GloFAS 30-day v2.0

GloFAS 30-day v2.0 was launched on the 14th November 2018

GloFAS 30-day v2.0 technical settings

GloFAS 30-day v2.0 uses the Lisflood river routing model (van der Knijff et al., 2010) to propagate along the river channel the surface and sub-surface runoff forecasts of the HTESSEL land surface model (Balsamo et al., 2009) of the ECMWF ENS medium- and extended-range forecasting system to produce daily discharge forecasts. In GloFAS 30-day v2.0, Lisflood was calibrated for better accounting for the variability in the hydrological processes (Hirpa et. al., 2018). More details can be found in the Lisflood calibration page.

ECMWF-ENS is the ensemble forecast of ECMWF and consists of 51 members at ~ 18 km resolution up to 15 days, increasing to ~36 km from day 16 to 30. ECMWF ENS runoff forecasts are resampled to 0.1 spatial resolution (~10 km) before being used as daily input to Lisflood to produce discharge up to day 15; from day 16 the latest available extended range forecast is used as forcing to produce seamless discharge forecasts for 30 days.

GloFAS 30-day v2.0 initial conditions (atmosphere and land surface states from which to start the ensemble discharge forecast) are based on a 5-day monitoring analysis of the latest proxy-observations, the ERA5 ECMWF reanalysis. Because of ERA5’s has a ~2 days latency (compared with real time), ERA5 is used only as long as it is available; for the remaining 2 to 3 days, prior to the GloFAS 30-day forecast, the day-1 of ECMWF-ENS-CNTL forecast is taken as input.

The GloFAS system
Schematic of GloFAS 30-day v2.0.

GloFAS 30-day v2.0 reference discharge runs

GloFAS 30-day v2.0 reference discharge runs are a collection of discharge simulations as close as possible to GloFAS 30-day v2.0 operational system.

GloFAS 30-day v2.0 hydrological reanalysis is based on ERA5, the latest atmospheric reanalysis product of ECMWF available in 2018. ERA5 has a horizontal resolution of ~32 km (see Hersbach and Dee 2016 for further reading). GloFAS 30-day v2.0 hydrological reanalysis covers 1981-2017. GloFAS 30-day v2.0 discharge thresholds are computed from GloFAS 30-day v2.0 discharge time series of GloFAS 30-day v2.0 hydrological reanalysis.

GloFAS 30-day v2.0 forecasts reruns are produced from the twice weekly, 20-year, 11-member ensemble ECMWF ENS reforecasts (1997-2016) and 1.5 years of 51-member ensemble ECMWF-ENS hindcasts (2017-June 2018), using GloFAS 30-day v2.0 operational settings and GloFAS 30-day v2.0 hydrological reanalysis as initial conditions. GloFAS 30-day v2.0 forecast reruns have 21.5 years of daily discharge ensemble data, and can be used to assess GloFAS 30-day v2.0 skill.

GloFAS 30-day v2.0 discharge thresholds are generated by fitting a Gumbel extreme value distribution to the annual maxima series extracted from GloFAS 30-day v2.0 discharge time series of GloFAS 30-day v2.0 hydrological reanalysis (1981-2017) using the method of L-moments. GloFAS 30-day v2.0 discharge thresholds are used to generate flood awareness forecast maps.

GloFAS 30-day v2.0 products

GloFAS 30-day v2.0 products are generated by comparing GloFAS 30-day v2.0 real-time forecasts with GloFAS 30-day v2.0 discharge thresholds. They include exceedance probability maps and forecast hydrographs and are available daily through the GloFAS map viewer.

GloFAS 30-day v2.0 available data

GloFAS 30-day v2.0 available data is summarised in the table below. It contains the period covered, the production frequency, the number of ensemble members and the ECMWF IFS versions used for each dataset. Note that GloFAS 30-day v2.0 datasets only include discharge data. Data can be requested from info@globalfloods.eu

GloFAS 30-day v2.0 Period Frequency Ensemble members ECMWF IFS version
Reanalysis 1981-2017 Daily 1 41r2
Forecast reruns Jan 1997 - Dec 2016
1 Jan 2017 - 30 Jun 2018
Twice per week
Twice per week
11
51
43r1: 1 Jan 2017 - 10 July 2017
43r3: 11 Jul 2017 - 4 Jun 2018
Real-time forecasts 1 Jul 2018 - real-time Daily 51 45r1: 5 Jun 2018 - real-time

GloFAS 30-day v2.0 skill

Overall probabilistic skill of GloFAS 30-day v2.0 is shown for 1, 5, 10, and 15 days lead time using the Continuous Ranked Probability Skill Score (CRPSS) metric. Forecasts are from GloFAS 30-day v2.0 reruns for 11 ensemble members from 1997-2016 benchmarked against a deterministic hydrological persistence forecast, and using GloFAS 30-day v2.0 discharge timeseries from GloFAS 30-day v2.0 hydrological reanalysis as reference. CRPSS values of 0 signify no skill, negative values for when forecasts are less skilful than the benchmark, and 1 for a perfect score.

Hover the map and zoom in with mouse wheel
Hover the map and zoom in with mouse wheel
Hover the map and zoom in with mouse wheel
Hover the map and zoom in with mouse wheel