Search

The prediction of water resource evolution is considered to be a major challenge for the coming century, particularly in the context of climate change and increasing demographic pressure. Water resources are directly linked to the continental water cycle, and the main processes modulating changes can be represented by global hydrological models. However, anthropogenic impacts on water resources, and in particular the effects of dams-reservoirs on river flows, are still poorly known and generally neglected in coupled land surface–river routing models. This paper presents a parameterized reservoir model, DROP (Dam-Reservoir OPeration), based on Hanasaki's scheme to compute monthly releases given inflows, water demands and the management purpose. With its significantly anthropized river basins, Spain has been chosen as a study case for which simulated outflows and water storage variations are evaluated against in situ observations over the period 1979–2014. Using a default configuration of the reservoir model, results reveal its positive contribution in representing the seasonal cycle of discharge and storage variation, specifically for large-storage capacity irrigation reservoirs. Based on a bounded version of the Nash–Sutcliffe efficiency (NSE) index, called C2M, the overall outflow representation is improved by 43 % in the median. For irrigation reservoirs, the improvement rate reaches 80 %. A comprehensive sensitivity analysis of DROP model parameters was conducted based on the performance of C2M on outflows and volumes using the Sobol method. The results show that the most influential parameter is the threshold coefficient describing the demand-controlled release level. The analysis also reveals the parameters that need to be focused on in order to improve river flow or reservoir water storage modeling by highlighting the difference in the individual effects of the parameters and their interactions depending on whether one focuses on outflows or volume mean seasonal patterns. The results of this generic reservoir scheme show promise for modeling present and future reservoir impacts on the continental hydrology within global land surface–river routing models.

Remote sensing products provided by satellite missions, airborne and unmanned aerial vehicle (UAV) campaigns have tremendously developed over the last decade. They undoubtedly offer opportunities to improve our ability to monitor and forecast flooding. The observation of inland waters benefits from several altimetry missions that provide along-track water surface elevation observation from nadir (e.g. TOPEX/Poseidon, Jason, SARAL/AltiKa, Sentinel-6) or large-swath altimeters (e.g. SWOT launched in December 2022), as well as other radar/optical missions (Sentinel-1, Sentinel-2) that provide high-resolution water masks. The limitations of each type of sensor are potentially circumvented when data from different satellite sensors are combined; the fusion of multi-source data has thus become one of the mainstream research topics in the remote-sensing community nowadays. Such fusion can be achieved with data assimilation algorithms applied to hydrodynamics models, namely MASCARET-1D and TELEMAC-2D. The present work focuses on the validation of water surface height (WSH) data from Sentinel-6MF with respect to in-situ gauge data, UAV and 1D/2D-hydrodynamics model outputs as shown in Figure 1. This work participates in a global effort that aims at combining various remote sensing products to represent and forecast flooding. The study is carried out over a dry period in June 2022 and over a flood event that occurred in December 2021-January 2022 over the Garonne catchment near Marmande, in the southwest of France. The WSH of the river is retrieved from Sentinel-6MF high-resolution fully-focused SAR data with an algorithm that relies on the estimation of the river width and the positioning of the river center line. The impact of these a priori data is investigated and the Sentinel-6MF-derived WSH observations are compared to the WSH simulated with TELEMAC-2D. It should be noted that due to the defection of Sentinel-1B (one of the two satellites in the Sentinel-1 constellation) in mid-December 2021, this flood event is only partly observed by Sentinel-1A and that the additional data from Sentinel-6MF with a 10-day revisit period are of great use. This study shows that hydrodynamic simulations and satellite altimetry time-series compare particularly well during the dry period whereas flooding events are often underestimated by satellite altimetry data. We believe that the combined use of satellite altimetry, hydrodynamics model simulations and independent UAV-borne data is key to a better understanding of the processes involved in the Garonne river flow dynamics and flooding events. We also take advantage of this study to improve our Sentinel-6MF data processing techniques. Figure 1- Representation of water level height and flood extent with remote sensing data and hydrodynamic models.

The prediction of water resource evolution is considered to be a major challenge for the coming century, particularly in the context of climate change and increasing demographic pressure. Water resources are directly linked to the continental water cycle and the main processes modulating changes can be represented by global hydrological models. However, anthropogenic impacts on water resources, and in particular the effects of dams-reservoirs on river flows, are still poorly known and generally neglected in coupled land surface – river routing models. This paper presents a parameterized reservoir model, DROP (Dam-Reservoir OPeration model), based on Hanasaki’s scheme to compute monthly releases given inflows, water demands and the management purpose. With its significantly anthropized river basins, Spain has been chosen as a study case for which simulated outflows and water storage variations are evaluated against in situ observations over the period 1979–2014. Using a default configuration of the reservoir model, results reveal its positive contribution in representing the seasonal cycle of discharge and storage variation, specifically for large-storage capacity irrigation reservoirs. Based on a bounded version of the Nash-Sutcliffe Efficiency (NSE) index, called C2M , the overall outflow representation is improved by 43 % in the median. For irrigation reservoirs, the improvement rate reaches 80 %. A comprehensive sensitivity analysis of DROP model parameters was conducted based on the performance of C2M on outflows and volumes using the Sobol method. The results show that the most influential parameter is the threshold coefficient describing the demand-controlled release level. The analysis also reveals the parameters that need to be focused on in order to improve river flow or reservoir water storage modeling by highlighting the difference in the individual effects of the parameters and their interactions depending on whether one focuses on outflows or volume mean seasonal patterns. The results of this generic reservoir scheme show promise for modeling present and future reservoir impacts on the continental hydrology within global land surface – river routing models.

