Climatic curves as predictors in MaxEnt niche modeling

Abstract

Environmental niche modelling (ENM) uses different types of variables to predict species occurrence. In widespread use are variables derived from climatic curves, i.e., average annual changes in some climatic parameter. This study shows how to use the climatic curves themselves as ENM predictors. The key step is projection of the curves constituent variables on a suitable spline basis, which preserves time-ordering of the variables and supports smoothness of predictions. Complexity of the model is controlled by sensible choice of the spline basis, followed by lasso regularization in model fitting.