Predicting Mangrove Carbon Stock in Riau Province: A Machine Learning Approach Using Sentinel-2 Satellite Imagery

Mangrove forests play a role in preserving biodiversity and capturing carbon, which helps combat climate change. Unfortunately, these vital ecosystems are facing increasing threats from deforestation in areas. This research focuses on developing efficient methods to monitor and safeguard mangrove forests by using satellite images and advanced machine learning techniques and also predicting mangrove carbon stock. Various machine learning models like Random Forest, Gradient Tree Boosting and K Nearest Neighbors were compared for their performance in distinguishing mangrove forests from land types in Riau Province, Indonesia. The objective is to classify each pixel in the imagery as mangrove or non-mangrove and estimate the associated carbon stock for conservation purposes. Among the models tested the Random Forest model showed the accuracy of $\mathbf{9 9. 3 9 \%}$ making it the preferred choice for calculating carbon stock in the mangrove areas. Based on this models’ predictions the estimated value of carbon stock was approximately IDR $\mathbf{1, 3 9 3}, 887,913.11$. The research shows how well integrating high resolution satellite imagery with machine learning techniques for mangrove monitoring and carbon stock estimation.