PMIP 2 Atmosphere Variables

Note: the list of required variables is not stable yet. You will find in the tables below the variables that are likely to be found in the database, and the variables that were suggested as useful but did not make it to the final list.

Required variables

Atmosphere variables required by specific studies, but NOT SAVED IN THE PMIP 2 DB

These variables require a lot of storage space and will not be stored in the PMIP 2 database, at least not at the beginning of the project. Each group should tell if they will be able to provide this data for about YY years of experience, and keep them in their lab until the data is needed.

High frequency atmosphere variables required by the Dust Model

The variables required at sub-daily timesteps are needed by an offline dust model to validate MPI-BGC LGM experiments.

The required atmosphere variables are: orog, ts, pr, prc, cl, clt, snc, uvas (this name, for surface wind speed, may change), ua, va, ps, evs, tas, ta, hus and FPAR.

The required vegetation variable is: mrso.

For the sub-daily values, indicated frequencies are the acceptable frequencies, the ideal frequencies being twice the accepted ones (e.g. when 6-hourly is acceptable and will work, 3-hourly would be ideal).

The fields should be saved for at least 10 model years, for OA and OAV simulations it might be better to have 20 years. To avoid getting data for some decadal-type anomaly, we could randomly sample the model years during the analytical period.

The following table lists the variables that either are not required at all at DAily frequency in the PMIP 2 database, or are not required at a high enough frequency (sub-DAily) for the dust model.

Note that the dust model also requires one variable from the vegetation variables: mrso.

Please get in touch with Sandy Harrison if you want more details about this!

High frequency atmosphere variables required by the NCAR Regional Model

The following table lists the variables that are required at sub-DAily frequency (at least two times daily : noon and midnight) by the NCAR RegCM regional climate model. The orography is also required, but it can be found in the regular variables of the PMIP 2 database.

Note that the RegCM model also requires a variables from the ocean variables: tos.

Please get in touch with Noah Diffenbaugh if you want more details about this!

Rejected variables

Notes

Michel Crucifix

At the Hadley Centre, we are not in a position to provide Potential Evapotranspiration for the time being. It is necessary to know whether this is an absolutely necessary variable for Sandy's model.

To be discussed:

• Soil moisture and temperature: is it over the first meter, the first 5 meters? First model layer (which will then depend on the model).
• Depth to soil thaw may pose a definition problem. In MOSES II, for example, a fraction of water is allowed to be frozen at each soil level. A suggestion: depth to 50% frozen water as an alternative.
• Surface air temperature: I propose 1.5 m as more relevant and internationnally accepted.
• Shall we include clear-sky radiative fluxes?
• I propose to prefer surface moisture flux to latent heat flux, in order to avoid the ambiguity related to snow melt and/or sublimation. (the latent heat provided to the atmosphere is the enthalpy of evaporation, only; but the soil had to provide the heat for fusion and evaporation).
• Should we decompose the vertically integrated transports into transports layer by layer (as was done with AMIP)?

XYZT fields in sub-daily mode will take a lot of place. I propose to provide them for the last year, only. A climatological year wouldn't be helpful because this would smooth out the extreme events.

Also, I would urge people to review the need for daily outputs (this may be useful for some studies, e.g. storm tracks). Can we decide to provide the daily outputs of all suggested fields for the last 5 years, only?

Rough estimate:

X * Y * Z = in HadAM3 about 131 K
one data = 2 bytes ==> 1 3-D field for one timestep = 262 Kb
6 3-D fields for one time step = 1.5 Mb

1 year at 3-hourly frequency = 360 * 8 = 2880 timestep.

2880 timesteps * 10 Mb = 4.3 Gb.... for one model

20 model years = almost 84 Gb my model

Do we have that much space? 20 years at daily frequency = 10.5 Gb

20 years at monthly frequency = 360 Mb

20 years sat seasonal frequency = 72 Mb

20 years at annual frequency = 30 Mb