Research Highlights

Building a Carbon Net-Zero Infrastructure for the U.S. by 2050

CMI Co-director and Principal Investigator: Stephen Pacala

At a Glance

Pacala’s lab focused on three areas in 2019: 1) They continued their work on the terrestrial biosphere and carbon cycle, and on the role of land use change in carbon mitigation. They completed a series of studies that improve the effects of drought on the carbon cycle in climate models, particularly in the tropics. Also, Pacala spent a large amount of his time completing the National Academy of Sciences report on negative emissions technologies. He chaired the effort and co-wrote the chapters on negative emissions from land use change. 2) They continued their work on the possibility that tropical forests may be spontaneously switching to vine-dominance, which would cause them to lose 95% of their carbon to the atmosphere. This is a possible new tipping point. They expect to have an answer by the 2020 CMI meeting. 3) They started a large project on net-zero-emitting infrastructure for the U.S., which is described below. Although organized initially by the CMI, it is now a collaborative effort involving the CMI, the Andlinger Center, the Princeton Environmental Institute, the Woodrow Wilson School and Princeton University’s central administration, as well as collaborators from the Environmental Defense Fund, The Nature Conservancy, the Natural Resources Defense Council, Exxon, and BP.


Research Highlight

To meet the 2 ̊C target of the Paris Agreement, global greenhouse gas emissions would have to decline from their current value of over 50 billion metric tons of carbon dioxide (CO2) equivalent per year (Global Carbon Project, 2018), to net-zero sometime between the middle and end of the current century. (IPCC 1.5 ̊ Special Report, 2018). Emissions in developed countries would have to drop even faster, and reach net-zero approximately at mid-century (IPCC, 2018). Would it be possible for the U.S. to build an energy system with zero net emissions in one billion seconds (just over 31 years)? Is the technology even available at a cost that the economy could bear?

The answers to these questions have changed dramatically over the past 10 years because of unprecedented breakthroughs in energy technology. The cost of solar electricity fell during the last decade by 76%, wind electricity by 69%, and lithium ion battery packs by 85% (Figure 1.1). Over the same decade, hydrofracking and horizontal drilling made natural gas inexpensive and abundant in the U.S., while carbon capture and storage technology both matured and came down in price. As a result, it would now be possible for the U.S. to build a non-emitting energy system at a cost only marginally higher than consumers pay for energy today. The system would be largely electric, with electricity from a roughly 50:50 mix of renewables and natural gas with carbon capture and storage (plus existing sources such as hydro and perhaps nuclear electricity), and with electric light- and medium-duty transport. Such a system would probably use more gas than is consumed today.

The U.S. is better positioned for this energy system than any other nation, because of its abundant wind and solar resources, abundant gas, well-characterized reservoirs for storage, and land for the negative emissions required to offset difficult-to-mitigate sources. The new CMI infrastructure project emerged to determine what it would take to build such a system, as well as alternatives. (“Infrastructure” is broadly defined here to include all plant, equipment, and services associated with energy resource extraction, conversion, transmission and distribution, and utilization.) The Pacala group proposes to describe qualitatively and quantitatively the engineering/industrial activities and financial flows required to decarbonize the U.S. economy, that is, to achieve net-zero greenhouse gas emissions, by mid-century.

Figure 1.1. Cost of renewable electricity and lithium ion batteries plummeted over the last 10 years. (Graph produced by J. Jenkins).

Deep decarbonization scenarios for the U.S. have been proposed by many others. In most cases, the objective is to minimize total cost, using a mix of top-down and bottom-up analysis. Capital and operating costs (and associated learning curves) are considered at varying levels of detail, and costs tend to be reported as aggregated and amortized values. But there is little attention to constraints related to rates of deployment. The lack of transparency limits the extent to which such exercises can inform actionable mitigation plans that identify spend-and-build schedules needed to achieve decarbonization targets.
In this project, bottom-up analysis will quantify the cost-and-build schedules for plausible mixes of investments across the energy system that deliver net-zero greenhouse gas emissions by mid-century. The group will emphasize expert engineering judgement. Analysis will be region-by-region, sector-by-sector, and major-project-by-major-project. Only technologies that have a reasonable basis for commercial-cost estimation today will be included, i.e., ones that arguably have already been demonstrated at industrial scales. The analysis will transparently report asset turnover rates and costs and time associated with pre-investment activities (feasibility studies, environmental impact assessments, community acceptance, and permitting). The analysis will include cost and schedule variances likely to be experienced for development of new natural resource capacity or construction of major projects, including balance-of- plant and supporting infrastructures.

Alternative decarbonization plans, each emphasizing a different technological pathway, will be articulated. For example, one plan might emphasize a balanced portfolio of low-carbon energy supply technologies and energy-use efficiency improvements. Another might emphasize aggressive energy efficiency improvements. Other plans might emphasize variable renewable electricity generation, nuclear energy, or fossil fuels with CO2 capture and storage.
A cost minimization will not be performed but a detailed and comprehensive energy accounting model will be used to ensure consistency across sectors and energy forms. Resource requirements over time – investment capital, workforce, major raw materials, manufactured equipment and bulk materials – will be quantified and contrasted with those for business-as-usual energy infrastructure development. For the near-term (2021-2025), annual capital commitments for each plan will be quantified. Capital commitments and resource requirements beyond 2025 will be estimated in five to 10-year tranches.



1 Anderson, C.M., R.S. Defries, R. Litterman, P.A. Matson, D.C. Nepstad, S.W. Pacala, W.H. Schlesinger, M.R. Shaw, P. Smith, C. Weber, and C.B. Field, 2019. Maximize natural climate solutions—and decarbonize the economy. Science, 363(6430): 933-934.

2 Martinez Cano, I., H.C. Muller-Landau, S.J. Wright, S.A. Bohlman, and S.W. Pacala, 2019. Tropical tree height and crown allometries for the Barro Colorado Natural Monument, Panama: a comparison of alternative hierarchical models incorporating interspecific variation in relation to life history traits. Biogeosciences, 16: 847-862.

3 Muller-Landau, H. C. and S.W. Pacala. What determines the abundance of lianas and vines? In A. Dobson, R. Holt, and D. Tilman, Eds., for the volume Unsolved Problems in Ecology, in press.

4 Weng, E., R. Dybzinski, C.E. Farrior, and S.W. Pacala, 2019. Competition alters predicted forest carbon cycle responses to nitrogen availability and elevated CO2: simulations using an explicitly competitive, game-theoretic vegetation demographic model. Biogeosciences Discussions,; in review Biogeosciences.

5 Zeppel, M., W.R.L. Anderegg, H. Adams, P. Hudson, A. Cook, R. Rumman, D. Eamus, D. Tissue, and S.W. Pacala, 2019. Embolism recovery strategies among species influenced by biogeographic origin and nocturnal stomatal conductance. Ecology and Evolution, in press.


Robotic Observing System Challenges Previous Estimates of the Southern Ocean Carbon Sink

Principal Investigator: Jorge Sarmiento

At a Glance

A robotic observation system that researchers have deployed in the Southern Ocean is providing year-round measurements of carbon fluxes and is changing our understanding of the ocean carbon sink. Previous results had revealed that significant outgassing of carbon dioxide (CO2) in the region was occurring in wintertime, reducing the region’s net uptake of carbon. Researchers are now trying to determine if part of the change in uptake may be due to Southern Ocean circulation changes and surface warming in response to a shift in the regional climate over recent years.


Research Highlight

Jorge Sarmiento directs the National Science Foundation-funded Southern Ocean Carbon and Climate Observations and Modeling (SOCCOM) project, a multi-institutional effort funded by the National Science Foundation to dramatically increase the number and variety of observations of the Southern Ocean through the world’s first large-scale deployment of biogeochemical (BGC) Argo floats – robotic floats that have been used to measure ocean temperature and salinity that can now be equipped with newly developed biogeochemical sensors to measure pH, nitrate, and oxygen. SOCCOM has 123 of these augmented Argo floats operating, which have collectively made nearly 5 million observations in the Southern Ocean (Figure 2.1) in all seasons and under ice.

