Workshop #4, January 21-25, 2002
In order to start up the AOMIP experiments and to discuss possible problems and details of spin-up simulations, a virtual workshop is scheduled during January 21-25, 2002, using an E-mail List. The main questions to be addressed are summarized in the Agenda prepared by Najda STEINER.
What else do we need to finally get started our common forcing runs?
no problems so far with new IBCAO-Bathymetry?!
I also didn't hear any concerns about PHC 2.0 for initial conditions, open boundaries and for restoring. One thing we realized is that PHC seems to be too fresh and too warm in Bering Strait (compared to e.g. Roach et al 1995, JGR) In our case this leads to problems with the Pacific water representation (PW is not dense enough to sink or build up interleaving structures).
III. ATMOSPHERIC FORCING:
1. Last May we agreed upon using OMIP as a spin up. The concern was, we wanted a climatological forcing in order to provide as nearly as possible a common spin-up, but that forcing spinup needed daily variability. That lead to OMIP: We do not yet have any comparable data set from NCEP. More about OMIP (mike):
2. The version 1.0 OMIP climatological daily forcing is derived from 15 years of ECMWF reanalysis. First it's lowpass filtered (about 1 month Gaussian window) and then daily averaged. Then the daily variance from the year 1982 is added back in. The year 1982 is used because it had the largest daily variance of these 15 years, mostly because of El Nino. This is all quite nice, but is this particular year best for our (high latitude) purposes.
3. The international OMIP was born last year at a CLIVAR meeting when the idea of an ocean MIP was raised and the Germans noted that they're already doing it. Frank was appointed the leader for international coordination. So OMIP is not just German at this point, in fact there's many groups from around the world and in the USA who have signed on. Actually, it's called P-OMIP now, for Pilot OMIP.
4. However, they (like some of us) have found the version 1.0 of the forcing data set to be hard to use. Thus just recently the version 2.0 was created. No one has used this yet, however. It has 365 days, but I'm not sure that the strange Arctic fluxes have been fixed. is there more info on this available, the new forcing report is not yet accessible ( Rudiger et al: Can you tell us more about this OMIP version 2.0 forcing data?)
5. Frank agreed that switching from OMIP climatological forcing based on ECMWF to NCEP interannual forcing might be problematic. He recommended using interannual forcing from ECMWF reanalysis, which he said is available for free from NCAR. (I checked the NCAR webpage and found ECMWF reanalyses data from 1979-1993 available, is that what we we are looking for, i.e 50year timeseries...) Did anybody already try to run timeseries with OMIP spinup, i.e Michael?
Nonetheless, here are some questions or comments to the OMIP1 version:
O1: (nadja) PRESSURE is needed for calculations of water vapor pressure and incoming longwave, and even more importantely to calculate windstress geostrophically, but it is not contained in the OMIP1-data set. It is contained in the OMIP2 data set, so the question what to do about that might not arise anymore.
O2: RUNOFF: (mike) The version 1.0 OMIP river runoff data are BAD. They only use "major rivers" and their runoff model (using precipitation fields plus a terrain model) underestimates the runoff quite severely. They recognize this, and somehow take "extra" precip and dump it into the Arctic. In short, this is not a very good data set. Nonetheless, we might feel obligated to use it to match with OMIP. R-ArcticNET is probably a much better data set. (nadja) If we all interpolate river runoff from OMIP-grid to our individual grid we'll probably all have a different annual mean river inflow. Would it make more sense to choose a common data set with seasonal information on the certain rivers? Or would that be a problem because we can't capture ungauged rivers adequately? It seems that it was already agreed upon not using OMIP runoff, but was not yet clear what we would like to use instead. what about the data Matthias Prange prepared, as greg recalled it was talked about this in the meeting last May?
O3: There were simlar discussion about PRECIPITATION and CLOUDS as well (?!) Especially if precipitation has been tuned to cover errors runoff underestimations Moreover PRECIPITATION is said to have some negative values (due to inconsistencies in the use of GRIB, as stated in the Readme file). I didn't find negative values within our model region, I don't know if this is because they are in some other regions, or if they had been masked out before already. I also don't know about that in OMIP2.
