Remote Sensing Retrieval Model For Chlorophyll-a Concentration Of Algae In South China Sea Coastal

Human activities often lead to deterioration of water quality , reduction of biological resources and frequent occurrence of red tide phenomenon in the marine coastal water environment. As chlorophyll-a is the main photosynthetic pigment cells of algae,dynamics of chlorophyll-a concentration reflects the abundance of algae in water,biomass and its variation. It is an objective biological indicator to reflect the nutritional status of ocean water and also the most important indicator to evaluate the eutrophication in the ocean. To built a model for retrieving the concentration of chlorophyll-a through satellite remote sensing data in study area is important to study the distribution of algae and primary productivity of marine ecosystems,and it has become an effective way to monitor the red tide.Relied on the NSFC-GD jointly fund project,the algae under the South China Sea coastal waters as research subjects,combined different remote sensing images to analysis the spectral of chlorophyll-a,BP neural network was used to study the remote sensing way for extracting the chlorophyll-a concentration of marine algae. Based on field observation data,radar image and satellite remote sensing data,we had established a spectral retrieval model of marine algae chlorophyll-a concentration,providing references for monitoring the marine environment.In this study,BP neural network was used to extract remote sensing information of chlorophyll-a. Research through the following steps to achieve:(1)analyze the spectral characteristics of algae based on TM image,and then find out the spectral parameters of TM image:TM1,TM2,TM 3,TM4;(2)analyze the potential relationship between the backscattering coefficient in radar parameters and the concentration of chlorophyll-a,and then extract the feature parameters of radar:the backscatter coefficient under HH and VV polarization;(3)Through analyze the different input layer nodes and hidden nodes,the structure of BP neural network model in this study was finally fixed,that is,above six parameters as the input parameters,the hidden layer nodes is 6,and chlorophyll-a concentration as the model output parameters;(4)Constructing different input parameters of linear regression model,comparing the accuracy between the measured and predicted values,the advantages of BP neural network has been confirmed.The results showed that:(1)when combined the optical remote sensing and radar remote sensing,the prediction accuracy was much higher than TM or SAR used alone;(2)the relationship between remote sensing spectral characteristics and chlorophyll-a was more complex nonlinear problem,so the statistical model seemed clearly insufficient,but the BP neural network model can achieve better retrieval results , especially had obvious advantages in the inversion of chlorophyll-a concentration in II water.