International audience

The upcoming Surface Water and Ocean Topography (SWOT) satellite mission will provide a global and high resolution measurement of river water surface elevation. This product issued from large swath interferometry altimetry will be combined with high fidelity hydrodynamics solvers thanks to data assimilation algorithms to allow for river discharge estimation and prediction. Before launch, the SWOT-HR hydrology simulator is used to produce synthetic SWOT observation at each overpass time, adding the radar measurement error, such as layover and thermal noise, to the water elevation issued from a golden run simulation. The ToolBoxSWOT is a chain of python scripts that formats the time varying hydrodynamic model outputs into a temporal sequence of water elevation raster data files used as inputs for SWOT-HR. A Digital Elevation Model, a geolocalized series of bathymetric profiles and river centerline are used to map the outputs of the hydrodynamic models onto the expected regular and high resolution 2D grid requested by SWOT-HR. The toolbox gathers strategies adapted to various levels of knowledge from well-known to unknown catchments. The ToolBoxSWOT is available on git and is provided with a container featuring the proper environment that embeds Python3, QGIS3, QGDAL and GRASS78. The ToolBoxSWOT was applied over a well-known reach of the Garonne river for which a DEM and gauge data are available as well as on the Brahmaputra river, using SWOT only-derived data. The toolbox allows the generation of various observation data sets available for the SWOT-hydrology community.

Flooding is one of the major natural disasters occurring in the world, with climate change increasing their occurrence and severity. Reliable flood forecasting models are needed to have better insurance in the emergency services’ actions. This work reflects the capacities of Machine Learning models to improve discharge prediction results from empirical lag and route models based on hourly measured water level at gauge stations on the Garonne River. With scarce flooding data (30000 points) over the last 15 years, several learning algorithms have been implemented to predict floods in Toulouse at a 6-hour lead time from upstream stations providing hourly observations. A Linear Regression, a Gradient Boosting Regressor (Machine Learning) and a MultiLayer Perceptron (Neural Network, Deep Learning) are compared, using various strategies for learning, validating and predicting. Preliminary results show that the various strategies score as well as the empirical lag and route model. Further improvements are being investigated regarding the constitution of learning and validation data bases. This paper highlights how AI algorithms allow to improve the reliability of flood forecasts and how the layout of the limited volume of data influences their performance.

As the severity and occurrence of flood events tend to intensify with climate change, the need for flood forecasting capability increases. In this regard, the Flood Detection, Alert and rapid Mapping (FloodDAM) project, funded by Space for Climate Observatory initiatives, was set out to develop pre-operational tools dedicated to enabling quick responses in flood-prone areas, and to improve the reactivity of decision support systems. This work focuses on the assimilation of 2D flood extent data (expressed in terms of wet surface ratios) and in-situ water level data to improve the representation of the flood plain dynamics with a Telemac-2D model and an Ensemble Kalman Filter (EnKF). The EnKF control vector was composed friction coefficients and corrective parameter to the input forcing. It is then augmented with the water level state averaged over several floodplain zones. This work was conducted in the context of Observing System Simulation Experiments (OSSE) based on a real flood event occurred in January-February 2021 on the Garonne Marmandaise catchment. This allows to validate the observation operator associated to the wet surface ratio observations as well as the dual state-parameter sequential correction implemented in this work. The merits of assimilating SAR-derived flood plain data complementary to in-situ water level observations are shown in the control parameter and observation spaces with 1D and 2D assessment metrics. It was also shown that the correction of the hydraulic state significantly improved the flood dynamics, especially during the recession. This proof-of-concept study paves the way towards near-real-time flood forecast, making the most of remote sensing-derived flood observations.

A surrogate model is developed to accurately approximate a two-dimensional hydrodynamics numerical solver in order to conduct a reduced-cost variance-based global sensitivity analysis of the hydraulic state. The impact of uncertainties in river bottom friction and boundary conditions on the simulated water depth is analyzed for quasi-unsteady flows. An autoencoder technique adapted to non-linear variable dimension reduction is used to reduce the multi-dimensional model output so that the formulation of the surrogate remains computationally parsimonious. In addition, following the divide-and-conquer principle, a mixture of local polynomial chaos expansions is proposed to deal with non-linearity in the hydraulic state with respect to uncertain inputs. Machine learning techniques are used to automatically partition the input space into clusters that are not affected by non-linearities and support accurate surrogates. This combined strategy is applied to a reach of the Garonne River where river and floodplains dynamics are simulated by the numerical solver Telemac-2D. The merits of this strategy are highlighted when the flood front reaches regions where the topography features a strong gradient and where, consequently, strong non-linearities occur between the water depth and friction as well as hydrologic input forcing. By applying this strategy, the $$Q_2$$ Q 2 metric improves by 90% compared to a classical polynomial chaos expansion surrogate, resulting in a much more reliable sensitivity analysis. This is particularly important in floodplain areas where human and economic activities are at stake.

Floods are the most common natural disasters all over the world. The space climate observatory (SCO)-FloodDAM project aims at utilizing the capabilities of new observing strategies in order to better alert, detect and map flood events globally. Leveraging from both aerial- and satellite-based platforms (Sentinel, TerraSar-X, SWOT) as well as in-situ based sensors, the main objective of the project is to develop an automatic system to better prevent flood events and assess their consequences. In this paper, we will demonstrate the strategy deployed for the selected test-sites in France.

