Over the past year, the Sarmiento, Pacala, and Bender groups have made substantial progress towards improving regional estimates of air-sea and air-land carbon fluxes using models and observations from multiple platforms, including aircraft observations, satellite retrievals, ship board measurements, and regular atmospheric observations at surface sites. In addition, these groups have conducted modeling and data analysis studies that have informed the measurement community about the optimal locations for new observations and the interpretation of data. Finally, further development and analysis of models has provided insight into the processes governing the carbon fluxes and how they are likely to respond under climate change.
Sarmiento’s group has contributed to the development of the strategy for a high-density aircraft observation network over the United States. This network, comprising 20 stations, combines regional measurements of CO2 and other trace gases. The first year of observations has recently become available, and these observations are being used to estimate carbon fluxes with the Direct Carbon Budgeting Approach (DCBA). Sarmiento’s group used this approach to analyze the strengths and weaknesses of the network, leading to the implementation of a new station at Berms (Montana) to complement the network in a region that was poorly sampled by the existing network.
In parallel, Sarmiento’s group has started to develop a multi-species approach to infer carbon surface fluxes through a combined atmospheric and oceanic inversion constrained by observations of multiple trace gases such as CO2, CO, CH4, O3, and O2. This approach takes advantage of the correlation existing between different species to add some constraint on the fluxes and the transport that affect trace gas distributions.
Exploiting aircraft measurements of SF6 made around the globe, Manuel Gloor and colleagues have shown that state-of-the-art atmospheric transport models underestimate the ventilation of the boundary layer during winter over the continent, which could lead to substantial errors in the interpretation of atmospheric CO2 data. To improve atmospheric model realism, Sarmiento’s group is evaluating new parameterization of vertical transport using observations of CO2 from space in collaboration with GFDL.
Air-sea fluxes/ Predicting carbon feedbacks in the ocean
Sarmiento’s group has published a series of four papers in the past year that use ocean interior observations and Ocean General Circulation models to estimate regional air-sea fluxes. Using this technique, we have estimated the oceanic sink of anthropogenic carbon with a greater degree of certainty than any previous work, 2.2 ± 0.2 Pg C yr-1.
Ocean modeling studies have also been employed to identify variability in natural air-sea CO2 fluxes and the rate of uptake of anthropogenic CO2. In particular, recent work has focused on the North Pacific Ocean, since the variability here has a significant impact on the carbon source and sink estimates in North America. Recent work with an ocean model has shown that interannual variability in uptake of anthropogenic CO2 in this region consists primarily of modulations of the uptake maximum in winter, rather than a slowly evolving change in the mean. There is a substantial decadal trend towards increased winter uptake over time, primarily due to decadal changes in ocean circulation.
In addition, Sarmiento’s group is engaged in modeling studies to aid in the interpretation of data from international repeat hydrography measurement campaigns that aim to sample selected transects once every ten years to detect changes in oceanic uptake of anthropogenic CO2. Analysis of simulations from three models suggest that the amplitude of the natural variability of column inventories of dissolved inorganic carbon is of the same order of magnitude as the anthropogenic signal as it changes over a decade. The large magnitude of natural variability will complicate the interpretation of repeat hydrographic data because differences between past and present measurements may not be directly attributable to increasing anthropogenic carbon uptake by the oceans.
Air-land fluxes/ Predicting carbon feedbacks in the land biosphere
Changes in climate have the potential to affect the geographic distribution of ecosystems, and the mix of species that they contain. However, little is known about how these ecosystems might respond to climate change, or how these changes might feed back on climate. To improve our predictive skill, Pacala’s group has continued to develop the Earth System Model. The team finished tuning the model, which is now the only model with a free running ocean and biosphere that produces a stable climate.
In addition, Sarmiento’s and Pacala’s group have started to develop a data assimilation scheme to constrain the new GFDL vegetation model LM3V in order to improve the estimation of carbon fluxes over North America. This effort, so far limited to the use of eddy flux tower and forest inventory data, is the first step towards the development of a full carbon observing system, which will use data on varied spatiotemporal scales (flux and tall towers, flask and aircraft atmospheric data, forest inventories, oceanic data), to better understand the key processes controlling carbon fluxes over the North American continent.
A particular focus concerns the modeling of carbon emissions by fires, because fire is responsible for most of the year-to-year variation in the growth rate of atmospheric CO2. As part of this effort, a fire model, relying on in-situ and remote sensing observations, has been designed for boreal regions. Provided realistic climate predictors, the model successfully reproduces the spatio-temporal evolution of burned area in the boreal forest, including seasonality and interannual variability. The model is the first that predicts both the location and timing of forest fires, which makes it suitable to quantitatively study the evolution of fire with climate. The team is now turning to the more difficult problem of simulating tropical fires.
Pacala and colleagues also made considerable progress on a difficult technical problem. A modern and powerful method to estimate the parameters in models is called the Monte Carlo Markov Chain. It requires that one run the model many times while adjusting parameter values until the model has the best possible agreement with all available data. The researchers figured out how to run the model of the biosphere in the Earth System Model quickly enough so that this method can be used, and should have results using the new method next year.
