Spatio-temporal Distribution Characteristics Analysis And Water Quality Prediction In Minjiang River Basin

Climate,topography,human activities,industrial and agricultural production in Minjiang River Basin have significant spatio-temporal differences.Under different time and space,runoff and pollutant concentration are different,water quality has spatio-temporal heterogeneity.Identify spatio-temporal distribution characteristics and Source of water pollutants is important for a comprehensive understanding of water quality and river management.Establish a water quality prediction model can find out the trend of water quality and sources of pollutants,provide a basis for timely and effective management,it's a basic work of river management and pollution control.According to 8 water quality indexes monitor data of Minjiang river basin from 2014 to 2017,multivariate statistical methods,wavelet transform and BP neural network are used to analyze spatio-temporal characteristics of water quality and sources of pollutants,build a water quality prediction model based on wavelet transform and genetic algorithm for improving BP neural network,the following are results:Multivariate statistical methods were used to identify the Spatio-temporal Distribution Characteristics of water quality in the Minjiang river basin and analyze the source of water pollutants.The year was categorized into two clusters,the T1 period was the month of April to December,during which water quality is better and with NH3_N as the main pollutants,the main pollution sources is agricultural contamination,and T2 period was the month of January to March with NH3_N and total phosphorus as the main pollutants.Minjiang river basin can be classified into three regions.The S1 region located in the downstream of Jianxi,Shaxi,Dazhangxi and main stream of Minjiang river basin where water quality is worst and with nutrients as the main pollutants and oxygen consumption organics is following,industrial wastewater,domestic sewage,agriculture and livestock wastewater from Sanming,Nanping and Fuzhou as pollution sources flow into the river.The S2 region located in the downstream of Shaxi,Futunxi basin where water quality is best,nutrients from non-point source pollution is the main pollution sources,self-purification of water body in the substream is good.The S3 region located in the upper and middle reaches of Jianxi,upper and middle reaches of Futunxi basin with agriculture non-point source pollution as main pollution sources.Eiectrical conductivity has positive correlation with impervious surface and negative correlation with woodland.Ammonia nitrogen has positive correlation with impervious surface and water and negative correlation with woodland.Total phosphorus and turbidity has positive correlation with cultivated land and bare land,and negative correlation with woodland.A water quality prediction model based on wavelet decomposition-GABP neural network is established aim at water quality monitor data time series of Nanpingshuifenqiao station with strong volatility and randomness.The wavelet decomposition is used to decompose the original time series into multiple subsequences,GA-BP neural networks are used to establish the prediction model respectively,the sum of each subsequence predicted values is taken as final prediction value of water quality indexes.BP neural network prediction model and ARIMA prediction model are established as contrast experiments to prove the excellent performance of wavelet decomposition-GABP water quality prediction model.The prediction model achieved high prediction accuracy at the Fuzhouyuancuo,Shunchangmowu and Jianoupengdun stations.The results show the model is suitable for the water quality prediction in Minjiang river basin.