Smurf is an open source modular system developed in Python for running and cycling data assimilation (DA) systems. It is organised around three super classes for numerical model management, assimilation schemes and observation instruments. Any new item can be easily plugged in by defining a child class that will override as many methods as necessary. Non intrusive, Smurf can be used in any applicative domain for numerical models written in any language.

International audience; This paper investigates the hydrometeorological chain with an ensemble approach. The objective is the generation of Hydrologic Ensemble Forecasts (HEF) on the Odet catchment (France, Brittany), using the Quantile Regression Forest (QRF) method usually applied for the ensemble calibration of meteorological forecats. First, a Global Sensitivity Analysis (GSA) in the distributed MORDOR-TS model is carried out taking into account uncertainty in forecasted rain with AromeEPS-RR1 and in model parameters. GSA highlights the role and importance of the different hydrologic model parameters during rain events and allows to only take into account the most influent parameters for the generation of an Hydrologic Ensemble Forecast (HEF). Three strategies for the generation of HEF are then compared. First (i), a raw ensemble is built with a model-based only approach using the deterministic forecast rainfall Expert-RR3 and perturbed model parameters, without further statistical calibration. Then, the QRF calibration method is used to generate two ensembles of quantiles: (ii) the observation-based approach uses only predictors that are independent from hydrology, whereas (iii) the combined model and observation approach combines these predictors with statistics of the raw hydrologic ensemble (mean, standard deviation). This latter approach was shown to outperfom the previous ones, enhancing the importance of the choice of the predictors in the QRF method. In the prospect of using the hydrologic ensemble as input for hydraulic simulation, the Ensemble Copula Coupling method (ECC) and a trajectory smoothing procedure is then applied on (iii). This step slightly deteriorates the reliability of hourly streamflows, yet Continuous Ranked Probability Score (CRPS) and forecast skills on the cumulated or maximum streamflows are improved.

In the context of the development and implementation of data assimilation techniques in Gironde estuary for flood forecasting, a Telemac 2D model is used to calculate water depths and velocity fields at each node of an unstructured mesh. Upstream, the model boundaries are respectively La Réole and Pessac on the Garonne and Dordogne rivers. The maritime boundary is 32 km off the mouth of Gironde estuary, located in Verdon. This model, which contains 7351 nodes and 12838 finite elements, does not take into account overflows. It was calibrated over 4 non-overflowing events and validated over 6 overflowing events.Uncertainty in hydraulic parameters as well as fluvial and maritime boundary conditions are quantified and reduced in this study. It is assumed that time-varying functional uncertainty in boundary conditions is well approximated by a Gaussian Process characterized by an autocorrelation function and an associated correlation length scale. The coefficients of the truncated Karhunen-Loève decomposition of this process are further considered in the control vector, together with the friction coefficients and wind influence factor, of Global Sensitivity Analysis based on variances decomposition to quantify uncertainty and an Ensemble Kalman Filter to reduce uncertainty. The performance of the data assimilation strategy in terms of control vector composition, length and cycling of the data assimilation window, size of the ensemble and mesh, was assessed on synthetical and real experiments.It was shown that uncertainty in water level predominantly stems from uncertainty in the maritime boundary condition and the friction coefficient in the mouth and in the central part of the estuary. Synthetical experiments showed that data assimilation succeeds in identifying time varying friction following tidal signal, as well as reconstructing the time-dependent maritime forcing even though the KL coefficients identification suffers equifinality. A resampling method based on the persistence of the initial background covariance matrix is used to avoid well-known ensemble collapse in the Ensemble Kalman Filter. Difficulties in estimating the friction parameter of the confluence zone, where the flows are the result of non-linear physical processes, were highlighted. Also, the equifinality problem for identification of the KL coefficients in the boundary conditions was shown to be enhanced, nevertheless, leading to the proper reconstruction of the maritime forcing and consequently to the expected water level in the estuary. In the real experiment, it was shown that water levels are significantly improved with error smaller than 10cm, along the estuary, except in the upstream sections of the Garonne and Dordogne rivers where model refinement should be improved.KEY WORDS2D hydrodynamic simulations, TELEMAC, Gironde Estuary, data assimilation, Ensemble Kalman filter, Karhunen-Loève decomposition, time-dependent forcings

In the context of hydrodynamic modeling, the use of 2D models is adapted in areas where the flow is not mono-dimensional (confluence zones, flood plains). Nonetheless the lack of field data and the computational cost constraints limit the extensive use of 2D models for operational flood forecasting. Multi-dimensional coupling offers a solution with 1D models where the flow is mono-dimensional and with local 2D models where needed. This solution allows for the representation of complex processes in 2D models, while the simulated hydraulic state is significantly better than that of the full 1D model. In this study, coupling is implemented between three 1D sub-models and a local 2D model for a confluence on the Adour river (France). A Schwarz algorithm is implemented to guarantee the continuity of the variables at the 1D/2D interfaces while in situ observations are assimilated in the 1D sub-models to improve results and forecasts in operational mode as carried out by the French flood forecasting services. An implementation of the coupling and data assimilation (DA) solution with domain decomposition and task/data parallelism is proposed so that it is compatible with operational constraints. The coupling with the 2D model improves the simulated hydraulic state compared to a global 1D model, and DA improves results in 1D and 2D areas.