O4: (mike) The version 1.0 OMIP upward and downward longwave forcings are mysteriously combined, and they use a different parameterization than e.g. Parkinson and Washington). This might be ok. But then they multiply by a factor (1.045) in order to close their heat and freshwater budgets! (nadja) I had similar problems and ended up using our old calculation with incoming longwave after augstein and koenig-Langlo. OMIP 2 changed to longwave from ERA itself, but the forcing report is not yet available, so we do not yet know what that is. I sent off email to Frank, but there was no response so far.
O5: (sirpa) TOTAL SHORTWAVE, if we use daily shortwave, we would lack interdaily variability (day and night) question: do we need to account for interdaily variability? It might be less important in the Arctic where we have night the whole winter and day the whole summer, but I never checked the difference, in the IOS-model we are currently averaging over 24h.
The question arose: Do we really want to use OMIP?
6. If we decided to abandon OMIP, what is our alternative? We need a strategy -- to be agreed upon! -- what to do about daily variability.
7. (nadja) My main concern is: we decided upon OMIP 8 months ago, and I don't know if there are more than two models that have run with OMIP already. If we now decide upon something else which is even not yet prepared I fear that it will just last forever until something happens. this is probably more a practical concern than a physical, but might be taken into account anyway.
8. AOMIP seems for most of the participants to be one of the rather smaller projects they are working on. So we have to keep the workload for each one as small as possible and I think data preparation is a rather time-expensive job. (we should employ a data manager who could provide the data we need in an easily accessible format and probably also transfers the individual output on a common grid... nice dream, isn't it?)
9. Our model run for 30 years with OMIP1 forcing (still using ncep pressure and different longwave, see above). The results show a really cold Arctic Ocean, cooler shelves and at river mouth higher SSS and lower SST. We are already familiar with problems of lower (too low) heat contents reduced or missing Atlantic Layer...with other windforcing (E.G. NCEP TAU instead of stress from sea level pressure) AWI results show too low heat contents in some areas as well and also cooler shelves.
10. However, at the current stage we may think of using OMIP data despite some concerns (with a few changes, e.g. river runoff). Since most of us are still at a preliminary stage we might swich to OMIP2, with better readable netcdf-files, surface pressure and other longwave formulation. This will also give us a year with 365 days, since we had some concerns with a 360 day year.
IV: In general: How coordinated should we be concerning model parameters? At the moment we are at the stage, where everybody is using whatever he/she had before. Should we be more organized with that, coordinate at least some of them, e.g. albedo?? We could try to define a set of common (coordinated) parameters versus a set of "internal" parameters for each model. Then if any internal parameter is subsequently brought into the common set, we will see the dependence upon that parameter.
Yellow questions from Davids workshop report: ( I listed all, althougho some might be postponed to be discussed at the spring meeting, e.g. data storage, common grid. we should focus on what is essential to perform the runs.)
D1. DO WE WANT TO WORK ONLY WITH A 365-DAY OMIP FORCING? THAT DATA IS NOT YET AVAILABLE BUT ONLY CURRENTLY UNDER DEVELOPMENT AT AWI/MPI.
OMIP 2 is now available with 365 ( see above)
D2. DO WE PREFER TO USE SURFACE PRESSURE TO DERIVE WIND STRESS (RATHER THAN USE WINDSTRESS DIRECTLY)? IN THAT CASE WE HAVE TO AGREE ON HOW SURFACE PRESSURE IS CONVERTED TO WIND STRESS.
(the question may arise for OMIP2 as well) Problem with geostrophically derived stress: basic stress-from-SLP has probs nearing major topography like greenland where important effects of arctic consequence occur.
Different Schemes are available. Thomas, 1999 (JGR, 104, C6, p.13627..) discusses several approaches ( e.g. Ip, 1995, Overland & Colony, 1994 . This might be useful for us to check out. I compared the two schemes Andrey suggested as well as Overland & Colony scheme. The results are showing significant differences, mainly in windstress magnitude leading to ice thickness differences of 1m in the Beaufort Sea ( I will try to provide more accurate information on that). I did not yet manage to do compare with the results with observed ice drift to see which one is more accurate.
