ANALYSIS OF CHL-A MODELS FOR THE NORTHERN COAST OF SÃO PAULO - BRAZIL

Código

XXVI-SBRH0533

Autores

Júlio César Pimenta dos Santos, Bruno Rech, Daniel Andrade Maciel, Rogerio Flores Junior, Élcio Hideiti Shiguemori, Cláudio Clemente Faria Barbosa, Evlyn Márcia Leão de Moraes Novo, Áurea Maria Ciotti

Tema

STE110 - Sensoriamento remoto da água: avanços técnicos-científicos e aplicações na nova era de disponibilidade de dados

Resumo

This study evaluated different algorithms for estimating chlorophyll-a (Chl-a) concentration using remote sensing reflectance (Rrs) data from the northern coast of São Paulo, Brazil. The evaluated models included: a) Normalized Difference Chlorophyll Index (NDCI), b) Two Bands Algorithm (2BDA), c) Ocean Color 3 (OC3), d) Ocean Color 4 (OC4), e) Random Forest (RF), and f) Convolutional Neural Network (CNN). The NDCI and 2BDA models showed the poorest performance, with R² values around 0.05. The OC3 and OC4 models performed slightly better, with R² values around 0.12. The best results were achieved by the machine learning models, RF and CNN, both reaching R² values close to 0.60. These results demonstrate that, even with a limited dataset, the RF and CNN models provided satisfactory performance, outperforming the empirical models evaluated in this study.