Figure 2.1. Locations and trajectories of 124 SOCCOM floats operating as of January 28, 2019. White dots are locations of operating floats; cyan are inoperative floats; yellow dots are pre-SOCCOM floats deployed before 2014. Black lines indicate float trajectories since deployment. (Credit: SOCCOM)

Analyses made since the inception of the project in 2014 have suggested a significant outgassing of CO2 in the high-latitude Southern Ocean that had escaped previous ship-based observation. This outgassing occurs during the wintertime when deep waters enriched in carbon are brought to the surface, and when hardly any ships travel into this harsh and remote region. Analyses of float data alone had indicated that outgassing cancelled CO2 drawdown elsewhere in the Southern Ocean, resulting in a net uptake of only -0.08 Pg C yr-1 (petagrams of carbon per year) in the Southern Ocean, a significant reduction from prior ship-based uptake estimates of over -1 Pg C yr-1 (negative into the ocean). To put this in context, the ocean as a whole takes up approximately -2 Pg C yr-1.

In order to reconcile the differences between ship and float-based estimates, this year SOCCOM researchers at Princeton have worked with scientists involved with the Global Carbon Project – an international research collaboration that seeks to fully understand the carbon cycle - to merge SOCCOM pCO2 estimates with the shipboard and mooring observations that form the backbone of our understanding of the ocean’s role in carbon uptake. The group has found that the addition of SOCCOM observations reduces annual Southern Ocean carbon uptake estimates by approximately one-third relative to estimates based on ship-board observations alone, yielding an annual uptake of -0.75 Pg C yr-1 (Figure 2.2). SOCCOM observations thus appear to be capturing an important signal previously missed by the shipboard-only dataset.

Figure 2.2. Southern Ocean (south of 35oS) carbon fluxes calculated from ships and floats. Air-sea Southern Ocean CO2 fluxes over the past 30 years are shown as calculated from the two mapping products used in the Global Carbon Project (blue line +/- 1 s.d.: Landschützer et al. 2013 neural network; green line: Rödenbeck et al. 2013 Jena CarboScope interpolation scheme). 2015-2017 mean fluxes for the three pCO2 products used in this study are overlaid. The combined SOCAT (shipboard dataset) and SOCCOM (new float observations) estimate of the Southern Ocean carbon flux is -0.75 ± 0.22 Pg C yr-1 (black triangle), a decrease in the Southern Ocean uptake of 0.4 Pg C yr-1 from the SOCAT-only estimate of -1.14 ± 0.19 Pg C yr-1 (orange circle). Removing shipboard measurements south of 35oS and from 2014 onwards further decreases the Southern Ocean uptake to -0.35 ± 0.19 Pg C yr-1 (purple square). (Bushinsky et al., submitted)

The group is now working to understand how much of the difference between older and new estimates of uptake is due to the impact of the new observations, and whether any of the decrease may reflect a change in the Southern Ocean’s state during the observational period that could foreshadow conditions in a warmer climate.

Figure 2.3. Time-series of sea-ice area (upper panel; red) and sea-surface temperature (lower panel; black) anomalies from 1982 to 2018. The anomalies are averaged over the region south of 55° S as indicated by the gray area on the small map. The blue dashed line (2014) indicates the start year of the SOCCOM observational period during which a major shift in the regional climate occurred. (Credit: A. Haumann)

In recent decades, Southern Ocean surface waters have been cooling slightly, but this trend was abruptly and unexpectedly interrupted in 2016 and 2017, when southern hemisphere sea-ice cover shrank to a record minimum and the ocean’s surface warmed strongly (Figure 2.3). This warming occurred as a circumpolar high-latitude heat wave during austral summer 2017 that exceeded the average mean surface temperature by more than four standard deviations. Analysis of the float data shows that these events took place due to a destratification of the water column that might have switched the Southern Ocean climate system to a new state over the last few years, enhancing natural CO2 release. Because sea ice plays an important role in controlling stratification, future changes in sea ice and ocean circulation might change these fluxes. While the anomaly in the surface ocean was only a short-time departure from the long-term mean, it may foreshadow the consequences of a potentially warmer future Southern Ocean for biogeochemistry and ecosystems.

In the coming year, the Sarmiento group will continue to scrutinize carbon fluxes in the Southern Ocean and identify causes of their variability. Work in progress includes analysis of float data to identify the subsurface source of high-CO2 waters and the comparison of float and ship-based flux estimates with observations from NASA’s Orbiting Carbon Observatory, which measures atmospheric CO2 concentrations from space.

Sarmiento is also involved in two promising efforts to expand the biogeochemical observing system to all the world’s oceans. The National Science Foundation has announced a competition for “mid-scale” infrastructure projects in the range of $20-70 million and Sarmiento is co-principal investigator on a proposal to develop a global biogeochemical float network of ~400 floats. He is also involved in developing the community plan for the next stage of the global Argo system, which similarly proposes incorporating biogeochemical floats as part of the existing global network.



1 Arteaga, L., M. Pahlow, and J.L. Sarmiento, 2018. Mesopelagic remineralization and Surface nutrient limitation of export production in the Southern Ocean. Geophys. Res. Let., submitted.

2 Bronselaer, B., J.L. Russell, M. Winton, N.L. Williams, R.M. Key, J. P. Dunne, R.A. Feely, and J.L. Sarmiento, 2018. Impact of wind and meltwater on recent observed physical and chemical evolution of the Southern Ocean. Nature, submitted.

3 Bushinsky, S.M., P. Landschützer, C. Rödenbeck, A.R. Gray, D. Baker, M.R. Mazloff, L. Resplandy, K.S. Johnson, and J.L. Sarmiento. Revisiting Southern Ocean air-sea CO2 flux estimates with the addition of biogeochemical float observations, submitted.

4 Chen, H, A.K. Morrison, C.O. DuFour, J.L. Sarmiento, S.M. Griffies, P. Zhai, and M. Winton, 2018. Deciphering patterns and drivers of anthropogenic heat and carbon storages in the Southern Ocean. Geophys. Res. Let., submitted.


Effects of Urbanization and Volcanoes on Tropical Cyclone Activity

Principal Investigator: Gabriel Vecchi

At a Glance

A modeling study indicates that the urbanization of Houston acted to enhance the flooding from Hurricane Harvey (2017) in the city, both because less of Harvey's rainfall was able to infiltrate the soil due to an increase in impervious surface coverage and because the increased surface "roughness" of the urban landscape acted to locally enhance Harvey's rainfall. An observed 28-year increase in rapid intensification rates of North Atlantic hurricanes is unusual and may already include a signal from human radiative forcing; however, uncertainties in the hurricane data record over the rest of the world preclude a confident assessment of recent changes in rapid intensification. A modeling study indicates that explosive volcanic activity, such as the 1963 Mount Agung eruption, could impact global tropical cyclone (TC) activity in the years that follow, but that the response will be fundamentally different for different volcanoes. The goal of this work is to improve the understanding of the mechanisms behind and limits to the predictability of TC activity over the past few and next centuries. The work connects to broad questions in the climate science community, such as uncertainty over what TC changes are likely to occur over the coming century, and the extent to which intrinsic climate variability and natural forcing may be dominant over the impact of greenhouse forcing.


Research Highlight

The goal of this work by the Vecchi group is to improve the understanding of the character of, mechanisms behind and limits on the predictability of, variations and changes in the statistics of TC activity over the past few and coming centuries. Key tools in these studies are climate and atmospheric model studies, along with analyses of the observed record, to better understand the extent to which observed multi-decadal to centennial changes in TC activity have been driven by large-scale factors (such as ocean temperature changes, greenhouse gases, volcanic eruptions, and El Niño) versus random atmospheric fluctuations. Furthermore, the group has explored aspects of the extent to which natural and human-induced climate drivers (such as volcanic eruptions and greenhouse-induced warming) and non-climatic factors (such as urbanization) have influenced the character and impact of TC activity.

