Days of Future Passed: Best Calibration Does Not Guarantee Reliable Forecasts of Future Tick Distributions

Forecasting species distributions under future climate scenarios has become a standard application of Species Distribution Models (SDMs), yet their temporal transferability is rarely evaluated. We assessed the reliability of future projections using four widespread tick species (Amblyomma americanum, Dermacentor variabilis, Ixodes ricinus, and Ixodes scapularis) with occurrence records spanning from 1961-2025. Models were calibrated sequentially using decade-specific records and projected onto subsequent decades, for which independent observations were available. Climate predictors derived from TerraClimate using harmonic regression were modelled with Boosted Regression Trees (BRT) and Maxent. Transferability was evaluated using the Boyce Index, Schoener's D, and niche dynamics. Both algorithms achieved satisfactory calibration, but temporal projections frequently failed to reproduce the climatic niche represented by independent records. Although BRT consistently outperformed Maxent, forecasting performance depended primarily on the proportion of the realised climatic niche represented during calibration rather than on the number of occurrence records. Models with excellent calibration statistics often showed poor temporal transferability, whereas models trained on progressively more complete niche representations produced substantially more reliable forecasts. Our results demonstrate that reliable model calibration does not guarantee reliable forecasts of future tick distributions, suggesting that incomplete niche representation acts as a major constraint on long-term SDM projections under the evaluated framework.