AQUACULTURE MAPPING: A LITERATURE REVIEW ON CLASSICAL METHODS AND MACHINE LEARNING APPROACHES

Resumo

The aquaculture sector has been expanding rapidly due to the growing demand for animal-based food and the pressure on natural fisheries resources. This expansion is driven by advances in farming technologies and more sustainable practices. In this context, geotechnologies have proven essential for monitoring and mapping aquaculture areas. This article presents a literature review integrating bibliometric and systematic approaches to mapping aquaculture areas using remote sensing, Geographic Information Systems, and machine learning techniques. Based on 355 publications between 2007 and 2025, the bibliometric analysis revealed a significant increase in scientific output, particularly after 2018, with notable contributions from Asian institutions. From a systematic perspective, 35 studies were analyzed, identifying the frequent use of satellite images from LANDSAT 8, Sentinel-2, and ZY1-02D, combined with indices such as NDVI, NDWI, and SAVI. Classical methods such as threshold segmentation, edge detection, and Tasseled Cap are still employed. Still, there is a growing adoption of machine learning algorithms, including Random Forest, SVM, LVQ, and neural networks. Despite advancements, challenges remain in result validation, and research in tropical regions is scarce. The article concludes that combining traditional and modern methods can enhance the accuracy and applicability of mapping efforts, contributing to the sustainable management of aquaculture on a global scale.