Carbon Assimilation and Modelling of the European Land Surfaces

CCDAS

by Wolfgang Knorr and Peter Cox

CAMELS is an EU funded project on "Carbon Assimilation and Modelling of the European Land Surface", forming part of the CarboEurope cluster of projects.  The aims of CAMELS are to produce:

- best estimates and uncertainty bounds for the contemporary and historical land carbon sinks in Europe and elsewhere, isolating the effects of direct land-management.

- 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".

It is designed to address important questions concerning the global and European carbon cycle from a combined data and modelling point of view. As one of the CarboEurope projects starting towards the end of the 5^th Framework Programme, it capitalizes on the rich data findings of the project cluster by integration into a consistent modelling framework, using several European state-of-the-art ecosystem models. Scientific questions addressed include:

·  Where are the current carbon sources and sinks located on the land and do European sinks compare with sinks of other large continental areas?

·     Why do these sources and sinks exist, i.e. what are the relative contributions of CO₂ fertilisation, nitrogen deposition, climate variability, land management and land-use change?

·      How could we make optimal use of existing data sources and the latest models to produce operational estimates of the European land carbon sink?

The first question has up to now been addressed by two major approaches, one with a bottom-up, and another with a top-down view. Bottom-up modelling uses data from the field and basic process understanding in carbon cycle and plant physiology to compute CO₂ fluxes between the land and the atmosphere. The top-down approach uses atmospheric measurements in combination with an inverse atmospheric transport model to infer the same quantity. In addition, CarboEurope and other international projects have created a wealth of information on locally measured CO₂ and water exchange fluxes at the stand level.

The advantage of the top-down, atmospheric inversion approach is that it relatively little prior information on flux patterns enters the calculation (but biosphere model results are often used to generate prior, "first guess" fluxes), and that it works on large scales. Its main disadvantage is relevant to the second question above: there is no information gained concerning processes. The advantages of the bottom-up approach are complementary: it can make use of process knowledge, is thus able to distinguish between natural and management effects (required under the Kyoto protocol, for example), and can be used prognostically. Its disadvantages: it is often uncertain, there are large data gaps, and it cannot make optimal use of large-scale constraints.

CAMELS uses a novel approach, termed Carbon Cycle Data Assimilation System (CCDAS), that combines both views and adds a few additional elements. An additional innovation is that CAMELS produces consistent uncertainty bounds on carbon fluxes that are essential for policy purposes. It starts from flux measurements at the stand scale, which are used to improve and best parameterise a number of ecosystem models. The exercise also yields uncertainty bounds for ecosystem model parameters, and, by using data from all major biomes, a notion of the representativeness of the models and parameterisations.

The assumption used in CAMELS is that the best way to spatially extrapolate the results from the flux measurements is not through fluxes, but through parameter values that describe the underlying processes. Hence, the parameter values optimised from the site data are used as a priori values in a global carbon cycle data assimilation system (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 (Scholze, 2003). During El Niño (warm) pacific conditions, large parts of the tropical ecosystem come under water stress with reduced photosynthesis (see Figure 5.6.1). Of the 58 parameters that enter the optimisation, considerable reduction in uncertainty is found for about 12 (Figure 5.6.2). 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 (Scholze et al., 2002, Figure 5.6.3).

Building on the experience gained with CCDAS, CAMELS is currently working on a series of historical ecosystem model simulations that span the entire 20^th century and that include further processes, such as land management and nitrogen deposition. The final aim is to present a concept for an operational system that is able to optimally combine all relevant large-scale observations to deliver the best possible estimates of European and global CO₂ fluxes on a routine basis. Further information about CAMELS is available form http://www.bgc-jena.mpg.de/public/carboeur/projects/camels.htm; for CCDAS please check the website http://www.ccdas.org. Members of the the CCDAS consortium are Marko Scholze, Wolfgang Knorr, Heiner Widmann (Max-Planck Institute for Biogeochemistry, Jena), Peter Rayner (CSIRO, Melbourne),Thomas Kaminski and Ralf Giering (FastOpt, Hamburg).

Figure 5.6.1: Time series of global monthly fluxes prognosed from CCDAS smoothed with a five month running mean filter, after subtracting average seasonal cycle. Red/blue arrows: exception El Niño/La Niña events, yellow arrow: Pinatubo eruption.

Figure 5.6.2: Reduction in uncertainty, expressed as 1-s_prior/s_optimized., for the 58 parameters of BETHY used in CCDAS.

Figure 5.6.3: Estimated biosphere carbon sink strength and associated uncertainties compared to fossil fuel and land use emissions for 5 regions.