Species Distribution Modeling

Species distribution modeling is perhaps the most fundamental task in natural resource management and conservation practice. This task dates to the earliest days of ecology but has undergone a rapid evolution in the past several years, driven by new statistical techniques and the increasing availability of large-scale species distribution data. Here, we illustrate a workflow for this task, which weaves together ecological principles, data considerations, and statistical models. The statistical modeling itself follows a sequence that flows from exploratory data analysis to model fitting, model evaluation and calibration, model validation, and to applications. The process is first illustrated with a generalized linear model and then extended to alternative tools, including generalized additive models, tree-based models (random forests), and maximum entropy modeling (maxent). The results of species distribution modeling can provide the basis for an initial site prioritization (Chap. 8 ). But we begin with SDM because it introduces most of the empirical challenges that arise in working with landscape-scale data: ecological data are multivariate (and so redundant), noisy, and spatially structured. We address these challenges in the next few chapters.