Remote Sensing Modeling and Applications in Drought Monitoring Based on XGBoost and Fusion of Multi-dimensional Spatiotemporal Data

Southwest China is an important ecological conservation area in the country. Its complex climatic and geographical conditions frequently lead to drought events. Accurately understanding the spatial distribution and trends of drought is essential for protecting the ecological environment in this region. This study utilizes the eXtreme Gradient Boosting algorithm to construct a remote sensing drought monitoring model (eXtreme Gradient Boosting Drought Monitor, XGBDM). The model considers vegetation status, land surface conditions, climate variables, and environmental factors using remote sensing drought indicators, which characterize multidimensional feature variables. The model was used to monitor drought conditions in Southwest China from 2001 to 2020. Typical drought events and soil moisture data were selected to evaluate the accuracy of the model. Combining Theil- Sen Median trend analysis, Mann- Kendall significance test, Hurst index, and the center of gravity shift model, this study reveals the spatio-temporal variation characteristics of drought, future trends, and the shift in drought center of gravity in Southwest China. Results indicate that: (1) The XGBDM model accurately monitors drought events in Southwest China across different seasons, with model accuracy indicators R2 ranging from 0.816 to 0.897, MAE from 0.200 to 0.283, RMSE from 0.296 to 0.424, and correlation between the model and soil moisture ranging from - 0.60 to 0.86. Compared to the station- based SPEI- 3 monitoring method, the XGBDM model shows a higher correlation with soil moisture and can more accurately reflect the spatial distribution details of drought conditions. (2) Temporally, from 2001 to 2020, the annual average XGBDM values in Southwest China exhibit an overall fluctuating declining trend, indicating an exacerbation of drought conditions, particularly in spring and summer, while alleviating in autumn and winter. Spatially, the changing rate of XGBDM values in Southwest China presents a pattern of "high in the north and low in the south" in spring and summer, and "high in the south and low in the north" in autumn and winter. The proportion of areas with increasing drought varies by season, with 69.17% in spring, 76.02% in summer, 34.43% in autumn, and 47.5% in winter, respectively. (3) The XGBDM values in Southwest China generally exhibit weak anti-persistence changes. In spring, summer, and winter, future drought conditions are mainly alleviated, with the proportion of areas transitioning from exacerbation to alleviation ranging from 28.44% to 63.82% across different seasons. The area with persistently exacerbating drought conditions is highest in spring(17.97%), while the highest proportion of persistently alleviating conditions is in winter (15.92%). (4) The center of drought conditions from 2001 to 2020 is mainly located in the central part of the study area, following a northwest-to-southeast distribution pattern. In the future, there is a higher probability of the drought center migrating in the northwest- to- southeast direction. These findings can serve as a theoretical basis for drought monitoring and management in Southwest China. © 2024 Science Press. All rights reserved.