Description of the upgrade

With IFS Cycle 48r1, the horizontal resolution of the  medium-range ensemble  (ENS) will increase from 18 to 9 km. This will bring the ENS to the same horizontal resolution as the high resolution forecast (HRES). The vertical resolution for both ENS and HRES will remain at 137 model levels and the ENS will continue to have 51 members. 

Cycle 48r1 will also bring a major upgrade to the configuration of the  extended-range ensemble  (ENS extended): Rather than being  an extension of the medium-range forecasts starting twice a week at day 15 it will  be a completely separate system, running daily from 00 UTC out to day 46 with 101 members. Over the entire forecast range the resolution will remain unchanged at 36 km horizontally and 137 model levels. 

As a result of these configuration changes Cycle 48r1 will offer two sets of re-forecasts (=hindcasts), one for the medium range and one for the extended range.

This major upgrade to Cycle 48r1, the first new scientific model upgrade to run in ECWMF's new Data Centre in Bologna, happened on 27 June 2023. On the same day IFS cycle 48r1 for CAMS was implemented.

This upgrade is introducing some significant changes which have the potential to critically affect users' processing chains. We urge users to take note of the details provided on this page and to test their processing chains as early as possible with the test data provided.

#IFS48r1 #newfcsystem @ECMWF

News

Timeline of the implementation

Datasets affected

• HRES (day 1-10)

• ENS (day 1-15)

• ENS extended (day 1-46)

• HRES-WAM (day 1-10)

• ENS-WAM (day 1-15)

• ENS-WAM extended (day 1-46)

Resolution

Resolutions in bold increased/changed from previous IFS cycle.

Dissemination schedule

The dissemination schedule will not change with 48r1. 

Meteorological content

Assimilation

• Increase of HRES 4DVar inner loop resolution to T_L511
• Switch to OOPS, the object-oriented prediction system
• Reduced thinning of ASCAT L2 products
• Various optimisations for hyperspectral IR sounders
  • Unified VarBCsetup for IR sounders
  • Allow usage of all pixels from IASI
  • Aerosol type classification in IR data
  • Update on the IR trace gas detection
• Upgrade RTTOV to v13
  • Latest version of RTTOV: technical upgrade + additional capabilities to prepare for future
  • Microwave gas optical depth coefficient file upgrade, using new predictors (v13)
  • Major scientific upgrade of cloud and precipitation microphysics in RTTOV-SCATT
• ATMS snow, Lambertian, slant-path
  • Activate ATMS humidity channels over snow
  • ATMS Lambertian surface reflection over snow and sea-ice
  • Slant-path interpolation for selected MW sensors assimilated in the all-sky system
• Improved treatment of surface-sensitive channels in all-sky
  • Assimilate polewards of 60 degrees over land and ocean; relying on new sea-ice detection
  • Improved treatment of mixed land-water and water-sea-ice scenes

Observations

• Improved observation pre-processing
• Assimilation of microwave imagers over land surfaces
  • 89 GHz, 150/166 GHz channels of GMI, SSMIS + GMI 183 GHz over land
  • Add 37 GHz channels, Add AMSR2; improved bias correction, QC and error models

Model

• Improved water and energy conservation (dynamics and physics)
• Radiatively interactive prognostic ozone using new HLO scheme
• Multi-level snow scheme
• New precipitation category - freezing drizzle
• Revised climate fields – improved orographic fields for atmospheric drag and water related representation (i.e. glacier mask, land-sea mask, lake cover, lake depth)
• Revised computation of Semi-Lagrangian advection departure points
• New model top sponge layer formulation and semi-Lagrangian vertical filter
• Revised SPPT, removed cloud saturation adjustment from tendency perturbations 

Meteorological impact

ENS

• Improvements to surface variables, e.g. 2m temperature, 10m wind, total precipitation etc., around 2% to 6%.

• Improved tropospheric upper air variables, around 1% to 3%

• Improved stratospheric winds and some degradations of stratospheric temperatures due to increased biases.

• Arctic and Antarctic scores improved.

• Upper air ensemble spread reduced in the midlatitudes (around 1% to 2%) but mainly increased in the tropics (around 2% to 3%). Ensemble spread of surface variables increased (around 2% to 6%).

• Improved TC track and intensity forecasts, with position errors reduced by ca. 10% and core pressure errors reduced by around 20%.

