Integration of Spatial Analysis, Random Forest, and Clustering for Mapping and Predicting Sustainable Fisheries Activities

Abstract

This research integrates fish catch detection considering the ocean environment using the three primary disciplines of spatial science, machine learning, and cluster analysis. The scope of this research will encompass the analysis of four species of tuna and specific environmental metrics such as water temperature, chlorophyll-a, salinity, sf, and sea level. Spatial techniques will be used to analyze geographic the geographic diversity of the region. Environmental variables to be used with the Random Forest algorithm to make predictions for the catch while K-Means is to be used to characterize the oceanographic dominant variables clusters. The most variable parameters for the sea level and Chlorophyll were the most variable parameters used for the catch predictions especially for the AL and YF species. The catch of BE, AL and YF species too has shown significant declining trends over time thus suggesting the probable overfishing of these species. A comprehensive system of this kind provides the framework to monitor fisheries systems with model predictions that aid in determining which areas in these fisheries systems require sustainable management.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint