Attached are the long-awaited minutes of our February meeting. Much was covered, and I did my best to record it faithfully. Jim Orr has been through the minutes and has put his seal of approval on it all--except for how C-14 should be simulated in OCMIP. Ken Caldeira has made some figures pertinent to this and I will make them available to you shortly, at which time I suggest we discuss the issue in more detail.
In attendance: Jorge Sarmiento, Reiner Schlitzer, Roger Hanson, Anne Mouchet, Ken Caldeira, Mick Follows, Helge Drange, Patrick Monfray, Jean-Claude Dutay, Ferial Loanchi, Chris Sabine, Nicholas Gruber, Yasu Yamanaka, Fortunat Joos, Ian Totterdell, Matthew Hecht, Martin Heimann, John Bullister, Rick Murnane, Phil Duffy, James Orr, Raymond Najjar
We had several brief (thought not as brief as planned!) presentations by several OCMIP participants, interweaved and followed by much discussion.
Jim Orr opened the meeting by stating that OCMIP is an open international project. Any modeling group can join and have access to OCMIP protocols (subroutines, forcing functions, etc.), but there may be an upper limit as to how many models can actually be analyzed.
OCMIP is a project of the GAIM (Global Analysis, Interpretation and Modeling) task force, which is part of the IGBP (International Geosphere-Biosphere Program).
As it now exists, OCMIP consists of several groups that are separately funded. The European groups are funded by the European Commission under a project called Global Ocean Storage of Anthropogenic Carbon (GOSAC). US-OCMIP is funded by NASA as part of the JGOFS Synthesis and Modeling Project (SMP). There is also an effort headed by Yasu Yamanaka in Japan, as well as an Australian effort headed by Richard Matear. Ray Najjar noted that some other US and Canadian groups have expressed interest in participating in OCMIP. Jim Orr noted that there are differences between GOSAC and US-OCMIP in terms of the simulations that will be done, but there is much greater overlap.
OCMIP-1, the first phase of OCMIP, could now be considered over, Jim noted. Four groups participated in this project: Hadley Centre, Princeton/GFDL, IPSL/LSCE and Max Planck/Hamburg. Most of the published work from this largely unfunded effort is in reports and newsletters. A number of manuscripts are either complete or nearly so, and will be submitted to journals shortly. A few primary results from OCMIP-1 were noted:
Jim Orr noted that this meeting should be considered the end of OCMIP-1 and the beginning of OCMIP-2. OCMIP-1 received minimal funding and, understandably, output from simulations was sometimes not submitted on time. OCMIP-2 is quite different in that it is well funded (i.e., model groups are now supported to provide output). Jim emphasized that it is important for the success of the project that model results are submitted in a timely manner. It is also important that the same version of the circulation model is used for all simulations throughout the duration of OCMIP-2.
Ray Najjar then discussed US-OCMIP plans for biological modelling.
US-OCMIP will have a two-tiered approach:
We had some discussion as to how long these runs would be. US plans were to conduct long nutrient restoring runs and relatively short "individual investigator" runs. Some noted that in order to compute steady-state air-sea CO2 fluxes, longer runs would be needed. Helge Drange noted that very long runs with complex ecosystem models could be highly computationally demanding. Sarmiento added that plans since the beginning of OCMIP were to focus on runs to steady state. Further discussion is needed here concerning what the individual investigator runs will be.
Ray Najjar also noted that there are now global seasonal data sets for oxygen and nutrients that could be used, for the first time, to evaluate global marine carbon cycle models.
There then followed some discussion as to how to model CaCO3 production and dissolution. US-OCMIP plans are to use a simple scheme that guarantees agreement with the global mean alkalinity distribution.
After this presentation, Roger Hanson, the executive director for JGOFS, noted that there appears to be a large "disconnect" between OCMIP and JGOFS. There was some discussion as to why this might be so. Jorge noted that there will be a close link between US-JGOFS and US-OCMIP, since PI meetings were included in the SMP. Patrick Monfray noted that there was a formal discussion at the International JGOFS modeling meeting in Oban, Scotland about the role that OCMIP could play as a bridge between GAIM and JGOFS, both of which are programs of the IGBP. There was a sense that more interaction is needed between JGOFS and OCMIP, and Roger suggested that if OCMIP felt this was so, that it could send a statement to the next international JGOFS steering committee meeting this April 25-28.
Patrick Monfray then discussed atmospheric oxygen as a powerful tracer for marine biogical processes, both in terms of seasonal and meridional variations in atmospheric O2. Ken Caldeira said that he would be willing to provide O2 flux fields for any atmospheric transport modeler, but did not want to get involved directly with such modeling. Ray Najjar suggested that TRANSCOM, a project for intercomparing atmospheric transport models (mainly for CO2), might be interested in ocean model air-sea O2 flux fields. Martin Heimann stated that the next part of TRANSCOM will be devoted to atmospheric inversion.
