Carbon Assimilation and Modelling of the European Land Surfaces


Fifth Consortium Meeting - Minutes 
Scientific Questions
Scientific Objectives
Project Work Plan
Partner Institutes
Peer Reviewed Papers
Book Chapters
Annual Reports
Workpackage Reports
Miscellaneous Reports
Request for data
Useful Links
Meetings  and Actions
Forthcoming Events
Contact US
Minutes of the 5th CAMELS meeting
EFI, Joensuu, Finland
26th January 2005


Met Office: Richard Betts, Venkata Jogireddy
LSCE: Pierre Friedlingstein, Roger Dargaville, Diego Santeren
MPI-BGC: Jens Kattge
ALTERRA:  Mart-Jan Schelhaas
UNITUS: Dario Papale
EFI: Andreas schule, Gert-Jan Nabuurs
JRC: Nadine Gobron


Peter Cox (CEH),Wolfgang Knorr (MPI-BGC), Philippe Peylin, Andrew Friend, Philippe Ciais (LSCE), Martin Heimann (MPI-BGC), Ronald Hutjes (Alterra), Riccardo Valentini (UNITUS), Richard Harding (CEH), Thomas Kaminski (FastOpt), Michel Verstraete, Bernard Pinty (JRC)

Date of next meeting:  Unitus, Italy, End of October 2005

1.    ACTIONS from 4thmeeting (20-21/06/04):

ACTION 3(e): Hadley (RB) to check for consistency between inventory data and land-use
Done. The results were discussed. Differences were observed between Hyde dataset and forest inventory data. Further analysis is planned between Alterra and HC.

ACTION 3(k): LSCE (PF) to design the protocol for WP3 simulations, based on the discussion summarised in the minutes of the 3rd meeting. 
Done. Completed

ACTION 3(s): MPI-BGC(JK) to lead on paper on Use of eddy flux data to reduce uncertainties in TEM parameterisation (for submission in December 04).
Ongoing. As different modelling groups adapted different optimisation approaches. MPI-BGC(JK) will assess whether to write collaborative paper on Use of eddy flux data to reduce uncertainties in TEM parameterisation (oct 05)

ACTION 3(t): CEH(CH) and ALTERRA (IvdW) to lead on paper on Robustness of land-surface parameterisation within plant functional type (for submission March 05).
Ongoing. Currently, paper is being written and will be submitted by March-2005

ACTION 4(a): EFI (AP) to provide link to grid data on fractional cover of forests (June 04)

ACTION 4(b): EFI (AP) to provide detailed land use inventory data (July 04)

ACTION 4(c): ALL MODELLING GROUPS to provide EFI with specific requirements from inventory data. (Aug 04)
Ongoing. Modelling groups still decide the format need for inventory data.

ACTION 4(d): JRC (NG) to provide FAPAR data in netCDF format (July 04)
Ongoing. NG will provide FAPAR dataset in netCDF format by Feb-2005.

ACTION 4(e): MPI-BGC (JK) to provide protocol on optimisation (July 04)
Done. Further protocol is not needed

ACTION 4(f): ALL MODELLING GROUPS to complete optimisation runs for LeBray and Loobos (October 04)
Ongoing. ALL MODELLING GROUPS to complete optimisation runs for LeBray and Loobos if appropriate for collaborative paper (May 05)

ACTION 4(g): EFI (M-J S) to provide VJ with web pages on inventory data and methods

ACTION 4(h): UNITUS (RV) to coordinate brochure for COP10 based on WP reports for 2nd Annual report (Oct 04)

ACTION 4(i): LSCE (PF) to circulate protocol for WP3 simulations (Sept 04)

ACTION 4(j): Hadley/CEH (VJ/CH) to circulate climatic driving data for WP3 simulations (July 04)

ACTION 4(k): LSCE (RG) to provide flux estimates from inverse models for comparison with forward models (Mar 05)

ACTION 4(l): ALL MODELLING GROUPS to perform offline simulations of 20th century carbon balance with a priori and a postiori parameter sets (Mar 05)

ACTION 4(m): EFI (AP) to provide flux estimates and uncertainties from inventory data for CCDAS

ACTION 4(n): Hadley (RB) to circulate link to CAMELS Technological Implementation Plan (June 04)

ACTION 4(o): JRC (NG) to write 1-2 page report on D1.5 and forward to RH for WP1 final report (Aug 04)

ACTION 4(p): Hadley (RB) to write 1-2 page report on consistency between historical land use datasets and inventory data and forward to RH for WP1 final report (Aug 04)