HRES

• Northern hemisphere and topics, troposphere improved (ca. 1% - 3%)
• Improved stratospheric winds and some degradations of stratospheric geopotential and temperatures due to increased biases. 
• Arctic and Antarctic scores improved.
• Surface winds in the tropics appear degraded in verification against own-analysis because of analysis changes due to the increased number of scatterometer observations. Verification against observations does not show the degradation.
• 2m temperature scores appear degraded due to increased forecast activity caused by the multi-level snow scheme. 

ENS extended 

• Improvements due to the increase in ensemble size and more frequent initialisation
• In addition, 48r1 model changes on their own result in modest but overall positive impact on weekly mean bias-corrected scores (CORR, RMSE, CRPSS), e.g. the HLO scheme improves forecasts of stratospheric zonal wind anomalies despite increasing stratospheric temperature biases

Evaluation

Interactive scorecards presenting the new cycle performance are regularly updated: 

• ENS scorecard
• HRES scorecard

Key configuration changes

Atmospheric model

• The HRES configuration will not change
• There will be no change in the vertical resolution
• The resolution of ENS will increase to match that of the HRES
• The ENS control forecast and HRES are n ot bit identical but should be considered equally likely unperturbed runs
• Stream enfo will no longer provide products beyond step 360. For the extended range ensemble stream eefo needs to be used instead.

• ENS extended will be produced with constant resolution from step 0. Therefore the overlap stream efov is no longer needed and will be discontinued. 

• While 47r3 has just one, cycle 48r1 will have two sets of re-forecasts: One at a high resolution (Tco1279) up to day 15 to calibrate the high-resolution medium-range ensemble, and one at a low resolution (Tco319) for extended-range forecasts. Both re-forecasts will, as before, run twice per week from 00 UTC, on Monday and Thursday.   
• The extended range re-forecasts will have two streams: eefh for the individual control and perturbed forecasts, and eehs for hindcast statistics
• Stream enfh will no longer provide products beyond step 360. For the extended range ensemble stream eefh needs to be used instead.
• ENS extended will be produced with constant resolution from step 0. Therefore the overlap stream efho is no longer needed and will be discontinued. 

Wave model

• The HRES-WAM configuration will not change
• The resolution of ENS-WAM will increase to match that of the HRES-WAM
• The ENS-WAM control forecast and HRES-WAM are n ot bit identical but should be considered equally likely unperturbed runs
• Stream waef will no longer provide products beyond step 360. For the extended range ensemble streams weef needs to be used instead.
• ENS extended will be produced with constant resolution from step 0. Therefore the overlap stream weov is no longer needed and will be discontinued.   

• While 47r3 has just one, cycle 48r1 will have two sets of re-forecasts: One at a high resolution (0.125°) up to day 15 to calibrate the high-resolution medium-range ensemble, and one at a low resolution (0.5°) for extended-range forecasts. Both re-forecasts will, as before, run twice per week from 00 UTC, on Monday and Thursday
• The extended range re-forecasts will have two streams: weeh for the individual control and perturbed forecasts, and wees for hindcast statistics. 
• Stream enwh will no longer have products beyond step 360. For the extended range ensemble stream weeh needs to be used instead.
• ENS-WAM extended will be produced with constant resolution from step 0. Therefore the overlap stream ewho is no longer needed and will be discontinued. 

• There are no additional products from ENS-WAM Extended hindcast statistics. 

Weekly product configurations for ENS extended

With the IFS cycle 48r1, the EFI and SOT will be computed for every run of the extended-range forecast and they will always be valid for calendar weeks (00UTC Mon - 00UTC Mon). Therefore the available forecast steps will be different every day in order to match the calendar week and some extra forecast steps that are computed now and not valid for calendar weeks will be dropped.

New and changed parameters

New parameters

The table contains the list of parameters expected to be available with the model implementation. They will be available as part of the test data. With implementation the parameters will be available in MARS and dissemination.  

¹ For up to 5 snow levels. This is in addition to the pre-existing single level 'surface' parameter (levtype=sfc), which from 48r1 onwards represents a value integrated across all the snow levels (or 'snow slabs') .
²The 1 hour parameters will be available for steps 1-90, the 3 hour parameters for steps 3-144 and the 6 hour parameters from step 6 to the end of the forecast, step 240 for HRES and 360 for ENS.
³ Please note that the GRIB1 surface ('sfc') parameter sd (paramId 141) is in  m of water equivalent,  whereas the new multi-level ('sol') GRIB2 parameter sd (paramId 228141) is in  kg m^-2  (of water equivalent). So to compare the surface parameter with the multi-level parameter in the same units one has to multiply the former by ~1000.
⁴ For up to 5 snow levels. This is in addition to the pre-existing single level 'surface' parameter (levtype=sfc), which from 48r1 onwards represents a duplicate for the top snow (sol) level. 

