Abstract. This paper compares event-based and continuous hydrological modelling approaches for real-time forecasting of river flows. Both are compared using a lumped hydrologic model (whose structure includes soil moisture accounting (SMA) store routing store) on data set 178 French catchments. The main focus this study was to investigate the actual impact initial conditions performance flood models possible compensations with updating techniques. rainfall-runoff assimilation technique we...

Abstract Different approaches used in hydrological modelling are compared terms of the way each one takes rainfall data into account. We examine errors associated with accounting for variability, whether (distributed vs lumped models) or computing catchment rainfall, as well impact approach on representativeness parameters it uses. The database consists 1859 events, distributed 500 basins, located southeast France areas ranging from 6.2 to 2851 km2. study uses reference hydrographs computed...

Abstract. To date, long short-term memory (LSTM) networks have been successfully applied to a key problem in hydrology: the prediction of runoff. Unlike traditional conceptual models, LSTM models are built on concepts that avoid need for our knowledge hydrology be formally encoded into model. The question, then, is how we can still make use domain and practices, not build themselves, as do but them more effectively. In present paper, adopt this approach, investigating information concerning...

This article presents a comparison between real-time discharges calculated by flash-flood warning system and post-event flood peak estimates. The studied event occurred on 15 16 June 2010 at the Argens catchment located in south of France. Real-time warnings were provided AIGA (Adaptation d'Information Géographique pour l'Alerte en Crue) system, which is based simple distributed hydrological model run 1-km2 resolution using radar rainfall information. timing (updated every min) was compared...

Abstract. The smash software is a differentiable and regionalizable framework enabling modular high-resolution hydrological modeling data assimilation, from catchment to regional country scales, for water research operational applications. combines various process-based conceptual operators vertical lateral flows, which can be hybridized with descriptors-to-parameters neural network regionalization. features an efficient, Fortran solver using Tapenade that supports CPU parallel computing...

Abstract Floods are major natural disasters that, in several occasions, can be responsible for life losses and severe economic damages. Flood forecasting alert systems needed to anticipate the arrival of these events mitigate their impacts. They particularly important risk management response nowcasting flash floods. In this case, precipitation fields crucial is consider uncertainties coming from observed used as input data system. One approach take into account generate an ensemble possible...

Occurring at small temporal and spatial scales, flash floods (FF) can cause severe economic damages human losses. To better anticipate such events mitigate their impacts, the French Ministry in charge of Ecology has decided to set up a national FF warning system over territory. This automated will be run by SCHAPI, service flood forecasting, providing warnings for fast-responding ungauged catchments (area ranging from ~10 ~1000 km2). It therefore complementary SCHAPI's "vigilance" which...

Abstract. This contribution presents a regionalization approach to estimate spatially distributed hydrologic parameters based on: (i) the SMASH (Spatially Modelling and ASsimilation for Hydrology) hydrological modeling assimilation platform (Jay-Allemand, 2020; Jay-Allemand et al., 2020) underlying French national flash flood forecasting system Vigicrues Flash (Javelle 2019); (ii) variational algorithm from (Jay-Allemand 2020), adapted high dimensional inverse problems; (iii) spatial...

Abstract Most often, flood frequency analysis describes a event only by its peak. However, the true severity is also defined volume and duration. This paper presents an approach allowing events to be considered in more complete way: flood‐duration–frequency (QdF) approach. In similar manner rainfall intensity–duration–frequency analysis, averaged discharges are computed over different fixed durations d . For each duration distribution of maximum studied. Finally, continuous formulation...

Abstract. This paper compares event-based and continuous hydrological modelling approaches for real-time forecasting of river flows. Both are compared using a lumped hydrologic model (whose structure includes soil moisture accounting (SMA) store routing store) on data set 178 French catchments. The main focus this study was to investigate the actual impact initial conditions performance flood models possible compensations with updating techniques. rainfall runoff assimilation technique we...

The dreadful floods of 1999, 2002 and 2003 in South France have alerted public opinion on the need for a more efficient further generalized national flood-forecasting system. This is why Irstea Meteo-France implemented new warning method flash floods, including small watersheds, using radar rainfall data real-time: AIGA method. modelling currently provides real-time information magnitude but doesn't take into account elements at risk surrounding river streams. Its benefit crisis management...

Abstract. Every year in France, recurring flood events result several million euros of damage, and reducing the heavy consequences floods has become a high priority. However, actions to reduce impact are often hindered by lack damage data on past events. The present paper introduces new database for collection assessment flood-related damage. DamaGIS offers an innovative bottom-up approach gather identify from multiple sources, including media. study area been defined as south France...

Abstract. Reducing uncertainty and improving robustness spatio-temporal extrapolation capabilities remain key challenges in hydrological modeling especially for flood forecasting over large areas. Parsimonious model structures effective optimization strategies are crucially needed to tackle the difficult issue of distributed calibration from sparse integrative discharge data, that is general high dimensional inverse problems. This contribution presents first evaluation Variational Data...

Abstract. Calibration of a conceptual distributed model is challenging due to number reasons, which include fundamental (model adequacy and identifiability) algorithmic (e.g., local search vs. global search) issues. The aim the presented study investigate potential variational approach for calibrating simple continuous hydrological (GRD; Génie Rural involved in several flash flood modeling applications. This defined on rectangular 1 km2 resolution grid, with three parameters being associated...

Abstract. In the field of Deep Learning, long short-term memory (LSTM) networks lie in category recurrent neural network (RNN) architectures. The distinctive capability LSTM is learning non linear term dependency structures. This makes a good candidate for prediction tasks time dependent systems such as runoff catchment. this study, we use large sample 740 gauged catchments with very diverse hydro-geo-climatic conditions across France. We present regime classification based on three...

Abstract Estimating spatially distributed hydrological parameters in ungauged catchments poses a challenging regionalization problem and requires imposing spatial constraints given the sparsity of discharge data. A possible approach is to search for transfer function that quantitatively relates physical descriptors conceptual model parameters. This paper introduces Hybrid Data Assimilation Parameter Regionalization (HDA‐PR) incorporating learnable mappings, based on either multi‐linear...

<p>The prediction of extreme hydrological events at high resolution is a tough scientific challenge linked to major socio-economic issues. Accurate numerical models are crucially needed perform reliable and meaningful operational predictions. In this context, modern efficient modeling tools required integrate progress advances, take advantage the wealth information provided by new generations satellite sensors in complement situ data, meet needs diverse end...

Abstract. Flash floods have dramatic economic and social consequences, efficient adaptation policies are required to reduce their impacts, especially in the context of global change. Developing more flash flood forecasting systems can largely contribute these requirements. The aim this study was assess ability a new seamless short-range ensemble quantitative precipitation forecast (QPF) product, called PIAF-EPS (Prévision Immédiate Agrégée Fusionnée prediction system) recently developed by...