contact  |  support  |  links
There will continue to be three levels to the TransCom 3 experiment with central coordination performed by CSU. The first level will be an annual mean inversion, the second will include seasonality (using 12 one month pulses), and the third will incorporate different inversion methods into a comparison.
Basis function map
It was agreed to make some changes to the existing basis function map. Africa will be split at the equator to account for the different vegetation seasonality. This will add one basis region. The Tropical Pacific will be split at 160 W, adding one basis region. The Southern Intermediate Ocean and the Southern Ocean will be merged into one basis region. This will remove a basis region. This leaves 22 basis regions. The N=2 smoothing case seems adequate (see the document that explained the boundary construction and smoothing).
Some minor adjustments
It was suggested that the northern coast of Africa be added to the Europe basis region. This small amount of northern Africa will get eliminated during the smoothing routine used. Unless, it is felt that this is important, we recommend that this be left as it was originally. The Polar oceans should include ice masking. This issue needs further discussion as there is quite a bit of seasonality in the ice cover of the Polar oceans - one possibility is to have 12 basis function maps reflecting the month-by-month changes in Polar ocean ice cover. The split between the Southern Indian Temperate Ocean and the South Pacific Ocean will be moved to run down the center of Tasmania.
Pre-subtracted fossil-fuel emissions
It was agreed that there would be no seasonal or interannual variability included in the pre-subtraction of the fossil-fuel emissions input field. We will use a 0.5 x 0.5 degree resolution map. The two most recent years available are 1990 (Andres et al 1997) and 1995 (Brenkert 1998) which exhibit some southward migration of emissions. Gurney will investigate how significant these differences are. It may be interesting to check the sensitivity of the results to these two fossil-fuel pre-subtraction fields by running a four year pre-subtraction centered on 1990 and 1995, respectively.
Pre-subtracted neutral biosphere carbon exchange
It was agreed to use a satellite-driven CASA NEP map. No interannual variability - just an annually repeating biosphere exchange map. These input data will be supplied as 12 monthly maps.
Pre-subtracted oceanic carbon exchange
It was agreed to perform a pre-subtraction of the oceanic fluxes. This will be accomplished with a forward run with Takahashi et al 1997/W map. No interannual variability - just an annually repeating ocean flux map. These input data will be supplied as 12 monthly maps. The ocean data reflects "climatological" (1990-era) oceanic exchange.
Spatial pattern of the oceanic carbon exchange within basis regions
It was agreed to use a satellite-driven CASA NPP map for the spatial pattern. Jim Randerson will test the viability of an NPP x tc (NPP weighted by the residence time of carbon) to see if this is meaningfully different from a straight NPP map. It was agreed that there will be no seasonality to the spatial pattern.
SF6 forward runs
There will be one forward run of SF6 for each of the 11 terrestrial basis regions. An emissions map with time variation will be supplied.
Observational data
It was agreed that almost all questions about observational data can be put off until after the forward runs are performed. For example, which stations, whether to use extrapolated data, what subset of GlobalView, whether to avail of aircraft data, etc.?
The only issue that comes up prior to the forward runs is at what locations would we want high-frequency CO2 concentration output. These are yet to be determined (more on this below).
Analyzed winds
It was agreed that those using off-line models (Heimann taking the lead) would propose an appropriate year for the analyzed winds used (this year of winds would be repeated for the multiple year forward runs). If the spatial pattern of fossil-fuel emissions turns out to be significantly different between 1990 and 1995, it may be best to choose a year of analyzed winds that is centered on either of these years. Since many groups may be using a single model (TM2), some sensitivity to this issue could be explored by having different groups use different analyzed wind time periods (assuming at least one group runs the same year as other non-TM2 off-line models).

Editorial Thought
TransCom 3 is being constructed to account for errors in carbon budget inversions due to transport and inversion methodology. The aim is not to generate a consensus 'answer' on the missing sink. For this reason, it is not absolutely necessary to require that the pre-subtraction fields, the analyzed winds, and the observational data all conform to the same time period. What matters is that all the modelers use the same datasets.
However, to the extent that it is possible, it may prove prudent to generate as much temporal conformance between these different input/analysis datasets as possible. In other words, if we can, we might as well.
For example, using a fossil-fuel pre-subtraction run with a spatial pattern representative of the mid-1990s, analyzed winds from the late-1980s, and observational data representative of the mid-1980s might be best to avoid if possible because, while fine for intercomparison, would miss the opportunity to weigh in on the sink debate. Generating an average, physically-based estimate of carbon sources and sinks would be a useful addition to the generation of carbon budget errors due to model and inversion methodology. This should not be viewed as a top priority but, if it can be done, it seems prudent to do.
Output datasets
  • Modelers will use NETCDF for input/output. The central coordinator will generate a 'manual' on how to use NETCDF in the context of TransCom, and provide sample code and documentation, binary libraries, and any other details appropriate to facilitate this.
  • Model output will be submitted on standard pressure levels (not sigma or other model coordinate levels). Gurney will propose a set of standard levels.
