A. Piacentini

Correction of upstream flow and hydraulic state with data assimilation in the context of flood forecasting

The present study describes the assimilation of river water level observations and the resulting improvement in flood forecasting. The Kalman Filter algorithm was built on top of a one-dimensional hydraulic model which describes the Saint-Venant equations. The assimilation algorithm folds in two steps: the first one was based on the assumption that the upstream flow can be adjusted using a three-parameter correction; the second one consisted of directly correcting the hydraulic state. This procedure was applied using a four- day sliding window over the flood event. The background error covariances for water level and discharge were repre- sented with anisotropic correlation functions where the cor- relation length upstream of the observation points is larger than the correlation length downstream of the observation points. This approach was motivated by the implementation of a Kalman Filter algorithm on top of a diffusive flood wave propagation model. The study was carried out on the Adour and the Marne Vallage (France) catchments. The correction of the upstream flow as well as the control of the hydraulic state during the flood event leads to a significant improve- ment in the water level and discharge in both analysis and forecast modes.

On the Merits of Using a 3D-FGAT Assimilation Scheme with an Outer Loop for Atmospheric Situations Governed by Transport

Abstract Three-dimensional variational data assimilation (3D-Var) with the first guess at appropriate time (FGAT) appears to be an attractive compromise between accuracy and overall computing time. It is computationally cheaper than four-dimensional (4D)-Var as the increment is not propagated back and forth in time by a model, yet the comparison between the model and the observations is still computed at the right observation time. An interesting feature of the 4D-Var is the iterative process known as the outer loop. This outer-loop approach can also be used in conjunction with 3D-FGAT. But it requires the application of the 3D-FGAT analysis increment at the beginning of the assimilation window. The pros and cons of using this unusual 3D-FGAT variant are illustrated in this paper on two applications focused on the transport, one of the main phenomena governing the atmospheric evolution. The first one is the one-dimensional advection of a passive tracer. By three representative situations, it shows the ben...

Variational Data Assimilation With Telemac. Proof Of Concept For Model State Correction On The Berre Lagoon 3D-Model

TELEMAC is a component of the open-source integrated suite of solvers TELEMAC-MASCARET for use in the field of free-surface flow that solves the Reynolds Averaged Navier-Stokes equations. Generally speaking, uncertainties in the model formulation itself due to simplified physics and also in the input fields to the model such as the boundary conditions, initial conditions and hydraulic parameters translate into errors in the simulated hydraulic variables. In spite of significant advances in numerical schemes, description of geographical data (topography, bathymetry) and environmental conditions (hydrologycal and meteorological fields), the representation of the true state of a system as well as its forecasted state remains imperfect and some of these limits can be overcome combining observations with simulation via data assimilation techniques. This paper presents the implementation of a 3D-Var FGAT variational data assimilation algorithm as a proof of concept for improving TELEMAC simulations and forecast. The demonstration is made on the Berre lagoon application with TELEMAC-3D: the salinity state is sequentially corrected assimilating in situ salinity measurements.

Estimation of Lateral Inflows Using Data Assimilation in the Context of Real-Time Flood Forecasting for the Marne Catchment in France

The present study describes the assimilation of discharge in situ data for operational flood forecasting. The study was carried out on the Marne River (France) catchment where lateral inflows’ uncertainty is important due to karstic areas. This source of error was partly accounted for using an Extended Kalman Filter (EKF) algorithm built on the top of a mono-dimensional hydraulic model. The lateral inflows were sequentially adjusted over a sliding 48 h time window. The correction leads to a significant improvement in the simulated water level and discharge in re-analysis and forecast modes. These results pave the way for the operational use of the data assimilation (DA) procedure for real-time forecasting at the French flood forecasting service.