The continental part of the water cycle is commonly represented with hydrological models. Yet, there are limits in their capacity to accurately estimate water storage and dynamics because of their coarse spatial resolution, simplified physics, and an incomplete knowledge of atmospheric forcing and input parameters. These errors can be diminished using data assimilation techniques. The model’s most sensitive parameters should be identified beforehand. The objective of the present study is to highlight key parameters impacting the river-routing scheme Total Runoff Integrating Pathways (TRIP) while simulating river water height and discharge as a function of time focusing on the annual cycle. Thus, a sensitivity analysis based on the decomposition of model output variance (using a method called ANOVA) is utilized and applied over the Amazon basin. Tested parameters are perturbed with correcting factors. First, parameter-correcting coefficients are considered uniform over the entire basin. The results are specific to the TRIP model and show that geomorphological parameters explain around 95% of the water height variance with purely additive contributions, all year long, with a dominating impact of the river Manning coefficient (40%), the riverbed slope (35%), and the river width (20%). The results also show that discharge is essentially sensitive to the groundwater time constant that makes up more than 90% of the variance. To a lesser extent, in rising/falling flow period, the discharge is also sensitive to geomorphological parameters. Next, the Amazon basin is divided into nine subregions and the sensitivity analysis is carried out for regionalized parameter-correcting coefficients. The results show that local-region parameters impact water height, while upstream-region parameters affect discharge.

River hydraulic models are used to assess the environmental risk associated to flooding and consequently inform decision support systems for civil security needs. These numerical models are generally based on a deterministic approach based on resolving the partial differential equations. However, these models are subject to various types of uncertainties in their input. Knowledge of the type and magnitude of these uncertainties is crucial for a meaningful interpretation of the model results. Uncertainty quantification (UQ) framework aims to probabilize the uncertainties in the input, propagate them through the numerical model and quantify their impact on the simulated quantity of interest, here, water level field discretized over an unstructured finite element mesh over the Garonne River (South-West France) between Tonneins and La R´eole simulated with a numerical solver, TELEMAC-2D. The computational cost of the sensitivity analysis with the classical Monte Carlo approach is reduced using a surrogate model instead of the numerical solver. The present study investigates one of the machine learning algorithms: A surrogate model based on Gaussian process. This latter was used to represent the spatially distributed water level with respect to uncertain stationary flow to the model and friction coefficients. The quality of the surrogate was assessed on a validation set, with small root mean square error and a predictive coefficient equal to 1. Sobol’ sensitivity indices are computed and enhance the high impact of the input discharge on the water level variation.

The SWOT (Surface Water and Ocean Topography) mission, to be launched in 2021, will provide water surface elevations, slopes, and river width measurements for rivers wider than 100 m. In this study, synthetic SWOT data are assimilated in a regional hydrometeorological model in order to improve the dynamics of continental waters over the Garonne catchment, one of the major French catchments. The aim of this paper is to demonstrate that the sequential assimilation of SWOT-like river depths allows the correction of river bed roughness coefficients and thus simulated river depths. An extended Kalman filter is implemented and the data assimilation strategy was applied to four experiments of gradually increasing complexity regarding observation and model error over the 1995&ndash, 2000 period. With respect to a &ldquo, true&rdquo, river state, assimilating river depths allows the proper retrieval of constant and spatially distributed roughness coefficients with a root mean square error of 1 m1/3 s&minus, 1, and the estimation of associated river depths. It was also shown that river depth differences can be assimilated, resulting in a higher root mean square error for roughness coefficients with respect to the true river state. Finally, the last experiment shows how one can take into account more realistic sources of SWOT error measurements, in particular the importance of the estimation of the tropospheric water content in the process.

The use of ensemble streamflow forecasts is developing in the international flood forecasting services. Ensemble streamflow forecast systems can provide more accurate forecasts and useful information about the uncertainty of the forecasts, thus improving the assessment of risks. Nevertheless, these systems, like all hydrological forecasts, suffer from errors on initialization or on meteorological data, which lead to hydrological prediction errors. This article, which is the second part of a 2-part article, concerns the impacts of initial states, improved by a streamflow assimilation system, on an ensemble streamflow prediction system over France. An assimilation system was implemented to improve the streamflow analysis of the SAFRAN-ISBA-MODCOU (SIM) hydro-meteorological suite, which initializes the ensemble streamflow forecasts at Météo-France. This assimilation system, using the Best Linear Unbiased Estimator (BLUE) and modifying the initial soil moisture states, showed an improvement of the streamflow analysis with low soil moisture increments. The final states of this suite were used to initialize the ensemble streamflow forecasts of Météo-France, which are based on the SIM model and use the European Centre for Medium-range Weather Forecasts (ECMWF) 10-day Ensemble Prediction System (EPS). Two different configurations of the assimilation system were used in this study: the first with the classical SIM model and the second using improved soil physics in ISBA. The effects of the assimilation system on the ensemble streamflow forecasts were assessed for these two configurations, and a comparison was made with the original (i.e. without data assimilation and without the improved physics) ensemble streamflow forecasts. It is shown that the assimilation system improved most of the statistical scores usually computed for the validation of ensemble predictions (RMSE, Brier Skill Score and its decomposition, Ranked Probability Skill Score, False Alarm Rate, etc.), especially for the first few days of the time range. The assimilation was slightly more efficient for small basins than for large ones., JRC.H.7-Climate Risk Management

