ECOMIP

Experimental protocol

Simulations

Two simulation types are envisaged:

1. 48-hour forecasts initialized at midnight every day between 00 UTC on 9 August 2024 and 00 UTC on 29 September (52 start dates). Comparison with EarthCARE and between models would be carried out on the second day of each forecast.  Optionally, to evaluate the degradation of forecasts as a function of lead time, longer simulations may be carried out but still initialized each day. Operational NWP centres may wish to simply provide their operational forecasts for the period.

2. A free-running simulation initialized at 00 UTC on 1 August 2024 and running until the end of September.

Since high resolution simulations are computationally intensive, some models may not be able to simulate the entire period, in which case the dates corresponding to the most interesting ORCESTRA case studies should be prioritized. The 18 dates with HALO aircraft flights are listed here. The most interesting of these in terms of deep convection are the following:

• 18 August

• 22 August

• 27 August

• 3 September: a golden case in terms of observations from EarthCARE, the HALO aircraft and the Meteor ship

• 19 September

Note that the simulations themselves should be initialized at the beginning of the previous day.

Note also that no CPR data were taken on 29 August and 22 September, and the CPR collection was incomplete for 28 and 30 August, and 2, 3, 12, 20, 21 and 23 September.

Initial data

Ideally the simulations would be initialized from ECMWF's operational 9-km analyses, which may be obtained by making an archive order here. Note that research access is free of charge - if this is questioned, please contact Robin Hogan. A second option is to initialize from ECMWF's ERA5 reanalysis, which is also freely available. Generally this is an inferior option as it uses an older version of the model and is at lower resolution. If your model has its own analysis system or is normally initialized from different analysis dataset to the above, then if easier, this may be used instead.

Output data

Two output datasets are envisaged from each model:

1. Model slices corresponding to each EarthCARE "frame", which is an eighth of an orbit. The Mini JSG files here define the times and locations of each frame in the ECOMIP period. The model data should be taken from the closest archiving time of the simulation and interpolated to the longitude and latitude grid from the JSG file, which is around 1-km resolution. The models own levels should be used in the vertical. One file per frame should be produced.

2. Three-dimensional model fields extracted over the Tropical Atlantic domain of ORCESTRA, bounded in longitude between 64°W and 8°W, and latitude between 4°S and 24°N. Ideally the output frequency would be 30 minutes for the 3D fields and 10 minutes for the 1D fields, but it is recognised that some models (and certainly archives of operational forecasts) are tricky to archive more frequently than every hour. Focus should be placed on the most interesting ORCESTRA case studies listed above, but storing the entire 24 hour period starting at 00 UTC. Output from ICON-MPI at 1.25 km for this region can be visualized here.

Optionally, the full global dataset could be shared, which would provide a dataset similar to previous global storm-resolving model intercomparisons such as DYAMOND. Such a dataset is particularly requested by JAXA and would benefit the satellite community, who could use it for future applications of satellite projects.

The output fields should ideally include:

•  Meteorological variables on model levels: height above the surface, pressure, temperature, specific humidity, and prognostic variables for all hydrometeor species, e.g. the mixing ratios of cloud liquid, cloud ice, rain and snow. If available also store cloud fraction, precipitation fraction, gridbox-mean vertical velocity and convective mass flux. 

• Surface variables: surface elevation, skin temperature, rain rate, snowfall rate (split into convective and large-scale, if available)

• Top-of-atmosphere variables: upwelling and downwelling shortwave and longwave radiative fluxes

• Aerosol variables on model levels: for models with prognostic aerosols, they should output the aerosol prognostic variables, e.g. the mixing ratio of each species. If available, additional diagnostics could facilitate comparison with EarthCARE, for example the 355-nm extinction and backscatter coefficients.

The output format should be netCDF (either netCDF3 or netCDF4/HDF5). While we are not prescriptive on variable names, dimension orders etc, some sample netCDF files from ECMWF's Integrated Forecasting System may be found here.

Data analysis

Once the model data are shared, participants may intercompare models and evaluate using EarthCARE as they see fit, although care needs to be taken to ensure the correct microphysical interpretation of the model fields, and indeed to the strengths and weaknesses of the various EarthCARE products. Generally speaking we may compare model fields against retrievals of the same quantities, as illustrated in the preliminary model evaluation shown here, or compare in observation space by "forward modelling" the observations from model fields. An important part of ECOMIP will need to be an intercomparison of the various satellite simulators in order to understand their differences. This is challenging as they need to make assumptions about the particle size distributions and particle shapes and densities, as well as sub-grid cloud structure. These assumptions are typically different in each model. Satellite simulators used by the individual ECOMIP participants are:

• The Joint Simulator for Satellite Sensors

• PAMTRA: Passive and active microwave radiative transfer tool

• ECSIM: used to simulate EarthCARE observations for the pre-launch special issue of AMT

• RTTOV: used for simulating microwave, infrared and most recently solar radiances for data assimilation, used by European NWP centres. A radar simulator is under development.

• COSP: used to simulate radar, lidar and radiometer data targeted at low-resolution climate models used in IPCC.

• JEDI/CRTM: used to simulate radiances for data assimilation.

• ZMVar: used by ECMWF for assimilating radar and lidar backscatter profiles.

• The Spaceborne Radar Simulator (SR-SIM)