mundssonAbstract
Carbon exchange between the terrestrial biosphere and the atmosphere is one of the key processes that need to be assessed in the context of the Kyoto Protocol1. Several studies suggest that the terrestrial biosphere is gaining carbon2, 3, 4, 5, 6, 7, 8, but these estimates are obtained primarily by indirect methods, and the factors that control terrestrial carbon exchange, its magnitude and primary locations, are under debate. Here we present data of net ecosystem carbon exchange, collected between 1996 and 1998 from 15 European forests, which confirm that many European forest ecosystems act as carbon sinks. The annual carbon balances range from an uptake of 6.6 tonnes of carbon per hectare per year to a release of nearly 1 t C ha -1 yr-1, with a large variability between forests. The data show a significant increase of carbon uptake with decreasing latitude, whereas the gross primary production seems to be largely independent of latitude. Our observations indicate that, in general, ecosystem respiration determines net ecosystem carbon exchange. Also, for an accurate assessment of the carbon balance in a particular forest ecosystem, remote sensing of the normalized difference vegetation index or estimates based on forest inventories may not be sufficient.
The terrestrial sink for carbon is estimated to be of the order of 2 
1 Gt C yr-1 (ref. 1). In the Northern
Hemisphere, the terrestrial biosphere is currently absorbing carbon according
to several studies2, 3, 4, 5, 6, 7, 8. These studies use different
techniques dependent on indirect estimates of the carbon fluxes, like isotopic
analysis and inversion methods from CO2 concentration measurements5, 6, 7, remote sensing8, growth trend analysis2, 3, 4 and modelling. All these methods provide the necessary global
and continental scale perspective for carbon balance calculations. However,
these studies suffer from uncertainties in the assumptions used. For instance,
in the inverse modelling studies the anthropogenic sources and sinks are frequently
prescribed a priori and they lack adequate representation of the carbon balance
at local scales. Their use in addressing small temporal and spatial changes
in the carbon balance is therefore rather limited.
The net carbon exchange of terrestrial ecosystems is the result of a delicate balance between uptake (photosynthesis) and loss (respiration), and shows a strong diurnal, seasonal and annual variability. Under favourable conditions, the net ecosystem flux is dominated by photosynthesis during daytime, and by respiration at night and for deciduous ecosystems in leafless periods. The influence of climate and growing-season length can in some cases shift a terrestrial ecosystem from a sink to a source of carbon9, 10, 11.
Global- and continental-scale techniques are of limited use in addressing one of the key questions raised by the Kyoto Protocol, namely how to calculate the changes in "carbon stocks" associated with land use changes and forestry activities during the commitment period. Indeed, one of the major effects of land-use changes, including the afforestation, reforestation and deforestation of land, is to change soil organic matter (SOM), by both build up and decomposition12. For most ecosystems, the changes in stocks of soil carbon in a 4–5 year period are unfortunately within the errors of the survey techniques used. Remote sensing approaches also appear inadequate for such purposes, because they have limited capability for estimating below canopy processes such as soil respiration.
In this context, the direct, long-term measurement of carbon fluxes by the eddy covariance technique13 offers the possibility of assessing on a local scale the carbon sequestration rates of forests and of different land-uses. The technique can also provide a better understanding of the vulnerability of the carbon balance of ecosystems to climate variability, and can be used to validate ecosystem models and to provide data for land surface exchange schemes in global models14.
Automated eddy covariance measurements of CO2 fluxes have been made routinely over 15 forests in Europe since 1996 within the EUROFLUX network15. In 1998, the network approach was expanded in the US (AMERIFLUX) and plans exist to implement similar networks in Brazil (the Large Scale Biosphere Atmosphere Experiment in Amazonia), South East Asia (the GEWEX Asian Monsoon Experiment) and Siberia. These tower sites are now forming a global network, FLUXNET16, with standard measurement protocols, data quality control and storage systems15. The flux stations measure the net flux of carbon entering or leaving the ecosystem. This is the flux which provides a measure of net ecosystem exchange (NEE), and, if summed annually, provides a direct estimate of the annual ecosystem carbon balance (excluding disturbances by harvest and fire which give rise to net biome productivity)17. In several studies the accuracy of annual sums has been estimated to be about 5%, or typically 0.3 t C ha-1 yr -1 (refs 18, 19), with the error influence decreasing with increasing size of the flux data set20.
To reduce the uncertainty associated with site-to-site variation in flux measurement methods and calculations and to make comparisons between sites, the EUROFLUX network was designed with the same hardware and software specifications at all sites15. The EUROFLUX results for 1996–98 show a sink strength of up to 6.6–6.7 t C ha-1 yr-1 for two forests in Southern Europe and for a Sitka spruce plantation in Scotland, and indicate that European old boreal forests are close to equilibrium and may switch from being a carbon source one year to a carbon sink the next (Table 1). Within the same biome, younger stands still gain carbon, although at a lower rate than temperate forests, Mediterranean forests or fast-growing plantations. Despite the wide range of species composition, stand structure, soils, tree age, site disturbance history and year-to-year variability, a consistent latitudinal trend in NEE is found ( Fig. 1). Indeed a multivariate statistical analysis on the effect of the single factors (latitude, precipitation, ecosystem type, elevation, mean annual temperature, age, management type, leaf area index) on NEE, showed that latitude is the most significant single variable model (r 2 = 0.55, P < 0.001). Latitude is not a phenomenological driving variable per se, however it is a good proxy for the actions of a multiplicity of factors (for example, radiation balance, length of growing season, frost events, disturbance regime).
