Page History

GloFAS v2.1 reanalysis and initialisation run

GloFAS v2.1 has a common daily river discharge reanalysis and an initialisation run shared between GloFAS v2.1 30-day and GloFAS v2.1 Seasonal.

GloFAS v2.1 river discharge reanalysis was produced using the ERA5 ECMWF reanalysis, the atmospheric reanalysis product of ECMWF. ERA5 has a horizontal resolution of ~32 km (see Hersbach at al., 2018 for further reading). GloFAS v2.1 river discharge reanalysis covers 1979 to near-real-time and is run at the model setup. GloFAS v2.1 discharge reanalysis uses the officially released ERA5 data throughout the January 1979 to near-real-time. The spin-up period (1978) uses pre-released ERA5 (ERA5pr), a preliminary estimate of atmospheric fields. GloFAS v2.1 river discharge reanalysis is used to compute the GloFAS v2.1 30-day river discharge thresholds. While GloFAS v2.0 used some preliminary ERA5 data, the GloFAS v2.1 river discharge reanalysis uses only officially released ERA5 dataset as forcing.

GloFAS v2.1 river discharge initialisation run is used to provide initial conditions (atmosphere and land surface states from which to start the ensemble river discharge forecasts) for both the GloFAS v2.1 30-day and GloFAS v2.1 Seasonal. It is based on a near-real-time or 'timely' version of ERA5, called ERA5T, and for the remaining days prior to real time, based on day-1 ECMWF-ENS-CNTL forecast. Currently, ERA5T is available with a latency of 2-5 days (compared with real time) and does not include additional quality checks.

Available GloFAS v2.1 30-day 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 only daily river discharge time series are available to users.

GloFAS v2.1 30-day

GloFAS v2.1 30-day real-time river discharge forecasts are produced by the GloFAS v2 modelling chain initialised from the GloFAS v2.1 river discharge initialisation run and forced with data from the ECMWF ENS medium- and extended-range forecasting system. ECMWF ENS is ECMWF's ensemble forecast system and consists of 51 members at ~18 km resolution up to 15 days, increasing to ~36 km from days 16 to 30 in the monthly extension, which is available only for Monday and Thursday runs. In GLoFAS, the ECMWF ENS runoff forecasts are resampled to 0.1° spatial resolution (~10 km) before being used as daily input to the Lisflood model to produce river discharge. In the first 15 days the medium range runoff can be used, while from day 16 the latest available extended-range forecast is used, with up to 3-day lag to real time (to the last Monday or Thursday run), to produce seamless river discharge forecasts out to 30 days. Results are presented as daily values.

GloFAS v2.1 30-day river discharge thresholds are generated by fitting an extreme value distribution (Gumbel using method of L-moments) to the annual maxima series, extracted from the river discharge reanalysis of 1979-2018. These thresholds are used to generate the 30-day flood forecasts. Compared with GloFAS v2.0, GloFAS v2.1 30-day river discharge thresholds are based on GloFAS v2.1 river discharge reanalysis generated using the officially released ERA5 data. Another change is that the drier years are not filtered in the annual maximum computation, hence using all the 40 available years in 1979-2018 period.

GloFAS v2.1 30-day flood forecasts are generated by comparing the GloFAS v2.1 real-time forecasts to the thresholds. This provides flood signals as maps of probability of threshold exceedence and forecast hydrographs with the river discharge evolution in the next 30 days. GloFAS v2.1 30-day flood forecasts are available daily through the GloFAS map viewer from the launch date. Compared with GloFAS v2.0, GloFAS v2.1 30-day flood forecasts benefit from the threshold update based on the GloFAS v2.1 river discharge reanalysis using the officially released ERA5 data.

