POSTER SESSION 3
15:30–17:30, Wednesday, September 16
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
ABSTRACT 1179 | POSTER 143
BRAZA: A BIO-OPTICAL DATASET FOR LARGE-SCALE MONITORING OF BRAZILIAN INLAND AND COASTAL WATERS USING SATELLITE REMOTE SENSING
Monitoring water quality is essential for understanding aquatic biogeochemical cycles and assessing anthropogenic impacts on ecosystems. Brazil hosts approximately 180,000 km² of freshwater systems across highly heterogeneous environments, including turbid Amazonian waters and reservoirs with varying trophic conditions. These systems are increasingly affected by stressors such as illegal mining, deforestation, dam construction, agriculture, and urbanization. Although in situ monitoring programs are critical, they lack sufficient spatial and temporal coverage, highlighting the need for satellite-based approaches. However, the wide variability in optical properties across Brazilian waters challenges the development of robust and transferable algorithms. To address this, we leveraged the BRAZA dataset, a comprehensive bio-optical database comprising 2,895 stations across 128 aquatic systems. It includes concurrent measurements of remote sensing reflectance (Rrs) and key water quality variables such as chlorophyll-a, Secchi disk depth (Zsd), CDOM absorption (aCDOM), suspended particulate matter, and dissolved organic carbon. The dataset integrates contributions from 17 institutions, providing a strong foundation for algorithm development in optically complex waters. To demonstrate dataset robustness, we evaluated machine learning models (MDN, Random Forest, Support Vector Machines, and XGBoost) to estimate Zsd from Sentinel-2/MSI imagery. Models trained with in situ data achieved median symmetric accuracy errors of 11.61% (MDN) and 13.74% (Random Forest) (N = 779; Zsd: 0.1–30 m). Applied to Sentinel-2/MSI data within the Brazil Data Cube infrastructure, the best model achieved 23% accuracy. These models are being integrated into the open-access MAPAQUALI platform, enabling near-real-time water quality monitoring and large-scale environmental assessment in Brazil.
Daniel Maciel, INPE, [email protected], https://orcid.org/0000-0003-4543-5908
Claudio Barbosa, INPE, [email protected], https://orcid.org/0000-0002-3221-9774
Evlyn Novo, INPE, [email protected], https://orcid.org/0000-0002-1223-9276
Aurea Ciotti, CEBIMAR/USP, [email protected], https://orcid.org/0000-0001-7163-8819
Vitor Martins, Mississippi State University, [email protected], https://orcid.org/0000-0003-3802-0368
Poster Session | 1 | 2 | 3 | 4 | Schedule at a Glance
