples, substantially reducing uncertainty in model predictions when projecting beyond the range of some of the original data sources. Finally, we discuss the application of our method and its accessibility to conservation biologists and land managers. Although the method maybe inaccessible to users without extensivestatistical programming experience, we anticipate the re sults will be of wide application and interest, andwe therefore encourage collaboration between modelers and practitioners in applying our framework.
Plain Language Summary
Current interest in modeling potential changes to species ranges have resulted in a great diversity of approaches to species distribution modeling. In general, individual approaches include only a small subset of the available information about a species. Correlative models often predict presence or absence as a function of climate, but ignore smaller-scale processes such as growth, fecundity, and dispersal. Furthermore, different approaches often produce divergent predictions, with no simple method to reconcile them. We present a flexible framework for integrating models at multiple scales using hierarchical Bayesian methods. Our method results in a metamodel that produces probabilistic estimates of species presence, reports uncertainty from both data and process error from all sub-models, and incorporates all of the information used as input for the original scale-specific sub-models. We illustrate our approach using two examples, and demonstrate that the framework can substantially reduce uncertainty when projecting beyond the range of the original data. We conclude by discussing the application of our method and its accessibility to conservation biologists and land managers. Although implementation of our method can be technically challenging, we anticipate the results will be of wide interest, and we encourage collaboration between modelers and practitioners in applying our framework.