Fortunat Joos then presented the utility of conducting pulse simulations. He showed that ocean GCM results for anthropogenic CO2 uptake could be reconstructed very well from pulse experiments. Other scenarios for anthropogenic CO2 could easily be conducted without running the GCM again, thereby saving vast amounts of computer time. He explained that "mixed-layer" pulse simulations, in contrast to the more common atmospheric pulse simulations, solve the problems with non-linearity of the carbonate equilibria. Fortunat suggested that the pulse simulations be conducted much sooner than planned.
Jim Orr then discussed simulations of purposeful CO2 sequestration. He emphasized that GOSAC will only address the potential for the global ocean to temporarily store additional CO2 from fossil emissions. With global models one cannot address the environmental, economic, legal, or political concerns related to this issue. There was some discussion about potential biological impacts, and Helge Drange noted that some of his own research into the literature revealed that modest pH changes (0.1) could have significant impacts. Such pH changes would occur even after dispersion of CO2 plumes to the grid sizes used in ocean models.
Chris Sabine then summarized investigations into what set of equilibrium constants are the best for the purposes of modeling in OCMIP. He noted that the carbonate constants in the DOE handbook (the Roy constants) do not represent a community consensus. There was a meeting of many inorganic carbon chemists (Sabine, Wanninkhof, Takahashi, Goyet, Wallace, and others) to discuss the equilibrium constants in which participants "agreed to disagree." That is, there is no consensus. Chris presented calculations using different sets of equilibrium constants. It was clear that the most consistent set, when computations involved alkalinity, pCO2 and DIC, was the Mehrbach constants. Specifically, these constants should be used when Alk and pCO2 are used to compute DIC, and when DIC and Alk are used to compute pCO2. These computations are generally the most relevant for the purposes of OCMIP. It was then decided that the Mehrbach constants (K1 and K2) would be adopted by OCMIP. Chris emphasized that using the Mehrbach constants for any calculations involving pH could be disastrous. Uncertainties in constants for water, borate, or CO2 solubility tend to be either small or unimportant.
Chris Sabine then shifted gears and talked about the data collected by WOCE and the CO2 survey. C-14 is rather densely sampled in the Pacific and Indian, but there are no new data in the Atlantic. TTO, SAVE and GEOSECS data must be used for evaluating models in the Atlantic. He showed a remarkable figure of deep C-14, which revealed the deep circulation quite clearly. For DIC and Alk, all of the oceans are well sampled. Gruber-type computations of anthropogenic CO2 have been made in some regions.
Nicki Gruber then discussed some of these calculations, showing how uncertainty in the C:O ratio can translate into a large uncertainty in the computed CO2 uptake. The Southern Ocean is where these calculations are probably the most uncertain. Ray Najjar suggested that these calculations could be tested using simulated data from ocean GCMs. Jim Orr then showed a few transparencies revealing large differences between models and "data" for anthropogenic CO2 uptake.
Rick Murnane suggested that OCMIP should include C-13 in its simulation suite. The air-sea flux of 13-CO2 is critical to understanding the 13-CO2 distri- bution in the atmosphere, and therefore for using this distribution for constraining sources and sinks of CO2.
Ken Caldeira then discussed how increased in atmospheric CO2 will cause C-14 in the atmosphere to increase. This is because CO2 increases are changing the pH of the ocean and therefore the speciation of C-14, such that the concentration of 14-CO2 in the ocean, and therefore in the atmosphere, will increase. The C-14 concentration in the atmosphere has been going down since the bomb tests, but Ken's predictions are that it will go up and eventually exceed the bomb peak for the business-as-usual scenario. The 14C/C12 ratio, however, will keep going down.
Jean-Claude Dutay then presented some general information about the utility of CFCs for evaluating ocean circulation models. He showed an evaluation of the OPA model with CFCs which revealed that the model does a reasonable job generally, but does not capture Antarctic Bottom Water formation. Chris Sabine noted that CFCs and anthropogenic CO2 are quite different due to their different histories and equilibration times, so that one needs to be careful in making assessments of one tracer based on the other.
Jean-Claude also discussed the utility of Helium-3 as a tracer of the deep circulation, and suggested that this tracer would be useful for providing a reality check of the purposeful CO2 sequestration simulations. He-3 is produced at mid-ocean ridges. There were some questions about contamination from bomb tests, but this is apparently only relevant in the thermocline of the Northern Hemisphere. Modeling the source is tricky, but current thinking is that the production rate is proportional to the spreading rate at the ridges, thereby making the Southeastern Tropical Pacific the greatest source. The only sink is gas transfer to the atmosphere. Data are available from GEOSECS and some from WOCE, but the latter does not have coverage comparable to C-14. Some suggested that Bill Jenkins would be the person to contact for such data.
