Most parts of Xinjiang are in the arid area, which caused people become accustomed to the drought problems in the region, so insufficient attention was paid to its extreme drought problem. However, as the largest inland province in China, Xinjiang occupies around 1/6 of China's land mass. The regional extreme drought in Xinjiang province persistently induces important influences on the agriculture and stock, the water resources，the national unity, the regional stability and the prosperity of Silk Road Economic Belt. It is obviously that studying the extreme drought problems in Xinjiang province will not only has an important scientific value, but also has an essential practical significance.
This paper selects the Weibull distribution as the most suitable distribution for detecting the extreme drought events in Xinjiang province, improve the original SPI indexes and make the model more suitable and useful to study Xinjiang’s extreme drought problems. By using the spatial distribution of improved SPI indexes, we divide the extreme droughts into three spatial distribution types: Northern Xinjiang type, Southern Xinjiang type and Eastern Xinjiang type, and clarify the corresponding circulation situations and physics characteristics.The numerical simulations results show that KUO scheme is good for the extreme drought events of Northern Xinjiang type and Eastern Xinjiang type, and Grell scheme is more matched to the extreme drought events of Southern Xinjiang type.Accordingly, we can optimize the parameterization schemes and distinctly improve the capability of the model system for regional extreme drought climate simulation in Xinjiang province. Also, the drought prediction models of different regions and seasons are establish based on PSO_LSSVM. All of these can provide the remarkable scientific basis and technical support for the detection,forecast and disaster prevention of the extreme drought events in Xinjiang province.