1. HydroPredicT_Extreme: A probabilistic method for the prediction of extremal high-flow hydrological events.
- Author
-
Molina, Jose-Luis, Espejo, Fernando, Zazo, Santiago, Molina, María-Carmen, Hamitouche, Mohamed, and García-Aróstegui, José-Luis
- Subjects
- *
FLOOD control , *CAUSAL models , *ARTIFICIAL intelligence , *HYDROLOGIC models , *GREEN roofs - Abstract
• A novel method for assessing prediction of extremal hydrological events. • Bayesian causal modelling tool for predictive assessment. • Hybridization of Classical and new Hydrology based on Artificial Intelligence. • HydroPredicT_Extreme as a new tool for the prediction of potential flooding events. • Applicability to different time scales of extreme and sudden events. Disastrous losses related to high-flow events have increased dramatically over the past decades largely due to an increase in flood-prone regions settlements and shift in hydrological trends largely due to Climate Change. To mitigate the societal impact of hydrological and hydraulic extremes, knowledge of the processes leading to these extreme events is vital. Hydrological modelling is one of the main tools in this quest for knowledge but comes with uncertainties. For that it is necessary to deeply study the impact of hydrological models' structure on the magnitude and timing of extreme rainfall-runoff events. This paper is mainly aimed to show the development of a method called "HydroPredicT_Extreme" based on Bayesian Causal Modelling (BCM), a technique within Artificial Intelligence (AI). This method may enhance predictive capacity of extreme rainfall-runoff events. "HydroPredicT_Extreme" follows an iterative methodology that comprise 2 main stages. First one comprises a mixed graphical/analytical method from Hydrograph. This stage is conditioned by two initial constraints which are, a) pluviometry station is representative of hydrograph downstream flow behaviour; b) there must be independence of events. This first stage comprises sub-phases such as: 1.1. Calculation of Response Time (RT) through a mixed graphical/analytical approach, 1.2 Subtraction of RT from the flow series to remove the Rainfall-Flow delay; 1.3 Calculation base flow rate; 1.4 Subtraction base-flow from flow series to work on absolute inputs. Second man stage is called Bayesian Causal Modelling Translation (BCMT) that comprises the 2.1 Learning, 2.2 Training, 2.3 Simulation through BCM modelling, 2.4 Sensitivity Analysis-Validation. This whole methodology will become a digital application and software that could be extrapolated to several similar case studies. This may be coupled with posterior devices for the prevention of catastrophic flood consequences in the form of MultiHazard-Early Warning System (MH-EWS) or others. [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF