| Under the combined influence of climate change and human activities,extreme droughts occur frequently in the Yellow River basin,often causing regional water crises,threatening the safety of agricultural production,and the contradiction between supply and demand of water resources has become increasingly prominent.Under the influence of the changing environment,the hydrological elements are no longer consistent,and thus the occurrence mechanism of drought in the basin has changed,which makes the drought research method under the consistent condition questioned.Since the drought index is a basic tool for characterizing drought,it is important to construct a reasonable and accurate drought index to characterize the changing environment,which will provide a basis for decision-makers to establish timely and effective drought and disaster reduction programs,and provide guarantee for ecological protection and high-quality development of the Yellow River Basin.The paper takes the Yellow River basin from 1960 to 2016 as the research area,divides it into 7 secondary divisions of water resources,and adds 8 important tributaries at the same time,taking hydrological drought as the research object.The research is carried out from the aspects of index construction,index rationality verification,drought characteristics evolution,model prediction and impact factor analysis.The purpose of constructing stationary and non-stationary hydrological drought indices and analyzing their applicability is to characterize hydrological drought in changing environments more accurately and reasonably,in order to provide a scientific basis for precise drought prevention in the Yellow River Basin.The main results of the paper are as follows:(1)Based on the GAMLSS model,a non-stationary Gamma distribution model that changes over time was constructed,and the non-stationary hydrological drought index SRIt was created,and compared with the traditional hydrological drought index SRI,the capture results of the two indexes for the historical typical drought in the Yellow River Basin were analyzed,and verifying the superiority of SRIt in characterizing hydrological drought.The results show that SRIt is more sensitive to hydrological drought and can capture hydrological drought well,while traditional SRI underestimates the severity of hydrological drought.(2)The run-length theory was used to identify the characteristics of hydrological drought,and the spatiotemporal evolution of hydrological drought characteristics in the Yellow River Basin in January,March,and December was statistically analyzed.The research shows that:the whole Yellow River Basin shows a trend of aridification,especially in the section from Longyangxia to Lanzhou and the area below Huayuankou;the severe areas of drought duration and intensity migrated from west to east,and the high-value areas of intensity concentrated in Huangshui,Datong River and the middle and lower reaches;the frequency of drought in most areas of the basin showed a trend of first increasing and then decreasing,and the peak frequency of severe drought was concentrated in the area below Huayuankou.With the increase of scale,extreme hydrological drought showd a trend of aggravating first and then relieving,but the frequency of extreme hydrological droughts is increasing in the areas below the Huayuankou of the Yellow River.(3)Based on the lagged data of meteorological data and remote sensing data,SVM and LSTM drought prediction models were established,and the LSTM model with the strong applicability was selected.Construct an LSTM model with a forecast period of 1 month,and add circulation factors for drought prediction.It is found that the drought prediction result is better when the circulation factor is added.Compared with the prediction results without adding the circulation factor,the relative increase of NSE is 0.29%-25.78%,the relative decrease of RSR is 0.08%-9.43%,and the relative increase of R2 is 0.23%-29.34%.(4)Quantitative analysis of factors influencing hydrological drought on monthly and annual scales using random forest importance scoring criteria.It was found that under the monthly scale,hydrological drought in the first 1 month had the strongest influence,with an importance range between 30.0%and 50.0%;followed by meteorological drought in the same period,with an importance range between 10.0%and 40.0%;followed by soil water moisture in the same period,accounting for between 5.0%and 30.0%of all factors,with a smaller role for the circulation factor;and under the annual scale,meteorological drought had the greatest influence in the range of 10.0%-60.0%,followed by rainfall and soil water humidity,both in the range of 10.0%-40.0%,and the importance of GDP and population in the range of 2.0%-27.0%.(5)In the simulation of the drought prediction scheme,the simulation results of meteorology,initial conditions,hydrology,and circulation factors screened by the Spearman correlation coefficient method are used as the baseline scheme,and based on the control variable method,one of the variables in the baseline scheme is replaced by a new variable that deletes the interdecadal impact of the variable as a comparison scenario,and LSTM is used to simulate and predict the comparison scenario.The study found that the NSE of the simulation results of different previous runoffs was reduced by 2.7%-39.3%compared with the baseline scheme,indicating that the impact of different previous runoffs was the strongest;the NSE of hydrological drought under different previous conditions is mostly reduced,compared with the baseline scenario,the range is 2.2%-40.4%,followed by soil moisture under different previous conditions,meteorological drought under different previous periods,and finally circulation factor. |
PDF Full Text Request