This paper presents a global sensitivity analysis study applied to a TELEMAC2D nu- merical flood forecast model of Gironde estuary which aims at identifying which input variables should be better described for water levels to be better simulated and forecasted. A variance sensitivity study (ANOVA) was carried out, by calculating Sobol’ indices for all numerical parameters (wind influence coefficient, Strickler friction coefficients for 4 zones) and time-dependent forcings of the model (rivers discharges and maritime boundary con- ditions). It led to the identification of parameters and forcings to which the model is most sensitive for each area of the estuary. Sobol’ indices for 2003 event show a predominance of the influence of the maritime boundary conditions and of Strickler coefficients all along the estuary. A mesh convergence study shows that the results don’t depend on the mesh. Moreover, a special focus on the eigenvalues of the tide signal correlation error function shows no predominance of one mode on the other. These first results are currently used to implement a sequential ensemble Kalman filter improving both the state of the system and the maritime boundary condition and optimizing the friction coefficients over the Gironde estuary.

Assessing epistemic uncertainties is considered as a milestone for improving numerical predictions of a dynamical system. In hydrodynamics, uncertainties in input parameters translate into uncertainties in simulated water levels through the shallow water equations. We investigate the ability of generalized polynomial chaos (gPC) surrogate to evaluate the probabilistic features of water level simulated by a 1-D hydraulic model (MASCARET) with the same accuracy as a classical Monte Carlo method but at a reduced computational cost. This study highlights that the water level probability density function and covariance matrix are better estimated with the polynomial surrogate model than with a Monte Carlo approach on the forward model given a limited budget of MASCARET evaluations. The gPC-surrogate performance is first assessed on an idealized channel with uniform geometry and then applied on the more realistic case of the Garonne River (France) for which a global sensitivity analysis using sparse least-angle regression was performed to reduce the size of the stochastic problem. For both cases, Galerkin projection approximation coupled to Gaussian quadrature that involves a limited number of forward model evaluations is compared with least-square regression for computing the coefficients when the surrogate is parameterized with respect to the local friction coefficient and the upstream discharge. The results showed that a gPC-surrogate with total polynomial degree equal to 6 requiring 49 forward model evaluations is sufficient to represent the water level distribution (in the sense of the $\ell _2$ norm), the probability density function and the water level covariance matrix for further use in the framework of data assimilation. In locations where the flow dynamics is more complex due to bathymetry, a higher polynomial degree is needed to retrieve the water level distribution. The use of a surrogate is thus a promising strategy for uncertainty quantification studies in open-channel flows and should be extended to unsteady flows. It also paves the way toward cost-effective ensemble-based data assimilation for flood forecasting and water resource management.

In a context of development and implementation of data assimilation techniques in Gironde estuary for flood forecasting, a TELEMAC2D model is used to compute water depth and velocity fields at each node of an unstructured mesh. Upstream, the model boundaries are, respectively, La Reole and Pessac on the Garonne and Dordogne rivers. The maritime boundary is 32 km off the mouth of Gironde estuary, located in Le Verdon. This model, which contains 7351 nodes and 12,838 finite elements, does not take into account overflows. This paper presents a global SA study in the context of flood forecasting in the Gironde estuary. It aims at identifying which input variables should be better described for water levels to be better simulated and forecasted in the estuary. On the one hand, a propagation and quantification of uncertainties by a unidirectional analysis method was carried out. On the other hand, a variance sensitivity study (ANOVA) was carried out, by calculating the total and partial sensitivity Sobol’ indices for all numerical parameters (wind influence coefficient, Strickler friction coefficients for four zones) and time-dependent forcings of the model (rivers discharges and maritime boundary conditions). It led to the identification of parameters and forcings to which the model is most sensitive for each area of the estuary, as well as the identification of very low interdependencies, in order to choose the variables to assimilate later. The standardized variation coefficients for 1981 event as well as Sobol’ indices for 2003 event show a predominance of the influence of the maritime boundary conditions all along the estuary and of the Strickler coefficient corresponding to the zone considered for the estuarine part and the confluence, to which must be added the Garonne discharge as a predominant parameter for the latter. Unsurprisingly, the upstream part of river zones is influenced primarily by the friction coefficient and the respective river flows of Garonne and Dordogne rivers.

