Fishery forecasting is an important component of fisheries science. It has vital significance for fisheryproduction and management. is an important target for Chinese squid jigging fleets in the southwestAtlantic Ocean. Some previous studies employed various approaches to forecast optimal fishing groundsbased on environmental factors, such as sea surface temperature (SST), sea surface height (SSH), and chlorophyll-aconcentration (Chl-a). These approaches use experiential knowledge obtained from historical fisheries andenvironmental data to forecast fishing grounds, but there is no research on how to select the most appropriate spatialand temporal scales or environmental data to forecast models. In this study, models were constructed based on differentenvironmental factors with various spatial and temporal scales to better forecast optimal fishing grounds inthe southwest Atlantic Ocean.In this study, historical fishing data from Chinese mainland squid jigging fleets from 2003 to 2011, sea surfacetemperature (SST), sea surface height (SSH), and chlorophyll-a (CHL-a) data were divided into different temporal andspatial scales. Temporal scales included monthly, ” 0.25°°° × environmental factors were divided into four categories, including I (SST), II (SST and SSH),III (SST and Chl-a), and IV (SST, SSH, and Chl-a). A total of 24 models were constructed using error backpropagationartificial neural network; model training, validating, and testing were completed in Matlab. Mean square error andaverage relative variance (ARV) were used to evaluate accuracy, and sensitivity analyses were used to evaluate theinterpretation of models for fishing grounds. The results indicated that the fishery forecasting model with maximum accuracy and minimum ARV wasconstructed by two models, one was with a 1.0° ×SST”monthly” 0.25°”environmental factor. Sensitivity analyses using those two models showed that models with different temporal andspatial scales expressed different habitat suitability. This research revealed that when models had the same temporal scales, there were no proportional or inverserelationships between spatial scale and model accuracy, when models had same spatial scales, there was no proportionalor inverse relationships between temporal scale and model accuracy. Additionally, more environmental factors were notalways better; sometimes more environmental factors increased the difficulty of model fitting. In summary, consideringthe temporal and spatial scale of fishing and environmental data was needed to construct fishing ground forecastingmodels for