Remote Sensing Application in Estimating Marine Water Quality

With the development of coastal marine area, water body pollution is becoming an outstanding environmental problem. Considering the characteristics that different water quality has different spectral signature, remote sensing technique has great advantage and potential in assessing water quality whereas direct measurements are expensive and time-consuming. This paper concentrates on sea water quality estimation of coastal marine area in Hong Kong in terms of four parameters. They are chlorophyll, Suspended Solids (SS), turbidity, Secchi disk depth (SDD). An empirical neural network algorithm is chosen to estimate the correlations between digital data and water quality parameters. This method is better than traditional regression model in modeling transfer functions with higher accuracy. The network is divided into three layers—input layer, hidden layer and output layer. Landsat TM data and water sample data are input layer whilst four variables are treated as output layer. Hidden layer which is consisted of a varying number of neurons is used for the summation and activation functions. Usually, the optimal size of the hidden layer is usually between the size of the input and size of the output layer so that there are 4 neurons in the hidden layer in this study. Samples data which are collected by Environment Protection Department (EPD) in 2011 are separated into two sets. One set is used for training the network while the other is used for testing the retrieval result. At last, through comparison between estimated water quality parameters and real measurement data, the accuracy of this method can be derived. The correlation coefficient can be obtained by forming a 2D scatter plot. Hence, this research may benefit for monitoring the change of water quality and have a significant on water resources protection issue.