As follow up of its predecessor FloodDAM, FloodDAM-DT aims to provide an automated service to reliably detect, monitor, assess and predictfloods on a global scale within a Franco-American collaboration. This Digital Twin development is an international effort to devise an Earth System digital twin based on the water cycle and focused on floods.



The SCO FloodDAM Digital Twin (DT) project aims to devise an Earth System Digital Twin (ESDT) for alert systems and flood risk maps on local and global scales using space technologies. This is a joint effort that associates the SCO FloodDAM-DT project led by CNES to the IDEAS (Integrated Digital Earth Analysis System) project led by NASA within the Advanced Information Systems Technology (AIST) program on water cycle applications, and more specifically on floods. This international collaboration will enable researchers and decision-makers to visualize, analyze and evaluate different scenarios for environmental issues where risk assessment is at stake.

SChema FLoodDAM DT

▲ Figure 1. High-level diagram of the IDEAS and FloodDAM-DT digital twin for the water cycle and flooding© NASA JPL

Objectives and methodology

The DT architecture for hydrology aims to dynamically connect continental water data (in situ, airborne, remote sensing (RS) data sources, e.g. radar and optical) with data processing algorithms and hydrological/hydrodynamic models from both agencies (NASA and CNES) to produce large scale flood predictions at global and local scales.

This work entails a multi-agency effort to provide a scalable open-source platform focused on the distribution of hydrological products such as water resources and flood risks analysis. Two case studies will serve as demonstrators.

The FloodDAM-DT demonstrator is based on an ever-growing volume of heterogeneous Earth observation data, multi-physics and multi-scale fluid dynamics modelling tools enhanced by data assimilation algorithms, Artificial Intelligence (AI) algorithms and advanced analyses for data processing. The project benefits from the standardized and normalized interfaces of the French and American space agencies to facilitate the use of data, as well as the considerable computing resources available to the consortium.

The project uses open-source software and data. The main processing blocks are:

  1. flood detection and alert,
  2. rapid flood extent mapping, and monitoring of on-going flood events ,
  3. short-term forecasting of flood surfaces and  free-surface elevation maps using high-fidelity hydrodynamic models over local areas, and
  4. real-time and post-event estimation of the financial risk associated with flooding.

Pipeline FloodDAM

▲ Figure 2. FloodDAM Digital Twin project pipeline © CNES

The FloodDAM-DT processing chain will be integrated into the national hydrological platform, which is open to all types of users.

Innovative contributions and logical links

Compared to its precursor, the FloodDAM-DT project has the following advantages:

  • Automation and demonstration of the complete chain (flood detection and warning, rapid flood mapping, hydrodynamic modelling, mapping and estimation of socio-economic impacts) on selected catchment areas (Garonne in France, Ohio in the United States) with the provision of an associated flood report for decision-making purposes.

  • Integration of the entire chain into the French platform (HYSOPE-II project) in order to centralize all open data related with hydrology and integrate  products requested by users.

  • Interoperability with other services and systems via the Platform and APIs using standardized interfaces and data homogenization formats.

  • Multi-scale chaining of NASA/JPL RAPID large-scale hydrology models and Telemac-2D river hydrodynamics with data assimilation (CERFACS).

  • Use of new satellite data for rapid flood extent mapping, short-term predictions and free surface elevation maps via data assimilation in hydrodynamic models.

  • A high-resolution, hydro-compatible digital elevation model (DEM) based on stereo acquisitions of Pleiades images.

  • Assessment of the socio-economic impact of flooding.

  • Towards an Earth System Digital Twin focused on hydrology by creating a digital replica of the state of catchment areas to understand hydrology and flooding at multiple spatiotemporal scales where risk assessment, security and financial issues are at stake.

