The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data

Andrew Fox; Mathew Williams; Andrew D. Richardson; David Cameron; Jeffrey H. Gove; Tristan Quaife; Daniel Ricciuto; Markus Reichstein; Enrico Tomelleri; Cathy M. Trudinger; Mark T. Van Wijk

doi:10.1016/j.agrformet.2009.05.002

The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data

Andrew Fox
, Mathew Williams
, Andrew D. Richardson
, David Cameron
, Jeffrey H. Gove
, Tristan Quaife
, Daniel Ricciuto
, Markus Reichstein
, Enrico Tomelleri
, Cathy M. Trudinger
, Mark T. Van Wijk

Earth Systems Modeling

Research output: Contribution to journal › Article › peer-review

127 Scopus citations

Abstract

We describe a model-data fusion (MDF) inter-comparison project (REFLEX), which compared various algorithms for estimating carbon (C) model parameters consistent with both measured carbon fluxes and states and a simple C model. Participants were provided with the model and with both synthetic net ecosystem exchange (NEE) of CO₂ and leaf area index (LAI) data, generated from the model with added noise, and observed NEE and LAI data from two eddy covariance sites. Participants endeavoured to estimate model parameters and states consistent with the model for all cases over the two years for which data were provided, and generate predictions for one additional year without observations. Nine participants contributed results using Metropolis algorithms, Kalman filters and a genetic algorithm. For the synthetic data case, parameter estimates compared well with the true values. The results of the analyses indicated that parameters linked directly to gross primary production (GPP) and ecosystem respiration, such as those related to foliage allocation and turnover, or temperature sensitivity of heterotrophic respiration, were best constrained and characterised. Poorly estimated parameters were those related to the allocation to and turnover of fine root/wood pools. Estimates of confidence intervals varied among algorithms, but several algorithms successfully located the true values of annual fluxes from synthetic experiments within relatively narrow 90% confidence intervals, achieving >80% success rate and mean NEE confidence intervals <110 gC m^-2 year^-1 for the synthetic case. Annual C flux estimates generated by participants generally agreed with gap-filling approaches using half-hourly data. The estimation of ecosystem respiration and GPP through MDF agreed well with outputs from partitioning studies using half-hourly data. Confidence limits on annual NEE increased by an average of 88% in the prediction year compared to the previous year, when data were available. Confidence intervals on annual NEE increased by 30% when observed data were used instead of synthetic data, reflecting and quantifying the addition of model error. Finally, our analyses indicated that incorporating additional constraints, using data on C pools (wood, soil and fine roots) would help to reduce uncertainties for model parameters poorly served by eddy covariance data.

Original language	English
Pages (from-to)	1597-1615
Number of pages	19
Journal	Agricultural and Forest Meteorology
Volume	149
Issue number	10
DOIs	https://doi.org/10.1016/j.agrformet.2009.05.002
State	Published - Oct 1 2009

Funding

NERC funded much of this work through the CarbonFusion International Opportunities grant. MW, CT, AF, MR and AR were involved in the initial planning and discussions for REFLEX at the CarbonFusion meeting in Edinburgh, May 2006. AF undertook the main work on setting up the experiment, collating results, and undertaking analyses. MW initiated and managed the experiment, and led the production of the manuscript. AR was involved in defining the analytical process and the parameter estimation assessment. Other authors participated in the experiment and all contributed to the paper. We are grateful to A Granier and the Hesse research team, and to EJ Moors and the Loobos research team, for access to their data, and the FLUXNET team for data processing and preparation. We are grateful to Mike Raupach and Damian Barrett for their ideas and input at the start of the project, and in developing the protocol for the experiment outlined here. We also recognise Zhang Li's efforts and input to the analysis. Jens Kattge made useful comments on the manuscript. AR acknowledges support from the Office of Science (BER), U.S. Department of Energy, through the Northeastern Regional Center of the National Institute for Climatic Change Research. We thank two anonymous reviewers for their comments.

