Water Quality Management in the Estuary of Pearl River and Hong Kong'S Coastal Waters Based on Seawifs, Landsat Tm Sensor Data and in Situ Water Quality Sampling Data

XL Chen,YS Li,ZG Liu,ZL Li,OWH Wai,B King

DOI: https://doi.org/10.1117/12.467325

2003-01-01

Abstract:The Pearl River estuary and Hong Kong's coastal waters were selected to study the ocean color categories related to water quality. Three ocean color sensitive parameters: turbidity, suspended sediments (SS) and chlorophyll-a concentration (Chl-a), in 58 monitoring stations were selected to evaluate the water quality A dataset with 88 samples was picked up from the monitoring stations and the successfully retrieved points of SS and Chl-a from SeaWiFS, 66 of the 88 samples were used as training data and the other 22 as testing data. The normalized difference water index was extracted from the Landsat TM image on Dec. 22, 1998 and the threshold segmentation was used to retrieve the waters from the image for further analysis. The methods of maximum likelihood, neural network and support vector machine were employed for ocean color classification of the selected Landsat TM image. Five classes of water quality could be well interpreted for all the methods. The results showed spatial variation from the west turbid waters to the east relative clear waters and suggested that the turbid waters could be well classified using Landsat TM data.