Jorge Sarmiento ended the presentations by making two main points. First, the Southern Ocean is perhaps the most important region for controlling the air- sea partitioning of CO2. OCMIP-1 results show that anthropogenic CO2 uptake is greatest in the Southern Ocean, by virtue of its large area and vigorous vertical exchange. Simulations using coupled ocean-atmosphere GCMs show that changes in the Southern Ocean could be dramatic, due largely to increased stabilization resulting from lower salinities. The second point Jorge made was that knowledge of air-sea CO2 fluxes is critical for inverting atmospheric CO2 observations in order to determine terrestrial sources and sinks of CO2. His work shows a large sensitivity of inferred terrestrial sources and sinks depending on rather modest changes in the air-sea CO2 flux. Following this, Jorge presented some preliminary ideas about how to apply inversion techniques to oceanographic DIC observations in order to constrain air-sea CO2 fluxes.
Jim Orr then lead a discussion of the proposed OCMIP timetable, which was distributed electronically before the meeting. Additional copies (with slight modifications) were distributed at the meeting itself.
There was a long discussion about the use of a common biological model. US- OCMIP plans are to use a nutrient-restoring approach. It was suggested by some that another approach, such as the original Hamburg formulation, might be preferable. Advantages and disadvantages of both approaches were discussed for some time. Mick Follows reiterated that the main point of using a simple model was to compare the different capabilities of the physical model to deliver nutrients to surface waters. After much discussion, it was decided that individuals could choose either model to run, but that it was preferable for every group to run both models. Jim Orr asked who would be running what regarding the common biology model.
We had considerable discussion about DOM modeling. The main two factors to consider vis a vis such modeling is the fraction of new production that goes to DOM (as opposed to POM) and the lifetime DOM. There was disagreement as to whether these parameters should come from the field (such as from JGOFS data sets) or whether parameters should be chosen to fit the observed DOC data. US-OCMIP plans were to use the parameters chosen by the Princeton/GFDL group, which are based on fitting to DOC observations. It seemed clear that field observations show a much shorter DOC lifetime (months) than that used in most models (years). Ray Najjar said he would review the situation and come up with a recommendation. One point noted by Ray was that JGOFS data seem to show that the fraction of new production that goes into DOC varies spatially: in the Southern Ocean the fraction is much smaller than at Bermuda.
There was also considerable discussion about how to model radiocarbon in ocean GCMs. The debate was whether to use the "quick" method adopted by Toggweiler, in which, effectively, the 14C/12C ratio is advected around and that 12C is NOT carried as a separate tracer. This shortcut was contrasted with carrying separate tracers for 14C, 13C and 12C. 13C would be needed in order to compute Big Delta C-14, which is most commonly reported in observations. The debate was effectively settled when Anne Mouchet showed a comparison of the two approaches. The differences were large enough, generally greater than 10 permil (even for natural C-14), that we were convinced that the quick technique should not be adopted by OCMIP. Rather, the carbon isotopes should each be modeled separately. This choice forces the biological simulations to have high priority, because C-13 and C-14 concentrations have substantial biological sources and sinks. Adding C-13 to the suite, and modeling C-14 concentration, is a break from OCMIP-1 protocols, so there is a need for clear specification of how to model C-13 and C-14 concentrations. Ken Caldeira agreed to propose specific algorithms for such modeling.
It was stated that for OCMIP-2 anthropogenic CO2 simulations, the natural carbon cycle is needed for the initial conditions. This places even more priority on getting the natural carbon cycle runs done sooner than others.
An urgent need was to have seasonal nutrient fields for conducting the nutrient-restoring simulations, and to have such fields for evaluating more prognostic models. Najjar and Louanchi agreed to provide such fields in about two months.
We had some brief discussion about adding new "numerical" tracers, such as age and dye tracers, to the OCMIP suite of simulations. It was agreed that there was enough work to do for now and to postpone discussion of such simulations to a later time.
There was one mistake in the timeline distibuted: the "solubility + arbitrary biology" set of simulations should not appear.
Purposeful CO2 sequestration and natural He-3 were not proposed as part of US-OCMIP. Jim Orr asked individual US-OCMIP groups and Yasu Yamanaka whether they plan to do these simulations.
There were some changes made to the timetable handed out. These revisions are to be redistributed to all OCMIP groups. Also to be distributed is the table showing which groups will be doing what simulations.