Research On Comprehensive Drought Index Construction And Drought Prediction In The Middle Section Of The North Foot Of Tianshan Mountains

Drought disaster has the characteristics of long duration and great destructiveness.Due to the serious consequences caused by drought disaster,it is very necessary to carry out timely monitoring,analysis and assessment of drought,so as to realize rational management and efficient utilization of water resources.In this paper,standard evapotranspiration index and standard runoff index are calculated in the middle section of the north foot of tianshan mountain.The variation trend of the comprehensive drought index was analyzed by using non-parametric trend analysis and periodic analysis.Copula function is used to construct the multivariate joint distribution of hydrometeorological drought,and the probability and recurrence period of the characteristic variables of hydrometeorological drought are analyzed.Combined with atmospheric circulation factors,BP neural network prediction model and multiple linear regression prediction model are constructed.The main conclusions are as follows:(1)Hydrologic drought(SRI)and meteorological drought(SPEI)in the Jiangjunmiao hydrology station have an alternating trend,and generally hydrologic drought is the main trend.Similar to the fluctuation of the hydrology station of Kensiwate and the hydrology station of Yingxiongqiao,there is an obvious transition process from hydrology drought to meteorological drought,and the change range of the comprehensive drought index CI is between SRI and SPEI.Through Pearson correlation coefficient analysis,CI has a good correlation with precipitation,temperature and runoff,so it is necessary to select CI for drought analysis in the middle section of the northern foothills of Tianshan Mountains.(2)The comprehensive drought index CI of Jiangjunmiao hydrology station varies within the confidence interval on the whole,and the change trend is not obvious.The CI at Kensiwate hydrology station detected a mutation at 2000 a.Yingxiongqiao hydrologic station CI underwent a mutation in 1988 a.The first main period of CI change of Jiangjunmiao hydrology station is 27 a.The first main period of CI change in the Kensiwate hydrologic station is 48 a.The first main period of CI change of Yingxiongqiao hydrology station is 19 a.(3)The correlation between the drought characteristic variables of the three hydrologic stations in the middle section of the north foot of Tianshan Mountains is good,which passes the significance test of 0.01.The Weibull distribution was the best fitting distribution of the drought duration of the Jiangjunmiao station,and the optimal fitting distribution of the drought intensity was log-logistic(3P).The univariate optimal fitting distribution of drought duration and drought intensity in Kensiwate station is Gamma(3P)and generalized pareto distribution,respectively.The univariate optimal fitting distribution of the drought duration and drought intensity of Yingxiongqiao station is Weibull and lognormal distribution,and after k-s test,the fitting of the edge distribution function of the univariate passes the significance level of 0.05,indicating that the distribution fitting of the drought duration and drought intensity of each station is good.(4)The optimal Copula functions of the 2-d joint distribution of Jianhjunmiao station,Kenswate station and Yingxiongqiao station are Gumbel function,Gaussian function and Joe function respectively.The probability of drought events with drought intensity greater than 4 and lasting for more than 8 months in Jiangjunmiao station is 0.1,that of Kensiwate station is less than 0.1,and that of Yingxiongqiao station is 0.2.At the level of the same single variable recurrence period,the drought duration and drought intensity of Yingxiongqiao station are both greater than those of the other two stations,which indicates that the drought degree of Yingxiongqiao station is the most serious,and the drought events of Yingxiongqiao station are mainly manifested as long duration and high intensity.(5)The atmospheric circulation factors were screened by remote correlation analysis.The results of remote correlation analysis and the comprehensive drought index CI of each station were introduced into the BP neural network prediction model and the multiple linear regression prediction model for drought prediction.The prediction accuracy of the BP neural network prediction model is better than that of the multiple linear regression model.BP neural network prediction model was used to predict the CI of 2015-2025 a,and the prediction results showed that the Jiangjunmiao station had a high risk of drought events in the next decade,Kensiwate station had the lowest degree of drought,and Yingxiongqiao station had the lowest fluctuation range.Drought disasters occur from time to time in Xinjiang,so drought prediction is of great practical significance.