Spatial Pattern Analysis On Suspended Solids Concentration Based On The CCD Data In The Coast Water Of Hongkong

The Spatial patterns of suspended solids concentration (SSC) in coastal waterrelated to the port construction, ecological water quality assessment, ocean climateanalysis. In the ecological field,it was an great issue to the people. The data ofRemote sensing provided a advantage to the research in the different scale in thevariability quantitative analysis of SSC.This article studied the spatial distribution and the variability of SSC in the coastwater of Hongkong leveraging the HJ-1A/B CCD data of2009~2012and themeasured data. The remote sensing inversion model had been set up. Based on which,the analysis was conducted by means of the geostatistical method. The work andconclusions have six sides.i. This study used the RBF neural network algorithm and established the modelbetween the reflectance and the real value of SSC, the reflectance as the input data,the real value of SSC as the output data. The parameter of the model is set. The inputdata is the remote reflectance of B2、B3、B4, the size of hidden layer is9, the outputdata is the value of SSC, Goal is0.01, Spread is1.ii. The retrieval model has a high fitting precision, the correlation coefficient is0.89, the relative error of1.77, the measured squared error is1.54. Comparing to theLinear and logarithmic model, the error is reduced10times nearly. It suggested thatRBF neural network algorithm is superior to traditional linear inversion model andlogistic regression model in retrieving the SSC in coastal waters of Hong Kong, whenchoosing appropriate parameters precision. It completely meet the need in thequantitative analysis of spatial pattern in the study area.iii. This paper divided the study into four blocks, using the RBF to retrieval theSSC in different seasons from2009to2012. We analysis the spatial distribution ofSSC in the study area through the parameters as Nugget\Basis\the variationcoefficient.iv. According to the statistical data from2009to2012, the average value isbetween0.01~26.1mg/L all the four blocks. The value is higher in the north than the south. The variability characters is that the north part is strong,and the north part isweak. The proportion between Nugget and Basis is among0.01to0.49, it means astrong autocorrelation, and the value is lower in the north than that in the south allthrough the area.