ECMWF model output for the SHEBA column

Chris Bretherton, University of Washington

As a member of the FIRE Science Team, ECMWF generously provided hourly forecast model output for the SHEBA column to the FIRE/SHEBA Science Team, including many non-standard fields (such as radiative fluxes, surface temperature and winds, and vertical distributions of cloud and ice water and precipitation) at the full 31-level vertical resolution of the model. The model output is taken from 12-35 hour forecasts of the ECMWF operational model, which were sent daily to the University of Washington from 22 Oct 97-30 Sept 98. The twice-daily soundings and routine surface observations of pressure, wind, temperature, humidity from the ice camp were assimilated into the model to help initialize each daily forecast cycle. Statistics for Nov 97- Jan 98 suggest that roughly 85% of the soundings reached ECMWF and entered the analysis; click here for a technical discussion of model bias and error statistics at SHEBA over this period. Overall, this discussion suggests that the model wind and temperature fields are quite close to the sounding measurements. However, other fields are generated by the model itself with little direct constraint from SHEBA observations, and MUST NOT BE REGARDED AS OBSERVATIONS. One can see sharp, usually small, jumps at periodic intervals representing the effects of corrected initial conditions on the forecasts.

The model output has been improved (18 Feb 00) from its original version through the considerable efforts of Chris Jakob and lots of cycles of ECMWF's computers, by removing large position errors that existed in the original dataset and also by using the same forecast model version (13R4) to produce the equivalent of a 'reanalysis'. It is now for the grid column nearest to the exact (time-varying) ice camp location, which is at all times less than 50 km away from the ice camp location.

The ECMWF model physics are far too complex to fully describe here. In summary, they include prognostic equations for cloud fraction, liquid water, ice, and precipitation, a convection scheme that has a separate treatment of stratocumulus cloud which only occupy a single grid level that can be active in the summertime Arctic boundary layer, first-order closure for turbulence including stability corrections, a state-of-the-art radiation scheme incorporating cloud effects, and a single layer sea ice scheme in which the thickness and extent of sea ice are specified, but the ice slab temperature interacts with the atmosphere.

We encourage all of you to use and reality-check the ECMWF output vs. SHEBA observations that you may have. Two papers describing aspects of this comparison have been submitted to JGR:

Beesley, T. A., C. S. Bretherton, C. Jakob, E. L Andreas, J. M. Intrieri, and T. A. Uttal, 2000: A comparison of the ECMWF forecast model with observations at SHEBA, J. Geophys. Res., submitted 6/99, revised and accepted 12/99.
Bretherton, C. S., S. R. de Roode, and C. Jakob, 2000: A comparison of the ECMWF forecast model with observations over the annual cycle at SHEBA. J. Geophys. Res., submitted 12/99.

If you make use of this dataset, please acknowledge ECMWF as follows: 'We thank Christian Jakob and his coworkers at ECMWF for producing the ECMWF column dataset for SHEBA.' Here, we describe what model fields are being archived here, we provide a suite of plots which summarize the model output, and we describe how FIRE/SHEBA Science Team members can access it. My contribution to this work has been sponsored by NASA as part of the FIRE program under grants NAG1-1711 and NAG1-2072. Chris Jakob and Martin Miller of ECMWF have been instrumental in making this extraordinary suite of model outputs available to our community. Chris Jakob has an even more extensive set of model diagnostics archived at ECMWF, so if you do not see what you need, you might want to email him at paj@ecmwf.int .

Model Fields Archived

Basic fields

pressure (Pa)
zonal wind component (m/s)
meridional wind component (m/s)
temperature (K)
specific humidity (kg/kg)
specific cloud liquid water content (kg/kg)
specific cloud ice content (kg/kg)
cloud fraction (percent/100.)
relative humidity (percent/100.)
omega=vertical velocity in pressure coordinates (Pa/s)

Fluxes

net shortwave flux (W/m2)
net longwave flux (W/m2)
sensible heat flux (W/m2)
turbulent moisture flux (kg/kg * kg/(m2*s))
x turbulent momentum flux (= x wind stress, at bottom level) (kg/(m-s))
y turbulent momentum flux (= y wind stress, at bottom level) (kg/(m-s))
turbulent moisture flux (kg/kg * kg/(m2*s))
convective rain flux (kg/(m2*s) = mm/s)
convective snow flux (mm/s)
large-scale rain flux (mm/s)
large-scale snow flux (mm/s)

Tendencies

dudt_total (m/s2)
dudt_nonadiabatic (m/s2)
dvdt_total (m/s2)
dvdt_nonadiabatic (m/s2)
dTdt_total (K/s)
dTdt_nonadiabatic (K/s)
dqdt_total (1/s)
dqdt_nonadiabatic (1/s)

Surface variables

surface pressure (Pa)
2-m temperature (K)
2-m specific humidity (kg/kg)
10-m wind u component (m/s)
10-m wind v component (m/s)
skin temperature (K)
surface roughness length (m)
surface roughness length for heat (m)
surface albedo (%/100.)
downward surface solar radiation (W/m2)
downward surface thermal radiation (W/m2)

Plots of ECMWF output

Based on ECMWF output starting 97/10/22

Accessing the ECMWF model output

The ECMWF model output for each day is stored as a compressed unix tar file SHEBA_yymmdd.tar.gz. Here yy = 97 or 98 is the year, mm = 01-12 is the month, and dd is the day. All of these files can be downloaded from the anonymous ftp directory pub/breth/SHEBA on eos.atmos.washington.edu. The help file readme_ARM_ddh_sheba in this directory explains what to do with these daily files and what output is included.

I have also accumulated this output in a netcdf file for the entire period 22 Oct 97-30 Sep 98, which is a considerably more convenient way to access the data, especially if you use a graphics program that has a good interface to netcdf (such as matlab or NCAR graphics). The plots above were created with matlab using the netcdf file. The advantage of this format is that it is very easy to extract the subset of data that you wish to look at without having to parse the rest, and that the format is self-documenting so that it is easy to figure out what data is included and by what names. This netcdf dataset (60 MB!) is for use by any FIRE/SHEBA Science Team member.

Chris Bretherton <breth@atmos.washington.edu>