Finally, Pacala’s group cracked a problem that should lead to the next generation of biosphere models. Two years ago, they achieved a mathematical breakthrough with models of vegetation and can now analyze and understand them mathematically, whereas before they could only run them on a computer. The researchers have now exploited the mathematical breakthrough to develop a statistical method that estimates all the parameters in the biosphere model using the information in national forest inventories that have been established in many countries in response the Kyoto Accord. They can, for example, parameterize the model for every tree species in the US in every location. Because the inventories record how the biodiversity of forests change as forest age, the behaviors of the models can also be checked to make sure that their predictions are correct. This is the package that has always been missing in ecology: enough data to build quantitatively accurate models that can deal with realistic levels of biodiversity over large regions plus the capacity to analyze the models to extract general rules. The group has applied the new method to modeling forests of the northern Midwest, and can now predict which species occur where and how those forests will change through time. The researchers’ next steps will be to apply their methods to mid-Atlantic, German, and Panamanian forests.
Inversion modeling and the CO2 fertilization sink
One of the science group’s major findings resulted from combining the inversion modeling team’s large oceanic data set with atmospheric data in order to jointly estimate air-sea and airland fluxes. The group’s latest work has resulted in significant changes in the estimated terrestrial fluxes compared with previous analyses of atmospheric data, especially in tropical and southern hemisphere regions that are not well sampled by the atmospheric observations alone. The joint estimate finds that the tropical and southern land regions are a statistically significant source of carbon to the atmosphere, with a 77% chance that their total source exceeds 1 billion metric tons of carbon per year. This flux represents the sum of emissions due to tropical deforestation and a potential natural sink from plant fertilization via increasing atmospheric CO2 levels. The net source estimated by the joint inversion is of about the same magnitude as independent estimates of the tropical carbon source due to deforestation and land use change alone. This suggests that CO2 fertilization is less important a player in the tropics, and thus in the global carbon cycle, than previously thought.
Most of the current generation of climate models used to estimate the effects of climate change, including those being used for the International Panel on Climate Change (IPCC) Assessment Report 4, incorporate a substantial CO2 fertilization sink. This sink reduces the amount of mitigation required to stabilize atmospheric CO2 concentrations to about half what it would be without CO2. The science group’s studies challenge the existence of a substantial CO2 fertilization sink.
Monitoring atmospheric oxygen
Bender’s core CMI research has involved measurements of the atmospheric O2/N2 ratio. This ratio provides a primary constraint for partitioning the sequestration of fossil fuel CO2 between ocean uptake and land biosphere uptake. The oceanic uptake of CO2 has no effect on atmospheric O2, while the land biosphere produces approximately one O2 for every CO2 it consumes. Therefore, measurements of the changing O2/N2 ratio of air allow us to determine the part of CO2 sequestration due to the land biosphere, and the part due to the oceans.
This year, Bender’s group has been using observations of Ar/N2 ratios to interpret their O2/N2 observations. The quality of these data has improved markedly with the deployment of new sampling systems, and the Bender Group has produced new climatologies for the annual cycle of this gas at their 7 remote observing sites (Barrow, Sable Island, Samoa, Amsterdam, Cape Grim, Macquarie and Syowa). These values agree well with simulations, by Galen McKinley, of a model with active upper ocean gas exchange and atmospheric tracer transport.
Ocean model development
Sarmiento’s group has been working to improve representations of two fundamental ocean processes in computer models – the return of deep ocean water to the ocean’s surface and the release of nutrients from decaying marine organisms.
Understanding the processes that determine the return flow of deep water into the upper ocean is a central problem in oceanography. The potential of Helium-3 emanating from midocean ridges as a tracer for the pathways of water mass conversion has led Sarmiento’s group to initiate an investigation of the distribution of Helium-3 by a combination of model simulations and observations.
The available measurements of Helium-3 and Helium-4 from the WOCE program have been integrated to produce a three-dimensional gridded dataset, which provides an accurate estimate of the tracer inventories. Furthermore, Helium-3 has been implemented in a suite of simulations using GFDL models. The simulations from the suite of ocean models produced substantially different steady state distributions of the tracers. Comparisons between these model simulations and the new Helium-3 data set suggest that the currently accepted amount of Helium-3 injected into the interior of the ocean from mid-ocean ridges might be substantially overestimated. These Helium-3 injection estimates have been used widely in oceanography; therefore, this result has far-reaching implications for ocean circulation and the dynamics and geochemistry of Earth’s mantle.
Remineralization ratios, e.g. the stoichiometric ratios associated with the remineralization of organic material into dissolved nutrients, are used to derive anthropogenic carbon storage in the ocean and are included in many models that are used to simulate ocean biogeochemistry. The Sarmiento group has developed a novel technique to compute remineralization ratios that does not rely on several of the assumptions that have been used in previous studies. Analysis of model simulations indicates that this method is more accurate than the alternative approaches. Preliminary results, suggest that the variability of remineralization ratios is small in the deep ocean, confirming previous results.