They have also explored controls on the extreme rainfall and flooding associated with TCs, such as the devastating flooding in Houston during Hurricane Harvey in 2017. It was found that urbanization in Houston contributed both to locally enhance/concentrate the extreme rainfall over the city during Hurricane Harvey, and to increase the amount of flooding from the rainfall (Zhang et al., 2018). A very surprising result of this is that the urban surface characteristics of Houston, largely its enhanced “surface roughness” coming from the many tall buildings, lead to a concentration of the rainfall induced by Harvey (Figure 3.1). In addition, the study showed that urbanization changed surface characteristics so that the rainfall was less able to infiltrate the soil, and led to enhanced flooding compared to the natural state. It was found that urbanization led to an approximately 20-fold increase in the likelihood of such flooding – with both increase in rainfall and flood conversion playing a role.

Figure 3.1. Including Houston’s urban surface characteristics enhances rainfall over Houston from Hurricane Harvey. Rainfall in the Houston area (a) observed during Hurricane Harvey (25-30 August 2017), and from two ensembles of atmospheric model experiments in which (b) urban conditions were prescribed over Houston and (c) in which the land surface conditions over Houston are artificially set to be grassland. Notice the increased rainfall over Houston in the middle panel relative to the bottom panel. (Zhang et al., 2018).

The group continues to focus on the rapid intensification of TCs, in which TC intensity increases rapidly in less than a day such as with hurricanes Harvey, Irma, and Maria in 2017. Rapidly intensifying TCs present a particular challenge to society since they tend to be poorly forecast on the weather scale and a rapid change in TC intensity can leave society with little time to prepare for TC impacts. A manuscript was seen through publication (Bhatia et al., 2018) that shows that projected warming over the 21st century is expected to increase the potential intensity (theoretical upper bound) of hurricanes and therefore increase the global proportion of storms making rapid intensification, with changes relative to the late- 20th century expected as soon as the next couple of decades. Another manuscript, in press (Bhatia et al., 2019), finds that the fraction of Atlantic hurricanes undergoing rapid intensification has increased over recent decades faster than expected from internal climate variations, but that observational uncertainties preclude us from making reliable statements yet about the changes in the rate of rapid intensification across the rest of the globe.

In addition, the team has researched the impact of volcanic forcing on global-scale TC and hydroclimate variations – with a particular focus on the similarities and differences between forcing from different 20th century volcanoes (Yang et al., submitted). These questions were explored through targeted climate model experiments in order to test the hypothesis that the hydroclimate and TC sensitivity to volcanoes depends fundamentally on the hemisphere in which the volcanic plume is most pronounced. The studies found fundamentally different TC and hydroclimate responses to Pinatubo (1991), the most intense volcanic eruption in the 20th century with a stratospheric plume symmetric about the equator; Santa María (1902), whose plume had a northern hemisphere maximum; and Agung (1963), which had its plume primarily in the southern hemisphere. For asymmetric volcanoes, rainfall and TC activity shift away from the hemisphere of strongest stratospheric aerosol forcing. And even though Pinatubo was the strongest volcano in terms of global radiative forcing and surface temperature change, both Santa María and Agung drove larger regional rainfall and TC differences due to displacements of climatological wet/ dry regions. These experiments help provide a baseline to help interpret climate changes over the past millennium, in which volcanos were one of the main drivers. They also help assess the extent to which the impact of cooling from volcanic forcing and the impact of surface warming from greenhouse gas buildup can serve as useful analogues when looking at regional climate changes and extremes.



1 Bhatia, K., G.A. Vecchi, T. Knutson, H. Murakami, J. Kossin, K. Dixon and C. Whitlock, 2019. Recent increases in tropical cyclone intensification rates. Nature Communications, 10(1): 635. s41467-019-08471-z.

2 Yang, W., G.A. Vecchi, S. Fueglistaler, L.W. Horowitz, D.J. Luet, A.G. Muñoz, D. Paynter, and S. Underwood, 2019. Climatic Impacts from Asymmetric Large Volcanic Eruptions in a TC-Permitting Climate Model. Geophys. Res. Lett., submitted.


The Contribution of Ocean Waves and Bubbles to Global Air-Sea Carbon Dioxide Flux

Principal Investigators: Brandon Reichl and Luc Deike

At a Glance

Breaking waves induce bubbles of air in the upper part of the ocean that enhance air-sea gas exchange. This research utilizes new theory, data, and model simulations to quantify the contribution of bubbles to the air-sea carbon dioxide (CO2) flux in a global context. Our results indicate that bubbles contributed roughly 40% of the total air-sea CO2 flux over the timeframe 1982–2015, indicating they play a critical role in this important process.


Research Highlight

A major component of understanding how the Earth’s systems will respond to increasing greenhouse gas emissions is knowing where emitted compounds reside in the system. It is estimated that approximately 20–30% of the CO2 emitted in a given year is transferred from the atmosphere into the ocean (IPCC, 2014; Le Quére et al., 2018). This has important implications for the atmosphere, since it mitigates the greenhouse warming effect, and also for the ocean, where increased carbonic acid concentrations are lowering the ocean’s pH to the detriment of marine ecosystems.

The current generation of Earth System Models that predict the evolution of Earth’s climate system use empirical parameterization to estimate the air-sea CO2 flux (FCO2), which are represented as the product of a gas transfer velocity (kw), a gas solubility (K0), and the partial pressure difference of CO (ΔpCO2):

FCO2 = kwK0ΔpCO2

The ΔpCO2 component of the problem is resolved by the model (or can be derived historically, e.g., Landschützer et al., 2014) and K0 is well-constrained. However, models or empirical methods are needed to estimate kw, which is highly uncertain. The most common empirical relationships use the gas diffusivity and the wind-speed as the sole input to determine kw (e.g., Wanninkhof, 2014). This reflects the fact that the relevant processes to air-sea gas transfer (turbulence, micro-breaking, and bubbles produced by breaking waves) are highly correlated with the local wind speed. However, bubbles in particular are also strongly dependent on variability in the local surface wave-field that dictates the tendency for wave breaking, which is only partially correlated with the local wind speed. In recent years, several methods have been developed to estimate kw through decomposition into bubble and non- bubble components, where the bubble component is then estimated using additional information from the surface waves.

For this research, the team employed the recent ocean surface gas transfer model DM18 (Deike & Melville, 2018), historical time-series, and model simulations of the global wind (Kobayashi et al., 2015; Tsujino et al., 2018) and wave fields (using wave model WAVEWATCH-III) to estimate FCO2 and thus to better understand the importance of bubbles and waves in a global context. Figure 4.1 shows global maps of the yearly mean flux of CO2 computed from this method and averaged for the time period 1982–2015. This method can isolate the contribution of bubbles (right panel) from the total flux (left panel). This comparison shows that the bubbles are not the majority control on air-sea CO2 flux, but can still provide a sizeable contribution to the total flux.

Figure 4.1. The yearly mean CO2 flux predicted by the DM18 model for the total flux (left panel) and the component of the flux due to bubbles (right panel). By convention, positive is an upward flux (out of the ocean).

To look further into the importance of waves and bubbles in air-sea CO2 flux, the team looked at the time-series of the globally integrated CO2 flux, also decomposed into its bubble and non-bubble fractions (Figure 4.2). Here it can be seen clearly that the bubble contribution to the total flux is much less variable than the total flux, and can contribute between 40 and 55% of the total flux (with a mean flux of roughly 40% over the time-series). The previous estimate of the contribution of bubbles to air-sea CO2 flux was 30% (Woolf, 1997), which is revised up here based on the new data and theory. This result implies that a more advanced understanding of the role of bubbles in air-sea gas flux may improve understanding of air- sea CO2 exchange. Future research efforts toward improved and better constrained wave-state dependent kw models offer physically appealing methods that may improve CO2 flux estimates on global scales.

Figure 4.2. Left panel: Time series of the 12-month running mean flux of CO2 out of the ocean, where the black line is the DM18 estimate and the green line is the DM18 estimate for bubbles alone. The shaded region encompasses yearly minimum and maximum values. Right panel: The ratio of the bubble flux to the total flux for monthly means (gray markers) and 12-month running mean (black lines).



