Authors: Paul Palmer, Mathew Williams and Shaun Quegan

1 Introduction

The twentieth century saw a rapid rise in atmospheric CO₂ concentrations, due largely to anthropogenic activity. This trend is set to continue in the twenty-first century with increased emissions from deforestation, fossil fuel burning and cement manufacture due to increased global economic activity. The current century is likely to see a range of responses from land and ocean carbon pools to the associated climate warming (e.g., net release of carbon from peat reservoirs due to melting of permafrost). It is therefore critical to have a capability to monitor and quantitatively understand the rate of change in land and ocean carbon pools.

Observed variations in atmospheric CO₂ concentrations measure the net surface-atmosphere exchange. Inferences on the surface fluxes giving rise to these concentrations currently rely on inversing models of atmospheric transport and mixing, combined with CO₂ concentrations measured by the ground-based network. This ill-posed problem requires regularization, typically provided by prior CO_­2 concentrations derived from estimates of surface CO₂ fluxes inferred from in situ data, and has several problematic features: a) the sparseness of the data network, particularly over tropical landmasses; b) biases introduced by inaccurate modelling of atmospheric transport and mixing processes; c) covariance between surface fluxes and boundary layer meteorology (rectification artefacts). Diffusive atmospheric transport amplifies these uncertainties. Recent improvements in the sampling density from ground-based and aircraft-based instrumentation have improved flux estimates over northern hemisphere landmasses. However, the tropical landmasses, thought to be large sinks of CO₂, remain one of the most under-sampled regions in the world for carbon cycle research. Space-borne measurements of CO₂ have the potential to fill these gaps, providing global coverage and daily observations. Although less accurate than in situ measurements and challenging to use, these data seem certain to offer new quantitative insights into the working of the global carbon cycle.

2 Outstanding Questions and Uncertainties in Carbon Cycle Science

Global coupled carbon-cycle climate models disagree on the strength of carbon-climate coupling (Friedlingstein et al 2006), reflecting fundamental gaps in our knowledge of C cycling. These differences between models are clear indicators of major uncertainty in predictions of the future of the Earth system. Most information about the C cycle is gathered at either global scales or local scales (e.g. eddy flux towers). Regional datasets, lying between these scales, are sparser and more uncertain (by regional, we mean observations that sample the seasonal activity of an ecoregion, such as the Amazon basin, Siberian taiga, Alaskan tundra and African deciduous woodlands (~10⁶ km²)). Lack of regional data is one of the major issues preventing severe testing and subsequent winnowing of the coupled models.

Table 1. Characterizing the strength of various constraints on the global C budget (M. Gloor).

• 
  What is the relative importance of the land-biosphere CO₂ fluxes over different latitudes? There is still argument about the location of hypothesized terrestrial C sinks.
  
• 
  Which processes control land-biosphere fluxes in different ecoregions, how do these processes operate and how will they respond to climate change?
  
• 
  How much carbon is emitted from biomass burning, deforestation and land use change?
  
• 
  What is the magnitude and distribution of air-sea exchange of CO₂?
  
• 
  How much will large stores of C in northern high latitudes contribute to future carbon emissions?
  

The overarching objective of this document is to define a UK programme to resolve these questions using novel measurements of atmospheric CO₂ from OCO/GOSAT. This programme will provide the tools to exploit OCO/GOSAT data and combine these with state-of-the-art models and surface and aircraft observations. The outcome will be improved understanding of processes that control the distributions of CO₂ and ensuing improvements in attribution and prediction of atmospheric CO₂. Such a programme will focus a wide range of NERC expertise on unlocking the potential of the OCO and GOSAT missions, thus enhancing the scientific return of NERC-funded carbon cycle research and the profile of NERC in the international arena. In this document we outline the science, the existing UK expertise, and the current gaps and opportunities associated with using CO₂ measurements from space.

3 The OCO and GOSAT Missions

The OCO mission is dedicated to measuring space-based CO₂ columns at sufficient precision to identify CO₂ sources and sinks on regional scales over the globe and quantify their variability over the seasonal cycle. It has a 2-year design lifetime with launch scheduled for December 2008. OCO will join the EOS Afternoon Constellation (A-Train), flying in a sun-synchronous polar orbit with a constant 1:26 p.m. local solar time flyover, a 16-day (233 km altitude orbit) repeat cycle and near global sampling.

The Japanese GOSAT mission is led by JAXA, the National Institute for Environmental Studies (NIES) and the Ministry of the Environment (MOE), and will be launched in 2008 with a nominal lifetime of four years. Its main objective is to halve current sub-continental scale CO₂ annual flux estimation errors. GOSAT will be launched into a sun-synchronous orbit at 666 km height and an equator crossing time of 1 pm, similar to that of OCO. As compared to OCO, GOSAT offers both CO₂ and CH₄ with greater coverage of the Earth, but at slightly lower spatial resolution (the issue of resolution is discussed further below).

