A New Drought Monitoring Index on the Tibetan Plateau Based on Multisource Data and Machine Learning Methods
Meilin Cheng,Lei Zhong,Yaoming Ma,Xian Wang,Peizhen Li,Zixin Wang,Yuting Qi
DOI: https://doi.org/10.3390/rs15020512
IF: 5
2023-01-16
Remote Sensing
Abstract:Drought is a major disaster over the Tibetan Plateau (TP) that exerts great impacts on natural ecosystems and agricultural production. Furthermore, most drought indices are only useful for assessing drought conditions on a coarse temporal scale. Drought indices that describe drought evolution at a fine temporal scale are still scarce. In this study, four machine learning methods, including random forest regression (RFR), k-nearest neighbor regression (KNNR), support vector regression (SVR), and extreme gradient boosting regression (XGBR), were used to construct daily drought indices based on multisource remote sensing and reanalysis data. Through comparison with in situ soil moisture (SM) over the TP, our results indicate that the drought index based on the XGBR model outperforms other models (R2 = 0.76, RMSE = 0.11, MAE = 0.08), followed by RFR (R2 = 0.74, RMSE = 0.11, MAE = 0.08), KNNR (R2 = 0.73, RMSE = 0.11, MAE = 0.08) and SVR (R2 = 0.66, RMSE = 0.12, MAE = 0.1). A new daily drought index, the standardized integrated drought index (SIDI), was developed by the XGBR model for monitoring agricultural drought. A comparison with ERA5-Land SM and widely used indices such as SPI-6 and SPEI-6 indicated that the SIDI depicted the dry and wet change characteristics of the plateau well. Furthermore, the SIDI was applied to analyze a typical drought event and reasonably characterize the spatiotemporal patterns of drought evolution, demonstrating its capability and superiority for drought monitoring over the TP. In addition, soil properties accounted for 59.5% of the model output, followed by meteorological conditions (35.8%) and topographic environment (4.7%).
environmental sciences,imaging science & photographic technology,remote sensing,geosciences, multidisciplinary