Modeling River Runoff Temporal Behavior through a Hybrid Causal–Hydrological (HCH) Method

Abstract

The uncertainty in traditional hydrological modeling is a challenge that has not yet been overcome. This research aimed to provide a new method called the hybrid causal–hydrological (HCH) method, which consists of the combination of traditional rainfall–runoff models with novel hydrological approaches based on artificial intelligence, called Bayesian causal modeling (BCM). This was implemented by building nine causal models for three sub-basins of the Barbate River Basin (SW Spain). The models were populated by gauging (observing) short runoff series and from long and short hydrological runoff series obtained from the Témez rainfall–runoff model (T-RRM). To enrich the data, all series were synthetically replicated using an ARMA model. Regarding the results, on the one hand differences in the dependence intensities between the long and short series were displayed in the dependence mitigation graphs (DMGs), which were attributable to the insufficient amount of data available from the hydrological records and to climate change processes. The similarities in the temporal dependence propagation (basin memory) and in the symmetry of DMGs validate the reliability of the hybrid methodology, as well as the results generated in this study. Consequently, water planning and management can be substantially improved with this approach.