Abstract
Inverse estimation of the flow resistance (or bed roughness) because of thick vegetation in the flooded rivers was conducted using variational and sequential data assimilation methods. An adjoint method and a particle filtering technique were effectively used with the shallow-water model for searching optimum values of distributed Manning's roughness coefficients in both space and time. These data assimilation methods were applied to the flooded Asahi River which occurred in September 2011. The assimilated data were water levels that were longitudinally observed in the target reach using 18 water-stage recorders. Results showed that the optimum roughness coefficients analyzed by two methods present the similar trend in both space and time. They also showed such a trend is closely related with the establishment of type of dominant vegetation and vegetation wash-out phenomena during flooding.