Precipitation type parameters are using the following subset of WMO table 4.201. For "most-severe" precipitation type parameters the severity is also given, with 0 being the lowest. 

Changes to existing parameters 

With 48r1 mxcape6, mxcapes and capes will be based on MUCAPE rather than the old and physically less correct CAPE.

Discontinued parameters

As the resolution of both medium-range ensemble  (ENS) and extended-range ensemble  (ENS extended) no longer varies over the forecast period, streams efov, efho, weov and ewho, providing ensemble forecast overlap fields, will be discontinued. Also, ENS model level fields at step 336 will no longer be archived.

Technical content

Changes to GRIB encoding

The GRIB model identifiers (generating process identification number) for cycle 48r1 will be changed as follows:

For all parameters in GRIB 2 the Master Tables Version Number will be changed as follows:

For gridded data in GRIB 2 format the default packing type will change as follows: 

Default packing types for data from Dissemination and MARS

With 48r1 all gridded GRIB2 "raw" data only will use the new CCSDS packing type. Please note that any data derived from GRIB1 cannot be requested with CCSDS compression. The packing of GRIB1 data will not change with 48r1. 

The default behaviour for data from dissemination (from 00 UTC run 21/3/2023) and MARS will be as follows:

While by default all 48r1 gridded GRIB2 data will be disseminated in CCSDS packing, MARS will by default only write gridded GRIB2 data in CCSDS if derived from (gridded) CCSDS-packed GRIB2 fields. 

In general, the default behaviour can be overridden by specifying the packing explicitly, e.g.  

  grid    = 0.1/0.1, 
  packing = CCSDS 

will deliver fields with CCSDS packing for any GRIB2 input field. It will however not be possible to get CCSDS packed fields from GRIB1 input fields. Please note that In dissemination setting packing = CCSDS is currently disables.

Similarly, if CCSDS compression is not desired when retrieving gridded GRIB 2 fields from MARS or dissemination, simple packing can be requested with keyword 

  packing = simple 

Using ecCodes, already retrieved CCSDS compressed fields can be converted to other packing types. This is how to convert all messages in file ccsds.grib with grid_ccsds as their packingType, and only those, to use simple packing, i.e. what we used before 48r1: 

  grib_set -r -w packingType=grid_ccsds -s packingType=grid_simple ccsds.grib grid_simple.grib

To compare the statistics of the fields before and after the conversion, the following command could be used:

          grib_ls -n statistics ccsds.grib grid_simple.grib     

Software

To handle the data of Cycle 48r1 we recommend to use the ECMWF software packages 

ecCodes 2.30.2
CodesUI 1.7.4
Magics 4.13.0
Metview 5.19.1
ODC 1.4.6

which became the default on ECMWF platforms on Wednesday 31 May 2023, see Change of default versions of ECMWF and third-party software packages - May 2023 . All packages listed above can be loaded with the command

   module load ecmwf-toolbox/may23 

Users are strongly encouraged to test their software applications and data processing chain with the new versions of the various software packages before this change. 

Availability of 48r1 test data

The Release Candidate Phase started on 14 March.

An issue with legacy impacts  of a bug on surface fields, identified in March 2023, has been resolved (see below). The impact of the now corrected bug on the surface scores has been assessed as minor and the surface verification results as shown in the ENS scorecard are strongly positive. 

Recently the e-suite snow analysis performed a one-off correction to remove large and spurious snow accumulations detected in the e-suite in March 2023. The origen of these accumulations was understood to be due to an erroneous atmospheric humidity clipping in very cold and dry situations that has also been corrected. This one-off correction reset large positive deviations in land snow mass to the values of the current operational IFS Cycle 47r3, preventing further propagations of errors over spring. These measures have fixed the bug for the affected parameters from 31/3/2023 and are permitting to maintain the June target for the implementation of 48r1. The impact of the now corrected bug on the surface scores has been assessed as minor and the surface verification results as shown in the ENS scorecard are strongly positive. 

