Application of neural network and MODIS 250 m imagery for estimating suspended sediments concentration in Hangzhou Bay, China

Suspended sediments concentration (SSC) in surface water derived from bottom sediment resuspension or discharge of sediment-laden rivers is an important indication of coastal water quality and changes rapidly in high-energy coastal area. Since artificial neural networks (ANN) had been proven successful in modeling a variety of geophysical transfer functions, an ANN model to simulate the relationship between surface water SSC and satellite-received radiances was employed. In situ SSC measurements from the Hangzhou Bay and the Moderate-resolution Imaging Spectroradiometer (MODIS) 250 m daily products were adopted in this study. Significant correlations were observed between in situ measurements and band 1-2 reflectance values of MODIS images, respectively. Results indicated that application of ANN model with one hidden layer appeared to yield superior simulation performance (r (2) = 0.98; n = 25) compared with regression analysis method. The RMSE for the ANN model was less than 10%, whereas the RMSE for the regression analysis was more than 25%. Results also showed that different tidal situations affect the model simulation results to some extent. The SSC of surface water in Hangzhou Bay is high and changes rapidly due to tidal flood and ebb during a tidal cycle. The combined utilization of Terra and Aqua MODIS data can capture the tidal cycle induced dynamic of surface water SSC. This study demonstrated that MODIS 250 m daily products and ANN model are useful for monitoring surface SSC dynamic within high-energy coastal water environments.