Since the implementation of EFAS version 5.4 (XXX 2025) and GloFAS v4.3 (YYY 2025), the same design of products are used for the sub-seasonal range both in EFAS and GloFAS. This means, the previous EFAS sub-seasonal products were replaced by the new ones, while for GloFAS the sub-seasonal products are completely new. 

Sub-seasonal product configuration design

The new sub-seasonal products have calendar weekly periods (i.e. always Monday-Sunday) as lead time. The fixed calendar weeks as forecast lead times will allow the users to compare forecasts from different runs, as the verification period is fixed (as the calendar weeks). The generation of the sub-seasonal products rely on few major components:

1. Real time forecasts: This part is the hydrological forecasts produced in real time. This will give the actual predicted conditions for the sub-seasonal products that will be compared to climatologies to derive the forecast anomalies. In the following we describe the characteristics of these forecast simulations. Where appropriate, the difference between EFAS and GloFAS is specified. If there is no EFAS/GloFAS mentioned, then the method is identical between the two forecast systems:

• • Hydrological model: LISFLOOD (LINK!)
  • Meteorological forcing: the combination of the 9km (horizontal resolution) ECMWF ensemble forecasts (LINK!) and the 36km ECMWF sub-seasonal forecasts (LINK!) from the PREVIOUS RUN
    • The low-resolution sub-seasonal forcing is taken from the precious available forecast run simply because of timing, as the sub-seasonal meteo forecasts for the same day are only available with several hours delay after the high-resolution forecasts. 
    • The high- and low-resolution forcing is currently combined by a simple mechanical blending by using first the high resolution meteorological forcing for days 1-15, while the low-resolution meteorological forcing for days 16-45, for each of the 51 ensemble members. 
    • The blending can create inconsistencies locally over very complex orograpihcal areas, due to the resolution change at day15, and also there could be some smaller inconsistencies between the ensemble in the from 16 vs to 15 days, due to the mechanical blending which will mix different weather conditions from the low- and high-res meteo data, but overall as an ensemble of 51 scenarios, this will not be expected to lead to any larger discontinuities. 
  • Number of ensemble members: 51 ensemble members (one unperturbed and 50 perturbed) to represent the equally likely forecast scenarios of the future
  • River resolution: 3arcmin (~5km) in GloFAS and 1arcmin (~1.5 km) in EFAS
  • Run frequency: Forecasts are generated daily for 00 UTC
  • Lead time: 45 days (1080 hours). Only 45 days, as although the low-resolution sub-seasonal meteo forecasts run out to 46 days, we can only rely on the 1-day-old run, which means we can only run the simulations to 45 days ahead  
  • Forecast steps: 6-hourly in EFAS and 24-hourly in GloFAS
  • Forecast hydrological initialisation: From a fillup simulation forced with the shortest-range ENS-Control (unperturbed member of the ECMWF ensemble forecast) meteorological conditions

1. Reforecasts: The sub-seasonal products rely on range-dependent climatologies, that change with the forecast lead time. The climatologies are produced from a large set of hydrological reforecasts. In the following we describe the characteristics of these reforecast simulations. Where appropriate, the difference between EFAS and GloFAS is specified. If there is no EFAS/GloFAS mentioned, then the method is identical between the two systems:

• • Hydrological model: LISFLOOD (LINK!)
  • Meteorological forcing: the combination of the 9km (horizontal resolution) ECMWF ensemble forecasts (LINK!) and the 36km ECMWF sub-seasonal forecasts (LINK!) from the SAME RUN
    
    • The low-resolution sub-seasonal and high-resolution medium-range forcings are taken from the same forecast run date. The timing of the availability of the low-resolution forecasts is not an issue here, as these reforecasts are only produced retrospectively without any time constraints.
    • The high- and low-resolution forcing is currently combined by a simple mechanical blending by using first the high resolution meteorological forcing for days 1-15, while the low-resolution meteorological forcing for days 16-46, for each of the 11 ensemble members. 
    • The blending can create inconsistencies locally over very complex orographical areas, due to the resolution change at day15, and also there could be some smaller inconsistencies between the ensemble in the from 16 vs to 15 days, due to the mechanical blending which will mix different weather conditions from the low- and high-res meteo data, but overall as an ensemble of 11 scenarios, this will not be expected to lead to any larger discontinuities.
  • Number of ensemble members: 11 ensemble members (one unperturbed and 10 perturbed) to represent the equally likely forecast scenarios of the future
  • River resolution: 3arcmin (~5km) in GloFAS and 1arcmin (~1.5 km) in EFAS
  • Run frequency: Forecasts are generated always for the past 20 years. Before 12 Nov 2024, they were generated twice-weekly on Mondays and Thursdays (for 00 UTC), while from 12 Nov 2024 they are generated on given days of the months, as 1, 5, 9, 13, 17, 21, 25 and 29 (excluding 29 Feb).
  • Lead time: 46 days (1104 hours)
  • Forecast steps: 6-hourly in EFAS and 24-hourly in GloFAS
  • Forecast hydrological initialisation: From the monitoring hydrological simulation, forced with gridded meteorological observations in EFAS and ERA5 reanalysis meteorological data in GloFAS

