The Spatio-temporal Characteristics Analysis Of Hydrological Drought And Drought Prediction In The Upper Fenhe River

Drought is one of the most important natural disasters in the world.It is occurrence involves a wide range and has a long duration,which has a serious impact on human life.Droughts occurring in various temporal and spatial patterns will have different degrees of impact on the region,and there are more influencing factors on drought and more complex interactions,and the drought prediction is more difficult.In this study,the monthly runoff data are used to analyze at four hydrological stations in the upper Fenhe River,six distribution functions are adapted to fit long-term runoff series at various time-scales.The Probability Plot Correlation Coefficient(PPCC)and Quotient of Determination(QD)test are used to decide which distribution function could fit cumulative runoff series better.The standardized runoff index is calculated by the best distribution function,which analyzed the spatial-temporal characteristics of the hydrological droughts at various time scales.By analyzing the correlation between the standardized runoff index and the normalized difference vegetation index,the response mechanism of hydrological drought and normalized difference vegetation index was studied.And constructing a least square support vector machine(LSSVM)combination model based on Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)to predict hydrological drought.The main contents and results of the study are as follows:(1)The probability distribution of monthly runoff with various time scales at different stations is not consistence in the upper Fenhe River.The best-fitting probability distributions at the Shangjingyou,Fenhe reservoir,Zhaishang and Lancun hydrological station with the one-month time scale are all GEV distributions;At the three-month time scale,the best-fit probability distributions for the Shangjingyou,Zhaishang and Lancun stations are all W distributions,the best fitting probability distribution of the Fenhe reservoir is the distribution of GEV;On the 12-month time scale,the best-fit probability distributions for the Shangjingyou,Zhaishang and Lancun hydrological stations are W distributions,the best-fit probability distribution of the Fenhe reservoir is LN2 distribution.(2)The occurrence of drought events can be accurately identified using the standardized runoff index at various time scales,The hydrological drought had good consistency in the periods of drought on the monthly,seasonal,and annual scales at the four hydrological stations.On monthly scale,the standardized runoff index showed that the frequency of hydrological drought in the study area was higher in January,February,and December than in other months;On the seasonal scale,the standardized runoff index indicates that the drought in the study area is the most serious in winter,followed by the drought in spring and autumn,and the drought in summer drought is the smallest.On the annual scale,the standardized runoff indexes of various hydrological stations in the study area showed a decreasing trend in different degrees,indicating that the overall situation of the drought degree in the study area tends to be serious.(3)On the spatial scale,the evolution law of hydrological drought at different time scales in the study area in the upper reach of the Fenhe River is quite different.On the monthly scale,the severity of spatial distribution of drought in different months during the same year is different.The distribution of drought severity in different months has significant spatial distribution characteristics;On the seasonal scale,in the study area in the Upper Fenhe River,the spring shows that the drought in the Lancun Station is relatively serious compared to other stations in the entire region,while the drought near the Shangjingyou station is relatively light,the spatial distribution of the study area is more consistent.In summer and autumn,it is shown that the spatial distribution of severe drought is more consistent in some typical years.The spatial distribution of the winter shows that the area of drought in the entire study area is relatively large and the drought is severe;On the annual scale,the spatial distribution of drought severity shows that the severity of drought reduced in the northwest,and the severity of drought increased in the southeast.(4)In the study area in the upper reaches of the in the Upper Fenhe River,on the monthly scale,the changes in the NDVI during the year are roughly the same and periodic.And the NDVI value of the study area fluctuates between 0.1 and 0.53.The NDVI values in the four seasons under the seasonal scale were: summer> autumn> spring> winter;On the annual scale,the NDVI value of the study area increased slightly with time,vegetation coverage showed an improvement trend,but the increase was not significant;The correlation test between NDVI and SRI1,SRI3 at the four stations during the month indicated that NDVI is a major factor affecting SRI,and the extent of vegetation coverage in this area is significantly affected by monthly and seasonal changes,interannual changes in hydrological drought have weak interactions with NDVI;The correlation coefficient between NDVI and SRI was different in the four hydrological stations in different seasons,and the difference was greater.The correlation coefficients between NDVI and SRI12 in the four hydrological stations were different.The lagging one-year NDVI and SRI correlation coefficient results show no significant hysteresis effect.The linear phase characteristics of NDVI and SRI at different scales in the growing season at each station in the upper Fenhe River show that the SRI from 1 to 3 months are sensitive to NDVI.(5)After the CEEMDAN,the hydrological drought sequences at different time scales at each station in the Upper Fenhe River can obtain 9 IMF components and 1 trend item,the resulting IMF component and the trend term are basically the same as the original sequence.As the decomposition progresses,the magnitude of the IMF components in each order gradually decreases,and the IMF component periods of stations at different time scales are different.It can be seen from the trend items that the drought sequences of the four hydrological stations show a decreasing trend in different degrees,and the overall drought situation in the study area tends to be serious.(6)By predicting the results of hydrological drought series at various time scales in Shangjingyou hydrological station in the Upper Fenhe River,the results show that: single LSSVM model and CEEMDAN-LSSVM combined model were both able to fit and predict the hydrological drought series well,which proved the applicability of the model.The prediction results of hydrological drought sequences at different time scales show that the prediction accuracy of the combined model is different from that of the single model.The SIR1 hydrological drought series combined model prediction of the Shangjingyou is better than the single model prediction.It can be used for predictive analysis of hydrological drought sequences and has a high potential for application.