Prediction And Sensitivity Analysis Of Chlorophyll A In Dam Area Of Small And Medium-sized Riversin North China

The water bodies of small and medium-sized rivers in north China are characterized by strong closure,poor hydrodynamic conditions,low self-purification capacity and small water environmental capacity,therefore,algae outbreaks are easily caused under appropriate environmental conditions.Taking typical small river dam in the north China—Qingshui Dam Area of Zhangjiakou district as an example,to understand its temporal and spatial variation characteristics based on the monitoring of water environmental factors.Then I established the prediction model of Chl-a based on BP neural network,and analyzed the local sensitivity of Chl-a in order to explore the main limiting factors of Chl-a and provide a certain guidance for preventing and controlling the outbreak of phytoplankton in dam area of small and medium-sized rivers in north China.The main research contents and conclusions were as follows:1?The time-scale variation characteristics of environmental factors in water bodies were as follows: dissolved oxygen in summer was the highest,while it decreased slightly in spring and autumn;the concentration of total phosphorus was lower in the whole year,showing a slightly higher but not significant characteristic in summer;the concentration of total nitrogen was higher in the whole year,which was higher in autumn,followed by spring and lowest in summer;the concentration of ammonia nitrogen was lower in the year,and the seasonal variation was not significant.2?The spatial scale variation characteristics of environmental factors were as follows: pH value,dissolved oxygen and total phosphorus showed a significant increase trend with the flow direction of the water bodies;the concentration of total nitrogen showed a significant decreased trend with the flow direction of the water bodies;the spatial scale characteristics of ammonia nitrogen concentration were not obvious.3?The seasonal variation characteristics of Chl-a concentration were as follows: highest in summer and slightly decreased in spring and autumn;the overall spatial scale showed an increasing trend with the water flow direction.Phytoplankton was dominated by green algae and diatom,the dominant algae species were Scenedesmus,Navicula,Pediastrum and Chlorella,and the community distribution was not uniform.4?Through Q-type clustering analysis,the nine monitoring points in Qingshui River Dam Area were divided into two categories.The first category included 2#,5#and 8# dam area,and the second category included 15#,19#,23#,25#,27# and 29# dam area.Mapminmax function was used to normalize the environment factors to solve the influence of dimensionality and numerical difference on the neural network.According to the correlation analysis,the noise factor of pH were filtered out for both types of neural network models,and the input variables were determined as: water temperature,dissolved oxygen,ammonia nitrogen,total nitrogen,total phosphorus,COD and rainfall.5?According to the empirical formula,the number of neurons in the hidden layer of the neural network prediction model was determined to be 10,and the transfer function was determined as the combination of tansig-purelin;the correlation coefficient of the first kind of model training results of BP neural network prediction model was as high as 0.984,and the relative error of test was 0.067;the correlation coefficient of the second kind of model training results was 0.907,and the relative error of test was 0.085,which has a good fitting degree and accuracy.6?The first type of neural network was used to analyze the sensitivity of 2# dam area as an example,the overall variation trend of single sample sensitivity analysis was that Chl-a had the highest sensitivity to total phosphorus from August to October,while water temperature had the highest sensitivity except from August to October,the sensitivity of other environmental factors were low and varied greatly.The water temperature and total phosphorus of the whole sample sensitivity analysis had the greatest influence on Chl-a,with the contribution rates of 27.4% and 25.5% respectively;the main limiting factor of Chl-a was total phosphorus,followed by water temperature.7?The second type of neural network was used to analyze the sensitivity of 29# dam area as an example,the single sample sensitivity analysis showed that Chl-a had the highest sensitivity to water temperature in April and November,and the highest sensitivity to total phosphorus from May to October,the sensitivity of other environmental factors varied greatly in different time periods,and the law was not obvious.The main limiting factor of Chl-a in the whole year was total phosphorus,followed by water temperature,with the contribution rates were 44.1% and 28.1% respectively.