ABSTRACT Species distribution models (SDMs) are widely used to predict the spread of invasive species, yet their accuracy over time and the influence of climate data resolution remain unclear. Here, we examine the capacity of SDMs to predict the distribution and short‐term expansion of the invasive gall wasp Dryocosmus kuriphilus , and compare the performance of locally sourced, annually resolved climate data with global climatic datasets. We modelled the annual distribution and expansion of D. kuriphilus using SDMs. Three climate datasets: regional interpolations, temporally explicit CHELSA time series, and long‐term CHELSA averages, were compared to test their influence on model accuracy. Habitat suitability was estimated with GLM, MaxEnt, and Random Forest, and model performance was evaluated with the Boyce index. Temporal transferability was assessed by projecting early‐year models to subsequent years and analysing accuracy patterns via multifactorial ANOVA. Model accuracy, measured with the Boyce index, improved over time, surpassing 0.7 for most datasets and algorithms. Differences among climate datasets were minor, although regional data slightly enhanced early predictions. MaxEnt consistently achieved the highest and most stable performance, while Random Forest and GLM were more variable. Models projected beyond their calibration years rapidly lost predictive power, showing limited temporal transferability. ANOVA confirmed that neither climate dataset nor algorithm significantly influenced accuracy, though MaxEnt performed best overall. SDMs are valuable tools for invasive species management when used adaptively. Accuracy improves with more occurrence data but declines when projecting future expansion, highlighting the need for regular updates. Incorporating fine‐scale, temporally explicit data enhances early detection, monitoring, and rapid‐response interventions. Continuously updated SDMs support timely management decisions, including invasion monitoring, prioritisation of actions, and integration of biological control or other mitigation strategies.
Pérez‐Girón et al. (Fri,) studied this question.