Understanding how the BWI hydrological forecasting system works
BWI brings together information from different sources: Earth observation satellites (in particular EU space data and assets such as those of the Copernicus program), hydrometric stations, forecasts and meteorological data, and local observations. This diversity ensures a solid and representative foundation. This step is detailed below.
All data is centralized and organized in a secure warehouse located on French territory. This structure makes it possible to maintain a complete history and to facilitate large-scale exploitation.
Next, BWI implements advanced hydraulic and hydrological models on this data to compile an initial set of results. The collected data is ingested into models capable of simulating river behavior for each watershed that has been digitized. These models translate natural complexity into actionable results and require prior training before the system goes into production.
BWI also uses machine learning, through neural networks, to refine the results produced over time, so that the forecasts take into account the recent acceleration of climate change and its impacts on water resource availability.
This combination of data, modeling, and machine learning enables BWI to provide forecasts that are as reliable as possible, on a continuous basis. The forecasts thus become, for its users, an additional tool to anticipate, manage, and secure water use.
BWI uses satellite systems to gather multiple data sets and define the natural parameters of the digitized area: digital elevation model, land use and soil moisture, etc. More specifically, BWI uses Copernicus EU-Hydro River Network Database for European river topology and drainage. BWI employs CORINE Land Cover and EU-DEM for Hydrological Response Units parameterization and catchment delineation. BWI integrates ECMWF ERA5/IFS meteorological forcings with semi-distributed modeling at virtual stations.
Radars and weather stations are also used to collect information on local weather forecasts, both upstream and downstream of the observation point.
In addition, physical hydrometric stations — electronic devices located along rivers — measure in situ the flow and water level of the watershed. These data are necessary for the training and validation of the BWI model.
BWI’s virtual stations make it possible to deliver hydrological data digitally by leveraging both spatial and local data. This tool, developed by BWI, is presented in more detail below:
A virtual station refers to a point on the hydrographic network where hydrological information (flows & water levels) is estimated. The virtual station is freely placed on the hydrographic network by its user, but it does not physically exist in the field and does not require any physical instrumentation on site. It therefore makes it possible to extend the coverage of the hydrographic network with ease.
Each virtual station delivers projections as reliable as possible over a horizon of up to 10 days. This short-term visibility helps stakeholders anticipate risks and optimize water resource management.
Virtual stations can reconstruct hydrological knowledge and provide flow duration curves. These curves and their time series reflect the frequency and intensity of flows over several years or even past decades. This information facilitates the comparison of hydrological regimes and the monitoring of their evolution over time.
Thanks to an easy-to-use and well-documented API, data from BWI’s virtual stations integrates directly into clients’ existing tools (SCADA, GIS, etc.). No need for new software or major changes to your systems — making optimal use of your hydrological forecasts simple and seamless.
The use of BWI virtual stations and watershed digitization enables new applications, including the implementation of early warning systems, decision-making support for resource allocation (for example, between irrigation and energy) during periods of resource stress, as well as securing construction sites for hydraulic works.
Virtual stations eliminate the constraints of hardware devices: they are deployed instantly, provide real-time and forecast data, and benefit from machine learning. In addition, virtual stations are not exposed to the risks faced by field-based stations: breakdowns, drift, damage, theft, etc. Their freedom of placement by users on the hydrographic network and their remote operation make them a modern and effective alternative to traditional physical stations. However, virtual stations do not replace physical stations, which remain essential: virtual stations complement existing infrastructures by densifying them and enhancing their functionalities, in a competitive, agile, and coordinated way.
Contact BWI
for any questions