Keywords

Carbon cycle
Confidence intervals
Data assimilation
Ecosystem modelling
Eddy covariance
Kalman filter
Metropolis
Monte Carlo
Parameter optimisation
REFLEX project

Access to Document

10.1016/j.agrformet.2009.05.002

Cite this

Fox, A., Williams, M., Richardson, A. D., Cameron, D., Gove, J. H., Quaife, T., Ricciuto, D., Reichstein, M., Tomelleri, E., Trudinger, C. M., & Van Wijk, M. T. (2009). The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data. Agricultural and Forest Meteorology, 149(10), 1597-1615. https://doi.org/10.1016/j.agrformet.2009.05.002

@article{298a4e9331944f7b82c0a84005ac2aac,

title = "The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data",

abstract = "We describe a model-data fusion (MDF) inter-comparison project (REFLEX), which compared various algorithms for estimating carbon (C) model parameters consistent with both measured carbon fluxes and states and a simple C model. Participants were provided with the model and with both synthetic net ecosystem exchange (NEE) of CO2 and leaf area index (LAI) data, generated from the model with added noise, and observed NEE and LAI data from two eddy covariance sites. Participants endeavoured to estimate model parameters and states consistent with the model for all cases over the two years for which data were provided, and generate predictions for one additional year without observations. Nine participants contributed results using Metropolis algorithms, Kalman filters and a genetic algorithm. For the synthetic data case, parameter estimates compared well with the true values. The results of the analyses indicated that parameters linked directly to gross primary production (GPP) and ecosystem respiration, such as those related to foliage allocation and turnover, or temperature sensitivity of heterotrophic respiration, were best constrained and characterised. Poorly estimated parameters were those related to the allocation to and turnover of fine root/wood pools. Estimates of confidence intervals varied among algorithms, but several algorithms successfully located the true values of annual fluxes from synthetic experiments within relatively narrow 90\% confidence intervals, achieving >80\% success rate and mean NEE confidence intervals <110 gC m-2 year-1 for the synthetic case. Annual C flux estimates generated by participants generally agreed with gap-filling approaches using half-hourly data. The estimation of ecosystem respiration and GPP through MDF agreed well with outputs from partitioning studies using half-hourly data. Confidence limits on annual NEE increased by an average of 88\% in the prediction year compared to the previous year, when data were available. Confidence intervals on annual NEE increased by 30\% when observed data were used instead of synthetic data, reflecting and quantifying the addition of model error. Finally, our analyses indicated that incorporating additional constraints, using data on C pools (wood, soil and fine roots) would help to reduce uncertainties for model parameters poorly served by eddy covariance data.",

keywords = "Carbon cycle, Confidence intervals, Data assimilation, Ecosystem modelling, Eddy covariance, Kalman filter, Metropolis, Monte Carlo, Parameter optimisation, REFLEX project",

author = "Andrew Fox and Mathew Williams and Richardson, \{Andrew D.\} and David Cameron and Gove, \{Jeffrey H.\} and Tristan Quaife and Daniel Ricciuto and Markus Reichstein and Enrico Tomelleri and Trudinger, \{Cathy M.\} and \{Van Wijk\}, \{Mark T.\}",

year = "2009",

month = oct,

day = "1",

doi = "10.1016/j.agrformet.2009.05.002",

language = "English",

volume = "149",

pages = "1597--1615",

journal = "Agricultural and Forest Meteorology",

issn = "0168-1923",

publisher = "Elsevier",

number = "10",

}

Fox, A, Williams, M, Richardson, AD, Cameron, D, Gove, JH, Quaife, T, Ricciuto, D, Reichstein, M, Tomelleri, E, Trudinger, CM & Van Wijk, MT 2009, 'The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data', Agricultural and Forest Meteorology, vol. 149, no. 10, pp. 1597-1615. https://doi.org/10.1016/j.agrformet.2009.05.002

TY - JOUR

T1 - The REFLEX project

T2 - Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data

AU - Fox, Andrew

AU - Williams, Mathew

AU - Richardson, Andrew D.

AU - Cameron, David

AU - Gove, Jeffrey H.

AU - Quaife, Tristan

AU - Ricciuto, Daniel

AU - Reichstein, Markus

AU - Tomelleri, Enrico

AU - Trudinger, Cathy M.

AU - Van Wijk, Mark T.