1 Deike, L., and W.K. Melville, 2018. Gas transfer by breaking waves. Geophysical Re- search Letters, 45(19): 10482–10492.

2 IPCC, 2014. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, 151 pp.

3 Kobayashi, S. et al., 2015. The JRA-55 reanalysis: general specifications and basic characteristics. J. Meteorolog. Soc. Jpn., 93(1): 5–48.

4 Landschützer, P., N. Gruber, D.C.E. Bakker, and U. Schuster, 2014. Recent variability of the global ocean carbon sink. Global Biogeochem. Cycles, 28(9): 927–949.

5 Le Quére, C. et al., 2018. Carbon Budget 2018. Earth Syst. Sci. Data, 10(4): 2141–2194. essd-10-2141-2018.

6 Reichl, B.G. and L. Deike. Contribution of Bubbles to Carbon Dioxide Flux from Global Wave Simulation. In preparation.

7 Tsujino, H. et al., 2018. JRA-55 based surface dataset for driving ocean—sea-ice models (JRA55-do). Ocean Modell. 130: 79–139.

8 Wanninkhof, R., 2014. Relationship between wind speed and gas exchange over the ocean revisited. Liminology and Oceanography Methods, 12(6): 351–362.


Modeling the Wind-Driven Formation and Extent of Coastal Polynyas

Principal Investigator: Howard A. Stone

At a Glance

Climate changes involve atmospheric motions, ocean flows, and evolution of ice on land and in the sea. These dynamics are closely interrelated; insights into individual processes can help to illuminate poorly understood aspects of global climate dynamics, such as factors affecting the maintenance of sea ice cover in the Arctic basin. Sea ice cover can impact fresh water fluxes, local ecology and ocean circulation. The Stone group is providing simplified models for understanding the movement and distribution of ice during the formation of polynyas, which refer to localized regions of water surrounded by ice, and through narrow straits, which can affect flow, mixing, and ecology in the ocean. The approach seeks to draw generalizations valid for various geometric and climate conditions.


Research Highlight

A polynya is a region of persistent open ocean water surrounded by sea ice and/or rigid boundaries, such as a coastline, land-grounded ice or ice shelves (Figure 5.1); effectively it is a “hole of ice.” Polynyas remain open from a regional balance between the rate of ice production (due to freezing seawater) and the rate of ice depletion, for example, due to flow. Such polynyas may exist either in the open ocean or close to coastal boundaries. The latter are formed when winds sweep ice away from the coast, exposing open seawater that freezes to form new ice. Thus, these coastal polynyas, especially along the Antarctic coast and some regions of the Arctic, are an important source of new sea ice, are crucial to ocean-atmosphere energy, momentum and moisture exchanges, and are thought to regulate thermohaline circulation, i.e., the circulation of temperature and salinity, in the ocean. In addition, phytoplankton and other marine life thrive in polynyas, especially in summer months, and so these water-ice structures are important to ecology.

Although some qualitative mechanisms of polynya formation have been identified, modeling their extent precisely has proved challenging. Previous attempts have relied mostly on high-resolution numerical simulations, where a clear connection between polynya formation dynamics and the mechanical stresses due to the ice motion is lacking. Recently, the Stone group has succeeded in developing simplified descriptions of ice motion due to wind in the context of ice bridge formation in straits, taking into account the frictional stresses in ice. These simplified models represent the mechanical behavior of ice and ice flows, agreeing both with available measurements and with numerical simulations. This approach to seeking physically relevant simplified descriptions forms the basis for our more recent investigation of coastal polynyas.

Figure 5.1: (a) Natural-color image of the wind-driven, latent-heat coastal polynya formed near Ross Island in the Antarctic, captured by NASA's Aqua satellite. Image is from (b) Distribution of the main Arctic polynyas. (c) Schematic of the formation of a wind-driven, latent-heat coastal polynya. Here u(x,t) denotes the speed of the ice and c(x,t) is a measure of the concentration of the ice floe.

The Stone group’s current ice-related research efforts are focused on mathematical modeling wind- driven polynya formation in coastal regions, including islands and fjords. As with the group's previous studies of ice flows, they interacted with Dr. Michael Winton at the Geophysical Fluid Dynamics Laboratory (GFDL). These studies of polynyas quantify the formation of new ice by freezing seawater, while incorporating findings from the group’s previous work to quantify the stresses and motion of the formed ice in response to wind. The study includes a fully resolved numerical model, consistent with more sophisticated models published in the literature, as well as a simplified model, designed to capture important physical characteristics, to predict the roles of freezing (ice production), flow, and ice accumulation in determining the extent of coastal polynyas. The team has demonstrated that the theoretical predictions agree quantitatively with the results of direct numerical simulations with and without considering the curvature of the coastline. The combination of modeling approaches provides clear connections between the mechanics of sea ice motion and the thermodynamics of sea ice production. Consequently, these modeling efforts not only explain a complex geophysical phenomenon but also provide a means to refine the modeling of sea ice in the more general context of Arctic and Antarctic ice flows near land boundaries.



1 Zhu, L., B. Rallabandi, M. Winton, and H.A. Stone, 2019. An analytical model of wind-driven formation of coastal polynyas. Preprint.


Understanding Mineral Protection of the Soil Carbon Sink

Principal Investigators: Ian Bourg, Amilcare Porporato, Howard Stone, and Xinning Zhang

At a Glance

The objective of this project is to understand a key control on the stability of soil carbon: its protection by fine-grained minerals. Field experiments suggest that roughly half of the organic carbon present in soils is protected from microbial respiration by certain fine-grained minerals, but the mechanism of this mineral protection remains unknown. The Bourg, Porporato, Stone, and Zhang groups are using a unique combination of experimental and simulation approaches, spanning spatial scales ranging from molecules to landscapes, to elucidate this mechanism. These results will enable more accurate Earth System Model predictions of soil carbon dynamics and inform practical strategies for enhancing the soil carbon sink.


Research Highlight

Soils are the largest pool of carbon near the Earth’s surface, roughly as large as the atmosphere, biosphere, and surface ocean combined. Each year, soils take up about 61 GtC (gigatons of carbon), primarily as plant residues and root exudates, while emitting about 59 GtC of carbon, primarily as microbial carbon dioxide (CO2) and methane (CH4) (Lehmann & Kleber, 2015). Because of the imbalance between soil carbon uptake and emissions, soils presently act as a net carbon sink that absorbs about 2 GtC per year (~20% of anthropogenic CO2 emissions). This carbon sink is sensitive to changes in temperature, rainfall, and soil management, but a detailed predictive understanding of its magnitude and evolution remains elusive (Sulman et al., 2014; Hicks Pries et al., 2017; Mayer et al., 2018).

Recent meta-analyses of field-scale data reveal that a key predictor of soil carbon storage is the abundance of certain fine-grained minerals, in particular, smectite clay minerals in temperate soils and Fe and Al oxides in tropical and boreal rainforest soils (Rasmussen et al., 2018). A primary goal of this project is to decipher the fundamental mechanisms that cause this correlation in order to inform new soil carbon sequestration approaches. The mechanism likely involves couplings between soil microbiology, clay surface geochemistry, and the hydrologic cycle at various spatial and temporal scales that remain largely unexplored. Progress towards this goal is enabled by a unique combination of field-scale models (Porporato), microfluidic experiments (Stone), microbial ecology techniques (Zhang), and atomistic simulations (Bourg), which are detailed below.

The Porporato group is developing field-scale models of the evolution of soil hydrology, microbiology, and geochemistry. The models consist of ordinary differential equations governing the time evolution of the mass of different species in different soil layers. The processes are modeled through specific kinetic laws calibrated through data from laboratory experiments and field observations. This effort is enhanced by the Porporato group’s access to unique datasets of laboratory and field measurements (Barcellos et al., 2018) through the network of long-term field experiments of the National Science Foundation- funded Critical Zone Observatories. The research focuses particularly on understanding, first, how the formation and transport dynamics of soil clays are coupled to soil hydrology and carbon decomposition and, second, how the redox cycling of iron is used in microbial carbon decomposition in humid tropical soils that experience frequent fluctuations in soil O2. The second topic is also relevant to CMI efforts on CH4, as illustrated by preliminary estimates from field sites in Puerto Rico showing that Fe cycling can strongly inhibit the microbial production of CH4 in tropical soils (Hall et al., 2013).

