Meeting Minutes, DOE Project, Boulder, Colorado, May 15-16

Meeting Minutes, DOE Project, Boulder, Colorado, May 15-16

Present

Bill Gutowski, Iowa State

Wieslaw Maslowski, NPS

John Cassano, U. Colorado

Jaromir Jakacki, Institute of Oceanoglogy

Anton Kulchitsky, ARSC

Mark Seefeldt, U. Colorado

Justin Glisan, Iowa State

Dennis Lettenmaier, U. Washington

Chunmei Zhu, U. Washington

Jackie Kinney, NPS

Juanxiong He, IARC & ARSC

Gabriele Jost, NPS

Andrew Roberts, IARC & ARSC

Thursday, May 15

John Cassano – Work on identifying appropriate physics from WRF

Summary:

-WRF/MM5 Comparisons provided

-Problem identified on the ability of the models to simulate liquid clouds at low temperatures that affects downward longwave fluxes significantly. There is no immediate fix for this.

Bill Gutowski - Work toward understanding high resolution model response

Summary:

-Evaluation of atmospheric model behavior, looking at links between atmosphere & surface that influence circulation and error propagation.

-Summary of work by Justin Glisan and Brandon Fisel.

-RMS Differences: Pan-arctic MM5 vs. ERA-40

-There is a particular interest in understanding the behavior of the real and simulated atmosphere during periods of divergence and convergence in the pack.

-Effect of Permafrost Changes shown

Juanxiong He – IARC & ARSC report - effort to combine CPL6 with WRF

Summary:

-Working on allowing CPL6 to have a different land and atmospheric model resolution.

-Highlighted differences between CCSM atmospheric model coupling and WRF Coupling.

-Provided algorithms for the parallel execution of WRF coupled to CPL6

-Change in time format to allow decimal seconds so that model coupling is correctly synchronized.

-Specific coupling differences provided, including spectral (CCSM) verses bulk (WRF) albedo for coupling CCSM as against coupling WRF.

-Should CPL6 or CPL7 be used?

Discussion:

Bill – There is no particular need for the land model to have a lower resolution than the atmospheric models. Allowing different atmospheric and land model grids is a flexibility issue that Dennis and Bill suggest is not necessary at the moment, and may in fact introduce errors.

Wieslaw – We should move to CPL7 because CPL6 is being abandoned.

Dennis – How different is CPL6 from CPL7?

Wieslaw – The software has changed, but what the coupler requires and how it requires it is probably the same between CPL7 and CPL6. CPL7 is a major upgrade, and he thinks this should be adopted to keep us ahead of the curve. More discussion later.

Bill - Suggested that we need to define the variables needing to be passed to the coupler.

Group discussion on what couplers pass to one another

Summary:

-The input/output list from CPL6 documentation was worked through to look for potential problems.

Atmosphere

John - How do we split visible and infrared albedo? It is fine to pass information as arguments that do not fit the CPL definition, but these must be clearly defined in the group to avoid errors.

Bill – Rough breakdown could be made between different albedo wavelengths. Each of the models should have code to alter what passed through the coupler if it differs from what the component model uses. The best policy is likely to be keeping the coupler untouched.

Dennis – The only problem he sees with the land/atmosphere interface is with the albedo passing.

Cassano – We definitely need to make sure that what the atmosphere sees albedo as it is also seen by the land model to ensure that fluxes are conserved. Also need to be careful of spectral representation of the shortwave down.

He – Has found a radiation scheme incorporated into WRF that uses four types of albedo.

Wielsaw – Variables being passed need to be tabulated.

John - Need to be sure that evaporation is being passed from the ice/ocean to the atmosphere.

Mark - Need to be careful because the coupler calculates the ocean-atmosphere fluxes itself, which uses specific bulk formulations which may be different from WRF.

Ice/Ocean

Wieslaw - Spectral albedo is also questionable – it is not used by CICE as it is currently set up in Wieslaw’s model.

John - Need to be sure that sea ice thermodynamics model is passing evaporation to the coupler for the atmosphere.

