GloFAS v5 reanalysis dataset¹ covers a 46-years time period, from 01/01/1980 to 31/12/2025 (simulations started on 01/01/1975 to allow model spin-up); all the variables have 0.05 degrees (or 3 arcmin) spatial resolution and daily temporal resolution.

Key hydrological fluxes and aggregated variables are available from the Copernicus Early Warning Data Store (EWDS). These variables are:

• Average daily discharge (m3/s) 
• Soil wetness index (root zone) ( - )
• Snow water equivalent  (kg/m²)
• Runoff water equivalent (surface plus subsurface) (kg/m²)

The page GloFAS available data provides the definition and explanation of each variable; instructions on how to download the data can be found in the page GloFAS data access - EWDS.

GloFAS v5 reanalysis dataset was generated using the 0.05 degrees semi-global set-up and LISFLOOD v5.0.0 (for details, please visit https://confluence.ecmwf.int/spaces/CEMS/pages/673567675/placeholder+-+GloFAS+v5+hydrological+model ). Model parameters were estimated using a 4-steps protocol, in-situ time series of discharge observations, and reservoir in-situ observations and metadata (for more information, please visit https://confluence.ecmwf.int/spaces/CEMS/pages/673567680/placeholder+-+GloFAS+v5+calibration+methodology+and+parameters ).

Similarly to the previous versions, GloFAS v5 discharge reanalysis dataset (also known as climatology) was used to compute river discharge flood threshold values for selected return periods. Specifically, the Gumbel extreme value distribution is fitted over the 1980-2025 period's annual maxima. The code is available open source as part of OS LISFLOOD utilities; maps with threshold values for each pixel of the semi-global domain are available from the page GloFAS Auxiliary Data. River discharge threshold values with 1.5, 2, 5, 10, 20 years return period are used in the operational GloFAS set-up to raise awareness of potential upcoming flood events with different magnitudes (medium range forecasts):  the page GloFAS medium-range forecast products explains the use of river discharge thresholds to generate  GloFAS Reporting Points, Flood Summary, Threshold Level Exceedance, and Rapid Flood Mapping products.

Reliable flood threshold values require a reanalysis dataset representing as closely as possible the most recent configuration of the catchments. Reservoir construction and operations can heavily affect river flow and thus flood threshold values. For this reason, the 0.05 degrees OS LISFLOOD semi-global model set-up used to generate GloFAS v5 reanalysis included all the active reservoirs as of early 2026. In other words, differently from the calibration workflow, where reservoirs had been included dynamically on their year of construction, GloFASv5 reanalysis model set-up included all the 1486 reservoirs in a static manner since the first computational time step. This approach allowed simulating how the current reservoir configuration would react to a wide range (46 years) of meteorological conditions.

An example of the difference between river discharge simulation in calibration and in the reanalysis dataset is presented in Figure 1.

Calibration station ID 5709 (yellow circle) is located in the Amur basin, Bureya river (Asia). The available discharge time series (Global Runoff Data Centre) covers the years 1980-1988. The Bureya Dam (GRanD) (pink triangle), located upstream of station ID 5709, was built in 2009.

• Model calibration was performed without considering the reservoir: calibration outcomes are shown by the blue line; KGE’ performance metric of  0.836 demonstrates the model capability to reproduce rainfall-runoff-routing processes.
• GloFAS v5 reanalysis included the reservoir in the model set-up since the beginning of the simulation: the orange line represents the reanalysis dataset.

While the blue line (calibration) represents more accurately the actual flow conditions before 2009 (as expected), its use for the computation of discharge flood threshold values would lead to overestimated values compared to the current catchment configuration (more closely represented by the orange line), potentially leading to missed events in forecast scenario.

Figure 1 - Left: location of station ID 5709 (yellow circle) downstream of the Bureya Dam (pink triangle). Right: observed (black) and simulated discharge time series prior to the construction year (2009): calibrated time series (blue) and reanalysis (orange).

Known issues

GloFAS v5 reanalysis dataset aims to represent hydrological states and fluxes at the semi-global scale. Inaccuracies in input data and model structural limitations can lead to decreased performance in localized contexts.

For example, the 0.05 degrees local drainage direction is derived from a digital elevation model, which is inevitably affected by approximations. River bifurcations are currently not modelled. Precipitation input data has a prominent impact on model accuracy: the known reduced accuracy of C3S ERA5 in areas such as Central Africa (Zsoter et al. 2020) and South-West America (Hersbach et.al, 2020; Blanco et al. 2023) can lead to trends in soil moisture and bias in discharge values. Numerical instabilities can sporadically occur in low slope rivers (for more information, please visit this page).

A more detailed list of known issues will be published in the EWDS. Users are encouraged to report issues and provide their feedback via the GloFAS contact form.

References

¹Grimaldi, Stefania; Salamon, Peter; Russo, Carlo; Disperati, Juliana; Casado-Rodriguez, Jesus; Mohamed, Azhar; Mazzetti, Cinzia; Mastrantonas, Nikolaos; Moschini, Francesca; Schaffhauser, Timo; Ramos, Arthur; Carton De Wiart, Corentin; Zsoter, Ervin; Gomes, Goncalo; Garcia-Sanchez, Rafael; Serratosa-Marquez, Alejandro; Jimenez-Molina, Antonio; Gonzalez-Martin, Marina; Prudhomme, Christel (2026): GloFAS v5.0 hydrological reanalysis. European Commission, Joint Research Centre [Dataset] doi: 10.2905/JRC.KM1MF70 PID: http://data.europa.eu/89h/83cd8a58-765a-4aa5-8e74-5dfc429c3706