A key breakthrough in 2018 was the development of a model that captures the interaction between the hydrologic cycle and the pace of soil biogeochemical processes, in particular, the formation and transport of clays and the Fe-cycle (Calabrese et al., 2018; Calabrese & Porporato, 2018). Preliminary results shown in Figure 6.1 illustrate the ability of the clay transport model to predict the clay distribution profile. This profile originates from the balance between clay leaching from water percolation events and surface erosion processes, which tend to remove clay particles from above. The figure also illustrates how the time evolution of dissolved Fe2+, which results from the reductive dissolution of Fe3+ and associated carbon decompositions, is controlled by daily soil moisture fluctuations in a tropic soil.

Figure 6.1. (a) Soil clay content with depth (% weight of total clay) from texture measurements (red points) collected at the Calhoun Critical Zone Observatory, South Carolina, and from the model of clay translocation (solid line). The dashed line represents the estimated residence time of clays. (b) Time evolution of modeled soil moisture (top) and dissolved Fe2+ (bottom) for soils at the Bisley Watershed in Luquillo Experimental Forest, Puerto Rico. The model was parameterized with laboratory experiments conducted on soil samples from the field site.

The Stone group is applying its expertise in microfluidic technology to obtain direct evidence on how bacterial growth (current experiments utilize Pseudomonas aeruginosa) and activity are affected by clay minerals and aggregates. Preliminary results suggest an intriguing potential new mechanism, whereby the slowdown of soil carbon decomposition by clay may be caused by the clay-induced flocculation (clumping) and sedimentation of bacteria (perhaps because this flocculation may inhibit the transport of bacteria through porous soil). Preliminary experiments (Figure 6.2) and confocal microscopy observations support the hypothesis that clay minerals may induce bacterial flocculation and sedimentation. Future microfluidic experiments will involve tracking bacterial transport and growth in model soils with and without clay, and so will provide direct evidence to test the current soil-carbon-clay interaction hypothesis. Initial steps in the research have benefitted from frequent discussions with the Bourg and Zhang groups.

Figure 6.2. (Left) Evolution of optical density, which is proportional to bacterial cell density in the solution containing bacteria, Pseudomonas aeruginosa, with clay (circles) and without clay (crosses). (Right) Aggregation of bacteria in culture solution with clay particles, shown as bright red dots in the middle of the image. Scale bar: 25 microns.

The Zhang group uses microbial ecology, physiology, and biogeochemical measurement approaches to study the relationship between the production greenhouse gases such as CH4 and CO2, microbial communities, and reactive metal minerals in hydrologically dynamic, carbon-rich environments like wetlands. This work, funded by the CMI Methane project, is highly complementary to ongoing research in the Stone, Porporato, and Bourg groups, as it provides a contrasting view on carbon cycling in clay- poor environments to enable a full understanding of interactions between microbes, minerals, carbon, and water. An exciting discovery in 2018 was the identification of a metal-based, potentially new abiotic mechanism for CH4 production in wetlands. In addition to research on wetland methane (further outlined in the Methane portion of this CMI annual report), the Zhang group works closely with the Stone group to support their experimental investigations.

The Bourg group is using all-atom molecular dynamics (MD) simulations to gain fundamental insight into mineral-organic association. The simulations use the supercomputers of the U.S. Department of Energy to solve Newton’s equations of motion for mineral-water-organic systems of about 105 atoms using semi-empirical models of all relevant interatomic interactions. The research focuses particularly on understanding, first, how organic molecules interact with mineral surfaces and, second, how organic molecules influence the wettability of mineral surfaces by water versus non-aqueous fluids (e.g., air or CO2).

A key breakthrough in 2018 was the development and validation of an MD simulation methodology that accurately predicts the affinity of organic molecules for smectite clay surfaces (Willemsen et al., 2018).

Specifically, simulations and supporting experiments examined the adsorption of 10 different organic molecules with a range of molecular weights and hydrophilicities (phthalate esters, polycyclic aromatic hydrocarbons, and perfluorinated alkyl substances) on a stack of two smectite clay nanoparticles. Results showed that organic molecules consistently have a strong affinity for the clay surfaces (Figure 6.3). Contrary to the predominant theory of soil carbon protection by mineral surfaces, this affinity is primarily entropic, i.e., the tendency of soil carbon to become associated with mineral surfaces is determined by hydrophobicity rather than by the interaction of specific functional groups with the mineral surface.

Figure 6.3. (Top) Snapshot of a simulation cell containing two smectite clay nanoparticles (1 nm thick particles with 0.6 nm thick interlayer nanopores) in contact with bulk-liquid-like water (0.1 M CaCl2 solution). The clay structure is shown as red, yellow, and orange polyhedra (SiO4, AlO6, and MgO6, respectively); water molecules are shown as small blue dots; electrolyte ions (Ca, Cl) are shown as purple and green spheres; adsorbed organic molecules (in this case, phthalate esters) are shown as pink, blue, and white spheres. (Bottom) Molecular dynamics simulation reconstruction of the contributions of enthalpic interactions (i.e., attraction of specific organic moieties to the mineral surface) and entropic interactions (predominantly water’s tendency to expel hydrophobic solutes) to the overall free energy of adsorption are reported as blue diamonds and yellow circles for six phthalate esters with various shapes and sizes. Greater negative free energies of adsorption indicate greater affinity for the mineral surface.

Another important breakthrough in 2018 was the demonstration that MD simulations can predict details of the wettability of mineral surfaces by water versus non-aqueous fluids that are challenging to resolve using experiments (Sun and Bourg, 2018). In particular, MD simulations were used to predict the influence of organic matter and multiphase flow dynamics on water imbibition and drainage in small pores in the presence of supercritical CO2 (Figure 6.4). The results have implications for understanding the coupled fluxes of water and carbon in soils, but also broader implications for predicting the flow of water and CO2 in geologic formations during carbon capture and storage and CO2-enhanced oil recovery.

Figure 6.4. Snapshot of a simulation cell containing water (blue sticks), CO2 (gray spheres), and decanoic acid molecules (green spheres) between two parallel quartz surfaces. The simulations predict the existence of a stable adsorbed water film on the quartz surfaces and a water meniscus bridging the two quartz surfaces. Calculation of the local stresses (not shown) reveals that the capillary pressure difference between the two fluids is consistent with macroscopic-scale theories (i.e., the Young-Laplace equation). In the system shown in the figure, a uniform force fx applied to the fluid atoms drives fluid flow from left to right. As a result, the organic molecules accumulate at the trailing edge of the CO2 bubble (left insert) and the adsorbed water film thickness decreases with distance from the leading edge of the bubble (lower inserts).



1 Hall, S.J., W.H. McDowell, and W.L. Silver, 2013. When wet gets wetter: decoupling of moisture, redox biogeochemistry, and greenhouse gas fluxes in a humid tropical forest soil. Ecosystems, 16: 576. doi. org/10.1007/s10021-012-9631-2.

2 Hicks Pries, C.E., C. Castanha, R. Porras, and M.S. Torn, 2017. The whole-soil carbon flux in response to warming. Science 355: 1420.

3 Lehmann, J., and M. Kleber, 2015. The contentious nature of soil organic matter. Nature, 528, 60. doi. org/10.1038/nature16069.

4 Mayer, A., Z. Hausfather, A.D. Jones, and W.L. Silver, 2018. The potential of agricultural land management to contribute to lower global surface temperature. Science Advances 4: eaaq0932. aaq0932.

5 Rasmussen, C. et al., 2018. Beyond clay: towards an improved set of variables for predicting soil organic matter content. Biogeochemistry 137: 297.

6 Sulman, B.N., R.P. Phillips, A.C. Oishi, E. Shevliakova, and S.W. Pacala, 2014. Microbe-driven turnover offsets mineral-mediated storage of soil carbon under elevated CO2. Nature Climate Change 4:1099. doi. org/10.1038/NCLIMATE2436.


