OOS 15-9
Demographically driven distribution models; Advantages of using integral projection models to incorporate demography into species distribution models

Wednesday, August 7, 2013: 10:50 AM
101A, Minneapolis Convention Center
Cory Merow, Quantitative Ecology Group, Smithsonian Environmental Research Center, Edgewater, MD
Andrew M. Latimer, Plant Sciences, University of California Davis, Davis, CA
Adam M. Wilson, Ecology & Evolutionary Biology, Yale University, New Haven, CT
Sean McMahon, Quantitative Ecology Group, Smithsonian Tropical Research Institute, Edgewater, MD
John A. Silander, Ecology and Evolutionary Biology, University of Connecticut, Storrs, CT
Background/Question/Methods

Species Distribution Models (SDMs; e.g. Maxent, GARP, GLMs, etc.) are typically used to describe the correlation between occurrence patterns and environmental covariates. Often, these methods are used because species’ presence, and sometimes absence, at locations on a landscape are the only available population-level data. While this class of SDMs is useful for exploring spatial occurrence data, they offer limited insights into the underlying population biology that generates these patterns because occurrence patterns entangle a variety of processes such as demography, dispersal, biotic interactions and historical effects. Here, we show how to use a limited amount of demographic data to produce Demographically Driven Distribution Models (DDDMs) using Integral Projection Models (IPMs) for stage-structured populations. By modeling vital rate functions such as survival, growth, and fecundity using regression, they can interpolate across missing size data and environmental conditions to compensate for limited data. 

Results/Conclusions

DDDMs offer at least five advantages over occurrence-based SDMs: (1) DDDMs allow a mechanistic understanding of the demographic processes that generate spatial occurrence patterns; (2) DDDMs predict more biologically meaningful demographic summaries of population patterns such as population growth rate, life expectancy, or stage distributions; (3) DDDMs can avoid issues of aggregation by modeling how individuals respond to weather, rather than modeling how a species responds to climate; (4) DDDMs allow for more robust extrapolation under new environmental conditions (e.g. climate change) because one can more readily evaluate the consequences of extrapolating a vital rate representing single biological process than an occurrence probability that entangles multiple processes; (5) DDDMs can predict temporal dynamics, in contrast to the typically static predictions of SDMs. To illustrate these principles, we construct DDDMs for an overstory perennial shrub in the Proteaceae family in the Cape Floristic Region of South Africa by combining data from a variety of sources. We compare these models to SDMs that predict occurrence probability and illustrate inference about the demographic processes that drive differences in habitat suitability.