Monitoring Of Water Quality Parameters For Inland Waters By Multi-spectral Remote Sensing Data

Water quality influences water resource's ecological functions greatly, and quality of inland water is concerned with industry, agriculture and living quality. Inland water quality monitoring plays an essential role in developments of society and economic. The applications remote sensing techniques in inland water quality monitoring become more and more important. Since inland water variables (such as suspended matter, phytoplankton etc.) are spatially inhomogeneous parameters, corresponding synoptic information cannot be obtained from in situ monitoring network. This problem can be solved by the integration of water quality models, in situ data and remote sensing data, which provide spatial distribution information. The other two advantages of remote sensing techniques are its fastness and economy, which make it suitable for long-term dynamic water quality monitoring.This study aimed at developing and applying an operative method to estimate key parameters related to lake water quality, such as chlorophyll-a (chl-a) and total suspended particles (TSP), using multi-spectral remote sensing data. This paper focused on the development and applications of the remote sensing models of chl-a and TSP, using Lake Taihu, China as a case study. Landsat 5 TM data from 2003 and 2004, and synchronous in-situ measurements were used. Some key issues involved in applying multi-spectral satellite imagery to inland water quality, including atmospheric correction, methods for multi-spectral bands analysis were discussed. Regression models for chl-a and TSP were recommended, and the accuracies and the multi-temporal applicability of these models were analyzed.The results indicated that Landsat5 TM data were suitable for estimating the concentrations of chl-a and TSP in inland waters. TM4 has the strongest single-band relationship with chl-a (with r=0.721). The use of TM4/TM3 improved r to 0.819. Adding terms of TM1 and TM2 to TM4/TM3 ratio improved the mean adjusted r2 to 0.809. TM3 is the best band for detection of TSP for waters with high concentration of phytoplankton, while TM4/TM1 has the strongest relationship with TSP (with r=0.859) for waters with low concentration of phytoplankton (with concentration of chl-a less than 5 mgm" ).A four-coefficient regression model using TM4/TM3 ratio and TM1, TM2 was a reliable predicator of chl-a (r2 = 0.837) for waters with chl-a concentration between 5 mgm-3 and 100 mgm-3. This model will underpredict the values at very high chl-a concentration values (higher than 100 mgm-3), and overpredict the values at very low chl-a concentration (lower than 5 mgm-3). A two-coefficient regression model using TM3 was consistent and reliable predicator of TSP (r2 = 0.63). The multi-temporal applicability of the models has been demonstrated by the successful application to remote sensing data from different dates.The models were applied to estimate chl-a and TSP for the whole Taihu Lake, allowing for mapping and analysis of the spatial trends of water quality parameters.