The CMI Methane Project: Understanding the Biogeochemical Controls, Sources, and Sinks of Methane

Principal Investigators: Xinning Zhang, Elena Shevliakova, and Vaishali Naik

At a Glance

Methane (CH4) is the second most important anthropogenic climate forcer after carbon dioxide (CO2). Atmospheric methane has risen to levels roughly 150% above preindustrial concentrations due to human activities and continues to rise despite a short period of stabilization during 1999-2006 (Dean et al., 2018). Since the amount of methane in the atmosphere reflects the balance of chemical and biological processes that produce and consume the gas, efforts to decipher current trends and assess future emissions necessitate accurate accounting of different methane sources and sinks, as well as a clear understanding of their variability across temporal and spatial scales. These remain fundamental challenges for the scientific community. Since 2017, CMI has supported three complementary projects that address the largest unknowns in methane cycling: one experimental project focuses on the critical issue of wetland methane emissions (Project 1), and two global-scale modeling projects aim to quantify the individual sources, sinks, and variations of methane associated with land (Project 2) and the atmosphere (Project 3).


Research Highlight

Project 1: Biogeochemical controls on wetland methane emissions.

The CMI Wetland Project, led by Zhang, uses measurements to identify the biological and chemical mechanisms that promote methane emission from wetlands, dominant but highly variable sources of methane that are predicted to play a critical role in carbon-climate feedbacks (Dean et al., 2018). Wetland methane emission is shaped by a complex and poorly understood interplay of microbial, hydrological, and plant-associated processes, which vary in time and space. The set of conditions promoting the greatest methane emission from wetlands remain unknown.

CMI research has identified oxygen transitions related to hydrology as critical for methane production in wetlands (Wilmoth et al., in prep). Exposure of peat to oxygen dramatically enhances methane production compared to peat incubated under continuously anoxic conditions (Figure 7.1). Analyses of peat chemistry and microbiology indicate multiple mechanisms for oxygen-enhanced methane production. Thus, contrary to scientific dogma that microbial methane production is confined to water- logged, oxygen-free, deep wetland zones, CMI research identifies partially saturated wetland soils and peats, located near the water table and exposed to significant oxygen, as critical locations for wetland methane production ( the “emerging model” in Figure 7.1).

CMI research suggests a possible new feedback between increasing temperatures, hydrologic variability, and wetland methane emissions because oxygen levels in wetlands are intimately linked to hydrology (Reddy and Delaune, 2008). Ongoing work aims to evaluate this new carbon-climate feedback using a combination of observational and modeling approaches. Within a policy framework, plans to restore wetlands by rewetting need to consider the possibility that restored wetlands could produce significantly more methane relative to their pre-disturbed forms (Abdalla et al., 2016) due to oxygen-enhanced methane production. Furthermore, our work suggests that carbon mitigation strategies aimed at minimizing wetland water table variability may help limit wetland methane emissions.

Figure 7.1. Oxic-anoxic transitions promote methane formation. (A) Methane from peat incubated under continuously anoxic conditions (gray bar), under anoxic conditions after exposure to 5% oxygen (yellow bar) or 10% oxygen (orange bar). Error bars, standard error of pooled, replicate incubations of surface (n=3), at the water table (n=3), and deep peat (n=2 or 3). (B) Schematic of emerging and classical models of methane production in wetlands. (Wilmoth et al., in preparation.)


Project 2: Global Model of Methane Emissions from Wetlands

This project, led by Shevliakova, aims to 1) implement and evaluate the capability to simulate CH4 and CO2 emissions from different wetland ecosystems in the terrestrial component of the Geophysical Fluid Dynamics Laboratory (GFDL) Earth System Model (ESM), and 2) to explore how uncertainty in climate and in ecological characteristics affect uncertainty in CH4 and CO2 emissions from global wetlands in past and future climates.

The latest GFDL land model, LM4, includes a dynamic representation of vegetation (Weng et al., 2015). LM4 explicitly represents interactions between microbes and soil organic matter in a new vertically resolved soil biogeochemistry model CORPSE (Sulman et al., 2014) that also captures the effect of land use on vegetation and soil. Advanced hydrological features of LM4 include frozen soil dynamics, continuous vertical representation of soil water including water table depth, horizontal transport of runoff via rivers to the oceans, and a lake model (Milly et al., 2013). Soil moisture has a crucial control on soil carbon storage and methane production. Consistent with Project 1 findings, its impact on methane emissions is highly nonlinear due to complex interactions between levels of anoxia, microbes, and carbon in soils and wetlands.

We developed a new land component with an explicit treatment of the four microbial groups for use with GFDL ESMs, which simulate anaerobic decomposition and methane production (Figure 7.2). We integrated new methane consumption and production components, along with gas diffusion (e.g., O2, CH4 and CO2), through vertical soil layers within the GFDL vertically resolved soil hydrology model. New numerical approaches enable computation of changes in soil moisture, ice and gas concentrations under a wide range of environmental conditions. We are now evaluating the coupled soil carbon-water- methane model on individual observational sites and in global stand-alone land simulations.

A full-featured, vertically-resolved hydrological and biogeochemical soil model, which accounts for microbial dynamics, is a critical tool to resolve temporal and spatial variability of methane sources and sinks and make projections about future greenhouse gas under changing climate and land use. Interactive land–atmosphere methane fluxes will enable evaluation of new biogeochemical feedbacks between changes in wetlands and permafrost on climate, which have yet to be included in Earth system projections.

Figure 7.2. Structure of the new methane production and consumption component of the GFDL land model. DOC is dissolved organic carbon, simulated by LM4.


Project 3: Global Chemistry-Climate Modeling of Atmospheric Methane Cycle

This project, led by Naik, addresses the key question: what are the drivers of atmospheric methane trends and variability at the decadal to centennial time scales? An imbalance in methane sources and sinks leads to changes in atmospheric methane levels. Observations have revealed complex temporal variations in atmospheric methane growth over the past three decades that have been challenging and often controversial to attribute to specific methane sources or sinks (Crill & Thornton, 2017; Dean et al., 2018). The GFDL atmospheric chemistry group has applied a process-based global chemistry-climate model (GFDL-AM4) that simulates changes in methane sources as well as the primary methane sink within a unified framework to explore the contribution of individual sources and sinks on observed trends and variability in methane from 1980 to 2014.

GFDL-AM4 model results suggest that the methane stabilization during the period of 1999-2006 was mainly due to increasing methane emissions balanced by increasing methane sink, primarily due to its reaction with hydroxyl (OH) radical. Post-2006, increasing emissions outweighed any changes in sink, resulting in the renewed growth of methane as shown in Figure 7.3 (He et al., in preparation). Ongoing work using methane isotope observations aims to further pin down the rise in methane to specific source types (biogenic, anthropogenic, or pyrogenic).

A quantitative understanding of the roles of individual sources and sinks in driving methane variability is a crucial precursor to designing effective mitigation strategies to address near-term climate warming.

Model results imply the need for accurate bottom-up estimates of methane emissions to improve quantitative analyses of the global methane budget and prediction of atmospheric methane. The most important sources of uncertainty in emissions are wetlands and freshwater systems (Saunois et al., 2016). Future work involving the coupling of improved terrestrial wetland emissions model (Project 2 with input from Project 1) with GFDL’s chemistry-climate model will advance the characterization of the drivers of atmospheric methane variability.

Figure 7.3. Atmospheric methane trends driven by an imbalance in global methane budget. The left y-axis shows global average surface methane concentration from observations (NOAA Earth System Research Laboratory) and that simulated by the NOAA GFDL-AM4 model (black lines). The right y-axis shows the net methane flux calculated as the difference between optimized global total methane emissions and total methane sink in the model (red line).



1 Abdalla, M. et al., 2016. Emissions of methane from northern peatlands: a review of management impacts and implications for future management options. Ecology and Evolution, 6(19): 7080–7102. doi. org/10.1002/ece3.2469.

