5. Research on using Landsat 8 imagery and machine learning algorithms in monitoring changes in area and carbon absorption of mangrove forests in Thai Binh province, Vietnam, during the period 2021 - 2025

Abstract

Mangrove forests in the Red River delta play a crucial role in coastal protection, biodiversity, and carbon sequestration. This study uses Landsat 8 OLI/TIRS imagery and Random Forest classification to monitor mangrove forest changes in Thai Binh province, Vietnam, over three periods: 2021, 2023, and 2025. A multi-channel composite (B2 - B7) and vegetation indices (NDVI, NDWI, MNDWI, SAVI, NDBI) were used for classification. Overall accuracy ranged from 86.2 to 91.4 %, with Kappa coefficients from 0.83 to 0.89. Results show that the mangrove forest area increased by 250.56 ha from 2021 to 2025, with a transient decrease of 59.99 ha (2023 - 2025). Dominant species such as Kandelia obovata and Sonneratia caseolaris contributed an estimated annual carbon absorption of 349,411 tonnes of CO₂, totaling 1,747,055 tonnes of CO₂ over 5 years. This study demonstrates the effectiveness of open-access satellite data and machine learning in long-term mangrove forest monitoring and supports carbon credit certification within international frameworks.