Modelling Above-Ground Biomass Using Machine Learning Algorithms in Mangrove Forests of Peninsular Malaysia

¹ Earth Observation Centre, Institute of Climate Change, Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor Malaysia
² Forest Research Institute Malaysia, 52109 Kepong, Selangor Malaysia
³ Faculty of Science and Marine Environment, Universiti Malaysia Terengganu, 21030 Kuala Nerus, Terengganu Malaysia

^* Corresponding author: wanshafrina@ukm.edu.my

Abstract

Mangrove forests are crucial for carbon sequestration and biodiversity conservation but are threatened by anthropogenic effects and climate change. Although restoration efforts have been initiated, their effectiveness remains uncertain due to the absence of robust monitoring and evaluation mechanisms. This study investigates machine learning algorithms for modelling aboveground biomass (AGB) in mangrove forests across Peninsular Malaysia. Data on tree diameter at breast height (DBH) and species were collected in Sungai Pulai, Sungai Johor, and Sungai Merbok. Combined with remote sensing data, the performance of Extreme Gradient Boosting (XGBoost), Random Forest (RF), Boosted Tree (BT), and Support Vector Machine (SVM) was compared, chosen for their ability to analyze complex patterns and predict accurately. The results indicated that XGBoost outperformed the others by achieving higher R² values of 0.97, lower mean absolute error (MAE) of 6.80 Mg ha^-¹, and root mean squared error (RMSE) of 26.74 Mg ha^-¹, demonstrating superior accuracy and predictive performance. This study also shows similar above-ground carbon (AGC) values across the study areas and in comparison with previous studies. XGBoost’s robust capacity for estimating AGB and AGC highlights its potential to significantly improve mangrove forest management and conservation efforts on a broader scale.