2 Crill, P. M. and B.F. Thornton, 2017. Wither methan in the IPCC process? Nature Climate Change, 7(10): 678-680.

3 Dean, J.F., et al., 2018. Methane feedbacks to the global climate system in a warmer world. Reviews of Geophysics, 56(1): 207-250.

4 He., J., V. Naik, L. Horowitz, E. Dlugokencky, and K. Thoning, 2019. Evolution of the global methane budget over the 1980-2014 period using the GFDL-AM4, in preparation.

5 Milly, P.C.D., S.L. Malyshev, E. Shevliakova, K.A. Dunne, K.L. Findell, T. Gleeson, Z. Liang, P. Phillips, R.J. Stouffer, and S. Swenson, 2014. Enhanced Representation of Land Physics for Earth-System Modeling. J. Hydrometeorology 15:1739-1761.

6 Reddy, K. and R. DeLaune, 2008. Biogeochemistry of Wetlands. Boca Raton: CRC Press.

7 Saunois, M. et al., 2016. The global methane budget 2000-2012. Earth Syst, Sci. Data, 8(2): 697-751. doi. org/10.5194/essd-8-697-2016.

8 Saunois, M. et al., 2017. Variability and quasi-decadal changes in the methane budge over the period 2000-2012. Atmos. Phys. Chem., 17(18): 11135-11161.

9 Sulman, B., R. Phillips, A. Oishi, E. Shevliakova , and S. Pacala, 2014. Microbe-driven turnover offsets mineral-mediated storage of soil carbon under elevated CO2. Nature Climate Change 4:1099-1102. doi. org/10.1038/nclimate2436.

10 Weng, E.S., S. Malyshev, J.W. Lichstein, C.E. Farrior, R. Dybzinski, T. Zhang, E. Shevliakova, and S.W. Pacala, 2015. Scaling from individual trees to forests in an Earth system modeling framework using a mathematically tractable model of height-structured competition. Biogeosciences, 12(9): 2655-2694. doi. org/10.5194/bg-12-2655-2015.

11 Wilmoth, J.L., J. Schaefer, D. Schlesinger, J. Shoemaker, P. Hatcher, S. Myneni, and X. Zhang. Oxygen variability stimulates methane production by wetland peats. In preparation.


Carbon Capture and Storage Strategies in the U.S., China, and India

Principal Investigator: Michael Celia

At a Glance

In the past year, the Deep Subsurface research group has developed a research strategy focused on carbon capture and storage (CCS) opportunities in the United States, China, and India. For each, a targeted project has been identified, consistent with the current state of development in each location.


Research Highlight

United States. The recent study of Edwards and Celia (PNAS, 2018) focused on the impact of the 45Q Tax Credit on potential pipeline construction in the Midwest, with a focus on low-capture-cost sources like ethanol plants (Figure 8.1). That study is now being expanded in two ways. First, the study team has begun to collaborate with a group working on one of the U.S. Department of Energy CarbonSAFE projects. That project, headed by CMI alumnus Andrew Duguid, focuses on building a pipeline in the Kansas- Nebraska area, with a major component of the overall system being stacked carbon dioxide (CO2) storage involving a mix of CO2 enhanced oil recovery (EOR) and dedicated saline aquifer storage. This project provides local-scale detail on what is required to develop a real pipeline project. The team has also begun a second project, which will examine how development of low-cost capture technologies for the power sector, like NetPower’s Allam Cycle power plants, could influence the dynamics of pipeline construction and subsurface utilization over the next few decades. If the analysis is expanded to include the power sector as part of the low-capture-cost options, the amount of CO2 captured and stored increases by at least an order of magnitude. In this case, additional storage options along the pipeline route need to be identified and characterized, with the overall mix of storage involving both EOR and dedicated storage. The local-scale work with Andrew Duguid, and his analysis of stacked storage, will provide important inputs to this larger-scale study. The team expects to produce an updated report of the infrastructure development that includes low-cost capture for new power plants, with the captured CO2 used for both EOR and dedicated storage.

China. In China, the Celia group is focusing on the northwestern province of Xinjiang, where there appears to be substantial subsurface storage capacity as well as a range of CO2 sources. China’s policy to move coal activities from the populated east to the less populated west, together with the large coal resources in the Xinjiang region, make this a good target for large-scale CCS. One of the main challenges to a detailed CCS assessment in this region is availability of reliable data for subsurface characterization. Through contacts and visits to the China Geological Survey, the China Geosciences University, and the Chinese Academy of Sciences Institute for Soil and Rock Mechanics, the team has been able to access geological data that allows for an initial description of the Junggar Basin in Xinjiang. The team also has inventories of CO2 sources in the basin. Using this information, they are undertaking a preliminary analysis of dynamic storage capacity in the region. Because of the strategic location of the province along the Belt and Road corridor, the team is also looking into aspects of the Belt and Road Initiative (BRI), a development strategy adopted by the Chinese government, to try to understand CCS potential within the broader infrastructure development in the BRI.

India. Because large-scale storage in India is likely to require injection into the Deccan Traps, a large basalt formation, a project has been initiated to model CO2 injection and reactive transport in these highly reactive rocks. Celia and his team are beginning with an analysis of the completed field experiments in Iceland and in the northwest of the United States, and will develop appropriate modeling tools in collaboration with the subsurface computational group at the University of Stuttgart, Germany, where a center of excellence in high-performance computing provides outstanding computational options.

Figure 8.1. Blue lines show the optimal pipeline network to connect low-capture-cost sources, in this case, ethanol plants in the upper Midwest of the U.S., to the demand for CO2 (for EOR) in west Texas. The 45Q Tax Credit plays a central role in the economics of the proposed pipeline development. Details can be found in Edwards and Celia (2018).



1 Aslannejad, H., S.M. Hassanizadeh, and M.A. Celia, 2019. Characterization of the Interface between Coating and Fibrous Layers of Paper. Transport in Porous Media, accepted for publication.

2 Bandilla, K. and M.A. Celia, 2019. “Numerical Modeling of Fluid Flow during Geologic Carbon Storage” in Science of Carbon Storage in Deep Saline Formations: Process Coupling across Time and Spatial Scales. Newell, P. and A.G. Ilgen (Eds.). Elsevier (Amsterdam).

3 Edwards, R.W.J. and M.A. Celia, 2018. Infrastructure to enable deployment of carbon capture, utilization, and storage in the United States. PNAS, 115(38): E8815-E8824.

4 Riddick, S.N., D.L. Mauzerall, M.A. Celia, M. Kang, K. Bressler, C. Chu, and C.D. Gun, 2019. Measuring Methane Emissions from Abandoned and Active Oil and Gas Wells in West Virginia. Science of the Total Environment, 651: 1849-1856.

5 Tao, Y., B. Guo, K. Bandilla, and M.A. Celia, 2019. Vertically-integrated Dual-continuum Models for CO2 Injection in Fractured Geological Formations. Computational Geosciences, accepted for publication.


Wither Heat: A Rational Path to Fast Charge

Principal Investigator: Daniel Steingart

At a Glance

Recent research findings by the Steingart group indicate that a lithium ion battery may be charged considerably faster and with minimal degradation if the charging is done at a higher temperature than usually considered (40 ̊C to 50 ̊C).


Research Highlight

The rapid decrease in the cost of lithium ion batteries has increased interest in passenger electric vehicles (EVs) for both consumers and automakers, but it is unclear that cost alone can trigger the mass adoption of electric vehicles. At an average efficiency of 3.5 miles/kWh (Helms et al., 2010), an optimistic target gravimetric pack energy density of 300 Wh/kg (Wood et al., 2015), and a final pack cost of $100/kWh (Ciez & Whitacre, 2016), the average cost of a battery pack for a 400-mile vehicle would be over $11,000 and the average mass over 380 kg. The average daily distance driven is under 50 miles, and most drivers feel that their “range anxiety” is alleviated with a capacity of 200 miles (Neubauer & Wood, 2014). This indicates that except for a rare all-day drive, the typical 400-mile EV battery would be a factor of two to eight times too expensive and massive, adding a cost of $4,000-$7,000 and an excess mass of over 100 kg.

