Radiative Transfer Model (RTM)-based Water Pollution Parameters Retrieval From Landsat TM/ETM Data

Water quality pollution is a serious problem, which is needed to be solved immediately in China. Water quality remote sensing survey is an inversion algorithm by investigating the relationship of water reflected spectrum characters and water quality index concentration. There are three ways, which are customarily used to resolve the problem, including empirical, semi-empirical and mechanism modeling methods. The information of water pollution can be extracted by experiential expressions between some visible band or band ratio and pollution concentration. The method can be carried out very easily, but it is depended upon the measured data of water surface and synchronous data of remote sensing. However, the accuracy of this method will be significantly influenced by the changing of the water sediment concentration. The best way to the issue is to establish the physical model between reflection spectrum and contamination concentration. Thus, a radiative transfer model inbetween water is suggested for the water pollution investigation.Based on the radiative transfer process, a single scattering radiative transfer .model is first built under deeper water (>2m). The precision of single scattering radiative transfer .mode is lower if the water is turbid. Thus, a multiple scattering radiative transfer model is also presented for simulation of radiative transfer model. Taking Shitoukoumen reservoir for example, two models, single scattering and multiple scattering radiative transfer, are all applied to retrieve the water quality parameters.Three dates of Landsat TM/ETM data are acquired for these models, that is, Landsat TM on September 7, 1993 and Landsat ETM data on October 20, 2000 and September 21, 2001. Radiative calibration and geometry correction are first achieved. And the atmospheric correction is implemented for three dates of Landsat TM/ETM data under the FLAASH module of ENVI.The ETM+ data on September 21, 2001 is first employed to inverse water quality parameters. The results illustrates that heavier pollution in central area while lighter pollution in north and west areas of Shitoukoumen Reservoir. The density of suspended sediment tends to be decreasing from central part of reservoir toward north. There is a very high sand content in the southern shallows. The results from single and multiple scattering models are compared to show high accuracy of multiple scattering models. The maximum relative error is reduced from 20.49% to 20.45%, maximum absolute error from 13.75% to 13.69%, mean square error from 0.0136 to 0.0132 from single to multiple scattering models. Meanwhile, the water quality parameters are compared from three dates of Landsat TM/ETM data. It shows that the pollution is becoming more and more severe from 1993 to 2001.In order to verify the sensitivity of radiative transfer model, the solar azimuth angle and priori values of water pollution concentration are evaluated for the inverse. The result told us that the concentration of pollutants and suspended substance has changed a little with the changing of zenith angle. The variation range is about from 0.01 to 0.09. The priori data of the pollutants are among 0.01 to 0.09 and suspended materials is among 0.1 to 0.9. The results show that the inversed pollutant concentrations are very sensitive to the priori values and increases abruptly with the increase of the priori values.