Yield Prediction of Marine Algae and Shellfish Aquaculture Based on Multivariate Remote Sensing

The prevailing reliance on statistical yearbooks for current aquaculture yield data is problematic due to their inherent delays and biases, which significantly impede effective aquaculture management and optimal resource allocation. For the first time, this article proposes a sparse broad learning system (SBLS) for algae and shellfish yield based on multivariate remote sensing data, using extensive marine data analysis to enhance interpretability and accuracy. First, we extracted shellfish and algae aquaculture area in Dalian from 2013 to 2021 and obtained monthly data on six regional environmental variables—sea surface temperature (SST), chlorophyll concentration (Chl-a), organic carbon, turbidity, wind speed, and precipitation. Next, the correlation between these variables and yield is assessed using Spearman’s and Kendall’s correlation coefficients and distance coefficients, filtering the most relevant months and analyzing the biological growth characteristics of algae and shellfish. Finally, separate prediction models for shellfish and algae annual yields per unit area are constructed using the SBLS method with multiple windows, refining yield categories to obtain the density distribution of Dalian’s 2022 annual yield. The results show highly accurate predictions, with a mean absolute percentage error (MAPE) under 10%. In 2022, the total algal yield in the aquaculture area is 494044.58 tons, and the total shellfish yield is 391691.26 tons. This study represents a significant improvement over traditional methods relying on outdated yearbooks and can serve as an effective monitoring tool for aquaculture management in complex coastal environments.