If a 200-mile battery pack could be recharged as fast as the refueling of a petrol-fueled car, this might obviate the need for a 400-mile battery pack. For the average driver making few all-day (400-mile) drives per year, this would simply add one extra recharge stop to the full-day trip. But even with this compromise, there is a paradox. The energy density and cost profile assumes a battery designed to be discharged no faster than in three hours (200 miles/60 mph = 3.33 hours). The rate at which a battery can be safely discharged is the rate at which a battery can be safely charged, so this would indicate that an all-day trip requires two rest stops of three hours per day to fully recover all 200 miles of the battery, or an average recovery rate of one mile per minute of charge. In comparison, petrol refueling provides a range recovery of 100 miles per minute of charge.

For a traditional battery to enable the charge acceptance required to achieve even a recovery rate of 20 miles per minute of charge, the energy density of the battery would likely halve, and the cost would increase by at least a factor of two due to the thinner electrodes required per unit of current collector. And thus the battery capable of an ambient-temperature fast charge costs no less and weighs no less than a battery capable of driving 400 miles. So there would appear to be a fundamental zero gain in the engineering paradigm of practically trading cost for speed.

In the past year of this CMI effort, the Steingart group has been experimentally verifying and dissecting the aforementioned paradox, and they have confirmed this to be true for batteries held in an isothermal condition. But with further experimentation, they feel that there may be a practical methodology for recovery at 10 to 15 miles per minute of charge on a standard high energy-density pack design without the need for significant extra equipment. The key variable is temperature.

In a lithium ion battery, the most significant limitation to the safe and non-destructive recharge operation is the fundamental diffusion behavior of the slowest transport in the cell: the diffusivity of lithium in the solid state. And upon recharge, the diffusion limitation of the typical graphite anode can lead to the unwanted deposition of lithium metal atop the electrode, as opposed to the intercalation of lithium within the electrode (Figure 9.1). This is a lithium cobalt oxide cathode, graphite anode cell with an ethylene carbonate/dimethyl carbonate (EC/DMC) blend electrolyte utilizing lithium hexafluorophosphate (LiPF6). The cell is not designed to store lithium metal—and lithium, being the most reducing compound in existence, will react with its surrounds, at best significantly reducing the capacity of the battery and at worst creating a fire.

Figure 9.1. For a 210 mAh graphite-LCO cell with an LiPF6 EC/DMC electrolyte, the onset of unwanted lithium deposition during charging as a function of charge rate and temperature. The red dots are conditions where lithium definitely plates, yellow dots are light plating, and black dots are no plating.

The standard method to increase the diffusion rate of a material is simply to raise its temperature, as:

D(T) = D0e^(-EA/RT)

If only it were so simple. Within EV batteries, temperature is well controlled and maintained near 25 ̊C because batteries are designed in labs at 25 ̊C, and temperature excursion, high and low, can lead to unwanted behavior. Significant temperature excursions above 90 ̊C can lead to auto-catalytic thermal runaway of the cathode and subsequent ignition of the flammable organic carbonate solvents of the electrolyte (Koch et al., 2018). Slight temperature increases can trigger/enhance non-faradaic side reactions (Gyenes et al., 2015) within the cell, again following Arrhenius behavior

k(T) = Ae^(-EA/RT)

Thus it would seem that the damage present in allowing faster lithium transport at higher temperatures would be negated by the damage incurred by the increase in temperature itself.

Figure 9.2. For a 210 mAh graphite-LCO cell with an LiPF6 EC/DMC electrolyte: (a) diffusion vs. voltage and capacity as a function of temperature for a fresh cell; (b) maximum and minimum diffusion rates for a given cell as a function of temperature; (c) maximum and minimum diffusion profiles at various temperatures for a cell cycled at an hour rate at 20 ̊C; (d) diffusion profiles at various temperatures for a cell cycled at a hour rate at 60 ̊C.

The Steingart group believes, however, that they have found a window of operation where all of the competing factors above are simultaneously true yet a battery can be recharged quickly without incurring significant damage. Their hypothesis: if a cell is heated to a moderate temperature (40 ̊C to 50 ̊C) for a short period of time (20 minutes or less) and the cell is not held at high potential during this period of time, it can be quickly charge (10 minutes to 80% charge recovery) without incurring damage. This means that if a battery management system raises the temperature of the cell but avoids the constant voltage step, a cell can be charged at a fast rate while avoid voltage-driven degradation.

This is in contrast to traditional charging protocols, known as constant current, constant voltage (CCCV), which hold the cell near room temperature, but charge to a cutoff voltage at constant current, then hold at that constant voltage until a certain time passes or the current drops below given value. The CCCV protocol is designed to preserve the life of the cell while completely charging the battery. Our constant current, high temperature protocol (CCHT) preserves the life of the cell while charging the cell to 80% capacity. We trade maximum range for a fast effective range recovery in this protocol.

In the past year the Steingart group has collected significant evidence to support this hypothesis. The first is the determination of the change in diffusion coefficient as a function of state of charge and state of health. Figure 9.2 indicates that the diffusivity of a given cell (in this case, graphite vs. LCO in LiPF6:EC:DMC) increases by up to two orders of magnitude by raising its temperature from 20 ̊C to 40 ̊C, and that after 1,000 cycles without a high voltage constant-voltage step, the capacity of the cell moderately degrades (less than 10%) from the fresh case at a temperature as high as 60 ̊C.

This behavior was consistent and reproducible for over 100 cells, which included three different chemistries and three different form factors (210 mAh pouch, 1500 mAh pouch, and 2,400 mAh cylindrical).

Figure 9.3 illustrates a comprehensive relationship. When a given cell type, in this case the 2,400 mAh cell mentioned above, is metered on its total throughput, the protocol we have determined stands out.

Figure 9.3. For a 2,400 mAh graphite-NMC 532 cell with an LiPF6 EC/DMC electrolyte, capacity vs. total charge throughput at various temperatures.

The cells cycled at 60 ̊C without a constant voltage retain over 90% of their initial capacity at charge rates faster than one hour (as fast as 15 minutes) for over 1,000 cycles.

This is by no means authoritative or complete. Further ex-situ chemical analysis needs to be completed to study other forms of damage that might have occurred but have not yet presented as capacity fade. The implementation of this protocol across an entire pack of cells may be intractable. And the efficacy of this protocol on higher nickel content cathode needs to be confirmed.

From this study, however, the Steingart group is encouraged to further examine and probe the capability of fast charging a cell designed for a three-hour-plus discharge by tuning its charge temperature, and perhaps exploring electrode and cell engineering improvements that might exploit this asymmetric treatment. With further study and analysis, this may be a practical method for solving the paradox of a low-cost, high energy-density, 200-mile battery that can be recharged periodically at a rate recovery of over 15 miles per minute charged.

This work was done by Clem Bommier, funded by the CMI, and Andrew Kim, funded through the ACEE by American Tower. Andrew Kim was exploring damage as a function of environmental conditions such as temperature and humidity.



1 Ciez, R.E. and Whitacre, J.F., 2016. The cost of lithium is unlikely to upend the price of Li-ion storage systems. Journal of Power Sources, 320: 310–313.

2 Gyenes, B., Stevens, D.A., Chevrier, V.L., and Dahn, J.R., 2015. Understanding anomalous behavior in coulombic efficiency measurements on Li-ion batteries. Journal of the Electrochemical Society, 162(3): A278–A283.

3 Helms, H., Pehnt, M., Lambrecht, U., and Liebich, A., 2010. Electric vehicle and plug-in hybrid energy efficiency and life cycle emissions. 18th International Symposium.

4 Koch, S., Fill, A., and Birke, K.P., 2018. Comprehensive gas analysis on large scale automotive lithium-ion cells in thermal runaway. Journal of Power Sources, 398: 106–112.

5 Neubauer, J. and Wood, E., 2014. The impact of range anxiety and home, workplace, and public charging infrastructure on simulated battery electric vehicle lifetime utility. Journal of Power Sources, 257: 12–20.

6 Wood, D.L., III, Li, J., and Daniel, C., 2015. Prospects for reducing the processing cost of lithium ion batteries. Journal of Power Sources, 275: 234–242.