Identification of spatial and temporal contributions of rainfalls to flash floods using neural network modelling: case study on the Lez Basin (Southern France)

Abstract

Abstract. Flash floods pose significant hazards in urbanised zones and have important human and financial implications in both the present and future due to the likelihood that global climate change will exacerbate their consequences. It is thus of crucial importance to better model these phenomena especially when they occur in heterogeneous and karst basins where they are difficult to describe physically. Toward this goal, this paper applies a recent methodology (KnoX methodology) dedicated to extracting knowledge from a neural network model to better determine the contributions and time responses of several well-identified geographic zones of an aquifer. To assess the interest of this methodology, a case study was conducted in Southern France: the Lez hydrosystem whose river crosses the conurbation of Montpellier (400 000 inhabitants). Rainfall contributions and time transfers were estimated and analysed in four geologically-delimited zones to estimate the sensitivity of flash floods to water coming from the surface or karst. The Causse de Viol-le-Fort is shown to be the main contributor to flash floods and the delay between surface and underground flooding is estimated to be three hours. This study will thus help operational flood warning services to better characterise critical rainfall and develop measurements to design efficient flood forecasting models. This generic method can be applied to any basin with sufficient rainfall–runoff measurements.

The Impact of Rainfall Space‐Time Structure in Flood Frequency Analysis

“…It may be that for dry soils, subcatchment‐scale convective cells embedded within major storm events will produce at least some areas of fast surface runoff. When fine‐scale spatial structure is eliminated by averaging rainfall across the entire basin, it becomes less likely that fast surface runoff is produced (e.g., Darras et al, ).…”

Deep Multilayer Perceptron for Knowledge Extraction: Understanding the Gardon de Mialet Flash Floods Modeling

“…This supplementary layer is called "i" (linear hidden neurons) in Fig 2 . The second hidden layer (non-linear hidden layer) calculates a non-linear combination of the "locally added" rains. The KnoX method [7,8,9] allows calculating a "simplified" contribution of each input to the model output. This method is described for the general deep model (two hidden layers) shown in Fig.…”

“…For this reason and to enhance the understanding of the behavior of both the model and the physical processes, several works have been done to bring more transparency in the operating mode and introduced concepts of gray-box and transparent-box models [5,6]. Some other works have been conducted to make neural networks models more hydrologically meaningful [6,7,8].…”

Artificial Intelligence (AI) Studies in Water Resources

“…Recent studies Dam or lake water level Hipni et al, 2013;Üneş et al, 2015;Li et al, 2016 Evaporation and evapotranspiration Goyal et al, 2014;Karimi et al, 2016;Güçlü et al, 2017 Rainfall-runoff Talei et al, 2013;Darras et al, 2015;Londhe et al, 2015;Chithra &Thampi, 2016 Sediment Demirci andBaltaci, 2013;Güner and Yumuk, 2014;Droppo & Krishnappan, 2016;Talebi et al, 2016Streamflow Cigizoglu, 2003Huang et al, 2004;Nourani et al, 2012;Ashrafi et al, 2017 Water quality variables Ay, 2010;Akkoyunlu et al, 2011;Ay & Kisi, 2011;Ay & Kisi, 2012;Ay & Kisi, 2013a;Ay & Kisi, 2013b;Kisi & Ay, 2013;Ay, 2014;Ay & Kisi, 2014;Chang et al, 2014;Alizadeh & Kavianpour, 2015;Khan & Valeo, 2015;Ay & Kisi, 2017 …”