Runoff Refinement Simulation And Drought/Flood Projection Based On Digitwining Watershed Model

Under the global climate change and the increasing complexity of underlying suface,the hydrological cycle in the basin scale has been significantly altered.Water scarcity has been intensified,and the risks of drought and flooding have been increased significantly,causing serious socio-economic losses and human casualties.The construction of digital twin watershed is an important way to achieve high-quality water development in the new era,providing a new paradigm for the watershed governance and management under the impacts of climate change and human activities.The changing environment has led to diverse climatic conditions,complex topography,and significant spatial variation in hydrological characteristic parameters in a large-scale basin.The spatial heterogeneities of runoff generation and concentration mechanism have been significantly enhanced,and the original spatiotemporal distribution pattern of water resources in a basin has been changed.Therefore,a large-scale runoff refinement simulation method based on digitwining watershed model(DWM)was proposed and the precipitation scale adaptation technology between GCMs and DWM was explored,through integrating multiple technologies to address the difficulties of the runoff refined simulation and the problems of global climate models(GCMs)simulation in a largescale basin.The large-scale refined simulation and projection of runoff was carried out in the Hanjiang River Basin(HRB),China,and the hydrological drought and flood/waterlogging risk over the upper reaches of HRB in the future were analyzed.This paper can provide technical supports for the construction of large-scale digital twin watersheds.In addition,it can also provide scientific and effective decisions for flood and drought control in the HRB and water resources management of the Central Line Project of South-to-North Water Diversion.The followings are main works and contents of this study:(1)Based on the analysis of the difficulties in refined runoff simulation of large-scale basins,and considering the highly non-uniformity of the underlying suface,a method of largescale runoff refinement simulation based on DWM was proposed by integrating relevant technologies,which can store the massive spatial and temporal data of large-scale basins in a database for unified management.Based on the distributed datasets of DEM,precipitation grid points,land use,soil,leaf area index,and potential evapotranspiration,the upper HRB was discretized into 30,873 river units("slope-channel")and 76,361 slope surfaces,and the parameterization scheme of hydrological simulation for each unit was proposed.The DWM was calibrated and validated using daily average flow data during the period of 2015 to 2018 from hydrological stations in Hanzhong,Shiquan,Ankang,and Baihe in the upper HRB.After more than 640 times of calibration,the determination coefficient ranged from 0.68 to 0.83 and the Nash coefficients ranged from 0.64 to 0.81,and the percent of bias was controlled within25%,demonstrating the applicability of the above proposed method in the upper HRB.Based on the above,the spatial and temporal distribution characteristics of runoff from 1995 to 2014 were simulated and analyzed.(2)For the problem of scale matching between GCMs precipitation output and DWM precipitation input,a matching method was proposed by integrating spatial interpolation method,bias correction method and machine learning algorithms,and its downscaling effects in the HRB were compared and evaluated.Based on the comparison of the precipitation simulation capabilities of the 12 GCMs from coupled model intercomparison project phase 5(CMIP5)and CMIP6,6 GCMs from CMIP6 were downscaled in the HRB using the above method,and the historical precipitation simulation capabilities before and after the downscaling were compared and evaluated.The results show that the downscaled CMIP6 GCMs have significantly improved their abilities in simulating precipitation and capture spatio-temporal pattern over the HRB,with the correlation coefficient improving to 0.88 and the Taylor skill score improving to 0.764.(3)The precipitation and runoff changes in the future over the HRB under changing environments were projected by coupling DWM,CMIP6 GCMs,and PLUS model.Firstly,the future precipitation output data of 6 CMIP6 GCMs were downscaled in the HRB using the proposed matching method of DWM precipitation input and GCMs precipitation output to explore the future precipitation changes over the HRB under three SSP-RCP scenarios.The results indicate that the future precipitation changes in the HRB show a significant increasing trend from 2023 to 2100.The increase rates of annual precipitation changes under the SSP1-2.6,SSP2-4.5,and SSP5-8.5 scenarios compared to the baseline period(1995-2014)will be0.19%/a,0.38%/a,and 0.75%/a,respectively.Secondly,the land use distribution of the HRB in 2020 was simulated and compared with observed data based on the PLUS model.The Kappa coefficient was 0.796,proving the reliability of accuracy.After above,the land use distribution over the HRB in 2040 was predicted.Finally,future precipitation and land use data were used to drive DWM,and simulation projections of runoff changes in near-term(2023-2042)and mid-term(2043-2062)were explored over the upper HRB.The results indicate that the runoff changes in the upper HRB show an insignificant increasing trend from 2023 to 2062,and the increase rates of annual runoff changes under SSP1-2.6,SSP2-4.5 and SSP5-8.5 scenarios compared to the baseline period will be 0.43%/a,0.36%/a and 0.49%/a,respectively.Q5(low flow)of the HRB in the future under three SSP-RCP scenarios will decrease compared to the base period.Under SSP5-8.5 scenario,Q95(high flow)of the HRB will increase.Under SSP1-2.6 and SSP2-4.5 scenarios,Q95 of the HRB will decrease.(4)The trend of standardized runoff index(SRI),drought characteristics and frequency over the upper HRB in near-term and mid-term were analyzed by SRI and the run theory based on the above projection results.A risk assessment index system for flood/waterlogging in the upper HRB has been established based on comprehensive multi-source data.A flood disaster risk assessment model was constructed based on the risk expression paradigm of "dangervulnerability".The entropy method was used to calculate the flood/waterlogging risk in the upper HRB,and the hazard risk,vulnerability risk,and comprehensive risk of flood/waterlogging in near-term and mid-term were evaluated.The results of drought assessment indicate that the trend of the hydrological drought over the upper HRB is dominated by an insignificant trend in near-term and mid-term under the three SSP-RCP scenarios.The duration and intensity of drought in the HRB will be the highest in the near-term,followed by the mid-term,and the minimum in the baseline period.Overall,the magnitude of drought intensity in the future will be comparable to the baseline period,but stronger than the baseline period in mid-term under the SSP5-8.5 scenario.In terms of drought frequency,it will be comparable in the future to the base period.The results of flood/waterlogging risk assessment indicate that the high risk centers of hazard over the upper HRB in the future will be mainly concentrated in the southwest region under three SSP-RCP scenarios.The proportions of slightly high risk and high risk hazard areas will be higher than the baseline period,and the proportion of high risk areas in the near-term will be the highest under the SSP2-4.5 scenario,reaching 21.68%.The slightly high risk and high risk areas of vulnerability will be mainly distributed near the population gathering area in the east and the Hanzhong Plain in the west.The proportions of vulnerability areas with medium risk,slightly high risk,and high risk in the future will significantly increase compared to the baseline period.Under the SSP5-8.5 scenario,the proportion of vulnerability areas with high risk in near-term will be the highest,reaching7.64%.As to the comprehensive risk,the high risk centers over the HRB will be distributed near the southwest main stream,and a second high risk center will be formed near the Danjiangkou Reservoir in the east.The proportion of slightly high risk areas in the future is higher than that of the baseline period,and the proportion of high risk areas in the mid-term under the SSP5-8.5 scenario will be the hightest,reaching 21.62%.