Retrieval of water quality parameters by neural network and analytical algorithm in Guanting Reservoir in Hebei Province in China

Abstract

Based on the measured spectra in the research area of Guangting reservoir, we build the model to retrieve chlorophyll-a, suspended solid and yellow substance. The paper mainly achieved the following results: we adopted matrix inversion method and L-M & NN method to analyse the water quality parameters, the selection schemes of the spectral band include REF, DER, RAN method, and then use the Guanting Reservoir experiment data for inspection and comparative analysis. The results showed that: the retrieval accuracy of L-M & NN method was better than matrix inversion method for three ocean color elements, the RAN weighted method overall had better retrieval accuracy. For Cchla, acdom (440), the best scheme in band selection is RAN, REF showed better retrieval accuracy for Cs.