This paper is the second part in a series of two articles, which aims at presenting a data-driven modeling strategy for forecasting wildfire spread scenarios based on the assimilation of the observed fire front location and on the sequential correction of model parameters or model state. This model relies on an estimation of the local rate of fire spread (ROS) as a function of environmental conditions based on Rothermel's semi-empirical formulation, in order to propagate the fire front with an Eulerian front-tracking simulator. In Part I, a data assimilation (DA) system based on an ensemble Kalman filter (EnKF) was implemented to provide a spatially uniform correction of biomass fuel and wind parameters and thereby, produce an improved forecast of the wildfire behavior (addressing uncertainties in the input parameters of the ROS model only). In Part II, the objective of the EnKF algorithm is to sequentially update the two-dimensional coordinates of the markers along the discretized fire front, in order to provide a spatially distributed correction of the fire front location and thereby, a more reliable initial condition for further model time-integration (addressing all sources of uncertainties in the ROS model). The resulting prototype data-driven wildfire spread simulator is first evaluated in a series of verification tests using synthetically generated observations; tests include representative cases with spatially varying biomass properties and temporally varying wind conditions. In order to properly account for uncertainties during the EnKF update step and to accurately represent error correlations along the fireline, it is shown that members of the EnKF ensemble must be generated through variations in estimates of the fire's initial location as well as through variations in the parameters of the ROS model. The performance of the prototype simulator based on state estimation (SE) or parameter estimation (PE) is then evaluated by comparison with data taken from a reduced-scale controlled grassland fire experiment. Results indicate that data-driven simulations are capable of correcting inaccurate predictions of the fire front location and of subsequently providing an optimized forecast of the wildfire behavior at future lead times. The complementary benefits of both PE and SE approaches, in terms of analysis and forecast performance, are also emphasized. In particular, it is found that the size of the assimilation window must be specified adequately with the persistence of the model initial condition and/or with the temporal and spatial variability of the environmental conditions in order to track sudden changes in wildfire behavior. The present prototype data-driven forecast system is still at an early stage of development. In this regard, this preliminary investigation provides valuable information on how to combine observations with a fire spread model in an efficient way, as well as guidelines to design the future system evolution in order to meet the operational requirements of wildfire spread monitoring.

International audience; Sensitivity analysis techniques have been widely used in multitude of applications to quantify the impact of inputs variables imprecision on the accuracy of the model output variables. Depending on the problem at hand, an appropriate method of sensitivity analysis should be selected. Direct and adjoint sensitivity analysis are two complementary approaches known to be efficient. While the direct approach provides an assessment of the propagation of the error of a given input parameter in the studied system, the adjoint approach enables to identify the source of the uncertainty of a given output variable with respect to several input parameters. Direct methods have been extensively investigated in different geophysical applications, particularly in the context of the hydraulic modeling. In this work, several methods will be described and applied to the same benchmark during over-flooding events. The effect of uncertainties in the boundary conditions, the spatially distributed functions (bed level, river width, friction, etc.) and the numerical parameters on the model state variables (discharge, water surface elevation, etc.) is examined. This study has been carried out on the Garonne River test case, along a 50 km downstream reach, using 1D full Saint-Venant hydraulic models SIC2 (Irstea) or Mascaret (EDF), and 2D Telemac model (EDF). Results illustrate the influence of individual and combined contributions of input variables uncertainties.

This study presents the implementation and the merits of an Ensemble Kalman Filter (EnKF) algorithm with an inflation procedure on the 1D shallow water model MASCARET in the framework of operational flood forecasting on the “Adour Maritime” river (South West France). In situ water level observations are sequentially assimilated to correct both water level and discharge. The stochastic estimation of the background error statistics is achieved over an ensemble of MASCARET integrations with perturbed hydrological boundary conditions. It is shown that the geometric characteristics of the network as well as the hydrological forcings and their temporal variability have a significant impact on the shape of the univariate (water level) and multivariate (water level and discharge) background error covariance functions and thus on the EnKF analysis. The performance of the EnKF algorithm is examined for observing system simulation experiments as well as for a set of eight real flood events (2009–2014). The quality of the ensemble is deemed satisfactory as long as the forecast lead time remains under the transfer time of the network, when perfect hydrological forcings are considered. Results demonstrate that the simulated hydraulic state variables can be improved over the entire network, even where no data are available, with a limited ensemble size and thus a computational cost compatible with operational constraints. The improvement in the water level Root-Mean-Square Error obtained with the EnKF reaches up to 88% at the analysis time and 40% at a 4-h forecast lead time compared to the standalone model.

Data assimilation is widely used to improve flood forecasting capability, especially through parameter inference requiring statistical information on the uncertain input parameters (upstream discharge, friction coefficient) as well as on the variability of the water level and its sensitivity with respect to the inputs. For particle filter or ensemble Kalman filter, stochastically estimating probability density function and covariance matrices from a Monte Carlo random sampling requires a large ensemble of model evaluations, limiting their use in real-time application. To tackle this issue, fast surrogate models based on Polynomial Chaos and Gaussian Process can be used to represent the spatially distributed water level in place of solving the shallow water equations. This study investigates the use of these surrogates to estimate probability density functions and covariance matrices at a reduced computational cost and without the loss of accuracy, in the perspective of ensemble-based data assimilation. This study focuses on 1-D steady state flow simulated with MASCARET over the Garonne River (South-West France). Results show that both surrogates feature similar performance to the Monte-Carlo random sampling, but for a much smaller computational budget; a few MASCARET simulations (on the order of 10-100) are sufficient to accurately retrieve covariance matrices and probability density functions all along the river, even where the flow dynamic is more complex due to heterogeneous bathymetry. This paves the way for the design of surrogate strategies suitable for representing unsteady open-channel flows in data assimilation.

Scheduled for launch in 2021, the Surface Water and Ocean Topography (SWOT) mission will be a truly unique mission that will provide high-temporal-frequency maps of surface water extents and elevation variations of global water bodies (lakes/reservoirs, rivers, estuaries, oceans, and sea ice) at higher spatial resolution than is available with current technologies (Biancamaria et al. 2016; Alsdorf et al. 2007). The primary instrument on SWOT is based on a Ka-band radar interferometer (KaRIN), which uses radar interferometery technology. The satellite will fly two radar antennas at either end of a 10-m (33 ft) mast, allowing it to measure the elevation of the surface along a 120-km (75 mi)-wide swath below. The availability of high-frequency and high-resolution maps of elevations and extents for surface water bodies and oceans will present unique opportunities to address numerous societally relevant challenges around the globe (Srinivasan et al. 2015). These opportunities may include such diverse and far-ranging applications as fisheries management, flood inundation mapping/risk mitigation/forecasting, wildlife conservation, global data assimilation for improving forecast of ocean tides and weather, reservoir management, climate change impacts and adaptation, and river discharge estimation, among others.