The temporal and spatial resolutions quoted for the OCO and GOSAT sensors are for a single observation. However, the important quantities from the point of view of using the data are their effective resolutions in time and space after averaging or compositing the data. It is not obvious from the available literature what these resolutions will be and what will be their finest possible values, and we need greater clarity about this.

OCO Level 1 and Level 2 data will be delivered to a NASA Distributed Active Archive Center (DAAC) for open distribution no later than 6 months and 9 months after launch, respectively. Members of the OCO science team will have early access to the data for calibration and validation purposes. NASA Headquarters plans to release a NASA Research Announcement (NRA) for participation in the OCO science team within the NASA Roses 2008 program to support cal/val activities.  Such announcements usually require a designated PI for each proposed investigation, but often allow multiple investigators on each team.  A successful proposal to this NRA by a team from NCEO would provide early access to OCO data. 

3.1 The UK should identify individuals or groups charged with acting as the UK points of contact with OCO and GOSAT, who will know how to access the datasets and have detailed knowledge of their properties, especially instrument and retrieval errors properties, sampling and averaging issues. We must if possible ensure participation in the GOSAT science team.

4 Calibration and validation

The OCO team has installed a small network of Fourier Transform spectrometers (FTS) that observe the solar spectrum after passage of the direct beam through the atmospheric column. This network, called the Total Carbon Column Observing Network (TCCON, http://www.tccon.caltech.edu/), is dedicated to making accurate and precise measurements of CO2 and O2 in support of OCO and GOSAT. Figure X shows the sites of current, future and possible sites within TCCON.

The OCO science team has recently prepared a science case for the Carbon Observatory Validation Experiment (COVE). This will deploy in situ sensors on two aircraft: a high-altitude platform that can reach the tropopause (ER-2 or WB-57B) and an aircraft that will routinely measure trace gases in the boundary layer, focusing on CO₂ and other species measurable by the FTS (CH₄, CO, O₃). The high-altitude platform will also carry the GOSAT airborne simulator, a cloud imager and aerosol lidar, and a sun photometer capable of measuring in the near IR. These in situ concentration data will be used to sample atmospheric profiles of key species from the surface to the mid-stratosphere, with sampling anchored at the FTS sites at Park Falls, WI and the Oklahoma ARM site. The GOSAT simulator will provide a view of the spectra observed by the space-borne platforms, but with the geometry controlled by the experiment. The aerosol and cloud data will inform level-1 retrieval algorithm developers and testers. COVE will likely take advantage of aircraft deployments already planned for late 2009 or early 2010.

Recommendations

4.2 Use NCEO influence to raise the priority of OCO and GOSAT cal/val in NCAS activities, including inclusion in the EUFAR proposal, and formulate plans to link aircraft measurements to ground-based data, flux modelling and satellite observations of CO₂ and related species.

5 Relating atmospheric column CO₂ concentration measurements to surface fluxes

Surface flux models

The UK has great strengths in land surface flux models that estimate biospheric net primary production (NPP) and respiration fluxes, with a particular focus on the JULES land surface scheme that links the Hadley Centre, NCEO and QUEST. Although such models typically carry out ~hourly calculations for their biophysical calculations, they tend not to be used for estimating short-term dynamics, so do not routinely provide flux estimates at the hourly timescales needed for CTMs. Similar calculations are available from the CASA (Carnegie-Ames-Stanford) ecosystem model at a time resolution of 1 month, thus representing the seasonal cycle of NPP and respiration. However, CASA also provides datasets with a 3-hourly timestep by using high time-resolution ECMWF radiation and temperature data in the calculations, and daily values of NPP and respiration are available from the SiB-2 ecosystem model. The CASA and Sib-2 datasets are both at a spatial resolution of 1^ox1^o. (Typically, the spatial scale of biospheric C flux calculations is determined by the grid-scale of the climate data.)

The most recent air-sea exchange estimate, compiled from global sea-surface pCO₂ measurements, is at http://www.ldeo.columbia.edu/res/pi/CO2/carbondioxide/pages/air_sea_flux_rev1.html. The ocean element of the NCEO Carbon theme will derive models for both the physical and biological components of CO₂ exchange at the ocean surface.

Atmospheric transport modelling

Two world-class global tropospheric 3-D CTMs are supported by the NCEO: GEOS-Chem and TOMCAT (TM5 is also being used in the CCMAP project in Bristol: Marko Scholze). TOMCAT includes a detailed stratosphere and the GLOMAP aerosol scheme, and is driven by ECMWF meteorology, whereas GEOS-Chem includes a more complete aerosol-gas phase chemistry scheme in the troposphere and is driven by GMAO meteorology. Access to GEOS-Chem allows NCEO to benefit from extensive intellectual investment by US and European scientists in its development and the opportunity for continued co-operation. Both models can calculate the CO₂ dynamics needed for simulating OCO and GOSAT data.