ACTION 4(q): EFI (AP) to write 1-2 page report on work inventory data for WP1 (Aug 04)

ACTION 4(r): ALL MODELLING GROUPS to write 1-2 page report on work on parameter uncertainties and forward to JK for WP2 annual report (Aug 04)

ACTION 4(s): Hadley (VJ) to write 1-2 page report on climatic driving data for WP3 simulations, and forward to PC for WP4 annual report (Aug 04)

ACTION 4(t): ALTERRA (RH) to write WP1 final report and forward to RB and RV (Sep 04)

ACTION 4(u): MPI-BGC (JK) to write WP2 annual report and forward to RB and RV (Sep 04)

ACTION 4(v): LSCE (PF) to write WP3 annual report, including plans for WP3 and draft protocol for simulations, and forward to RB and RV (Sep 04)

ACTION 4(w): Hadley (PC) to write WP4 annual report, detailing plans for WP4, and forward to RB and RV (Sep 04)

ACTION 4(x): Hadley (RB) to edit 2nd Annual project report and submit to EU (Oct 04)
Done. The final report will be submitted during Feb-2005

ACTION 4(y): MPI-BGC (JK) to hold WP2 meeting/teleconference to prepare for joint paper. (end  Oct 04)

2.       Progress on WP1

a.    Biome-specific datasets to drive and validate TEMs (ALTERRA)

Dario Papale has presented Eddy covariance datasets: Quality and products. There have been different levels of flux products available. These products need to be assed before use by the model community. Level 3 and 4 products have been corrected and are good quality products which can be readily used. Semi automatic quality checks have been carried out in all these flux sites. For example, global radiation is compared with potential incoming radiation calculated using latitude and longitude. The linear regression between two variables is calculated and residual are estimated. The data is flagged as spike if the value with residual is higher than 4 times the standard deviation of all the residuals. Spike are also flagged using an outlier detection technique based on the median and the standard deviation of the section derivative of each point. Storage correction also been applied. Level 4 products have derived from USTAR filtering, gap-filling and flux partitioning applied to fluxes.  Dario informed that level 4 products are reliable and using of other level products in calibration of models results in incorrect results. All model groups agreed to provide the driving data/flux data used in models to Dario. He will check whether the data used by modelling groups varies with the quality of products available with him.

ACTION 5(f)c: VJ and PF to send driving data for Loobos and Bray to DP (feb 05)
ACTION 5(g): DP to compare driving data with new datasets and advise on differences, and send new data from other sites (except 2003) (feb 05)
ACTION 5(h): DP to provide new eddy covariance data + new variables for 2003 (may 05)

b.    Atmospheric CO2 dataset for use in nowcasting system (LSCE)

The CO2  dataset is available . The historical dataset enabling the models to be run from 1900 to present is available for use in the WP3.  The contemporary dataset based on the AEROCARB project database is also available.

c.    Land-use and nitrogen deposition historical datasets 1900-2000 (Hadley)
Consistency checks between inventory data and HYDE land-use dataset has been carried out. The variations in the extent of crop and forest area between these two datasets are discussed. It has been identified the forest/crop area has been varied highly. Everybody in the meeting felt that the comparison needs to be looked carefully by country by country. Mark, VJ and Richard will carry out the detail consistency analysis and will inform results to CAMELS members.

ACTION 5(a): Hadley (VJ+RB) and MJ to check methodology for comparing HYDE and inventory data (feb 05)

ACTION 5(b): Hadley (RB) to send information on HYDE grid orientation (feb 05)

d.    Datasets of recent change in European land carbon (EFI)

Jo van Brusselen has presented latest forest inventory map of Europe. The information of forest inventory data on country wise has been formatted on grid wise. He notified the difference in the forest cover from the inventory base approach and other forest information available. The forest inventories have been collected from the each country. There has been a discussion on the accuracy of forest inventories. Each country may not used a consistent definition of forests which might result in inaccuracies in the forest inventories collected. Andreas Schule informed there might be consistent in definition of forest in central Europe while the inaccuracies might exit in eastern Europe forest inventories.

ACTION 5(b)b: EFI to send map of forest areas (& growing stock & increment) on 0.5 degree res (mar 05)

e.    Dataset of fAPAR for Europe, and other remote sensing products (JRC)
NG presented fAPAR products available for CAMELS. Following products are available for use by CAMELS.
    Two years of European window data are delivered on DVD.
    20 Camels site windows extracted available on CDs
    Code for reading data available on CDs
    Prototyped global products @ 0.5 x 0.5 degrees available
    Time series corresponding to in-situ measurements sites are extracted.
NG presented details of FAPAR product available. She requested the data need to placed in the CAMELS website. NG will provide a weblink and this will be placed in the CAMELS site.