Data assimilation is a sophisticated technique, yet not available in combustion, that combines measurements to model simulation and account for uncertainties in order to improve the numerical prediction of a system. In the context of gas turbines, data assimilation may be used for example to improve the prediction of flame ignition and propagation by a smart analysis of images and measurements. A first illustration of data assimilation is given in a simple case, where synthetic time-evolving positions of the flame front are assimilated to calibrate parameters of a premixed flame model. Its successful application in the context of natural fire propagation assesses the predictive capacity of the technique and the resulting higher fidelity in the data-driven simulations.

This study presents an analysis of the potential benefits of data assimilation techniques for combustion applications through the example of wildland fire propagation. Data assimilation is an efficient strategy, initially developed for meteorological applications, that integrates sensor observations with computational models, accounts for observation and modeling errors, and thereby provides optimal estimates of poorly known variables and improved predictions of complex systems dynamics. In the present study, the effective wildfire rate of spread is calibrated using: measurements of the time-evolving fire front position; a level-set-based fire propagation solver with a description of the local rate of spread based on Rothermel’s model; and a simplified Kalman Filter algorithm. The prototype data-driven wildfire simulation capability is first evaluated in academic tests using synthetically-generated observations; the prototype capability is then evaluated in a more realistic validation test corresponding to a controlled grassland fire experiment. The results indicate that data assimilation has the potential of dramatically increasing fire simulation accuracy.

International audience; The present study explores the application of a data assimilation (DA) procedure to correct the radar rain- fall inputs of an event-based, distributed, parsimonious hy- drological model. An extended Kalman filter algorithm was built on top of a rainfall-runoff model in order to assimilate discharge observations at the catchment outlet. This work fo- cuses primarily on the uncertainty in the rainfall data and considers this as the principal source of error in the sim- ulated discharges, neglecting simplifications in the hydro- logical model structure and poor knowledge of catchment physics. The study site is the 114 km2 Lez catchment near Montpellier, France. This catchment is subject to heavy oro- graphic rainfall and characterised by a karstic geology, lead- ing to flash flooding events. The hydrological model uses a derived version of the SCS method, combined with a Lag and Route transfer function. Because the radar rainfall in- put to the model depends on geographical features and cloud structures, it is particularly uncertain and results in signifi- cant errors in the simulated discharges. This study seeks to demonstrate that a simple DA algorithm is capable of ren- dering radar rainfall suitable for hydrological forecasting. To test this hypothesis, the DA analysis was applied to estimate a constant hyetograph correction to each of 19 flood events. The analysis was carried in two different modes: by assimi- lating observations at all available time steps, referred to here as reanalysis mode, and by using only observations up to 3 h before the flood peak to mimic an operational environment, referred to as pseudo-forecast mode. In reanalysis mode, the resulting correction of the radar rainfall data was then com- pared to the mean field bias (MFB), a corrective coefficient determined using rain gauge measurements. It was shown that the radar rainfall corrected using DA leads to improved discharge simulations and Nash-Sutcliffe efficiency criteria compared to the MFB correction. In pseudo-forecast mode, the reduction of the uncertainty in the rainfall data leads to a reduction of the error in the simulated discharge, but un- certainty from the model parameterisation diminishes data assimilation efficiency. While the DA algorithm used is this study is effective in correcting uncertain radar rainfall, model uncertainty remains an important challenge for flood fore- casting within the Lez catchment.

Over the last few years, a collaborative work between CERFACS, LNHE (EDF R&D), SCHAPI and CE-REMA resulted in the implementation of a Data Assimilation (DA) method on top of MASCARET in the framework of real-time forecasting. This prototype was based on a simplified Kalman filter where the description of the background error covariances is prescribed based on off-line climatology constant over time. This approach showed promising results on the Adour and Marne catchments as it improves the forecast skills of the hydraulic model using water level and discharge in-situ observations. An ensemble-based DA algorithm has recently been implemented to improve the modelling of the background error covariance matrix used to distribute the correction to the water level and discharge states when observations are assimilated from observation points to the entire state. It was demonstrated that the flow dependent description of the background error covariances with the EnKF algorithm leads to a more realistic correction of the hydraulic state with significant impact of the hydraulic network characteristics

International audience; The Surface Water and Ocean Topography (SWOT) satellite mission planned for launch in 2020 will map river elevations and inundated area globally for rivers >100 m wide. In advance of this launch, we here evaluated the possibility of estimating discharge in ungauged rivers using synthetic, daily ‘‘remote sensing’’ measurements derived from hydraulic models corrupted with minimal observational errors. Five discharge algorithms were evaluated, as well as the median of the five, for 19 rivers spanning a range of hydraulic and geomorphic conditions. Reliance upon a priori information, and thus applicability to truly ungauged reaches, varied among algorithms: one algorithm employed only global limits on velocity and depth, while the other algorithms relied on globally available prior estimates of discharge. We found at least one algorithm able to estimate instantaneous discharge to within 35% relative root-mean-squared error (RRMSE) on 14/16 nonbraided rivers despite out-of-bank flows, multichannel planforms, and backwater effects. Moreover, we found RRMSE was often dominated by bias; the median standard deviation of relativeresiduals across the 16 nonbraided rivers was only 12.5%. SWOT discharge algorithm progress is therefore encouraging, yet future efforts should consider incorporating ancillary data or multialgorithm synergy to improve results.

