Modeling For Climate And Runoff Processes In Typical Basins Of The Tianshan Mountains

The Tianshan Mountains,also known as the‘Water Tower in Central Asia’,are the primary source of rivers in the arid region of Central Asia.Glaciers and snow are widely distributed in mountain areas,and precipitation is several times that of the plains.Precipitation,glacier and snow meltwater are the critical sources of runoff.In the context of global change,quantitatively simulating climate and runoff processes in the Tianshan Mountains is not only a theoretical requirement for understanding the mechanism of climate and hydrological processes but also a strategic need for regional water resource utilisation and sustainable development.However,the meteorological stations are few,and observed data are scarce due to the high-altitude and complex terrain,limiting the application of traditional hydrological models in the Tianshan Mountains.There is an urgent need to develop a model for simulating climate and runoff processes,to serve hydrological forecasting and water resources management in data-scarce basins of the Tianshan Mountains.This study developed a hybrid method by integrating climate downscaling,time series analysis and correlation analysis of climate and runoff,assessment of the impact of climate change on runoff and machine learning methods by selecting the Kashgar River Basin(Ka RB),Urumqi River Basin(URB),Manas River Basin(MRB),Toxkan River Basin(TRB),Kumaric River Basin(Ku RB)and Kaidu River Basin(KRB)in the Tianshan Mountains as typical representatives and based on Earth system data products and observed data.We simulated the runoff changes in the historical period(from 1979to 2020)and the future period(from 2021 to 2049)using the developed methods and then analysed the sensitivity of streamflow to climate change,providing references for hydrological forecasting and water resources management in mountainous basins.The main contents could be summarised as follows:(1)To solve the lack of observed data in mountainous areas,this study developed a batch gradient descent-nonlinear regression(BGD-NLR)downscaling model based on ECMWF Reanalysis Interim temperature,ECMWF Reanalysis v5 precipitation and Digital Elevation Model.The present study simulated the monthly temperature and precipitation data sets with high resolution from 1979 to 2019 in the Tianshan Mountains based on the developed model.The observed data from 30 meteorological stations indicate that the downscaled data sets have high accuracy and accurately reveal climate changes in high-altitude mountains and plain areas.In high-altitude mountains,the slope of the linear fitting(Slope)between downscaled and observed temperature is0.93;the Nash–Sutcliffe efficiency coefficient(NSE)is 0.95;the root mean square error(RMSE)and mean absolute error(MAE)are less than 3.2°C.The slope of downscaled and observed precipitation is 1.24,the NSE is 0.83,and the MAE and RMSE are less than 8.5 mm.In plain areas,the slope of downscaled and observed temperature is 0.90,the NSE is 0.93,and MAE and RMSE are lower than 3.5°C.The slope of downscaled and observed precipitation is 1.16,the NSE is 0.65,and MAE and RMSE are less than9 mm.According to the BGD-NLR model,longitude and latitude are the primary factors influencing the spatial distribution of climate in the Tianshan Mountains.Altitude contributes more to temperature and precipitation distribution than aspect and slope among topographic factors.(2)Based on the downscaled data sets mentioned above,this study found that the rate of temperature increase in the Tianshan Mountains from 1979 to 2019 was0.23°C/10a,and the rate of precipitation increase was 13.4 mm/10a.The temperature and precipitation increased the most in spring and summer.The east Tianshan Mountains experience the fastest rise in temperature,while the west Tianshan Mountains experience the fastest rise in precipitation.Temperatures in the Ka RB,URB,MRB,TRB,Ku RB and KRB all had an increasing trend.However,precipitation in Ka RB showed a decreasing trend but increased in other basins.Temperature and precipitation have multi-scale oscillation characteristics.On the seasonal scale,temperature and precipitation have a quasi-three-month cycle.However,on the inter-annual scale,the temperature has quasi-one-year,quasi-two-year,quasi-four-year and quasi-seven-year cycles.Further,precipitation has quasi-one-year,quasi-two-year,quasi-three-year and quasi-eight-year cycles.On the inter-decadal scale,the temperature has a quasi-17-year cycle,and precipitation has a quasi-24-year cycle.The spatial variation of climate shows that the high-altitude mountains have low temperatures,abundant precipitation and rapid increase in temperature and precipitation.In contrast,the river valleys and alluvial fan regions have high temperatures,little precipitation and a slow rise in temperature and precipitation.