1. Climatologies: The sub-seasonal products rely on range-dependent climatologies, that change with the forecast lead time, and which are produced from the hydrological reforecasts. The climatologies will give the reference point for the different anomaly categories applied in the sub-seasonal range. These reference points are some of the specific quantiles from the climate distribution, such as the 10th and 90th percentile values. In the following we describe the main characteristics of the climatologies. Where appropriate, the difference between EFAS and GloFAS is specified. If there is no EFAS/GloFAS mentioned, then the method is identical between the two systems:

• • We currently produce climate files for each reforecast run date, so in total 8*12-1 = 95 dates in a calendar year with 1/5/9/13/17/21/25/29 of each month, excluding 29 Feb.
  • For each of these climate dates, there will be climate files for each possible daily lead time of the sub-seasonal range. Here not only calendar weeks are considered, but all possible 7-day lead times, starting from days1-7, then days2-8, .., out to days 40-46. This way, the weekly mean climatology for all possible lead times will be available, and so for each real time forecast the right climatology can be used, with the correct lead time in days, depending on which day of the week the real time forecast run (i.e. which corresponding climate lead time to choose in order to get to the calendar weeks). 

EFAS/GloFAS are both handled in the same way, by the same structure and same scripts. The only differences are the source name variable in the suite and the copydis_seasonal.sms script, which has slightly different retrieval options for both EFAS/GloFAS, depending on the source variable setting. And also the slight difference is in the weekly averaging, as the steps are set according whether 24-hourly (GloFAS) or 6-hourly (EFAS) data is processed.

Currently the suite with the scripts sit here: /home/moi/ecflow/suites/cems-flood_subseasonal_seasonal

The definition file is generated by: gen_subseasonal_climate_def.py

1. The climatologies will give the reference point for the different anomaly categories applied in the sub-seasonal range. The climatologies will provide some specific quantiles from the climate distribution, such as the 10th and 90th percentile values.

• • Hydrological model: LISFLOOD (LINK!)
  • Mete

Sub-seasonal system

The sub-seasonal system forecasts are use the 

• • General: It should be based on the ECMWF ENS forecasts, primarily the 46-day, 101-member ENS-ext, produced for every 00 UTC, and potentially also the 15-day, 51-member higher resolution ENS (when blending).
  • Time frequency: We should use weekly means. We will have 6 lead times for the Monday, Thursday, Friday, Saturday and Sunday sub-seasonal runs, but only 5 for the Tuesday and Wednesday runs (as the 46-day lead time is not enough to cover 6 weeks).
  • Blending: Do we want to blend the low-res and high-res ensembles, or only use the low-res? We do blending in GloFAS currently. The blending can create inconsistencies locally due to the resolution change, and also there could be some smaller inconsistencies due to the mechanical 1-1 ensemble merging after day15, but these were deemed acceptable in a recent study (https://confluence.smhi.tds.tieto.com/display/ECC/48r1+impact+study+on+river+discharge). With the blending we can only have 51 ensemble members though, which again should be OK. In addition, the reforecasts seem to have the blending configuration for both EFAS and GloFAS, which makes it rather easy or maybe the only feasibly option for now to go for the blending.

• Seasonal:
  • General: It should be based on the SEAS5 data that runs out every month to 7 month with 51 ensemble members in the real time forecasts. So, we should really utilise the whole 7-month period in both EFAS and GloFAS.
  • Time frequency: We should use the (calendar) months as periods (not weeks as currently). It is really something that in every other external example is like that. We need to switch to monthly for sure. For each seasonal run, we would have 7 monthly lead times. 
  • Blending: Again, the question of blending. As, currently there is no blending considered in either reforecasts of real time seasonal. So, as the reforecasts are not using blending, so we should not consider thm for real time anyway. In the long term there will be this question though.