ACTION 4(d): JRC (NG) to provide FAPAR data in netCDF format (July 04). She informed she will provide the DVD/CDROM to RB.

3.        Progress on WP2
a.    TEM parameters and a priori estimates (MPI-BGC)
All modelling groups have completed the task of TEM parameterization. This information has been published on Second Annual Report.

b.    Parameter Optimization of Biosphere Model Using Flux Data (LSCE)
PF presented Diegos results on optimisation of ORCHIDEE using a Bayesian cost function. The optimisation used is Bayesian Cost function. He has used iterative BFGS algorithm derived by Nocedal et al which is adapted to nonlinear problems. Each iteration needs values of the cost function and of its derivatives wrt each parameter. It assumes local linear hypothesis. Initial uncertainties and bounds on each parameter are crucial in the Bayesian framework to prevent the inversion from retrieving unrealistic values. Nevertheless, he choose to retrieve the maximum information content of the data and thus deliberately assign large prior errors in order to minimize the influence of the Bayesian term in the retrieved values. He performed two different sets of optimization, depending on the period during which the eddy-covariance data are assimilated. First, he optimized against 3 weeks of data in summer 1997 in order to use the diurnal information at the peak of the Growing Season (GS optimization) and then he used the Full Year of data (FY optimization) to account for the seasonal cycle. The estimated values and uncertainties for each case allowed to improve the model performances, to separate well constrained parameters from those that are very poorly constrained by the data, and  to highlight some model deficiencies. The model is indeed not able to satisfactorily reproduce both the fast and slow processes that control the diurnal and seasonal cycles. The optimized model data fit is greatly improved compared to the prior one for the assimilation of the growing season data .The diurnal cycle for each data-type is correctly reproduced although several problems remain. These are phases mismatch between the modelled and observed NEE, and an overestimation of the sensible heat flux (H) during the night. When he used the parameters given by the optimization against the whole year of data, the model fails to reproduce the seasonal cycle of the observed NEE. The strong information imposed by all diurnal cycles prevents the optimization from fitting the other time scales feature (i.e. the seasonal cycle). In the future, he plans to optimize other parameters controlling slow carbon processes in order to properly simulated annual to decadal carbon sequestration.

c.    Monte Carlo simulations with BETHY  (MPI-BGC)
JK described work at MPI-BGC on the use of a Bayesian approach and the Metropolis Monte Carlo method to estimate parameter PDFs for BETHY. First inversions of parameter values of the terrestrial biosphere model BETHY against eddy covariance measurements at the Loobos site have shown that further application of this method requires a careful analysis of the prior uncertainties of model parameters, because results of the inversions depended on a priori parameter uncertainties and some parameters were only little constrained by the inversions.
He compared net ecosystem production (NEP) of the Loobos site calculated by the biosphere model BETHY with an inventory-based approach and results from eddy covariance measurements. BETHY calculates carbon exchange based on mechanistically assumptions of photosynthesis and respiration, using climate and site data for input. The calculation of NEP by BETHY was done with two sets of parameters: the standard set of parameters, which is derived from literature and which is used to calculate carbon exchange on a global scale (denoted as a priori set of parameters). The uncertainty of these parameters was assumed to be 25%. The second set of parameters and uncertainties was derived by inversion of a priori parameter values against seven days of half-hourly eddy-covariance measurements of net ecosystem exchange of CO2 (NEE) and latent energy (LE) at the Loobos site in 1997 and 1998 (denoted as optimized set of parameters). Thus the optimized set of parameters includes information from the eddy-covariance measurements and the NEP calculated with this set of parameters is not independent from results directly derived from eddy-covariance measurements.
Without being optimized the BETHY model produced much higher NEP than the two other approaches. Also these results had a huge uncertainty. After optimization NEP and uncertainties calculated by BETHY were well in the range of NEP derived with the inventory approach and directly derived from eddy covariance measurements. Given the methodological problems of eddy covariance measurements, the problems of comparison to not-optimized model results and the effectiveness of the optimization to include information from the eddy covariance measurements into the biosphere model, it seems to be a very good strategy to control the results from eddy covariance measurements and from the optimized model by independent measurements like inventories.

d.    Model Improvements, calibration and uncertainty analysis (MetO, CEH, UK)

