HW13 Hydrological Predictions in Ungauged Basins

IAHS (Hydrology)



23-Jun-2015, 13:30 - 15:00


 
Abstract content:

Improving the bayesian joint inference through the inclusion of hydrological state variables in the residuals dependence model

In hydrological modeling, the mixture of an unsuitable error model with the input errors and the hydrological model structural deficiencies is the main cause of yielding biased calibrated parameters and hydrological models which can seemingly works well but not for the right reasons. Biased parameters are an obstacle for proper regionalization processes with the aim of predicting the response at ungauged sites.

This research focuses on the Bayesian joint inference (BJI) of both the hydrological and error model parameters, considering a general additive (GA) error model that allows for correlation, non-stationarity (in variance and bias) and non-normality of model residuals. The joint inference approach presented here deviates from previous researches in two main points: i) non-stationarity in errors variance and bias is modeled, taking into account the Total Laws; ii) it is considered the possibility that the residuals contain some information which may be correlated with state variables of the hydrological model. This correlation would be caused by model structural problems which hinder the proper processing of the Forcing signals by the hydrological model. In this way, a multiple correlation analysis, embedded in the residuals autoregressive model (ARX model in literature), allows the extraction of remnant information from residuals, giving us clues of which processes (or state variables) can be responsible of the unprocessed information.

The results of this research show that the application of BJI with a GA with ARX error model (GAARX) outperforms the hydrological parameters robustness and improves the reliability of the streamflow predictive distribution, in respect of the results of GA error model which not take into account the dependence residuals-state variables.
 
Author(s):
M.R. Hernández1, F. Francés1.
1Universidad Politécnica De Valencia, Research Institute of Water and Environmental Engineering, Valencia, Spain.

 

Keywords:
Distributed Hydrology     Uncertainty     Bayesian Formal Inference     Error Models     Unbiased Parameters Estimation