GloFAS v2.1 30-day river discharge reforecasts are forecast simulations generated for a set of past dates. For the 1998-2017 period, they are produced from the twice weekly, 20-year, 11-member ensemble ECMWF ENS reforecasts. They are complemented by past forecasts obtained from 51-member ensemble ECMWF-ENS forecasts produced twice weekly from January 2018 to June 2019 and daily from July 2019 until real time. The reforecasts use the operational settings of the real time forecasts and are initialised from the river discharge reanalysis until July 2019 and from the initialisation run after that. Compared with GloFAS v2.0, the GloFAS v2.1 reforecasts (1998-2017) are produced using newer ECMWF ENS reforecasts which benefit from a major model upgrade (IFS cycle 45r1). The GloFAS v2.1 30-day real time forecast production started from July 2019 using the new GloFAS v2.1 initialisation run. The river discharge reforecasts, together with past real time forecasts, provide a long  period for GloFAS v2.1 30-day skill assessment.

Available GloFAS v2.1 30-day 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 v2.1 30-day datasets only include river discharge time series.

GloFAS v2.1 Seasonal

GloFAS v2.1 Seasonal real-time river discharge forecasts are produced by the GloFAS v2 modelling chain initialised from the GloFAS v2.1 river discharge initialisation run and forced with data from the ECMWF SEAS5 seasonal forecasting system. SEAS5 is ECMWF's ensemble seasonal forecast system and consists of 51 members at ~36 km horizontal resolution. SEAS5 runoff forecasts are resampled to 0.1° spatial resolution (~10 km) before being used as daily input to Lisflood to produce river discharge up to 16 weeks. GloFAS v2.1 seasonal real-time river discharge results are presented as weekly averages, starting every Monday, and are available on the 10th of each month.

GloFAS v2.1 Seasonal river discharge thresholds 

are river discharge magnitudes corresponding to fixed percentile values (20th and 80th for the low flow and high flow), derived from the 37-year (1981-2017) GloFAS v2.1 Seasonal river discharge reforecasts. A set of thresholds is produced specifically for each week of the year and each forecast lead-time. This is different to GloFAS v2.1 30-day, where one threshold set is used for all the forecast ranges in the 30-day forecast horizon. The GloFAS v2.1 Seasonal river discharge thresholds are used to produce the hydrological extreme forecasts. Compared with GloFAS v2.0, GloFAS v2.1 Seasonal river discharge thresholds are based on the GloFAS v2.1 river discharge reanalysis generated using the officially released ERA5 data.

GloFAS v2.1 Seasonal hydrological extreme forecasts

are generated by comparing the 51 scenarios of weekly-averaged river discharge forecasts to the thresholds. Both low flow (with 20th percentile) and high flow (80th percentile) probabilities are produced for each lead-time to 16 weeks ahead. This provides maps of probability of threshold exceedence/ non-exceedance and weekly forecast hydrographs for the next 16 calendar weeks. GloFAS v2.1 Seasonal hydrological extreme forecasts are available on the 10th of each month through the GloFAS map viewer from the launch month. Compared with GloFAS v2.0, GloFAS v2.1 Seasonal hydrological extreme forecasts benefits from thresholds based on GloFAS v2.1 river discharge reanalysis generated using the officially released ERA5 data. 

GloFAS v2.1 Seasonal historical river discharge forecasts are forecast simulations performed for a set of past dates, over January 1981 to October 2017, using ECMWF Seasonal forecasting system SEAS5. They use the operational configuration of the real-time forecasts and are initialised from the GloFAs v2.1 river discharge reanalysis. GloFAS v2.1 Seasonal forecasts from November 2018 to July 2019 were not generated, as they are identical to the existing GloFAS v2.0 forecasts. The GloFAS v2.1 Seasonal real time forecast production started from August 2019 using the new GloFAS v2.1 initialisation run. The reforecasts are used for producing the GloFAS v2.1 Seasonal river discharge thresholds and they also provide a long period for GloFAS v2.1 Seasonal skill assessment. Compared with GloFAS v2.0, GloFAS v2.1 Seasonal river discharge reforecasts benefit from initial conditions generated using the officially released ERA5 data.

Available GloFAS v2.1 Seasonal 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 v2.1 Seasonal datasets only include weekly-averaged discharge time series.