Generating realistic assemblages with a joint species distribution model

• Published in: Methods in ecology and evolution Vol. 6; no. 4; pp. 465 - 473
• Main Authors: Harris, David J., Warton, David
• Format: Journal Article
• Language: English
• Published: London John Wiley & Sons, Inc 01-04-2015
• Subjects: Animal breeding; Artificial neural networks; Associated species; birds; breeding bird survey; Ecological research; Ecologists; Geographical distribution; Grasslands; Heterogeneity; Introduced species; neural network; Neural networks; Predictions; probabilistic graphical model; species assemblages; species distribution model; Species richness; Stochasticity; North America
• ISSN: 2041-210X; 2041-210X
• DOI: 10.1111/2041-210X.12332

Summary
Species distribution models (SDMs) represent important analytical and predictive tools for ecologists. Until now, these models have either assumed (i) that species' occurrence probabilities are uncorrelated or (ii) that species respond linearly to preselected environmental variables. These two assumptions currently prevent ecologists from modelling assemblages with realistic co‐occurrence and species richness properties.
This paper introduces a stochastic feedforward neural network, called ‘mistnet’, which makes neither assumption. Thus, unlike most SDMs, mistnet can account for non‐independent co‐occurrence patterns driven by unobserved environmental heterogeneity. And unlike several recently proposed joint SDMs, the model can also learn nonlinear functions relating species' occurrence probabilities to environmental predictors.
Mistnet makes more accurate predictions about the North American bird communities found along Breeding Bird Survey transects than several alternative methods tested. In particular, typical assemblages held out of sample for validation were each tens of thousands of times more likely under the mistnet model than under independent combinations of single‐species predictions.
Apart from improved accuracy, mistnet shows two other important benefits for ecological research and management. First: by analysing co‐occurrence data, mistnet can identify unmeasured – and perhaps unanticipated – environmental variables that drive species turnover. For example, the model identified a strong grassland/forest gradient, even though only temperature and precipitation were given as model inputs. Second: mistnet is able to take advantage of outside information to guide its predictions towards more realistic assemblages. For example, mistnet automatically adjusts its expectations to include more forest‐associated species in response to a stray observation of a forest‐dwelling warbler.