| The source region of the Yangtze River is located on the Qinghai-Tibet Plateau,the birthplace of the Yangtze River Basin,and one of the important ecological protection areas and water sources in China.It is characterized by "high,cold and drought" and is highly sensitive to climate change,which will directly affect the hydrometeorological processes such as glaciers,snow lines,snowmelt,precipitation and runoff in the source region of the Yangtze River.This will have an important impact on the water resources and ecosystem of the Yangtze River basin.Therefore,studying the evolution law of runoff in the source region of the Yangtze River under the climate change environment and predicting the trend of future runoff change are important foundations for regional ecological security and water resources management,as well as requirements for water resources management,ecological environment protection and climate change adaptation in the Yangtze River Basin,and also urgent requirements for the sustainable development of human society.Based on the background of global climate change,this paper takes the source region of the Yangtze River as the research object,and selects the annual runoff data from 1965-2020 of Zhimenda hydrology Station,an outlet control station in the source region of the Yangtze River,as well as the monthly discharge observation data from 1965-1986 and 2006-2020.Combined with the daily measured meteorological data from 1965 to 2020 at five meteorological stations in the source region of the Yangtze River: Wudaoliang,Tuotuo River,Qumalai,Yushu and Qingshui River,the missing values of runoff meteorological data were interpolated using the K-proximity algorithm.Then,climate tendency rate,Mann-Kendal test,R/S analysis,wavelet analysis and Spearman rank correlation test were used to analyze and discuss the changing trend,abrupt characteristics,periodic evolution law and the correlation among hydrometeorological elements in the source region of the Yangtze River.Based on LSTM neural network and a variety of decomposition methods,a decomposition-prediction-reconstruction model is constructed to simulate and forecast the monthly discharge in the source area of the Yangtze River,and the advantages and disadvantages of each model are analyzed and compared.Finally,using the model data of the BCC-CSM2-MR climate system in the CMIP6 experiment,bilinear interpolation method was used to interpolate the model data of the monthly scale to a grid of 0.25°×0.25°,and then the site data used in this study was extracted.The future time series is divided into three phases: T1(2021-2045),T2(2046-2070)and T3(2071-2100).The change characteristics of precipitation and temperature in the source region of the Yangtze River in the next 80 years under four scenario models,SSP1-2.6,SSP2-4.5,SSP3-7.0 and SSP5-8.5,were analyzed,and the flow in the source region of the Yangtze River was simulated and predicted in the next 80 years under different scenario models.The specific research results of this paper are as follows:(1)From 1965 to 2020,runoff,precipitation and air temperature in the source region of the Yangtze River all showed a significant increase trend,and the annual distribution was uneven,showing a state of rain and heat in the same period,and the future change trend was opposite to the past,showing a negative continuity;The abrupt change point of runoff time series is 2008,the first main cycle is 56 a,the second main cycle is 14 a,and the third main cycle is 35 a.The abrupt change point of the precipitation time series is 2007,the first main cycle is56 a,the second main cycle is 14 a,and the third main cycle is 35 a.The abrupt change point of the mean temperature time series is 1998,the first main period is 42 a,the second main period is 55 a,and the third main period is 26 a.At the significance level of 0.05,there was a significant positive correlation between runoff,precipitation and air temperature,while precipitation and air temperature did not pass the significance test,and there was an insignificant positive correlation.(2)The measured monthly flow data from 1965-1986 were divided into the training set data of the model,and the measured monthly flow data from 2006 to 2020 were taken as the test set data of the model.The LSTM,EMD-LSTM,CEEMDAN-LSTM and PCACEEMDAN-LSTM models were constructed respectively.Through the analysis of the prediction results of the test set stage,it is found that these four models are all suitable for the runoff prediction of the Yangtze River source area,and each model can well simulate and predict the change trend of the discharge of Zhimen da hydrology Station in the Yangtze River source area,but the prediction effect of the annual peak discharge and the occurrence time is poor.The LSTM model has the worst prediction effect,followed by the EMD-LSTM model.PCA-CEEMDAN-LSTM model had the best prediction effect.RMSE index of PCACEEMDAN-LSTM model decreased by 29.4%,19.4% and 0.3%,and MAE index decreased by 35.6%,21.7% and 0.3%,respectively,compared with the other three models.The MAPE index was decreased by 69.6%,26.1% and 13%,and the fit R2 was increased by 7.3%,4.6%and 0.1%,respectively.(3)Under the four scenarios in the future,the future precipitation in the source region of the Yangtze River will show a trend of fluctuation and increase,and the annual precipitation will increase by 69.35%,69.75%,69.88% and 74.32%,respectively,compared with the historical period.The future precipitation will change gently in the early period,and the growth trend will be significant in the later period.Among the four scenarios,SSP3-7.0 is the one with the fastest precipitation growth rate,while SSP5-8.5 is the one with the highest annual precipitation.In the future,the annual distribution of precipitation in the source region of the Yangtze River will be roughly the same as that in the historical period,but the proportion of precipitation from June to September will decrease,while the proportion of precipitation from January to May and October to December will increase,indicating that the annual distribution of precipitation in the source region of the Yangtze River will be more uniform in the future.(4)Among the four scenario models,the temperature in the source region of the Yangtze River during 2021-2100 shows a rising trend on the whole.The model with the highest temperature and the most significant rising trend is SSP5-8.5 scenario model,followed by SSP3-7.0 scenario model.In the future,the lowest annual temperature in the source area of the Yangtze River will appear in January,and the highest annual temperature will appear in July.The annual distribution of temperature is roughly similar to the historical period.In the three different periods,the temperature increase rate of the four scenarios in T1 period and T2 period was large,and the temperature decreased in T3 period.(5)The precipitation and temperature data from the BCC-CSM2-MR climate system were input into the PCA-CEEMDAN-LSTM model with good simulation and prediction effect to simulate and predict the monthly discharge of Zhimenda hydrology Station in the source region of the Yangtze River in the next 80 years.The results show that the future flow of the source area of the Yangtze River will increase compared with the historical annual average flow under the four scenario models,and the increase is the highest under the SSP5-8.5 scenario model,which increases by 62.4%,62.1%,55.3% and 84.2%,respectively,compared with the historical period.In the future,the maximum discharge in the source area of the Yangtze River will occur in August,and the minimum discharge will occur in January.The annual runoff will still be concentrated in June to September,but its proportion will decrease,and the annual distribution will be more reasonable. |
PDF Full Text Request