PY - 2009/10/1

Y1 - 2009/10/1

N2 - We describe a model-data fusion (MDF) inter-comparison project (REFLEX), which compared various algorithms for estimating carbon (C) model parameters consistent with both measured carbon fluxes and states and a simple C model. Participants were provided with the model and with both synthetic net ecosystem exchange (NEE) of CO2 and leaf area index (LAI) data, generated from the model with added noise, and observed NEE and LAI data from two eddy covariance sites. Participants endeavoured to estimate model parameters and states consistent with the model for all cases over the two years for which data were provided, and generate predictions for one additional year without observations. Nine participants contributed results using Metropolis algorithms, Kalman filters and a genetic algorithm. For the synthetic data case, parameter estimates compared well with the true values. The results of the analyses indicated that parameters linked directly to gross primary production (GPP) and ecosystem respiration, such as those related to foliage allocation and turnover, or temperature sensitivity of heterotrophic respiration, were best constrained and characterised. Poorly estimated parameters were those related to the allocation to and turnover of fine root/wood pools. Estimates of confidence intervals varied among algorithms, but several algorithms successfully located the true values of annual fluxes from synthetic experiments within relatively narrow 90% confidence intervals, achieving >80% success rate and mean NEE confidence intervals <110 gC m-2 year-1 for the synthetic case. Annual C flux estimates generated by participants generally agreed with gap-filling approaches using half-hourly data. The estimation of ecosystem respiration and GPP through MDF agreed well with outputs from partitioning studies using half-hourly data. Confidence limits on annual NEE increased by an average of 88% in the prediction year compared to the previous year, when data were available. Confidence intervals on annual NEE increased by 30% when observed data were used instead of synthetic data, reflecting and quantifying the addition of model error. Finally, our analyses indicated that incorporating additional constraints, using data on C pools (wood, soil and fine roots) would help to reduce uncertainties for model parameters poorly served by eddy covariance data.

AB - We describe a model-data fusion (MDF) inter-comparison project (REFLEX), which compared various algorithms for estimating carbon (C) model parameters consistent with both measured carbon fluxes and states and a simple C model. Participants were provided with the model and with both synthetic net ecosystem exchange (NEE) of CO2 and leaf area index (LAI) data, generated from the model with added noise, and observed NEE and LAI data from two eddy covariance sites. Participants endeavoured to estimate model parameters and states consistent with the model for all cases over the two years for which data were provided, and generate predictions for one additional year without observations. Nine participants contributed results using Metropolis algorithms, Kalman filters and a genetic algorithm. For the synthetic data case, parameter estimates compared well with the true values. The results of the analyses indicated that parameters linked directly to gross primary production (GPP) and ecosystem respiration, such as those related to foliage allocation and turnover, or temperature sensitivity of heterotrophic respiration, were best constrained and characterised. Poorly estimated parameters were those related to the allocation to and turnover of fine root/wood pools. Estimates of confidence intervals varied among algorithms, but several algorithms successfully located the true values of annual fluxes from synthetic experiments within relatively narrow 90% confidence intervals, achieving >80% success rate and mean NEE confidence intervals <110 gC m-2 year-1 for the synthetic case. Annual C flux estimates generated by participants generally agreed with gap-filling approaches using half-hourly data. The estimation of ecosystem respiration and GPP through MDF agreed well with outputs from partitioning studies using half-hourly data. Confidence limits on annual NEE increased by an average of 88% in the prediction year compared to the previous year, when data were available. Confidence intervals on annual NEE increased by 30% when observed data were used instead of synthetic data, reflecting and quantifying the addition of model error. Finally, our analyses indicated that incorporating additional constraints, using data on C pools (wood, soil and fine roots) would help to reduce uncertainties for model parameters poorly served by eddy covariance data.

KW - Carbon cycle

KW - Confidence intervals

KW - Data assimilation

KW - Ecosystem modelling

KW - Eddy covariance

KW - Kalman filter

KW - Metropolis

KW - Monte Carlo

KW - Parameter optimisation

KW - REFLEX project

UR - https://www.scopus.com/pages/publications/62249145300

U2 - 10.1016/j.agrformet.2009.05.002

DO - 10.1016/j.agrformet.2009.05.002

M3 - Article

AN - SCOPUS:62249145300

SN - 0168-1923

VL - 149

SP - 1597

EP - 1615

JO - Agricultural and Forest Meteorology

JF - Agricultural and Forest Meteorology

IS - 10

ER -

The REFLEX project: Comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data

Abstract

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this