Abstract
Aim Joint Species Distribution Models (JSDMs) have become a critical tool in community ecology research, with a wide scope of application that is continuously expanding. However, inferring interspecies relationships from co‐occurrence data remains a challenge. This study examined the impact of varying spatiotemporal scales on JSDMs, with a focus on model stability and the evaluation of interspecies relationships. Location The northwest Pacific Ocean. Taxon Japanese sardine; Chub mackerel; Neon flying squid. Methods To comprehensively evaluate the impact of varying spatiotemporal scales on JSDMs, this study was designed using two temporal scales (monthly and annual), four spatial scales (0.25°, 0.5°, 1°, and 2°), and four different JSDMs (Bayescomm, HMSC, Boral, and Gjam). Using three economically important pelagic fish species from the northwest Pacific Ocean as examples—Japanese sardine ( Sardinops melanostictus ), chub mackerel ( Scomber japonicus ), and neon flying squid ( Ommastrephes bartramii )—we compared the performance of the models across 32 different spatiotemporal scales. Results Our results indicate that the spatiotemporal scale significantly affects the performance of JSDMs, with notable differences among the models. As spatial scales become finer and temporal scales longer, model simulation and prediction performance improve, and stability increases. Moreover, spatial scale has a substantial impact on the evaluation of interspecies relationships, as finer spatial scales can better evaluate interspecific relationships. Among the models, HMSC demonstrated better balancing performance, while the Boral model showed the least stability. Overall, the optimal JSDM identified was the HMSC model with an annual temporal and 0.25° spatial scale. Main Conclusions Spatiotemporal scales have a significant impact on JSDMs, particularly when inferring the strength of interspecies relationships. Therefore, it is recommended that researchers carefully design the model based on the spatiotemporal scales of their data and select the optimal model to enhance predictive performance and improve the interpretative validity of the results.