Impacts of data sources on the predictive performance of species distribution models: A case study for Scomber japonicus in the offshore waters southern Zhejiang, China

As our understanding of ecology deepens and modeling techniques advance, species distribution models have grown increasingly sophisticated, enhancing both their fitting and predictive capabilities. However, the dependability of predictive accuracy remains a critical issue, as the precision of these predictions largely hinges on the quality of the base data. We developed models using both field survey and remote sensing data from 2016 to 2020 to evaluate the impact of different data sources on the accuracy of predictions for S. japonicus distributions. Our research findings indicate that the variability of water temperature and salinity data from field suvery is significantly greater than that from remote sensing data. Within the same season, we found that the relationship between the abundance of S. japonicus and environmental factors varied significantly depending on the data source. Models using field survey data were able to more accurately reflect the complex relationships between resource distribution and environmental factors. Additionally, in terms of model predictive performance, models based on field survey data demonstrated greater accuracy in predicting the abundance of S. japonicus compared to those based on remote sensing data, allowing for more accurate mastery of their spatial distribution characteristics. This study highlights the significant impact of data sources on the accuracy of species distribution models and offers valuable insights for fisheries resources management.