Agenda :

  • 21 June 2022: 1st Copil (Steering Committee)
  • 15 November 2023: 2nd Copil
  • December 2023: pre-demonstration of IDEAS and FloodDAM-DT services
  • June 2024: end-to-end demonstration on scenarios in France and the United States

Application site(s)

  • Garonne basin, Tonneins-La Réole section (France)

Carte Garonne Marmandaise

▲ Figure 3. Garonne - Marmandaise catchment area (France) delimited by the red zone with an illustration of the flood extent maps from Sentinel-1 before and after the 2021 flood event. © CERFACS

  • Ohio River watershed (United States), Cannelton-Newburgh section 

Carte Ohio Cannelton-Newburgh

▲ Figure 4. Ohio River catchment (USA) delineated by the red zone with an illustration of flood extent maps from Sentinel-1 before and after the flood event in February 2018. © CERFACS



  • High-resolution (HR) data with global coverage for all test sites:
    • Optical (Sentinel-2, Landsat-8/9)
    • Radar (Sentinel-1)
    • Altimtry (Sentinel-6, Sentinel-3, SWOT)
  • Very high resolution (VHR) data on demand:

    • Radar (TerraSarX supplied by Airbus DS, Capella (X-band), Umbra (X-band), Saocom (L-Bande), NovaSAR-1 (S-Band), future NiSAR (L-Band)
    • Optical (Peiades/ Pleiades-Neo supplied by Airbus DS)
  • Derivative products

    • Spatial data portals : PEPS , IR DataTerra  (THEIA, AERIS, ODATIS, FORM@TER), etc.
    • Météo France data and models on the AERIS portal
    • Copernicus Services: C3S (Climate Change), Emergency Management Service (EMS) avec les service Rapid Mapping Service (RMS), Risk and Recovery Mapping (RRM) et European Flood Awareness System (EFAS), Atmosphere Monitoring Service (CAMS), Land Monitoring Service (CLMS)
    • DTM/DNR data: Copernicus DTM 30m, MERIT DTM 90m, IGN LIDAR (RGE Alti)


  • Data from in-situ stations in France by the VigiCrues network and in the United States by the USGS network
  • Data from in situ and drone stations supplied by
  • Numerous territorial data sets: land use maps (IOTA2 Land Cover, Corine Land Cover, ESA-WorldCover), population maps and densities, graphical parcel register (RPG), etc.

Données FloodDAM

▲ Figure 5. Overview of data used in the SCO FloodDAM-DT project.

Results – Final products

This 18-month project will result in a demonstration of the complete processing chain, showing its multi-scale aspect on at least 2 sites: the Garonne Marmandaise catchment in France and the Ohio catchment in the United States. This demonstration will provide access to in-situ and satellite/drone data, mapping of the extent of flooding resulting from AI data processing, CFD modelling output and estimates of the socio-economic risks associated with flooding.

Alert service

Floods are monitored by national in-situ networks such as VigiCrues in France and USGS in the United States. The FloodDAM-DT project completes these measurements with new in-situ stations in order to improve spatial resolution (particularly in flood plains and tributaries), diversify sensors types and consequently improve the quality of flood map predictions from hydrodynamic modelling with data assimilation.

In addition to these measures, a drone flight is planned during a flood event in the winter of 2024. The micro-stations and the drone carry a LIDAR instrument that provides water height with centimeter accuracy, a GNSS receiver and an 8MP camera that provides information on flow speeds and ortho-photos.



VorteX-io station

▲ VorteX-io micro-station © vorteX-io

PF Maelstrom

▲ Figure 6. The Maelstrom platform developed by provides a flood detection and warning service. The platform also provides real-time measurements of water height, surface speed and images from in-situ stations, as well as precise GNSS positioning. In this example, the in-situ station in Couthures-sur-Garonne is being monitored.  ©

Mapping the extent of flooding

The FloodDAM project produced impressive results on a global scale, covering an area of 10,000 km² in less than 5 minutes using the CNES open-source Flood-ML algorithm based on random forest change detection trained on past events. Following on from this work, FloodDAM-DT aims to produce a near-real time (NRT) flood monitoring product by generating flood extent maps after detecting water level anomalies in the time series of in-situ stations.

Flood extent maps can be generated on a global scale, at the times of satellite passes, as shown in Figures 7 and 8. More accurate flood maps will also be available thanks to hydrodynamic modelling with the assimilation of in-situ and remote sensing data, over the areas for which models have been developed.