A computational simplification of the Kalman filter (KF) is introduced – the parametric Kalman filter (PKF). The full covariance matrix dynamics of the KF, which describes the evolution along the analysis and forecast cycle, is replaced by the dynamics of the error variance and the diffusion tensor, which is related to the correlation length-scales. The PKF developed here has been applied to the simplified framework of advection–diffusion of a passive tracer, for its use in chemical transport model assimilation. The PKF is easy to compute and computationally cost-effective than an ensemble Kalman filter (EnKF) in this context. The validation of the method is presented for a simplified 1-D advection–diffusion dynamics. Keywords: data assimilation, Kalman filter, covariance dynamics, parameterisation of analysis (Published: 30 November 2016) Citation: Tellus A 2016, 68, 31547, http://dx.doi.org/10.3402/tellusa.v68.31547

Two Ensemble Streamflow Prediction Systems (ESPSs) have been set up at Météo-France. They are based on the French SIM distributed hydrometeorological model. A deterministic analysis run of SIM is used to initialize the two ESPSs. In order to obtain a better initial state, a past discharges assimilation system has been implemented into this analysis SIM run, using the Best Linear Unbiased Estimator (BLUE). Its role is to improve the model soil moisture by using streamflow observations in order to better simulate streamflow. The skills of the assimilation system were assessed for a 569-day period on six different configurations, including two different physics schemes of the model (the use of an exponential profile of hydraulic conductivity or not) and, for each one, three different ways of considering the model soil moisture in the BLUE state variables. Respect of the linearity hypothesis of the BLUE was verified by assessing of the impact of iterations of the BLUE. The configuration including the use of the exponential profile of hydraulic conductivity and the combination of the moisture of the two soil layers in the state variable showed a significant improvement of streamflow simulations. It led to a significantly better simulation than the reference one, and the lowest soil moisture corrections. These results were confirmed by the study of the impacts of the past discharge assimilation system on a set of 49 independent stations., JRC.H.7-Climate Risk Management

The Surface Water and Ocean Topography (SWOT) mission will provide free water surface elevations, slopes, and river widths for rivers wider than 50 m. Models must be prepared to use this new finescale information by explicitly simulating the link between runoff and the river channel hydraulics. This study assesses one regional hydrometeorological model’s ability to simulate river depths. The Garonne catchment in southwestern France (56 000 km2) has been chosen for the availability of operational gauges in the river network and finescale hydraulic models over two reaches of the river. Several routing schemes, ranging from the simple Muskingum method to time-variable parameter kinematic and diffusive waves schemes, are tested. The results show that the variable flow velocity schemes are advantageous for discharge computations when compared to the original Muskingum routing method. Additionally, comparisons between river depth computations and in situ observations in the downstream Garonne River led to root-mean-square errors of 50–60 cm in the improved Muskingum method and 40–50 cm in the kinematic–diffusive wave method. The results also highlight SWOT’s potential to improve the characterization of hydrological processes for subbasins larger than 10 000 km2, the importance of an accurate digital elevation model, and the need for spatially varying hydraulic parameters.

Several studies have illustrated how the univariate assimilation of temperature data can have a detrimental effect on the ocean-state variables (salinity, currents, etc.) not directly constrained by the data. In this paper, the authors describe how the salinity adjustment method proposed by Troccoli and Haines can be included as a multivariate temperature–salinity (T–S) constraint within a background-error covariance model for variational data assimilation. The method is applied to a three-dimensional variational assimilation (3DVAR) system for a tropical Pacific version of the Océan Parallélisé (OPA) ocean general circulation model. An identical twin experiment is presented first to illustrate how the method is effective in reconstructing a density profile using only temperature observations from that profile. The 3DVAR system is then cycled over the period 1993–98 using in situ temperature data from the Global Temperature and Salinity Pilot Programme. Relative to a univariate (T) 3DVAR, the multivariate (T, S) 3DVAR significantly improves the salinity mean state. A comparison with salinity data that are not assimilated is also presented. The fit to these observations is improved when the T–S constraint is applied. The salinity correction leads to a better preservation of the salinity structure and avoids the development of spurious geostrophic currents that were evident in the univariate analysis. The currents at the surface and below the core of the undercurrent are also improved. Examination of the heat budget highlights how the temperature increment must compensate for a perpetual degradation of the temperature field by abnormally strong advection in the univariate experiment. When the T–S constraint is applied, this spurious advection is reduced and the mean temperature increment is decreased. Examination of the salt budget shows that spurious advection is also the main cause of the upper-ocean freshening. When the T–S constraint is applied, the salinity structure is improved allowing for a better representation of the advection term and better preservation of the salt content in the upper ocean. The T–S constraint does not correct for all problems linked to data assimilation: vertical mixing is still too strong, and the surface salinity state and currents still have substantial errors. Improvements can be expected by including additional constraints in the background error covariances and by assimilating salinity data.