VJ described the work carried out at Met office along with CEH, UK. He presented the model improvements carried out to the existing MOSES Model. Comparison of MOSES against CO2 flux measurements from eddy-covariance towers showed that the model underestimate the high uptake rates in the middle of the day. The underlying reason for this bias was the assumption of optimal distribution of leaf N through plant canopy, which leads the canopy to behave like a big-leaf. This was resulting in the unrealistic flat-top behavior. He presented improvements in the model. The improvements are canopy radiation interception process based on analytical two-stream model, and an explicit scaling-up from leaf-to-canopy. The two- stream model will distinguish light-limited and non-light limited leaf classes. The new model produced better fit to the flux measurements from Loobos site and has additional benefit of distinguishing between diffuse and direct radiation effects on photosynthesis.

The ensemble of 200000 Monte Carlo simulation were carried out using different photosynthesis model parameters (leaf N, F0,DC, Quantum efficiency) had been carried out on both the old MOSES and new 2 leaf, two-stream model. The original MOSES model has a negative percentage of variance explained suggesting a very poor fit to the data. The fit is significantly improved even in the big leaf model by tuning parameters. However, this has achieved by moving the leaf nitrogen concentration and quantum efficiency values outside range. By contrast the new canopy model is able to get a reasonable fit to the data without requiring unrealistic parameter values.

e.    Flux Estimates and Uncertainty in NPP. (Alterra)
MJ presented inter annual variability and inter-site variability on parameter estimation. He explored the parameters and optimized these against the data. The total variance in carbon fluxes is up to 20 %. For evapo-transpiration the explained variance is generally but the total flux is generally under estimated. Year to year variations in cumulative flux are similar as for NEE.
ACTION 3(s): MPI-BGC(JK) to assess whether to write collaborative paper on Use of eddy flux data to reduce uncertainties in TEM parameterisation (oct 05)

ACTION 4(f): ALL MODELLING GROUPS to complete optimisation runs for LeBray and Loobos if appropriate for collaborative paper (May 05)

ACTION 4(m): EFI (MJ) to provide flux estimates and uncertainties from inventory data for CCDAS (May 05)

ACTION5 (b)c: EFI to send NPP estimates (Feb. 05) and uncertainties (may 05)

4    Progress on WP3

a.    Carbon balance of the terrestrial biosphere in the 20 th Century
PF presented the simulation carried for the carbon balanced for 20 th century. He has run the simulations using ORCHIDEE using climate forcing data taken from CRU dataset from 0.5 x 0.5, CO2 data taken from CSIRO and land-use maps from RB. He has presented the results of inter annual changes of GPP, NPP and NEP over the period between 1850 and 1900. The distribution generally agreed with the changes and extent of forest cover. The future work planed is carrying out the simulation with CO2 alone, with CO2 + climate and with CO2+climate+landuse. He presented the simulations carried for 2003. During summer 2003, GPP and NPP were lower depicting the dry season. These fluxes in Europe were lower than other continents. The variation in fluxes during 2003 was in the order of 3sigma lower than with other normal years 2002/1972.

b.    Imogen and CRU dataset creation for historical simulations.

VJ presented the features of Imogen and CRU dataset creation for historical simulations. Imogen is a global offiline model to test the land surface parameterisation which has been calibrated with HADCM3LC. This includes carbon cycle, MOSES and TRIFFID. He presented the model runs by imogen which were similar to the other GCMs. The CRU dataset has been procured from the University of East Anglia, this is available for  Camel group and cab be downloadable from the website. The short and long wave radiations were derived from the henning formulation using GCM simulations of Wind profile. A forward plan for the historical simulation using Triffid is also presented. Ochidee and Triffid models were used for estimation of carbon land balances. The consortium discussed use of BETHY in historical simulations. As the BETHY is main model in CCDAS, it would be necessary to carry out historical simulations.

