Discharge observations

Nearly 2500 river gauging stations with daily discharge data available were considered for the calibration of GloFAS v4.0. 

Daily observed discharge data for the period The time interval was 01/01/1980-31/12/2019 were considered for the calibration, thus allowing the tuning and verification of model parameters using for a potential time span of 40 years. This Moreover, this time period is consistent with the availability of both discharge data and the ERA5 reanalysis meteorological forcings (see par 3.2) in December 2021.

River gauging stations were geolocated on the GloFAS v4 0.05 degrees resolution drainage network: model river drainage network is derived from a digital elevation model, which is inevitably affected by approximations. Consistency between real and model drained area is essential to allow reliable comparisons between observed and modelled discharge values. A detailed explanation of the relevance of this step is provided here. Observed discharge data time series were then manually quality checked to exclude stations with obvious data issues (e.g. outliers). 

The calibration stations were selected based on the following criteria (with some exceptions):

1. Drainage area larger than 500 km².
2. A minimum number of daily discharge observations corresponding to 4 years within the period 01/01/1982-31/12/2019. The optimal calibration period started on 01/01/1982 to allow 3 years of spin-up of the hydrological model.
3. Stations located close to another station and having the same data quality but shorter time series were excluded.

Albeit some stations did not fulfil the criteria, they were still selected for calibration if they allowed to improve the spatial coverage of calibrated catchments in the global domain. Specifically, the following exceptions were allowed:

• Stations with nearly 4 years of data.
• Stations with only old observations: data from 01/01/1980 were used for 2 calibration points; data from 01/01/1981 were used for 49 calibration points. The adequate model spin-up was verified through numerical experiments prior to acceptance of these stations in the calibration data set.
• Stations for which information on the real drained area was not provided. Inclusion of these stations required careful verification through numerical experiments.

1996 quality checked time series were used for calibration. More specifically, 212 calibration points were in Europe, 250 in Asia, 61 in Oceania, 420 in Africa, 617 in Centre-North America, and 436 in South America. The drained area of these stations entailed 47.5 % of the quasi-global domain.

The spatial distribution of the calibration points and of their drained area is shown in Figure 2. The calibration of GloFAS v4 could benefit of a much larger number of observations compared to GloFASv3, with an increase of 770 points (see Alfieri et al. 2020 for a detailed description of the calibration of GloFAS v3).  An ongoing and constant effort is implemented to expand the global database of discharge data (few words to promote data collection + Share your data with GloFAS - Copernicus Emergency Management Service - CEMS - ECMWF Confluence Wiki). The higher spatial resolution of GloFAS v4 allowed to calibrate smaller catchment. The minimum drainage area of GloFAS v3 was 5000 km2: this threshold was reduced to 500 with GloFASv4. drained area of the calibration points, XXX (336/to check) calibration points have drainage area 500-5000

Figure 2 – points: calibration stations, red NEW, grey OLD. Blue: area covered in calibration, dark blue ADDITIONAL area 

The temporal coverage of the observed time series varies across the global domain. Figure 4 shows the length (total number of daily measurements in equivalent number of years) of the observation time series for each calibration point.

Figure 4 –calibration stations, length of the observation time series in years. Purple: exceptions in data scarce area.

Meteorological forcings

The meteorological forcings used for GloFAS v4 calibration were provided by C3S ERA5.

The input variables total precipitation, 2-metre temperature, 2-metre dew temperature, 10-metre U wind component, 10-metre V wind component, surface solar radiation downwards, surface thermal radiation were upsampled from their native resolution (~30km) to 0.05 degrees resolution using bilinear interpolation, implemented in the open-source pre-processor pyg2p.

2-metre temperature, 2-metre dew temperature, 10-metre U wind component, 10-metre V wind component, surface solar radiation downwards, surface thermal radiation are used to compute  reference values of evapotranspiration according to the Penmann-Monteith equation, which is implemented in the open source pre-processor LISVAP.

LISFLOOD OS then takes as input total precipitation, 2-metre temperature, and potential evapotranspiration. Similarly to all the previous GloFAS versions, the calibration of GloFAS v4 was completed with daily time steps.