Analysis Of Temporal And Spatial Distribution Characteristics Of Water Quality And Prediction Of Water Quality Change In Dongjiang Headwater Basin

The Dongjiang River is one of the tributaries of the Pearl River Basin and the main source of domestic water supply for the Hong Kong Special Administrative Region,the eastern part of Guangdong Province and the Guangdong-Hong Kong-Macao Greater Bay Area.As the source of Dongjiang River,with the increase of economic scale and the impact of human activities,the ecological environment of water resources is getting worse.Therefore,it is particularly important to understand the spatial and temporal distribution characteristics of water environmental quality in the Source basin of the Dong jiang River.Based on the measured water quality index data in the Source basin of the Dong jiang River,this paper takes the Source basin of the Dong jiang River as the research object.In combination with single factor analysis method and comprehensive pollution index method for watershed water quality evaluation.Descriptive statistical analysis and inverse distance weight interpolation were used to analyze the temporal and spatial variation characteristics of water quality indexes.The long and short term memory neural network and Mann-Kendall test were combined to predict the change trend of water quality indexes in watershed.（1）The results show that between 2015 and 2019,the river water quality success rate were 50%,61.11%,55.56%,55.56%,66.67%.（2）The contents of p H,DO,COD_Mn,BOD₅,NH₃-N and TP were significantly different in the basin.In time,p H,concentration of COD_Mn and TP continues to increase,DO content,BOD₅ and NH₃-N concentration continues to decrease,but the degree of change is not obvious;In space,Dingnan substation,Dingnan Tianjiu section and Xunwu Shangshipai section are the most serious water pollution.（3）Based on the LSTM neural network prediction results show that during the study period,p H value,CODMn,TP concentration increased continuously during the study period.DO,BOD₅and NH₃-N concentration showed a decreasing trend.The results of Mann-Kendall test showed that p H,COD_Mn and TP concentration increased,while DO content,BOD₅ and NH₃-N concentration decreased.The results of the two methods are basically the same.It can be concluded that the LSTM neural network can be used to predict the change of water quality in watershed.Combined with hydrologic meteorological data and social economic conditions,the causes and changes of Dong jiang source pollution were analyzed.Results show that the correlation between meteorological factors and Dingnan water region temperature,water temperature and precipitation have great influence on p H,COD_Mnand TP;Xunwu water area temperature,water temperature and precipitation on p H,DO content,COD_Mn,NH₃-N and TP concentration had a greater influence on the change.The average flow and average sediment transport rate in the whole basin have little effect on the changes of various water quality indicators.The results of correlation with socioeconomic status showed that the number of poultry in Dingnan water area had a greater impact on the changes of other water quality indicators except p H.The socio-economic indexes in Xunwu Water area had little effect on water quality,but the main crops area,tea and orchard area,fertilizer application amount,cattle number,pig number and poultry number had consistent changes on COD_Mn,BOD₅,NH₃-N concentration and TP concentration.Therefore,it can be concluded that the water quality of the river basin is affected by both meteorological factors and human activities,and reasonable suggestions for river pollution control are put forward.