The Monitoring And Analysis Of Chlorophyll-a And Turbidity By Remote Sensing In Minjiang River

The traditional fixed water sampling while the indicators of water quality monitoring,but it is time-consuming,high cost,and cannot reflect distribution in space of water quality indexes.With the development of remote sensing technology,the water quality parameters can be retrieved by remote sensing,which can effectively realize the change of water quality parameters in a wide range,low cost,fast monitoring of water quality.Minjiang River is used as the study area,CCD HJ-1A/B image as data source,using BP neural network model to monitor water quality parameters of chlorophyll a and turbidity for remote sensing monitoring.The main research contents and conclusions are as follows:(1)Adaptive threshold extraction of water body.In order to decrease the missing and wrong extraction rate of water body with single threshold extraction method in large-scale water body,a new method is put forward.This method determines the threshold value based on local maximum between-class variance(OTSU algorithm)adaptive law.The Minjiang river is chosen as our experimental area and the water from the HJ-1A/B CCD satellite images used this method is extracted.The result shows that our method can rapidly to extract the large scale water body,and the accuracy is 95.25%,while the global uniform threshold value is 90%.To some extent,it can also eliminate the influence of the terrain shadow and the buildings and the precision of small water extraction is also improved.(2)Construction and analysis of the inversion results of Minjiang River chlorophyll a concentration inversion model.In combination with the measured data of single band and band of Ch1 a concentration were analyzed by Pearson correlation analysis to obtain sensitive bands of chlorophyll a concentration inversion.Respectively.Chlorophyll a concentration inversion model is established by linear model and BP neural network model,which a linear regression model inversion of the average relative error for 127.8%,multiple linear regression the average relative error of the model inversion was 3.44%,BP neural network model to inverse the precision of reached 76.4%,which meets the requirement.Therefore,using BP neural network model of Minjiang River retrieved chlorophyll a concentration spatial distribution and spatial variation curve.The concentration of chlorophyll a there are high value areas are in the lower reaches of Minjiang,mainly affected by urban expansion,living sewage,upstream captive livestock led to the eutrophication.In the upper reaches of the Minjiang River in Jianxi,Shaxi,Futunxi in gathering near the points and convergence near the main chlorophyll a concentration value is higher,which Shaxi chlorophyll a concentration is high the river is longer.(3)Construction and analysis of the inversion results of Minjiang River turbidity inversion model.Linear conversion between turbidity and suspended matter concentration was obtained by linear fitting,and the coefficient of determination was 0.8125,which was significantly related to the 0.01 level.The average relative error of one element linear regression model with suspended matter concentration is 121.4%,and the inversion precision of the multiple linear regression model is very unstable,and the average relative error is 134.9%,which can not meet the needs of the Minjiang River water quality monitoring by remote sensing.29.8%of average relative error for the inversion result of BP neural network model,so the small error range,high precision of the BP neural network model inversion of suspended matter concentration and its conversion for turbidity and turbidity space distribution results and curves.In the lower reaches of the Minjiang River near the estuary turbidity in the 40-80NTU,wulongjiang turbidity in the 0-20NTU,showed high East and low in the West in the distribution trend and elsewhere including upstream and midstream turbidity in 0-30NTU scope changes.