Event-based Bayesian causal modelling for flood hydrograph prediction, Upper Andarax intermittent stream, Spain

Abstract

Study region: The study region comprises an intermittent stream named Upper Andarax, within a semi-arid watershed in SE Spain (Andarax river basin), characterised by uneven topography and its striking hydraulic heritage. Study focus: High flow extreme events are becoming more common and less exceptional, leading to devastating losses with very high associated economical costs. This paper presents the development of a predictive Bayesian Causal Model (BCM) for high-flow extreme hydrological events assessment, accounting for: extreme events characterisation and sub-modelling; base-flow consideration; and previous soil moisture influence. It comprises three main phases that are: 1) Data preparation; 2) Extreme events characterisation and analysis; and 3) BCM predictive modelling. This study could be a decision-making tool for flood risk management in the Andarax River basin. New hydrological insights for the region: Results suggest a complex and dynamic non-linear rainfall-runoff relationship in the Upper Andarax watershed, whereas, there is a strong similarity in the temporal distributions of flow discharges, being, in general, symmetric, as a result of symmetric hyetographs. The statistical dependency showed that the basin response depends much more on the rainfall intensity rather than on the soil antecedent moisture conditions. The simulated BCMs demonstrated that all events’ types may lead to river flooding, while the most hazardous are those of quartiles III and IV.

Keywords

• Hydrological extremes
• Artificial intelligence
• Bayesian Causal Modelling
• SE Spain
• Antecedent moisture condition
• Intermittent streamflow