(3)This study examined the correlation of the climate in the Tianshan Mountains with global temperature and typical atmospheric circulation indexes to better understand the causes of climate change at the global and regional scales.The findings indicate that the climate in the Tianshan Mountains is sensitive to global warming on a global scale.The correlation coefficient between temperature in the Tianshan Mountains and global temperature is greater than 0.95,indicating that the Tianshan Mountains are warming faster than the remaining world.On the regional scale,the Pacific Decadal Oscillation(PDO),North Pacific pattern(NP),North Atlantic Oscillation(NAO),Western Hemisphere Warm Pool(WHWP)and Atlantic Multidecadal Oscillation(AMO)were related to the increases in temperature and precipitation in the Tianshan Mountains from 1979 to 2019.The temperature is positively correlated with the AMO,NP and WHWP and negatively correlated with NAO.The precipitation is positively correlated with the AMO,NP and WHWP and negatively correlated with PDO.Climate and typical circulation indexes have multi-scale correlations and lag correlations.The resonance changes frequently and complexly on short time scales,whereas it is relatively stable on long time scales.During the present study,the temperature changes lag the Arctic Oscillation by 113days,AMO by 15–114 days,the North Tropical Atlantic Sea Surface Temperature Index(NTA)by 183 days,PDO by 60 days and the Tropical Northern Atlantic Index(TNA)by 180 days.The precipitation changes lag the AMO by 183 days,NTA by 137 days,PDO by 91–228 days and TNA and WHWP by 160 days.Overall,the lag time of precipitation relative to circulation changes is longer than that of temperature.(4)To simulate the response of runoff to climate change in mountainous basins,this study integrated an improved complete ensemble empirical mode decomposition with an adaptive noise-e Xtreme gradient boosting tree(ICEEMDAN-XGBoost)model to model the climate and runoff process at different scales by comparing machine learning algorithms including radial basis function artificial neural network,random forest,support vector regression and e Xtreme gradient boosting tree.The observed data from hydrological stations showed that the model could accurately simulate monthly runoff change at different time scales using monthly temperature and precipitation.The slope of the simulated and observed runoff in the six basins is between 0.94 and 1.05,the NSE is above 0.7 and the MAE and RMSE are small.The NSE of the simulated and observed runoff is greater than 0.9,the MAE is lower than 0.05×10~8 m~3 and RMSE is lower than 0.1×10~8 m~3 on the decadal scale.The simulation results indicate that temperature dominates the runoff changes in the Ku RB,while precipitation contributes more to the runoff change than the temperature in other basins.When the temperature changes by 1%,the annual runoff will change by 0.11%–0.40%,2.08%–3.90%,0.04%–0.06%,0.49%–1.12%,2.41%–3.38%and 0.22%–0.27%in the URB,MRB,Ka RB,TRB,Ku RB and KRB,respectively.When the precipitation changes by 1%,the annual runoff will change by 12.75%–13.63%,14.38%–17.00%,8.15%–8.21%,7.69%–8.16%,1.57%–1.68%and 12.95%–13.53%in the URB,MRB,Ka RB,TRB,Ku RB and KRB,respectively.Temperature and precipitation both contribute 27%and 73%to runoff changes in the URB,39%and 61%in the MRB,5%and 95%in the Ka RB,28%and72%in the TRB,80%and 20%in the Ku RB,12%and 88%in the KRB.Climate change caused the runoff in the Ka RB to decrease,while the runoff in the other five basins increased during the study period.(5)This study used the above ICEEMDAN-XGBoost model to simulate runoff change in typical basins of the Tianshan Mountains based on the RCP4.5 scenario of the CMIP5 multi-model ensemble and SSP245 scenario of the CMIP6 multi-model ensemble to further predict the runoff processes driven by climate change in the future.Here the temperature will rise by 0.3°C/10a,and precipitation will rise by 7.3 mm/10a in the Tianshan Mountains under the RCP4.5 scenario from 2021 to 2049.Precipitation and temperature will rise in typical basins,with the fastest warming in autumn and the fastest increase in precipitation in winter.Driven by climate change,the runoff in six basins will increase,and the rate from high to low will be the Ku RB with 5.2×10~8m~3/10a,Ka RB with 1.4×10~8 m~3/10a,TRB with 1.1×10~8 m~3/10a,MRB with 0.7×10~8m~3/10a,KRB with 0.5×10~8 m~3/10a and URB with 0.1×10~8 m~3/10a,respectively.The increase of runoff will be mainly from summer and autumn.From 2021 to 2049,the SSP245 scenario predicts a warming rate of 0.5°C/10a and an increased precipitation rate of 3.4 mm/10a in the Tianshan Mountains.Furthermore,the temperature will rise in six basins.Precipitation will increase in the Ka RB,MRB,KRB and URB while decreasing in the TRB and Ku RB.Driven by climate change,runoff in six basins will increase,and the rate from high to low will be the Ku RB with 3.5×10~8 m~3/10a,Ka RB with 1.7×10~8 m~3/10a,TRB with 1.5×10~8 m~3/10a,KRB with 0.9×10~8 m~3/10a,MRB with 0.9×10~8 m~3/10a and URB with 0.1×10~8 m~3/10a,respectively.