Document Type : Research Article
Real time updating is inevitable in flood forecasting systems due to uncertainty in hydrologic models and input data. In these systems, hydrological parameters are usually updated based on online recorded discharge in a telemetry system. Conventional methods-such as state space models have widely been used in real time updating of hydrological parameters in the context of flood forecasting systems. In this paper, however, we proposed a Monte Carlo (MC) based model in updating of flood forecasts. Generalized likelihood uncertainty estimation which is combination of MC and Bayesian theory, is applied to update rainfall runoff (RR) model parameters based on real time data. In this methodology, a large set of hydrological parameters is produced by random generation and then probability distribution of forecast is computed based on the results of RR simulation for the set of parameters. In real time updating of RR parameters, prior and posterior likelihoods are computed using forecast errors that are obtained from the results of behavioral models and real time recorded discharges. Then, prior and posterior likelihoods are applied to modify forecast confidence limits in each time step. In the updating process, probability distribution of peak discharge becomes stable prior to the time to peak and lead time of forecast is determined based on the stable time and real time to peak. The applied approach shows more flexibility in varying the parameters and modifying forecast limits in comparison with other conventional updating methods.