Discrete Dynamic Model of Watershed Sediment Yield

Abstract

The watershed fluvial system is hypothesized as a lumped system with runoff rate (log-transformed) as input and sediment yield rate (log-transformed) as output on a discrete time basis. The effects of errors in sampling, data reduction, discretizing etc. of input and output sequences are simulated by another linear system to which input is a white noise. A first order dynamic model with five parameters is demonstrated to provide an adequate model of monthly runoff-sediment yield process. A second order dynamic model with seven parameters adequately models the daily runoff sediment yield process. The noise component was found to be a white noise sequence in both cases, explaining 10% to 19% of the variance in the monthly sediment yields and 3% to 5% of the variance in the daily sediment yields. A linear linkage model explains 81% to 90% of the variance in the monthly sediment yields and 95% to 97% of the variance in daily sediment yields.