▼ FloodML automatic report

FloodML Garonne

▲ Figure 7. Rapid flood mapping production using FloodML for the flood event on 2021-02-03 on the Garonne Marmandaise using Sentinel-1 radar images. © CNES-CLS

FloodML Ohio

▲ Figure 8. Rapid flood mapping production using FloodML for a flood event on 2021-11-16 over Seattle (US) and Vancouver (CA) using Sentinel-1 radar images. © CNES-CLS

Re-analysis and short-term forecasting of floods

Modelling and data assimilation tools can also be used to provide maps of the extent of flooding between satellite passes, as well as forecast beyond the present time. It should be noted that the maximum forecast timeframe is highly dependent on the size and dynamics of the basin. The data assimilation algorithm makes it possible to combine in-situ water level data, Sentinel-1&2 flood extent maps, Sentinel-6 and SWOT altimetry data, as well as UAV acquisition data and in-situ surface velocity measurements. Finally, on the demonstration catchments, hydrodynamic models with data assimilation provide maps of free surface elevation over the entire simulated domain, at very fine time steps, in the past, for re-analysis, as well as for forecasting beyond the present.

It should be noted that the Telemac2D hydrodynamic model used for the Garonne marmandaise was made available by EDF under a research agreement with CERFACS. The Telemac-2D model for the Ohio selected area is currently being developed at CERFACS, in collaboration with NOAA and Ohio State University.

The results are illustrated in Figure 9 for the Garonne Marmandaise (Fig. 9a), for the January-February 2021 flood event. They show that the assimilation of in-situ data and remote sensing data reduces the quadratic deviation from water level observations at Vigicrue stations (Fig. 9b) and improves agreement with the flooded area observed by Sentinel1.

Carte hauteurs d'eau

▲ Figure 9. Map of water levels in the Garonne catchment for the 2021 flood. (a) Catchment area of the Garonne over a length of 50 km. (b) Comparison of CFD modelling with observations: the black dotted line represents the observed water level, the orange line represents the Telemac-2D open loop simulation and the blue line represents the Telemac-2D simulation using data assimilation. (c) Flood extent contingency map representing the flood extent prediction on the left by FR without data assimilation, and on the right by IGDA with satellite and in-situ data assimilation, for 03/02/2021. Correctly predicted flooded areas are shown in dark blue, correctly predicted non-flooded areas in light blue, under-predicted areas in yellow and over-predicted areas in red. © CERFACS

Socio-economic risk mapping

For areas with a high economic impact, a real-time product estimating the financial risks associated with flooding will be generated for different types of assets, such as industry, agriculture and regional planning.

Indeed, by combining the outputs of the modelling water elevation maps with a database of physical assets and social media geolocation data, and by using methods that link the height of water and the duration of flooding to the physical risk, as well as integrating textual and structured data (prices of agricultural raw materials, fuel, real-time inflation calculations for users, etc.), this approach will make it possible to estimate the financial risk of flooding directly at the level of a site and more widely at the level of a company. Directly accessible on the QuantCube environmental intelligence platform, the flood risk impact map and the financial risk assessment can then be transformed into a score.

Coût dégâts


◀︎ Figure 10. Map of costs in dollars per pixel for cultivated land in the Garonne Marmandaise for the flood of 02/02/2021. © QuantCube Technology

As part of its demonstrator, the FloodDAM-DT project will provide access to the physical risk maps and the financial impact of the flood risk assessed for the project's 2 case studies via the platform.

Hydrological visualization and processing platform:

With a view to further automation, the platform has been chosen to integrate the FloodDAM processing chain and its results on selected sites for the demonstration. The platform is open to all users and enables exchanges with other platforms and services through the use of standardized interfaces and normalized data via the interface or API.

▲ Figure 11. hydrological platform © CNES

Related projects

  • SCO FloodDAM project
  • IDEAS (Integrated Digital Earth Analysis System) project under the ESTO (Earth Science Technology Office) /AIST (Advance information system technology) programme

Project news


By ticking this box, I acknowledge that I have read and accept the Legal Notice.
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.