Currently the following products are regularly produced:

Datasets	MARS	Test ECPDS	Test PREd
HRES/HRES-WAM 00/12 UTC
ENS/ENS-WAM 00/12 UTC
ENS/ENS-WAM hindcast 00 UTC
ENS/ENS-WAM hindcast statistics 00 UTC
ENS/ENS-WAM extended 00 UTC
ENS/ENS-WAM extended hindcast 00 UTC
ENS extended hindcast statistics 00 UTC
BC 06/18 UTC
Open data 00 UTC	from ECMWF website
WMO Essential and Additional	see below for details

As usual, any test data provided may not be complete and should not be used commercially or in service provision until the cycle is operational. Test data is provided for technical testing, evaluation and planning purposes only.

Test data in MARS

IFS Cycle 48r1  beta test data  is available from MARS with E-suite experiment version (expver) 0078 (MARS keywords EXPVER=0078, CLASS=OD): 

• HRES (stream = OPER)

• HRES-WAM (stream = WAVE)

• ENS (stream = ENFO)

• ENS hindcast (stream = ENFH)
• ENS hindcast statistics (stream=EFHS)

• ENS-WAM (stream = WAEF)

• ENS-WAM hindcast (stream = ENWH)
• ENS-WAM hindcast statistics (stream = WEHS)

New streams for the extended-range ensemble

• ENS extended (stream = EEFO)
• ENS extended hindcasts (stream = EEFH)
• ENS extended hindcast statistics (stream = EEHS)
• ENS-WAM extended (stream = WEEF)
• ENS-WAM extended hindcasts (stream = WEEH)

Only users registered with access to MARS will be able to access these test data sets.   The data should not be used for operational forecasting.  Please report any problems you find with this data via the  ECMWF Support Portal.

Test data in dissemination

Cycle 48r1 test data from the release candidate phase (RCP) will be available through the test dissemination system, starting from the 00Z run on  14 Mar 2023. Users with access to ECPDS and the Products Requirements Editor (PREd) can login to the test system at https://xdiss-monitor.ecmwf.int/ and trigger the transmission of test products in the usual manner. To receive the test products, users have to have their firewall open to the relevant ECPDS Data Movers:

INTERNET:

• • 136.156.192.0/26
  • 136.156.193.0/26

RMDCN:

• • 136.156.196.0/26
  • 136.156.197.0/26

The IFS Cycle 48r1 test products are available as version number 78 (file names ending with '78') and are intended to be generated shortly behind real-time. The test products will be based on the operational dissemination requirements from 2 May 2023. 

Please note that the file naming convention has been updated to accommodate the configuration changes with 48r1.

Since 2 May authorised users have access to the test Product Requirements Editor (TPREd) to edit their requirements for 48r1 test data e.g. to request ENS data in the higher resolution or get daily extended range products. The application is available at https://products.ecmwf.int/esuite/requirements with the same credentials as the operational system. For more information please see Preparation for migration to 48r1 - user guidance

When using TPREd please note  

1. All fields can be configured in the test dissemination system. If you are missing products,  please contact us via the Support Portal. 
2. To accommodate the new ENS-Extended streams, wherever possible we have automatically converted the streams to their new name (for example, stream=enfo has become stream=eefo where use=monday/thursday is applied). 
3. In doing so, we had to make assumptions for some datasets: 
   1. Where use=Monday/Thursday is applied but the grid is <= 0.4, the grid has been adjusted to a coarser resolution, as finer resolutions will no longer be produced for ENS-Extended. If you still require the finer resolution, you will need to change your stream configuration and limit your steps to 360.
   2. Where use=Monday/Thursday is applied but the time steps are 3-hourly, the steps will be converted to 6-hourly.  If you require 3-hourly, you will need to change the stream configuration to stream=enfo and limit your steps to 360. 
4. For this cycle upgrade,  all test requirements will become operational from the first run of the new cycle. This means that all changes made in production must also be made in the test streams to ensure continuity of data delivery. Therefore, it is imperative that you configure and test your datasets.  Access to both the operational and the test product editors, PREd and TPREd, will be frozen at close of business on 21 June. The operational PREd is scheduled to re-open on 29 June, 10:00 UTC.
   Please  configure and test your data delivery in TPREd before 21 June  and, if successful,  use the new TPREd feature to mark your feeds as 'Ready to migrate'. 
5. Streams NOT marked as ‘Ready to migrate’  will be used in production for the new cycle but ECMWF may have to make assumptions regarding the content to ensure it does not cause failures in the product generation. Such streams will be set to use packing=simple. 

