Research overview: validating felid distribution models and evaluation metrics

Assessing whether a habitat model is any good requires careful evaluation, especially when projecting into the future where ground truth is unavailable. This overview surveys the metrics used, AUC, true skill statistic, and others, their limitations for presence-only data, and the value of spatially independent or temporal validation. It reviews why high apparent accuracy can mask poor transferability and how cross-validation strategies should reflect spatial structure. The synthesis frames model evaluation as essential due diligence in felid climate research, explaining why readers should weigh how a projection was tested, since a model that fits current data well may still fail to predict where cats will live as the climate changes.