内陆湖泊的水质遥感监测研究

THE STUDY ON WATER QUALITY OF INLAND LAKE MONITORING BY REMOTE SENSING

介绍了内陆湖泊水质遥感监测的特点及遥感监测水质的机理,总结了国内外近年来用于内陆湖泊水质参数反演的 3种常用方法:经验模型、生物光学模型和神经网络模型,并分析了 3种模型的优缺点;同时指出了影响内陆湖泊水质遥感监测精度的关键因素;提出了内陆湖泊水质遥感监测研究重点和方向。

The characteristic and theory of water quality of inland lake monitoring using remote sensing are addressed in this paper. Inland lake water quality remote sensing differs from ocean color remote sensing, which demands remote sensing data with high spatial & spectral resolution and complex inversion algorithm. The advantages and disadvantages of three common methods of water quality quantity inversion: empirical model, bio-optical model, artificial neural network model are discussed in this paper. Empirical model is a simple and convenient model, but not an universal model, the model only fit to the given region and lake, and to construct the empiric model needs a lot of sampling data, furthermore the empirical model just only precisely retrieve the water quality parameters in the given range, and the retrieve precision will fall greatly beyond the range. Bio-optical model is an universal and robust model, which can retrieve water quality parameters only from radiance or reflectance on the remote sensor without the support of in-site data, However, this is based on the comprehension of the absorption coefficient, scatter coefficient and volume scatter of the pure water , suspended substance , chlorophyll and yellow substances. Neural network model is an efficient inversion way, which can simulate complex relation and utilize various kinds of remote sensing data, and which can deal with vast data in the litter time, but the neural network model is dependent on the training data and the model construction needs a lot of time and much experience. The neural network model is a “classifier” not an “extractor”. The factors determining water quality retrieval accuracy are also analyzed, the atmosphere plays an important role in the water quality parameters inversion, and the accurate atmosphere correction model must be developed for water quality remote sensing. Finally, the future directions and the key points in this filed are proposed, the water quality parameters solar reflection response rule should be carefully investigated and the spectral reflection database of characteristic water of China is supposed to build in the near future. Radar remote sensing and hyperspectral technology should be emphasized in the future research on the water quality remote sensing. In China, researches on the theory of bio-optical model and the measurement of absorption coefficient and scatter coefficient of water quality parameters of Chinese characteristic lakes ought to be strengthen.