Land

Dennis – why is snow depth passed? This is not the most reliable snow cover variable to pass since it is based on parameterizations. Water equivalent is better.

John – Would rather the snow is seen as snow-water equivalent, providing this is well documented so both the land and atmosphere are ‘talking the same language’.

Dennis – define what is a coastal grid cell. A coastal cell will accumulate whatever is upstream down to the coastal grid cell.

Wieslaw – the land model will initially be operating on a lower resolution grid than the land model.

*Outcome: Each groups needs to document the variables and units being fed into‘data port’ of the coupler so that other models extract data correctly. Particular care must be given to spectral albedos and radiation to make sure the sent and received data is consistent with each model’s physics and consistent between models.

Dennis Lettenmaier – Land scheme report

Summary:

-VIC model has been heavily used for offline simulations, but seldom is it coupled to other models. The resolution of offline simulations has largely been dictated by data sparseness. The model runs at daily or, when snow is present, sub-daily timesteps. There was a substantial upgrade in version 4.04. VIC has been coupled to MM5.

-VIC only runs on a single processor. One year for the whole grid runs ‘fast’ but no exact number was given. There is no MPI coding in VIC and so running VIC through the coupler would need to only use one node, if not only one CPU. This might be an issue, and needs to be checked in case VIC becomes a bottleneck.

-Debugging of VIC coupled to CPL6 is underway, with complications because CPL6 required CCSM to be compiled, even though data was being run through the Coupler, not CCSM component models.

Discussion:

Wieslaw – How often does the atmosphere need information from VIC?

Bill – NOAH in WRF is interacting every timestep, however an hourly timestep is probably adequate.

Dennis – Does the atmosphere run before the land or after the land? This may be an issue.

Wieslaw – Would prefer the atmosphere to run first.

Chunmei – CLM is being run at the same timestep as CAM.

Dennis – It is an unintelligent use of time to run the model every timestep, and different answers will be output if the model is run every timestep than for every day.

Wieslaw- Suggests porting the code immediately to ARSC for debugging rather than running on the UW cluster?

*Outcome: It is probably best to port VIC immediately to ARSC and debug CPL/VIC there, and to gain a good understanding of whether or not VIC run times could cause a bottleneck by not executing in parallel.

Andrew Roberts and Anton Kulchitsky – Computing environment & possible coupler noise.

Summary:

-summary of directory structure for code and data sharing at ARSC, summarized at:

http://data.arsc.edu/asm.

-presented some considerations for the ice-ocean/atmospheric coupling in frequency of models talking to each other through the coupler.

-presented some ideas on potential problems in using atmosphere and ice/ocean grids with consistent, significantly different spatial resolution

Friday, May 16

Wieslaw Maslowski – DOE Progress Report on Modeling Oceanic and Sea Ice Variability in The Arctic Ocean, Comparisons between CCSM and NAME.

Summary:

-IPCC models are not projecting the same rapid decline as is being observed in Arctic Sea Ice

-Heat Fluxes through oceanic passages into the Arctic Ocean has an influence on sea ice thickness

-Changes in modal thickness of model mean sea ice thickness 1979-2003 means more thin ice is being exported. Mean thickness down to ~2.5m Basin-wide.

-Differences highlighted between CCSM and the NAME (NPS) model.

-CCSM is probably exporting too much ice through Fram Strait, and the oceanic heat flux is also probably too small in CCSM (considerably less than the NAME model).

-Volumes of water influx to Barent Sea similar between CCSM and NAME, however volume transport is less in the Barents Sea which may be a product of less wind stress in CCSM relative to what actually occurs.

-Summary table of volume and heat transport provided for NAME relative to CCSM.

-Conclusions on why there may be differences between CCSM and ‘reality’.

-Currently working on using ECMWF TOGA 2005-2008 data for updated work to go into a paper.

-Runtimes: About 60-70 hours running 512-576 Cores to complete 1 year of the POPCICE model.

