Analysis And Stochastic Simulation Of Runoff Dynamics In The Upper Yellow River Under Water System Connectivity Variability

River and lake systems are the primary carriers of water resources systems and play a key role in the water cycle of the basin.As a fundamental property of river systems,a smooth river system is a key condition for river systems to maintain important functions such as energy and material transport,biological migration and self-cleaning of the water environment.As the main water-producing area of the Yellow River Basin.its water resources and hydraulic resources play a crucial role in the ecological protection and high-quality development of the Yellow River Basin.However.the natural climate change and frequent human activities such as reservoir construction have changed the spatial and temporal distribution of river runoff in the Upper Yellow River Basin and reduced the level of water continuity in the region,resulting in the decline of the river network and the weakening of water system connectivity.The potential problems of river network structure pattern have aggravated the regional water resources and water ecology problems to some extent.Based on the evaluation of the water system pattern and connectivity indicators,this paper investigates the development of runoff and drought characteristics under the variation of water system connectivity in the study area,and constructs a medium-and long-term runoff simulation model in the study area,in order to reveal the intrinsic connection between water system connectivity and runoff sequences in the study area,and provide a reliable scientific basis for alleviating the increasing water shortage problem and promoting the sustainable social and economic development of the Yellow River basin.This paper provides a reliable scientific basis for mitigating the increasing water scarcity in the Yellow River basin and promoting sustainable social and economic development in the Yellow River basin.The main research contents and results of this paper are as follows:(1)Analysis of the evolution pattern of water system connectivity.Based on the spatial data from 1980s to 2020s in the upper part of the Yellow River basin,we extracted the water system data and constructed an index system to evaluate the water system pattern and connectivity in the study area in terms of water system pattern and connectivity pattern,and used the combination weight method to quantitatively analyze the change pattern of water system connectivity in the study area.The results show that the water system connectivity in the upper part of the Yellow River basin showed obvious phase changes from 1980s to 2020s,and the water system connectivity declined from 0.841 in 1980s to 0.334 in 1990s,and then recovered to 0.576 in 2020s,with an overall trend of "decreasing then increasing".The overall trend was"decreasing and then increasing",and there were obvious quantitative changes and mutualistic trends in the 1990s.(2)The analysis of runoff dynamics under the variability of water system connectivity.The statistical characteristics and components of the runoff series at each hydrological station in the study area were analyzed in the time-frequency domain using the runoff statistical characteristics analysis method,TFPW-Pettitt mutation test and the polar symmetric modal decomposition algorithm.The results show that there are quasi-periodic fluctuation scales and significant decreasing trends in the runoff series of 0.25～1a,2～5a,8a,15a and 30a in the study area,and the abrupt changes in the runoff series occurred mainly from 1985 to 1995.The trend of "decreasing and then increasing" in the period of water system connectivity variability(1990-2019),compared with the base period(1960-1989),the scale of the main cycle change is significantly reduced,and the periodicity of runoff fluctuation is enhanced.(3)Analysis of drought characteristics under the variation of water system connectivity.The changes of drought characteristics in the study area before and after the water system connectivity variation were analyzed quantitatively by using the standard runoff index(SRI)and travel range theory.The results show that the overall drought level in the study area is slowing down,and the rate of hydrological drought severity reduction under the water system connectivity variation is accelerated and the scale of hydrological drought cycle change is significantly reduced,and the phenomenon of alternating droughts and floods in small scale is becoming more and more significant.(4)Medium-and long-term runoff simulation model construction in the study area.The simulation results show that the LSTM series models with "gating mechanism"perform well in the medium-and long-term runoff simulation,while the CNNBiLSTM models with coupled convolutional neural network and bi-directional longand short-term memory neural network have better spatial and temporal feature extraction and data processing capabilities.The CNN-BiLSTM model with coupled convolutional neural network and bi-directional long-and short-term memory neural network has stronger spatio-temporal feature extraction and data processing capability.and shows stronger generalization capability in runoff simulation than other models,and the best simulation results.