Research · Terrestrial Ecology

Development of the AmeriFlux Data Assimilation System (ADAS)

Objective

The objective of the project begun in late 2003 is to develop an AmeriFlux Data Assimilation System to integrate eddy covariance flux tower observations, site complementary information and mechanistic land surface models to generate synthetic, value-added products (reanalysis datasets and non-observable variables) that are spatially and temporally uniform and biophysically and biogeochemically consistent to support AmeriFlux network-wide synthesis activities and the North American Carbon Program's multi-scale initiatives in carbon budget studies. This project will produce solutions for the best use of available resources and greatly enhance the values of data products being collected by the AmeriFlux network and land surface models that have been developed. The synthetic products from the project will benefit researches in a wide range of areas including land – atmosphere interactions, climate modeling, and terrestrial carbon and hydrological cycles.

Goals

To explore effective and efficient methodologies for integrated land surface flux data assimilation and develop an operational land surface flux data assimilation system. Our specific tasks include:

Automatic and objective nighttime flux filtering and gap-filling.
Partitioning NEE of CO2, latent heat, and sensible heat fluxes as well as net radiation into contributions from ecosystem components with particular emphases on gross primary production, net primary production, leaf and stem respiration, soil respiration, transpiration, physical evaporation, biomass enthalpy change and metabolic energy storage.
Conducting synthesis studies with these data to evaluate how environmental and ecosystem processes affect flux exchanges.
Taking advantage of numerical skills obtained in the data assimilation system and developing generic data processing software needed by the flux research community and providing technical support.

Approach

We use a variational framework for our flux data assimilation system. The forward model is the terrestrial Fluxes And Pools Integrated Simulator (FAPIS). The FAPIS model includes all processes that are known to affect land - atmosphere flux exchanges of CO2, water vapor, sensible heat, momentum, shortwave and longwave radiation. It separates diffuse and direct radiation and sunlit and shaded leaves for both carbon assimilation and energy balance calculations. It predicts vertical gradients in meteorological conditions within the canopy and flux exchanges in each layer are computed with local environmental conditions. It couples simulations of canopy and soil processes directly rather than explicitly. It has detailed treatments in both carbon and hydrological cycles. These rigorous representations of biophysical and ecophysiological processes are essential for successful flux data assimilation. The structure of the FAPIS model has been designed specifically for automatic generation of adjoint code for numerical optimization. Different optimization techniques are used in the assimilation.

The major observational variables assimilated into the system through a cost functional include flux exchanges of CO2, water vapor, sensible heat, momentum, reflected shortwave and outgoing longwave radiation. A nighttime flux filter is employed to detect the influence of low turbulence. The forward model is optimized for ecosystem state and critical parameters (e.g. LAI, Vcmax, Jmax, TPU, parameters describing their vertical distributions in the canopy, parameters in the soil respiration model, etc.). Because FAPIS predicts vertical profiles of CO2 concentration, water vapor concentration and temperature, these profiles can be assimilated to provide further model parameter constraints.

Results to Date

Data assimilation software development has been the main focus of the project so far. New codes consisting of tens of thousands of lines have been developed and tested for forward modeling and optimization. An objective and automatic nighttime flux filter (Moving Point Test, or MPT) has been developed and published. MPT have been adopted by several users in AmeriFlux as well as in CarboEurope and AsiaFlux and technical support has been provided to them. In collaboration with the Missouri Ozark AmeiFlux project, a new comprehensive optimization method has been developed for A/Ci curve analysis. This new method completely removes the need of a priori selection of critical internal CO2 partial pressure (Ci) values for estimating leaf biochemical parameters from A/Ci measurements in conventional A/Ci curve analysis methods, which can cause serious errors in estimated biochemical parameters and cause biases in ecosystem models that use these parameters. A software package (OptimalACi) has been developed for sharing with the community. Time series of soil respiration measurements have been used successfully to estimate soil carbon stocks through a multi-pool carbon model and its adjoint model. During our data assimilation effort, we have developed several novel approaches for analyzing flux data, particularly soil efflux data. These approaches have led to a few important findings regarding dynamics in soil respiration and its relationship to photosynthesis and soil carbon pool variations:

Photosynthetically-driven signals in soil efflux measurements are detected. We have developed a partitioning procedure to separate contributions of environmental and biological factors to soil efflux. We found that at a temperate deciduous forest site in Oak Ridge, TN, a substantial portion of the diurnal variations in soil efflux is caused by export of newly synthesized carbohydrates to roots during the growing season. This is in contrast to the non-growing season during which diurnal variations in soil efflux are controlled largely by environmental factors.
Discovery of fast carbon pool turnover obscuring sensitivity of soil respiration to temperature change. Soils contain huge amounts of carbon with different physical and chemical properties. How soil carbon pools respond to changes in environmental conditions, particularly temperature, critically influences future atmospheric carbon dioxide concentration. In a modeling study, we found that some components of soil carbon turn over rapidly and this fast turnover can obscure the sensitivity of soil respiration to temperature change. This finding is important because the estimated sensitivity of soil respiration to temperature change is often used in coupled carbon - climate models to predict future atmospheric carbon dioxide concentration and climate. Without properly taking into account the phenomenon discovered in this study, role of soil carbon in climate change can be underestimated.

Peer-reviewed papers

Liu, Q., N.T. Edwards, W.M. Post, L. Gu, J. Ledford, and S. Lenhart, 2005. Photosynthesis-driven signal in soil efflux observed from a deciduous forest FACE site, Global Change Biology (in review).

Gu, L., S. G. Pallardy, P. Hanson and S. D. Wullschleger, 2005. A comprehensive optimization method for objective estimation of parameters in biochemical models of leaf photosynthesis, Plant Cell and Environment (in review).

Gu, L., E.M. Falge, T. Boden, D. D. Baldocchi, T. A. Black, S. R. Saleska, T. Suni, S. B. Verma, T. Vesala, S. C. Wofsy, and L. Xu, 2005. Objective threshold determination for nighttime eddy flux filtering, Agricultural and Forest Meteorology, 128, 179-197.

Gu, L. W. M. Post, and A. W. King. Fast labile carbon turnover obscures sensitivity of heterotrophic respiration from soils to temperature: a model analysis, Global Biogeochemical Cycles. Global Biogeochemical Cycles, Vol. 18, Gb1022, Doi:10.1029/2003gb002119, 2004.

For more information, contact:
W. Mac Post (postwmiii@ornl.gov, 865-576-3431)

Revised: 8/03/05