Discussion:

John - North Atlantic Storm Track is too active in CCSM which can contribute to the distribution of sea ice on the Greenland coast in CCSM that is not evident in satellite retrievals nor in the NAME model.

Andrew – noted that ARSC is obtaining a Cray XT5 which may be coming online as early as October. Using the PGI compiler may be a good policy on Midnight to improve portability of the code to the new Cray. In addition, ARSC has access to another Cray XT5 if anybody is interested in running test cases on the Cray before that time.

Jaromir Jakacki – Combining CPL7/CCSM4

Summary:

-Differences highlighted between CPL5 and CPL7, where CPL7 can have many executables.

-CPL5 Directory structure provided.

-Plan for next step using CPL7: Run CPL7 with 50km and 9km data components.

-Several files provided for mapping that are needed for building that have been provided as ‘empty’ files but it is not known exactly what these files do and whether or not it is important that these files have ‘meaningful’ content.

-A PIPS3 server has been set up at http://hydrodata.iopan.gda.pl/pips3, password is ‘popcice’. This is a live server for getting output from the model. Data cannot be downloaded, but the model output can be explored graphically. Currently only ocean results are set up, but sea ice will be added in future.

Discussion:

John – How often are we going to test CPL7 with model data? Will we provide data, at what frequency and for how long from WRF for testing CPL7?

Maslowski – Communicate between coupler and models every 6 hours for the present. Run for about 1 month of data for testing I/O from the coupler. Preferred format for data to be provided from models us NetCDF.

Maslowski – A manual will be set up to help the group in using CPL7 once Gabriel and Jaromir have done initial work in setting it up on Midnight.

Jackie Kinney – NPS Website

Summary:

-RACM website at http://www.oc.nps.edu/NAME/name.html, password is ‘RACM’. Click on ‘submit’. Location for meeting, presentations, comments and discussions is at this website.

Discussion:

Andrew – The ASM google group is also a good way for communicating

Bill – The website is a good thing, especially for locating meeting minutes, presentations and model work. Perhaps a mailing list is easier at times for forums.

Group discussion on plans for the next 6 months.

John -A decision needs to be made on whether or not we will use CPL6 or CPL7.

Has concerns that CPL7 is somewhat of a black box without documentation.

Bill – We really need a working manual for CPL7 to make this viable.

Dennis – Suggests a conference call in a month to decide whether or not we go with CPL6 or CPL7.

Wieslaw – This is a good idea, lets have the teleconference at the end of June. Who will involved: PIs, Newby, Juanxiong, Jaromir, and anybody else who is involved in coupling. June 16/17 is a good time. All involved should send a reply email to Jackie and telling when they are NOT available of June 16/17.

*Outcome: Teleconference on June 16/17 with Jackie to email all PIs, Newby, Juanxiong, Jaromir and Gabriel to find the most convenient time on 16/17th of June for deciding whether or not to proceed in using CPL6 or CPL7.

Wielsaw – It is doable to have component models working with CPL within 6 months, and this would be good progress for the group.

John – Lets decide on what time periods we want data for from the models to run a test data CPL6/7 case.

Wieslaw - Suggested dates for data to feed to the coupler are September 1993, 6-hourly output. We need the list of data required for each model in the correct units and format.

Mark – 1997 would be better from the WRF side.

Dennis – Needs to regrid data to supply to the coupler for the data run.

Wieslaw - NPS will provide information on what is needed for the data-model run using CPL7 within the next week.

John - WRF/Land Group will decide how they will supply the data (on what grid etc.).

*Outcome: Groups to supply data to NPS group for running coupler with data September 1998 for test data at 6-hourly intervals, preferably in NetCDF format.

Greg – Wants to define who at ARSC is doing what on the project. Wants to define deliverables for ARSC.

John – From the atmospheric perspective, it would be good to have greater interaction between his group and ARSC (and Bill’s group).

Wieslaw – Need to integrate He and other atmospheric groups to work toward common goal.

Greg – The repositories will force groups to work together and should be set up as soon as possible.