Characterization of process-oriented hydrologic model behavior with temporal sensitivity analysis for flash floods in Mediterranean catchments

Abstract

Abstract. This paper presents a detailed analysis of 10 flash flood events in the Mediterranean region using the distributed hydrological model MARINE. Characterizing catchment response during flash flood events may provide new and valuable insight into the dynamics involved for extreme catchment response and their dependency on physiographic properties and flood severity. The main objective of this study is to analyze flash-flood-dedicated hydrologic model sensitivity with a new approach in hydrology, allowing model outputs variance decomposition for temporal patterns of parameter sensitivity analysis. Such approaches enable ranking of uncertainty sources for nonlinear and nonmonotonic mappings with a low computational cost. Hydrologic model and sensitivity analysis are used as learning tools on a large flash flood dataset. With Nash performances above 0.73 on average for this extended set of 10 validation events, the five sensitive parameters of MARINE process-oriented distributed model are analyzed. This contribution shows that soil depth explains more than 80 % of model output variance when most hydrographs are peaking. Moreover, the lateral subsurface transfer is responsible for 80 % of model variance for some catchment-flood events' hydrographs during slow-declining limbs. The unexplained variance of model output representing interactions between parameters reveals to be very low during modeled flood peaks and informs that model-parsimonious parameterization is appropriate to tackle the problem of flash floods. Interactions observed after model initialization or rainfall intensity peaks incite to improve water partition representation between flow components and initialization itself. This paper gives a practical framework for application of this method to other models, landscapes and climatic conditions, potentially helping to improve processes understanding and representation.

Parameter regionalization for a process-oriented distributed model dedicated to flash floods

“…Other regionalization studies find also correlation coefficients usually lower than 0.5 for continuous rainfall runoff models (Merz et al, 2006;Oudin et al, 2008). The highest correlations are found for C Z , the multiplicative constant of the soil depths (Table 5), which is also the most sensitive parameter of the model (Garambois et al, 2013(Garambois et al, , 2015. Indeed, the soil depth multiplicative constant explains 80% of model output variance when most hydrographs are peaking (Garambois et al, 2013).…”

“…MARINE model parameters present few interactions during peak flow simulations as shown with temporal variance analysis on 6 Mediterranean catchments (Garambois et al, 2013). This probably stems from model parsimony and physical formulation.…”

“…Several sensitivity analysis and calibration/validation (cal/val) of the model have been performed for catchments of the French Mediterranean region with areas ranging from about 100 km 2 to 700 km 2 (Garambois et al, 2013(Garambois et al, , 2015Roux et al, 2011). The results of these studies show that soil depth and lateral water transfer through the subsurface zone have a significant impact on soil saturation dynamics and flood hydrograph.…”

“…The highest correlations are found for C Z , the multiplicative constant of the soil depths (Table 5), which is also the most sensitive parameter of the model (Garambois et al, 2013(Garambois et al, , 2015. Indeed, the soil depth multiplicative constant explains 80% of model output variance when most hydrographs are peaking (Garambois et al, 2013). Catchment soil volume from pedologic data has to be adjusted with C Z which is the most influent parameter of MARINE model on average.…”

“…This approach has been chosen for three parameters, namely the distributed saturated hydraulic conductivity K, the lateral transmissivity T 0 and soil thickness Z. The calibration procedure consists in estimating: three coefficients of correction for spatialized data; one for the saturated hydraulic conductivities, named C K , another one C KSS for the lateral subsurface flow transmissivities (T 0 ), and the last one for the soil thicknesses, named C Z , the Strickler roughness coefficient of the main channel K D1 and the Strickler roughness coefficient of the overbanks of the drainage network K D2 (Garambois et al, 2013;Roux et al, 2011). Concerning the transmissivity K ss , the spatial variability is taken from the hydraulic conductivity map.…”