modelskill - Assessing and Visualising the Performance of Prediction Models

Provides tools for evaluating continuous predictions and their associated predictive uncertainty from statistical, machine-learning, geostatistical, and process-based models. It implements complementary measures of prediction error, association, agreement, efficiency, uncertainty calibration, and predictive-distribution performance, together with Taylor, solar, target, coverage, probability integral transform, and quantile-coverage diagnostics. Methods include the integrated evaluation approach of Wadoux, Walvoort and Brus (2022) <doi:10.1016/j.geoderma.2021.115332> and the uncertainty-validation framework of Schmidinger and Heuvelink (2023) <doi:10.1016/j.geoderma.2023.116585>.