ACTION 5(b): PF to send Orchidee land cover maps for comparison against inventory data (feb 05)
ACTION 5(c): VJ to compare Orchidee and IGBP maps against inventory data and forest maps (Apr 05)

5    Progress on WP4

a.    Development of Carbon Data Assimilation system (MPI-BGC)

RB presented the WP4 objectives and the forward plan. CAMLES. Its aims were best estimates and uncertainty bounds for the contemporary and historical land carbon sinks in Europe and elsewhere, isolating the effects of direct land-management. And  a prototype carbon cycle data assimilation system (CCDAS) exploiting existing data sources (e.g.  flux measurements, carbon inventory data, satellite products) and the latest terrestrial ecosystem models (TEMs), in order to produce operational estimates of "Kyoto sinks". CAMELS succeeded in development of the First Prototype of CCDAS system using BETHY.

b.    Carbon Data Assimilation system (MPI-BGC)

JK presented the CCDAS system developed using BETHY model. CCDAS system has the framework of optimizing parameter values from the site data as a priori values in global CCDAS. CAMELS has so far produced one prototype CCDAS based on the ecosystem model BETHY: in a first data assimilation step, BETHY takes satellite-observed values of greenness to optimise parameters related to water status, phenology, and total plant cover. Next, the adjoint (the first derivative of the code with respect to model parameters) of the physiological and energy balance part of BETHY coupled with the adjoint of the atmospheric transport model TM2 is used to optimise parameter values of BETHY. This is done by assimilation of atmospheric CO2 concentration measurements. Uncertainties of optimised model parameters can be derived from the Hessian (the second derivative) of the BETHY code with respect to the parameters. By using the Hessian of the BETHY code with respect to the parameters, uncertainties of optimised model parameters can also be derived. These uncertainties, that reflect both the prior information (in a Bayesian context), as well as the information from the large-scale inversion, can finally be translated into uncertainty bounds for CO2 fluxes and any other model diagnostic. Both the adjoint and Hessian codes are generated automatically using the compiler tool TAF, developed by FastOpt. Automatic generation ensures that improvements of BETHY can be used in the assimilation scheme without delay.

First results with CCDAS using 20 years of CO2 observation from the free atmosphere, while still somewhat preliminary, clearly show that interannual fluctuations of terrestrial CO2 fluxes are dominated by the El Niño-Southern Oscillation (ENSO) cycle, except for the time after the Pinatubo eruption.. During El Niño (warm) pacific conditions, large parts of the tropical ecosystem come under water stress with reduced photosynthesis. Of the 58 parameters that enter the optimisation, considerable reduction in uncertainty is found for about12. We find a terrestrial sink for Europe (excl. Russia) that is around a third of the fossil fuel emissions of the area, but with uncertainty bounds of the same size as the fluxes themselves. The country analysed that has the largest uncertainty in terrestrial CO2 fluxes is Brazil, mainly because of the lack of observation stations in that area.

ACTION JK ask WK about transient runs (feb 05)

6    Progress on WP5

a.    Plans for CAMELS brochure (UNITUS)

DP informed that Riccardo Valentini made a presentation, including CAMELS results at a side event organized by the Commission at COP10. The following text is presented by Ricaardo to EC The feed-back that I can give to the project is the need to include more in the project development also the discussion on how carbon data assimilation systems could provide a backbone of monitoring of terrestrial carbon for understanding its vulnerabilities and the capacity of the terrestrial biosphere to adapt to changing climate. 

The CAMELS group have discussed about the release of brochure. It ha been agreed that the release of brochure should be taken during Oct, 2005.

ACTION 4(h): UNITUS (RV) to coordinate brochure for COP10 based on WP reports for 2nd Annual report (Oct 04)
ACTION 5(p): all partners to inform HC about presentations etc on CAMELS (ongoing)
ACTIO 5(q):HC to update presentation info on website (ongoing)
ACTION 5(r):RV to send COP presentation to HC (feb 05)
ACTIOn 5(s):HC to put on website (feb 05)
ACTION 5(t):RB to discuss CAMELS-CARBOEUROPE IP participation with Martin Heimann (feb 05)
ACTION 5(u):DP to agree date for next meeting with RV and inform CAMELS partners (feb 05)
ACTION 5(v):HC to invite Commission and CARBEUROPE participants to meeting (feb 05)

7    Any other business

a.    Annual Report
RB stated that the second annual report is completed and will sent to EC by Feb 2005. The third annual report will be completed by Oct2005. All group leaders should produce reports by the end of August 2005 and submit them to WP leaders. WP leaders will consolidate these reports and produce a WP report by end of Sep-2004 and submit to RB. Hadley will write the management report, and put together the scientific report based on WP reports by end of Oct-2005.

ACTION 5(n):: All partners to send final annual reports (Aug 05)
ACTION 5(o):HC final report (oct 05)

b.    CAMELS Publications

There was a discussion on CAMELS publications. All members agreed to provide the presentations or reports/papers prepared from CAMLES project should be informed to HC and need to be placed in the CAMELS website.

ACTION 5(i): all partners to send CAMELS publications over the last year (feb 05).

Meeting closed:  5pm on 26th January 2005