Working Group on "Carbon Cycle Data Assimilation: How to Consistently Assimilate Multiple Data Streams"


Marko Scholze (Lund University, Sweden)
Martin Heimann (MPI-BGC, Jena, Germany)

Members (to be confirmed):

Thomas Kaminski, FastOpt, Germany
Peter Rayner, U Melbourne, Australia
Soenke Zaehle, MPI for Biogeochemistry, Germany
Anna Michalak, Carnegie Institution for Science, USA
Frederic Chevalier, LSCE, France
Bernard Pinty, JRC, Italy
Fortunat Joos, University of Bern, Switzerland
Andreas Oschlies, GEOMAR, Germany
Sander Houweling, SRON, Netherlands


Data assimilation (DA) objectively combines information from observations with information contained in a model of the evolving system taking into account errors in both observations and model. DA has evolved from Numerical Weather Prediction (NWP) where it is now routinely used to improve weather forecasts by improving the models’ initial conditions for the next forecast. DA is now evolving to play a key role in various other fields of Earth System Science to initialise model simulations or to estimate model process parameters. Data assimilation methods are also becoming an important tool to assess and prepare for future EO Missions. It is only over the recent past (~ 10 years) that in carbon cycle research substantial progress has been achieved in the availability of observations (including EO) as well as in the use of these observational data to constrain models, and here in particular on the development of carbon cycle data assimilation systems to optimally estimate model process parameters given these observations. These observations cover the whole range of characteristics of the carbon cycle: from very local (plant leaf level data) to large-scale (ground-based flask samples of background atmospheric CO2 concentrations) in the spatial domain and instantaneous (eddy-covariance CO2 fluxes) to multi-annual (ground-based carbon inventories) in the temporal domain.

Scientific Questions:

The above mentioned range of observations together with the advent of additional EO products relevant to the carbon cycle (total column CO2, chlorophyll fluorescence) raises new open questions on the use of multiple data streams in assimilation systems in carbon cycle research:

• How can we combine several data streams with different temporal/spatial characteristics into a data assimilation system?

• How do we weight the different data sets in a data assimilation system?

• How can we make optimal use of the different types of observations in carbon cycle DA systems?

• How do we handle biases and observational data dependencies in DA systems?



First Meeting: 15 - 17 May 2013

Second Meeting: 10 - 14 January 2014

back to the overview of the Working Groups >