Fully Stochastic Distributed Methodology for Multivariate Flood Frequency Analysis

Abstract

An adequate estimation of the extreme behavior of basin response is essential both for designing river structures and for evaluating their risk. The aim of this paper is to develop a new methodology to generate extreme hydrograph series of thousands of years using an event-based model. To this end, a spatial-temporal synthetic rainfall generator (RainSimV3) is combined with a distributed physically-based rainfall-runoff event-based model (RIBS). The use of an event-based model allows simulating longer hydrograph series with less computational and data requirements but need to characterize the initial basis state, which depends on the initial basin moisture distribution. To overcome this problem, this paper proposed a probabilistic calibration-simulation approach, which considers the initial state and the model parameters as random variables characterized by a probability distribution though a Monte Carlo simulation. This approach is compared with two other approaches, the deterministic and the semi-deterministic approaches. Both approaches use a unique initial state. The deterministic approach also uses a unique value of the model parameters while the semi-deterministic approach obtains these values from its probability distribution through a Monte Carlo simulation, considering the basin variability. This methodology has been applied to the Corbès and Générargues basins, in the Southeast of France. The results show that the probabilistic approach offers the best fit. That means that the proposed methodology can be successfully used to characterize the extreme behavior of the basin considering the basin variability and overcoming the basin initial state problem.

An Application of a Stochastic Semi-Continuous Simulation Method for Flood Frequency Analysis: A Case Study in Slovakia

“…Existing DFF methods could broadly be categorized into event-based (Lamb, 2005) and continuous simulation (Boughton et al, 2003) approaches. Methods combining the advantages of both approaches have also been proposed (see e.g., Li et al, 2014;Flores-Montoya et al, 2016).…”

“…When compared to continuous simulation approach, this method produced similar results with a much higher computational efficiency (a factor of 100-1000) and significantly reduced the biases in the estimation of design discharges, when compared to design storm approaches. Another semi-continuous method has been proposed by Flores-Montoya et al (2016). It uses a spatial-temporal stochastic rainfall model (RainSim) to generate continuous series of rainfall inputs and a distributed event-based rainfall-runoff model (RIBS) to create complete hydrographs.…”

The Influence of the Annual Number of Storms on the Derivation of the Flood Frequency Curve through Event-Based Simulation

“…The curve number method was applied for the rainfall-runoff transformation [52] and the hydrographs were generated by using the Soil Conservation Service dimensionless unit hydrograph procedure [52]. We adopted the simpler option of deterministic average initial conditions based on the results of Flores-Montoya et al [24]. For the propagation of the flood hydrographs throughout the stream channels, the Muskingum method was applied [53].…”

“…The antecedent soil moisture condition, in professional practice, is usually assumed as deterministic. Many authors have achieved good results considering the antecedent soil moisture condition as a random variable associated to a probability distribution function [16,[24][25][26]. Recently, Flores-Montoya et al [24] proposed a probabilistic approach that considered the initial state as a random variable characterized by a probability distribution determined through Monte Carlo simulation.…”

Reservoir Routing on Double-Peak Design Flood

“…Flores-Montoya et al [15] develop a new methodology to generate extreme hydrograph series using an event-based model, where a spatial-temporal synthetic rainfall generator is combined with a distributed physically based rainfall-runoff event-based model. Blazkova and Beven [16] apply a continuous simulation approach for the estimation of flood frequency for a dam in the Czech Republic; the methodology was implemented within a Generalized Likelihood Uncertainty Estimation framework.…”