Search

• We propose a new forest prediction model and use actual deforestation data for optimization and verification. • Use the XGBoost method to reconstruct deforestation intensity data. • Deforestation hotspots in China are mainly concentrated in the southern and southeastern regions. • Our model provides certain reference information for forest disaster prevention and forest management. Forests are important for the global carbon cycle, hydrothermal balance, and climate change. Human activities can exert a significant impact on forest ecosystems, thereby having the potential to alter their physical and chemical properties and thus affecting carbon, water, and heat budgets, and climate change. The historical reconstruction of the disturbance of global forests can help us understand the processes and patterns of human activities and global change. In this paper, we construct a deforestation prediction model using a Spearman correlation coefficient and implement the XGBoost method, using Python 3.6, for the reconstruction of deforestation intensity data from 2000 to 2019. Secondly, the selection of the driver indicators is done by using extreme difference regularization to unify the magnitude, and the potential deforestation area risk index is calculated in the form of equal weights. Finally, the actual deforestation data were used for optimization and validation. The model shows that the deforestation hotspots are mainly concentrated in the southern and southeastern regions of China and that there are large differences in deforestation in different provinces. In the future, the fine spatial and temporal patterns of deforestation in China during the historical period can be quantitatively reconstructed, which can provide some reference information for forest disaster prevention and forest management in China. [ABSTRACT FROM AUTHOR]