  • Modelers would report 3D fields for u, v, w (plus the 3D concentration fields).
  • The central coordinator would subsample the submitted concentration fields for those grid cells with monitoring stations so that smaller files can be generated for alternative analysis by others.
  • Modelers would report high-frequency concentration and wind components at a chosen set of locations. Conditional subsampling to match the sampling decisions at the monitoring locations can then be performed. Heimann and Ciais will propose locations with a mind to past, present, and future future flask locations, towers, and profiles plus "upwind" grid cells, as appropriate.
  • Inversion
    It was agreed that inversion sensitivity should be tested. This, however, is a post-forward run activity so decisions about this can be put off.
    What modelers do for level I
    (Not necessary if level II is completed)
    (back to top)
    Forward runs (each integrated for four years)
  • pre-subtracted fossil-fuel emissions (maybe separate 1990 and 1995 runs)
  • pre-subtracted terrestrial exchange
  • pre-subtracted oceanic exchange
  • 11 SF6 basis functions
  • 22 CO2 basis functions
  • Output
  • Monthly mean 3D mixing ratio of the three pre-subtracted forward runs from the last simulation year. This represents 36 (3 CO2 x 12 months) 3D fields ~ 7 MB of data (or 9 MB should we want to run both a 1990 and 1995 fossil-fuel pattern).
  • Monthly mean 3D fields of SF6 mixing ratio from the last simulation year for each of 11 terrestrial basis regions. This represents 132 (11 regions x 12 months) 3D fields ~ 25 MB of data.
  • Monthly mean 3D fields of CO2 mixing ratio for each basis region from the last simulation year. This represents 264 (22 regions x 12 months) 3D fields ~ 46 MB of data.
  • Monthly mean 3D fields of u, v, and w for the last simulation year. This represents 36 (3 winds x 12 months) 3D fields ~ 7 MB of data.
  • High frequency station location reporting for surface mixing ratio, u, and v. Example: 100 locations reporting every hour for 48 months = 100 locations x 768 hours/month x 48 months x 3 variables x 4 bytes/variable ~ 44 MB.
  • What modelers do for level II
    (this assumes level I was not run)
    (back to top)
    Forward runs:
  • Run 4 years of the pre-subtracted fossil-fuel emissions (maybe separate 1990 and 1995 runs)
  • Run 4 years of the pre-subtracted terrestrial exchange
  • Run 4 years of the pre-subtracted oceanic exchange
  • Run 4 years of the 11 SF6 basis functions
  • Run the 12 monthly CO2 basis function pulses for each region, and integrate transport for 3 years - 264 basis functions
  • Output
  • Monthly mean 3D fields of mixing ratio for the three pre-subtracted forward runs from the last simulation year. This represents 36 (3 CO2 x 12 months) 3D fields ~ 7 MB of data (or 9 MB should we want to run both a 1990 and 1995 fossil-fuel pattern).
  • Monthly mean 3D fields of SF6 mixing ratio ratio from the last simulation year for each of 11 terrestrial basis regions. This represents 132 (11 regions x 12 months) 3D fields ~ 25 MB of data.
  • Monthly mean 3D fields of CO2 mixing ratio from the pulsed runs. This represents 792 (22 regions x 36 months) 3D fields ~ 150 MB of data.
  • Monthly mean 3D fields of u, v, and w for all of the forward runs. This represents 108 (3 winds x 36 months) 3D fields ~ 21 MB of data {off-line models repeating one year of analyzed winds only need to report 36 (3 variables x 12 months) 3D fields}.
  • High frequency mixing ratio , u, and v at station locations. Example: 100 locations reporting every hour for 36 months = 100 locations x 768 hours/month x 36 months x 3 variables x 4 bytes/variable ~ 33 MB.
  • Provide the central coordinator with individual inversion results for collation and comparison.
  • Provide basis function map
  • Provide pre-subtracted fossil fuel input map
  • Provide pre-subtracted neutral biosphere map
  • Provide pre-subtracted oceanic exchange map
  • Provide maps of the spatial pattern for the terrestrial basis functions
  • Provide emissions maps for the SF6 forward runs
  • Provide information and assistance on using NETCDF for input/output
  • Make all model output accessible to all participants
  • Provide subsampled concentration fields from all model output and make accessible to all participants
  • Generate and maintain web page with complete experimental protocol, input and output data, progress reports, timelines, etc.
  • Jim Randerson - examine how sensitive the spatial pattern of the terrestrial exchange within basis regions is to the use of NPP x tc as opposed to simply NPP.
    Martin Heimann - lead a discussion about the choice of time period for the analyzed winds. With Phillippe Ciais, propose a list of potential locations for this to report high-frequency output. Analyze high-frequency output?
    Phillippe Ciais - with Martin Heimann , propose a list of potential locations for this to report high-frequency output.
    Rayner/Enting; Gloor - perform inversions on model output.
     
    site design by Micek Design
    All Images on this Website are the property of the stock.xchng, all rights reserved.
    Copyright 2011, Arizona State University all rights reserved.