| As one of the most destructive disasters in nature,drought can cause great damage to agricultural production,socio-economic development,and human life safety.In the context of intensifying global climate change,the evolution pattern and prediction of meteorological droughts have become a hot spot and difficult issue in the field of climate change.It is of great theoretical and practical significance to clarify the evolution pattern,spatial and temporal distribution characteristics,and future development trend of droughts,especially for arid and semi-arid regions with water shortage,for the reasonable regional water resources allocation and prevention and mitigation of droughts.Although a large number of studies have focused on drought prediction,most of them have been conducted separately in two dimensions,spatial and temporal,while the characteristics and changes of drought are analyzed from the overall spatial and temporal scales in the future.Therefore,this study takes China as the study area,used the station and grid-point data continuously observed in the past 40 years to calculate several meteorological drought indices and comprehensively analyzes the evolution pattern of meteorological drought,quantitatively analyzes the applicability of different time series prediction algorithms to the drought indices,and deeply reveals the characteristics of the spatial and temporal distribution of meteorological drought in seven sub-regions of China,to provide an important theoretical basis for future drought warning.The research contents and conclusions of this paper mainly include.(1)The spatial and temporal evolution characteristics of different meteorological drought indices in seven major sub-regions of China are studied,and the correlations among drought indices are analyzed using methods such as correlation analysis.The results show that the differences between SPI and SPEI on a monthly scale are larger in the north than in the south,and the correlations between PDSI and SC-PDSI are the highest,and the meteorological drought in the northwest desert region and the steppe region of Inner Mongolia became more and more severe during 1999-2004.In recent years,the eastern meteorological drought in the northeast humid semi-humid temperate region and north China humid semi-humid warm temperate region showed oscillating trends.(2)The applicability of six machine learning algorithms in the prediction of different meteorological drought indices time series is compared,and a combined model is proposed to predict the meteorological drought indices by combining the advantages of linear and nonlinear models.The results show that the prediction accuracy of a single model in each index is lower than the combined model,and for SPI-1 and SPEI-1 the prediction accuracy is lower than that of PDSI and SC-PDSI,indicating that the short-term drought prediction accuracy is difficult to predict,and it is difficult to find the changing pattern compared with Palmer drought index.The prediction accuracy of the six algorithms was higher in both PDSI and SC-PDSI.(3)Combining the spatiotemporal cube model with the method of spatiotemporal pattern mining,we explore the spatiotemporal heterogeneity of meteorological drought and make predictions on the future development trend of meteorological drought from the overall spatiotemporal scale.The results show that SPI-1 and SPEI-1 have the largest number of clusters,while PDSI,SC-PDSI,and Z-index have relatively fewer clusters,indicating that short-term drought monitoring varies widely across regions,while Palmer drought index varies widely from north to south and less in the rest of the regions.The analysis of emerging spatial and temporal hotspot analysis shows that in China there are oscillating cold spots and emerging cold spots in the northeast and southwest regions,which represent a great possibility of severe meteorological drought in this region in the future. |
PDF Full Text Request