Extraction of eucalyptus age and estimation of its aboveground biomass in China with the integration of empirical model and machine learning algorithm

Eucalyptus plantations are extensively distributed in tropical and subtropical regions and play important roles in timber production, economic development, and regional carbon cycles. Due to its fast growth and short rotation periods, mapping of high spatial resolution eucalyptus age and its aboveground biomass (AGB) distribution becomes an urgent task, but such products are unavailable due to the difficulty in distinguishing eucalyptus from other tree species and lack of suitable methods to accurately estimate eucalyptus age and AGB. This study aims to develop a new approach to extract eucalyptus age and a new procedure to estimate AGB through integration of an empirical model and machine learning algorithm in subtropical and tropical regions of China. The eucalyptus distribution was first developed using Sentinel-2 imagery and its forest age in unit of months was then generated with a continuous threshold-based decision strategy based on monthly median composites of normalized difference vegetation index (NDVI) and the difference between NDVI and the normalized burn ratio (NBR) (DIF) from Landsat and Sentinel-2 time series data. The Chapman-Richards function was used to build a growth model based on eucalyptus age, and SHapley Additive exPlanations (SHAP) approach was used to identify key environmental factors for use in the AGB modeling procedure. The results showed that a root mean square error (RMSE) of 1.54 years was obtained, much lower than existing age products. About 77% of eucalyptus plantations were four years or younger. The predicted eucalyptus AGB in China was 217.41 million tons in 2023, with RMSE of 21.18 t·ha⁻¹ and relative RMSE (RMSEr) of 22.41%. This study provided the first products of eucalyptus distribution with 10 m spatial resolution, the estimated age and AGB distributions with 30 m resolution in China in 2023. The proposed framework provides a new insight for age extraction and AGB estimation for other tree species. The results from this research provide a fundamental data source for eucalyptus forest resource management, carbon assessment, and policy-making.