Applying Remote Sensing, Google Earth Engine, and Machine Learning to Predict the Carbon Sequestration Potential of Sundarbans Mangroves

Abstract

Mangrove forests are pivotal in mitigating climate change impacts due to their exceptional carbon sequestration capabilities and their role as carbon sinks. Mangroves are designated blue zones, storing two to four times more carbon dioxide than global rates observed in mature tropical forests. This study utilizes both radar (Sentinel 1) and optical (Sentinel 2) remote sensing datasets. It compares various machine learning algorithms, such as Random Forest, SVM, and CART, to predict the mangrove distribution and allometric equations to estimate the aboveground biomass and the amount of carbon sequestration for the Sundarbans mangrove forest without requiring time-consuming field measurements. It was found that the Random Forest machine learning classifier yielded the highest accuracy of 0.984 with a Kappa metric of 0.962. Using this classifier, the aboveground biomass was calculated to be 125 Mg/ha and the amount of carbon sequestration was 257 Mg/ha per year. The aboveground biomass found was consistent with prior studies using traditional field work. The carbon sequestration values will aid in highlighting the importance of mangroves as blue-carbon storages and help in accurately monitoring and preserving these vital ecosystems efficiently on a global scale.