Application of true skill statistics as a practical method for quantitatively assessing CLIMEX performance

Abstract

The species distribution model (SDM) predicts the potential distribution of a species as a function of abiotic variables. CLIMEX is a SDM that mechanistically evaluates the climatic suitability of a species in a specific area. However, compared with other SDMs, the metrics for quantitatively evaluating CLIMEX performance are limited. Hence, this study quantitatively evaluated the performance of the CLIMEX models of six example species by applying the metrics used for machine-learning-based SDMs. The results showed that the performance of CLIMEX had a minimum true skill statistic (TSS) of 0.35 and a maximum of 0.78 depending on the species. Depending on the scale of the prediction area, number of occurrence coordinates, and classification threshold, CLIMEX performance can be evaluated by applying true skill statistics (TSS). Although this method is not perfect because of the characteristics of CLIMEX modeling, we expect it to be a useful option before conceiving the best method to evaluate mechanistic SDM.

Keywords

• CLIMEX
• Confusion matrix
• Model performance
• Species distribution model
• True skill statistics