Predicting future patterns, processes, and their interactions: Benchmark calibration and validation procedures for forest landscape models

Process-based Forest Landscape Models (FLMs) rely on first principles to simulate ecological patterns and processes, making them uniquely powerful for forecasting ecological dynamics under unprecedented climatic and disturbance regimes. Persistent challenges with any ecological forecasting model are calibration (“tuning” the model) and validation (“proofing” the model). As no actual future data exist from which to conduct a formal model validation, model credibility is established through numerous tests against empirical datasets and comparisons with other types of models. The purpose of this study was to establish more consistent and generalizable standards for calibrating and validating LANDIS-II, a widely used, open-source FLM. We reviewed methods gleaned from a wide variety of previous FLM studies and advance some new techniques for evaluating the credibility of the model outputs. We used publicly available data with full coverage for the United States (US) so that our methods will be generalizable to other landscapes in the US, and we developed an ecologically meaningful set of validation metrics for evaluating the credibility of new applications.