D3. DO WE PREFER TO USE THE NET QUANTITY P-E DIRECTLY (RATHER THAN COMPUTE E)?
Evaporation is a response of the model to the forcing, also depending on the ice cover. I would rather say : use P , compute E
D4. DO WE WANT PHC DATA TO BE MADE AVAILABLE IN NETCDF FORMAT?
(might be a minor AOMIP question)
D5. WHAT VOLUME FLUXES FOR DENMARK AND FAROE-SHETLAND PASSAGEWAYS? Note, those are not necessarily covered by all model regions.
D6. WHERE DO WE GET CLIMATOLOGICAL NETCDF FORMAT SEA-ICE CONCENTRATION, THICKNESS, VELOCITY, SNOW COVER, ETC (FOR INTIALIZATION AND VALIDATION OF THE CLIMATOLOGICAL RUN?)
D7. WHERE DO WE GET THE SAME AS INTERANNUAL DATA (FOR VALIDATION OF OUR INTERRANNUAL SIMULATIONS)?
D8. Report: The 30-Year Spinup will be initialized with a climatological sea-ice concentration, a uniform 2.0 m ice thickness cover, no snow cover, no flow velocity, and a -20 ºC surface ice temperature.
question: why no snowcover in January? wouldn't it make more sense to use some reasonable mean distribution? Or maybe everybody runs his/her model for 5 or 10 years starting with zero initial ice and then saves the result to use as ice initial condition for the OMIP spinup?
D9. ACTUALLY, WHICH RIVER-RUNOFF DATA SET TO USE (R-ArcticNET, OMIP, GRDC, AWI)?
D10.WHAT IS THE CURENTLY ACCEPTED ESTIMATE FOR FRESHWATER RUNOFF TO THE ARCTIC OCEAN? WHAT IS THE LITERATURE REFERENCE FOR THAT NUMBER?
D11. WHO WILL CREATE THE RUNOFF CLIMATOLOGY, AND INTERANNUAL DATA? WILL IT BE IN NETCDF? DID AWI (PRANGE) ALSO PREPARE INTERANNUAL DATA
D12. DO WE WANT MERGED IBCAO/ETOPO5 DATA TO BE MADE AVAILABLE IN NETCDF FORMAT? (minor AOMIP question)
D13. DO WE WANT TO ARCHIVE EACH MODEL ON ITS NATIVE-VERTICAL COORDINATE? (mike) I vote for NOT interpolating the vertical grids. My experience with PHC and other data sets is that vertical interpolation from low resolution data can produce spurious water mass properties. Just imagine a model that poorly resolves a T-max. Vertical interpolation will create water masses in T-S space that simply don't exist. Better to keep the original grid.
D14. ARE WE REALLY SERIOUS ABOUT PERFORMING AND STORING THE SHEBA-YEAR DATA? SHOULD THIS BE AN OPTIONAL EXERCISE ONLY FOR THOSE GROUPS SO INTERESTED? (mike) If possible, I would strongly encourage us to save daily output for the SHEBA year.
D15. report: The horizontal coordinate system is in spherical coordinates and is rotated by the Euler angle set (-30, -90, 0). This places the model coordinate system north pole along the real earth?s equator at a longitude of 30 degW. The AOMIP Grid horizontal spacing is 1 degree in latitude and longitude. The AOMIP Grid extends to +/- 30 deg N and +/- 30 deg E (Figure 3.2), and fully covers the geographic domain described above. All AOMIP models compute their basic prognostic quantities on staggered horizontal grids. The AOMIP grid, however, requires that all prognostic quantities be reported at collocated points. This arrangement is often referred to as the A-Grid (Arakawa, 1966).
D16. Do we all agree on that ? 1 degree in lon lat seems quite coarse. most models provide higher resolution. Also, did -30 have a special reason? wouldn't it might more sense to chose some an angle that alignes a little better with the Lomonosov Ridge?