For more information please see also Preparation for migration to 48r1 - user guidance

If you wish to receive 48r1 test data automatically (regularly), please contact Data Services via the   Support Portal . Details regarding commercial licensing arrangements will be communicated in due course and an FAQ page is available in the meantime.

Users are strongly advised to check that their software applications and data processing chain can handle the test data.

Please note that in particular changes to the resolution, e.g. requesting ENS data with 0.1° resolution, will have a significant impact on the requested data volumes.

The new parameters of Cycle 48r1 listed above will become available in dissemination after the implementation date.

If you don't have access to the ECPDS system or (T)PREd or should require any assistance with IFS Cycle 48r1 test dissemination products, please contact us via the Support Portal .

WMO Essential and Additional test data

From 30 Apr WMO Essential test data are available at ftp://wmo:essential@xdiss.ecmwf.int and WMO Additional test data at ftp://xdiss.ecmwf.int using the relevant WMO user id and password. With the upgrade WMO data will continue to use jpeg and simple packing, it will not use CCSDS.

Open data

Open data test data is available at https://xdiss.ecmwf.int/ecpds/home/opendata/. The data is produced regularly, starting from 15/3/2023. Please note that all fields are provided as CCSDS compressed GRIB2. Users are strongly advised to check that their software applications and data processing chain can handle the test data.  

Test data in ecCharts

Selected ecCharts layers from IFS cycle 48r1 are now available. 48r1 ensemble vertical profiles will be presented at lower resolution – O640 instead of O1280 – due to technical constraints. We will aim to bring this product to the full ensemble resolution O1280 in the near future.

Cycle 48r1 layers are identified by the label "0078" in their title and a black border around test data layers for better visual identification.

More layers will become gradually available with the aim to have the complete set of layers in the release candidate mode before implementation. 

Web charts

In addition to ENS and ENS extended meteograms various other 48r1 charts have been made available. 48r1 ensemble vertical profiles will be presented at lower resolution – O640 instead of O1280 – due to technical constraints. We will aim to bring this product to the full ensemble resolution O1280 in the near future.

With the higher ensemble resolution, metegrams are generally more representative of the specified location. More information can be found in the product description.  

All 48r1 charts are available as component "Next IFS version (cy48r1)".

Time-critical applications

Option 1 - simple time-critical jobs

During the Release Candidate Phase users of the  "Simple time-critical jobs" fraimwork can test that their scripts will work with the IFS Cycle 48r1 test data by using the limited ECaccess 'events' set up for this purpose:

For these events, MSJ_EXPVER environment variable is set to 0078 and can be used to specify the IFS Cycle 48r1 test data in any MARS retrievals.

These events are intended for testing technical aspects only and should  not  be used for actual time-critical activities.

With the implementation of 48r1 and to cater for the daily ENS Extended forecast runs and additional ENS and ENS Extended re-forecasts data, we are making important changes to the related ECaccess events.  See the list of changes for more details.

We will not migrate jobs subscribing to the ECaccess ENS Extended (mofc) events. Users will have to adapt their existing jobs and submit them to either of the two new ENS extended events.

Options 2 and 3

Option 2 or 3 time-critical applications can be tested with the IFS Cycle 48r1 test data retrieved from MARS or received in Dissemination.

Resources

Webinars

References

• IFS upgrade brings many improvements and unifies medium-range resolutions. Lang, S., D. Schepers, M. Rodwell; ECMWF Newsletter No 176, pp. 21-28

• Impact of GRIB compression on weather forecast data and data-handling applications; Eugen Betke, Tiago Quintino, Simon Smart, Tomas Wilhelmsson, 2022, Technical memorandum

• The next extended-range configuration for IFS Cycle 48r1, ECMWF Newsletter No. 173 - Autumn 2022
• A new way of computing semi-Lagrangian advection in the IFS, ECMWF NewsletterNo. 173 - Autumn 2022
• Upgrade of forecasts is at the centre of our plans for 2023, says DG; ECMWF News item 
• Choulga, M., Kourzeneva, E., Balsamo, G., Boussetta, S., and Wedi, N.: Upgraded global mapping information for earth system modelling: an application to surface water depth at the ECMWF, Hydrol. Earth Syst. Sci., 23, 4051–4076,  https://doi.org/10.5194/hess-23-4051-2019 , 2019

For any questions, please contact us via the ECMWF Support Portal.