Uncertainty Analysis a single event distributed rainfall-runoff model with using two different Markov Chain Monte Carlo methods

So far flood forecasting and simulation in hydrologic literature suffer from lack of explicit recognition of forcing and parameter and model structural error. However since any model is a simplification of reality there remains a great deal of uncertainty even after the calibration of model parameters. Often parameters in hydrologic models cannot be measured directly and can only be inferred by a calibration process. This work addresses the application and comparison of two parameter uncertainty methods and their effects on the prediction of streamflow in Abolabbas watershed (290 km2) located in Khuzestan Province. Two Markov Chain Monte Carlo methods، Shuffled Complex Evolution Metropolis (SCEM UA) and DiffeRential Evolution Adaptive Metropolis (DREAM) were used in this study to quantify parameter uncertainty in AFFDEF implemented in FORTRAN language programming، a single event distributed rainfall runoff model. Respectively، four and two historical floods with hourly time step for calibration and validation periods were selected. More than 45000 Simulation runs with 15 chains were done to indicate convergence to a stationary posterior distribution. For example، results showed that p-factor and d-factor values for event 31/1/1993 changed from 0. 92 and 2. 25 in SCEM-UA approach and reached to 0. 96 and 1. 56 in DREAM approach respectively. Finally posterior parameters distributions were created with samples from last 20 percent of chains. All of these posterior distributions properly become smaller than initial upper and lower bounds.