Interesting article about the Euro. And some changes for later this year.
"On August 1st, 1979, the first operational Deterministic forecasts from the European Center for Medium range Weather Forecast (referred to ECMWF hereinafter) were released at low spatial resolution (210 km) using Spectral Transforms out to 120 hours (5 days) with further upgrades to 168 hours (Seven days) on August 1st, 1980. [1]
Ensemble forecasts were first created on November 24th, 1992 at 210 km horizontal resolution with 32+1 members 3x a day while the deterministic model was 200km, this difference in resolution gave the deterministic its name, HRES. Incremental horizontal resolution increases have occured since 1992 with 41r2 in 2016 featuring the last horizontal resolution increase to the HRES and a new grid (Octahedral Gaussian, 16 km-->9 km HRES, 36 km--> 18 km ENS). Improvements to many surface, upper level variables including track & intensity of hurricanes were noted. [2]
In 2021 brought 2 upgrades to the IFS model with one in particular was 47r2 which brought Single Precision (32 bits) from Double Precision (64 bit) to the IFS HRES & ENS, this helped reduce computational cost to increase the vertical layers in ENS to 137L to match the HRES. [3]
In October 18th, 2022, a major supercomputer upgrade took place from the Cray XC40 cluster to the ATOS BullSequana XH2000 in Bologna, Italy. This has allowed ECMWF to introduce another significant step in unifying the ENS with HRES by introducing 48r1. With this upgrade came the increase in horizontal resolution to the ENS from 18 km to 9km. This is now in sync with the HRES and makes the ENS control meteorologically identical to HRES with the only changes being HRES runs before ENS control and that ENS Control goes out to 15 days to match the 50 Perturbed ECMWF EPS members. Significant improvements to track & intensity of hurricanes/typhoons were well noted as well as many surface, upper level. The analysis of HRES/ENS are exactly the same with differences in model outputs appearing with increased lead time. [4] The former twice a week ECMWF-Ext ENS changed to once daily with 101 members starting from day 0 instead of 15 which was an extension of the medium range forecast..the reforecast were also separated to be consistent. [5]
Plans For the Future?
In 49r1 (set to be implemented in October 2024) will still feature 2 separate deterministic systems 'HRES' 'ENS control' but the HRES will be extended from 10 days-->15 days and still come out earlier than ENS. [6]
In 50r1 (set to be implemented in June 2025) the 'ENS control' will be retired while the 'HRES' will be renamed to 'ENS control' but still come out earlier than the ENS 50 perturbed members. This will remain at least until 51r1. [6]"