1. Vertical transport, especially for biomass burning regions.

2. CTM differences cause significant differences in CO₂ estimates
Recommendations

5.1 Urgently develop an atmospheric transport simulation capability to support OCO and GOSAT. This should include access to and preparation for NRT ECMWF data.

5.2 Establish a clear link between atmospheric transport modelling and the land and ocean surface flux modelling activities in NCEO and elsewhere.

5.3 Establish (or help develop) a NRT nested-grid simulation of CO₂ mole fraction and column over Europe that complements the US CarbonTracker, possibly as a consortium bid. This could be based on CCDAS or the Dutch group responsible for CarbonTracker. Information on ECMWF plans for carbon flux estimates are needed to firm up any recommendation.

6 Level 2 Data Analysis

Recommendations

6.1 Establish a programme to improve CO₂ source attribution using co-emitted trace gases (NCEO Atmospheric Composition Theme).

6.3 Establish, with NCAS, an aircraft programme of trace gas measurements that can be used to help interpret observed X_CO2 observations over specific geographical regions.

7 Data assimilation of X_CO2

Two complementary assimilation approaches will be adopted in the NCEO: four-dimensional variational assimilation (4DVAR) and the ensemble Kalman Filter (EnKF). The EnKF is supported for CH₄ flux estimation by the NCEO DA Themes. An 18-month study to simulate CO₂ flux estimation from OCO using EnKF has already started at Edinburgh, funded through the NERC EO Enabling Fund. 4DVAR is supported by the NCEO DA Theme, with a focus on CO₂ flux estimation from OCO and GOSAT.

In the NCEO DA theme, 4DVAR will be used to assess the ability of OCO and GOSAT to estimate model parameters in the BETHY terrestrial biosphere model within the CCDAS framework.

Budgets for instrument, retrieval (see Recommendation 2.1) and model, errors need to form part of the DA scheme. In particular, reducing vertical transport error remains a key challenge for atmospheric transport models, especially in the boundary layer and over active convective regions. In theory, DA can account for model error but estimating this error is difficult.

Combining EO and in situ measurements of co-emitted trace gases and land surface properties will improve attribution of observed variations in CO₂ to their sources. For instance, variations in carbon-containing trace gases such as carbon monoxide (CO), methane (CH₄), carbonyl sulphide (COS) and formaldehyde (HCHO) provide additional constraints on net surface fluxes of CO₂ from the biosphere, biomass burning and fossil-fuel burning. Independent land-surface properties, such as burnt area and enhanced vegetation index, can be used to improve the prior information used in regularizing the inverse problem or used as independent verification for a posteriori flux estimates.

Recommendations

7.1 Establish a DA working group to consider (a) model error; (b) error covariance for successive measurements and different species; (c) compare DA techniques.

7.2 Obtain funding for OCO and GOSAT DA work after the current EO enabling grant expires.

8 National capability and international collaborations

• 
  Palmer - 18 month PDRA through EO enabling fund
  
• 
  Dance - project partner on Edinburgh OCO grant
  
• 
  Scholze - NCEO Theme 7 (data assimilation) PDRA
  
• 
  Woodward - NCEO Theme 2 (carbon cycle) PDRA
  
• 
  Gloor - NERC quota PhD 2007-
  
• 
  Monks/Boesch – PDRA (BNSC, Scottish Exec, UKMO); NERC quota studentship; and NCEO PDRA (tbc)
  
• 
  Bernath - NERC quota PhD (tbc)
  
• 
  Suntharalingam (RCUK fellow)
  
• 
  Engelen - GEMS
  

Instrument Science Teams

ECMWF is currently coordinating the GEMS (Global and regional Earth-system Monitoring using Satellite and in-situ data), an integrated project funded by the European Commission Framework 6 programme. GEMS includes 31 European partners and aims to build a global monitoring system for atmospheric composition as well as a regional air-quality forecasting capability (clarification needed: does this include CO₂?). The GEMS system has extended the NWP 4D-Var data assimilation system to enable assimilation of various satellite data sources to constrain concentrations of greenhouse gases, reactive gases, and aerosols. ECMWF is open to exchange of expertise and relevant data with research groups that are not directly involved in the project.

Our objectives are twofold. Partly, we have some deliverables under CarboEurope along the lines of constrained estimates of European carbon balance using data assimilation techniques. In the longer term we're interested in a full CCDAS-type system - maybe taking CCDAS itself and substituting JULES for BETHY. Initially we will use flask data, but have an eye on satellite CO₂, perhaps ogether with some sort of system to Monte Carlo the flux tower data to get prior info as input to CCDAS. Our DA approach: Var with adjoint is the initial aim, but again open for debate if other techniques with JULES look promising.

The National Centre for Earth Observation