Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

FP5

CAMELS Résumé de rapport

Project ID: EVK2-CT-2002-00151
Financé au titre de: FP5-EESD
Pays: Germany

D2.2 Biome-dependent ecosystem parameters plus uncertainty bounds for each TEM

MAX
We inverted the parameters of the terrestrial ecosystem model BETHY against eddy covariance measurements of Net Ecosystem exchange of CO2 (NEE) and Latent heat (LE) flux at different sites considering a priori information about parameters in a Bayesian context. Based on the experience with the method and constrained a priori uncertainties, we analysed the robustness of the inversion of model parameters against eddy covariance measurements with respect to differences in phenology, data selection within a given year, the seasonal cycle and inter-annual variability for the Hainich site (Germany), a broadleaved deciduous forest.

The inversion was sensitive to the seasonal cycle and phenology, but rather insensitive to the selection of data within a given year as long as the selected data were not too few and almost equally distributed throughout the year. Also the inversion was insensitive to different years. A number of about 12 days of half-hourly data (1 day per month), carefully chosen to select days with high quality data, was sufficient to represent the average parameter set and to constrain parameter uncertainties. The posterior parameter-sets for different years derived from 12 days each were then quite similar. Two parameters were consistently driven out of their prior SE ranges: the basal soil respiration (Rhet0) and the soil water content at the permanent wilting point (SWC). A comparison of different sites showed quite large differences between sites, even of the same plant functional type. Thus it will be important to use a sufficient number of sites per plant functional type to characterize its average behaviour.

LSCE
We applied the same optimisation procedure to 14 different FluxNet sites from mid to high latitudes. These sites cover three major types of ecosystems: Deciduous broadleaf trees (4 sites), temperate conifers (6 sites), and boreal conifers (4 sites). For each site we were able to assimilate on average three years of observations. Most of the parameters are considered to be constant from one year to the next with the exception of the Vcmax, b, Q10 and Albedo parameters. Compared to the first case defined for the Bray site, we added few parameters controlling essentially the phenology and the assimilation of each type of ecosystem within the ORCHIDEE model (beginning and end of the growing season, critical leaf age and temperature dependency of the carboxylation rates).

We also built a cost function integrating both seasonal and diurnal information of the data. The optimised model outputs appear to be significantly closer to the observations than the prior model outputs, for most sites. Figure 4 illustrates the NEE model data fits for one site of the three different ecosystems. The amplitude of the seasonal cycle is modified in each case to reproduce the observed amplitude each year and the phase of the seasonal cycle is also adjusted to match the observed growing season length. Nevertheless, we are still missing some synoptic events especially during the summer. These events are usually associated to drought or rainy periods and the rather simplified soil hydrology in ORCHIDEE (double bucket model) is not able to properly reproduce the water stress and its effect on the NEE or LE fluxes on a synoptic time scale. If we consider the monthly mean diurnal cycles, the optimisation also succeed to improve the model data fit and especially to increase the CO2 uptake during the day according to the observations. However the changes of the diurnal cycle amplitude between spring, summer and autumn is still not perfectly modelled.

MET OFFICE
The best fit of model is determined by selecting the highest percentage of variance explained (PVE) from multiple simulation using 105 sets of parameter combinations. The standard set of parameters, optimised set of parameters and their literature range are given in table 1. The optimised parameters fit the model well and the fitness improved from 32.07 to 60.12 for broad leaf trees, 31.88 to 45.65 for Needle leaf tree, 8.45 to 15.45 for C3 grasses, 35.46 to 58.46 for C4 grasses and 38.18 to 51.12 for shrubs. Of five parameters used in the optimisation exercise, the leaf nitrogen (nl0) is most sensitive and constrained well with the observations. The quantum efficiency parameter (a ) is sensitive to day time fluxes that night time fluxes. The co variation of the parameters is studied showed that the critical moisture content and critical humidity deficit are correlated. Inverting the model parameters against eddy covariance data has shown that the optimised parameters fits the observations better compared to using standard set of parameters.

We inverted the parameters of the terrestrial ecosystem model BETHY against eddy covariance measurements of Net Ecosystem exchange of CO2 (NEE) and Latent heat (LE) flux at different sites considering a priori information about parameters in a Bayesian context.

Based on the experience with the method and constrained a priori uncertainties, we analysed the robustness of the inversion of model parameters against eddy covariance measurements with respect to differences in phenology, data selection within a given year, the seasonal cycle and inter-annual variability for the Hainich site (Germany), a broadleaved deciduous forest. The inversion was sensitive to the seasonal cycle and phenology, but rather insensitive to the selection of data within a given year as long as the selected data were not too few and almost equally distributed throughout the year. Also the inversion was insensitive to different years. A number of about 12 days of half-hourly data (1 day per month), carefully chosen to select days with high quality data, was sufficient to represent the average parameter set and to constrain parameter uncertainties. The posterior parameter-sets for different years derived from 12 days each were then quite similar. Two parameters were consistently driven out of their prior SE ranges: the basal soil respiration (Rhet0) and the soil water content at the permanent wilting point (SWC).

A comparison of different sites showed quite large differences between sites, even of the same plant functional type. Thus it will be important to use a sufficient number of sites per plant functional type to characterize its average behaviour.

Informations connexes

Contact

Ernst-Detlef SCHULZE, (Professor)
Tél.: +49-364-1643642
Fax: +49-364-1643665
E-mail