A wide-ranging upgrade of ECMWF’s Integrated Forecasting System (IFS), implemented on 5 June 2018, brings better global weather forecasts, with particularly consistent gains in the extended range. A key plank of the upgrade is enhanced dynamic coupling between the ocean, sea ice and the atmosphere.
|Implementation date: 05 June 2018||See News article|
|Detailed description of changes|
|Key characteristics of our current forecasting system|
Data set affected
|Scorecard for 45r1|
New bathymetry in wave models
Upgraded bathymetry (water depth) used in the wave models based on ETOPO1 in:
See Implementation of IFS cycle 45r1 for details
The new cycle leads to improvements in HRES upper-air fields. Verified against the model analysis, a positive signal is seen throughout the troposphere for most parameters, except temperature in the lower troposphere at shorter ranges.
Weather parameters and waves
There is an overall improvement in 2m temperature both in the HRES and ENS particularly for Europe. 2m humidity is largely neutral for HRES, but positive for ENS, particularly in the tropics. 10 m wind speed is largely neutral in the HRES and slightly negative in the ENS. Precipitation in the HRES is improved in terms of categorical verification (e.g. SEEPS), and near-coastal precipitation in warm-rain dominated situations is significantly improved due to changes in the cloud physics. However, these changes also lead to more activity at higher precipitation rates in active regions such as the East Asian monsoon, and as a result error measures such as RMSE or CRPS (for the ENS) are increased. The negative signal for significant wave height against analysis is a result of changes to the analysis from a large increase in observation usage, and verification against observations (buoys) shows the results are neutral for both HRES and ENS.
The implementation of the ocean-atmosphere coupling in the HRES removes the overall negative bias in tropical cyclone central pressure and thereby reduces the mean absolute intensity error by about 10% in the short range, and about 20% from day 5 onwards. Evaluations so far show statistically neutral results for the position error.
Changes in scores for the monthly system are generally positive across the range of parameters, with significance in week 1 for tropical winds. The only indication of a degradation is precipitation in the tropics with a consistent negative signal across all 4 weeks. There is an indication of a positive effect on skill across all parameters in the European domain. The MJO Index was significantly under-spread, but changes in 45r1 to the SPPT scheme have brought the spread and error in close agreement throughout the 30 day forecast range. The underestimation of the MJO Index amplitude error has been significantly improved throughout the forecast.
|New disseminated model output||
(*) Not added to the Real-time Catalogue
|e-suite experiment number 0072|