A study on the remote sensing estimation and spatiotemporal distribution patterns of aboveground biomass in savanna grasslands of the Yuanmou dry-hot valley

Caiying CHEN¹Wenfei Xi^1*Guangxiong HE²Haidong FANG²Liangtao SHI²Yongzai ZHUANG¹Zitian Ding¹Junqi GUO¹Xuewen YUE¹Kunwu YANG¹

• ¹Yunnan Normal University, Kunming, China
• ²Tropical Eco-Agriculture Research Institute, Yunnan Academy of Agricultural Sciences, Kunming, China

The dry-hot valley is a typical ecologically fragile region, where savanna grasslands-one of the dominant vegetation types-fulfill dual roles in ecological regulation and productive utilization. Timely and accurate acquisition of aboveground biomass (AGB) information for such representative vegetation is crucial for formulating scientifically sound ecological conservation strategies and effective grassland management practices. However, conventional AGB estimation methods often fail to adequately address multicollinearity among predictor variables, and approaches that rely solely on vegetation indices are generally insufficient under complex surface conditions.In this study, the Yuanmou dry-hot valley was selected as the research area. A comprehensive training dataset was constructed by integrating remote sensing indices, topographic variables, meteorological factors, and biodiversity metrics. To identify the most informative predictors, three feature selection techniques were employed: Spearman’s rank correlation analysis, Recursive Feature Elimination, and Least Absolute Shrinkage and Selection Operator regression. These selected variables were then used to train and evaluate three representative machine learning models: Random Forest, Gradient Boosting Decision Tree, and Extreme Gradient Boosting, for the inversion of AGB.The results indicate that the RF model optimized via RFE exhibited the best performance, achieving a coefficient of determination of 0.6975, a root mean square error of 89.3436 g/m², and a mean relative error of 0.7282. From 2019 to 2024, the AGB of savanna grasslands in the Yuanmou dry-hot valley demonstrated an overall increasing trend, along with a more spatially uniform distribution.These findings suggest that the proposed approach is well-adapted to ecologically complex environments such as dry-hot valleys and can substantially enhance the accuracy of remote sensing-based AGB estimation in savanna ecosystems. The method offers significant potential for applications in ecological monitoring and sustainable resource management.