Effects of Different Spatial Extents of Occurrence Data on Biomod2-Based Species Distribution Modeling and Prediction: A Case Study of the Potential Distribution of Spodoptera frugiperda in China

Climate change is reshaping species distributions worldwide, making reliable prediction of invasive species increasingly important for ecological risk assessment and pest management. However, species distribution models (SDMs) calibrated with regional occurrence records may underestimate potential suitable habitats because of niche truncation. Here, we evaluated the effects of occurrence data extent on SDM predictions using the globally invasive agricultural pest Spodoptera frugiperda (J. E. Smith, 1797) (Lepidoptera: Noctuidae). Global occurrence records (1763 records) and China-specific occurrence records (946 records) were combined with climatic, topographic, and vegetation variables to construct ensemble SDMs using the Biomod2 platform under current and future climate scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5). Model performance, environmental variable contributions, habitat suitability patterns, and centroid shifts were compared between the two datasets. The global model showed excellent predictive performance (True Skill Statistic (TSS) = 0.851, Receiver Operating Characteristic (ROC) = 0.977) with balanced contributions from multiple environmental variables, whereas the China model achieved only acceptable performance (TSS = 0.690, ROC = 0.905) and relied predominantly on Bio06. Both models consistently identified southern China and the Huang-Huai Plain as the core highly suitable regions. Future projections revealed under higher emission scenarios (SSP5-8.5), particularly toward northern and western China. Centroid analyses further indicated consistent northwestward shifts under higher emission scenarios (SSP5-8.5), whereas low-emission scenarios produced less directional and more variable centroid movements. These results demonstrate that the spatial extent of occurrence records exerts a greater influence on SDM predictions than differences among modeling algorithms. Although models of China may underestimate invasion risk because invasive populations have not reached ecological equilibrium, the spatial stability of highly suitable habitats and centroid migration direction provide robust indicators for cross-model comparisons. This study provides methodological guidance for occurrence data selection and improves invasion risk assessment and management of S. frugiperda under climate change.