Figure 1: Net ecosystem exchange (NEE) of the EUROFLUX sites plotted against latitude.
Closed symbols, forest of natural origin and planted stands with traditional forest management; open symbols, intensively managed plantations. According to the eddy covariance theory, a negative sign indicates that carbon is absorbed by the forest, while a positive sign indicates that carbon is released by the forest to the atmosphere.
High resolution image and legend (11K)
The trend indicates that high-latitude forests generally show lower and more variable carbon sequestration rates than low-latitude forests. The several forests growing within 50° and 52° N show a pronounced variability, with NEE ranging from an uptake of less than 1 t C ha-1 yr -1 (site Germany 1, point 7) to 5.4 t C ha-1 yr -1 (site Germany 2, point 11). In this latitudinal band, the variability can be related to stand, soil and climate characteristics, ranging from continental to maritime. With its maritime proximity, the intensively managed and fertilized fast-growing spruce plantation (site United Kingdom 1, points 19, 20) falls off the latitudinal trend, with a higher uptake of carbon than more continental stands located at similar latitude. Despite the large variation of NEE, gross primary production (GPP) is rather conservative across sites and latitude, indicating that other components of the carbon balance are responsible for the observed variation in NEE (Fig. 2). It is noteworthy that the young spruce plantation has the largest values of GPP, indicating strong stimulation of photosynthesis, while the young poplar plantation (point 25) growing in a cold climate at 64° N shows the smallest GPP.
Figure 2: Gross primary production (GPP) of the EUROFLUX sites plotted against latitude.
Closed symbols, forest of natural origin and planted stands with traditional forest management; open symbols, intensively managed plantations.
High resolution image and legend (10K)The observed variation in NEE across sites can be explained by the relative importance of ecosystem respiration (RE) in relation to NEE. The ratio NEE/RE increases with latitude (Fig. 3) indicating that RE becomes more important for northern sites and can explain the decrease of NEE previously shown. Generally, while GPP tends to be constant across sites, annual ecosystem respiration increases with latitude, despite the general decrease of mean annual air temperature (Table 1). It is well known that temperature has a strong effect on soil and plant respiration. For single sites our data also show a significant relationship between temperature and ecosystem respiration for both short and annual timescales. However, when a plot of RE versus temperature is drawn across all sites the relationship is not significant, indicating that mean annual air temperature may not be an important contributing factor to forest ecosystem respiration on a broader scale.
Figure 3: The ratio of net ecosystem exchange (NEE) and total ecosystem respiration (RE) plotted against latitude.
Closed symbols, forest of natural origin and planted stands with traditional forest management; open symbols, intensively managed plantations.
High resolution image and legend (10K)In forests, total ecosystem respiration tends to be dominated by root and microbial soil respiration. Boreal soils contain a larger amount than temperate soils of soil organic matter (SOM) in a labile form12, 21, 22, 23 that is prone to rapid decomposition21, 22. The effective temperature sensitivity (Q10) of SOM decomposition is much higher in colder than in warmer climates and temperature increases in cold regions are likely to affect decomposition rates more than net primary productivity23. There is also evidence that northern latitudes have warmed by more than 4 °C, while southern latitudes have warmed less24. This may have resulted in non-steady state conditions for SOM which could explain relative enhancement of respiration in the north compared to the south.
In this respect, land-use change and site history could also be important. For example, site Sweden 1 (points 21–23) is losing carbon as a result of past soil drainage, while the high respiration rates of the maritime spruce plantation may be linked to preparation of the site by ploughing, the favourable maritime climate and fertilization. Furthermore the relatively low rates of respiration of the southern sites may be the result of drought limitations to soil respiration25, 26, 27.
The carbon balance is ultimately a delicate equilibrium between the two large fluxes of photosynthesis and respiration, and this appears to be particularly true for boreal European ecosystems, making them very vulnerable to disturbances in climate. Indeed, annual variability for these high latitude sites is very pronounced, as shown by the remarkable variation in NEE from year to year: warm winters tend to switch old boreal stands from a sink to a source of carbon by increasing the annual amount of respiration9 (site Sweden 1; Table 1). In other boreal ecosystems, year-to-year changes in timing of the thawing of the soil in the spring are important for the carbon balance10.
The direct flux estimates of carbon exchange provide a useful tool for understanding the overall carbon balance processes of terrestrial ecosystems. Indeed, partial accounting of carbon dynamics can easily lead to erroneous conclusions. For example, plant biomass is currently increasing in all the EUROFLUX sites, even though some of these sites have a carbon budget close to neutral and one is losing carbon on a yearly basis. Similarly, the increases in plant growth at northern latitudes estimated by remote sensing of the normalized difference vegetation index (NDVI) must be examined critically in the light of these results, confirming the need to consider ecosystem respiration8. Also forest inventory-based carbon balance estimates should be carefully examined in relation to comprehensive carbon budget accounting. Furthermore, flux tower networks can provide at local scale realistic constraints on the global carbon balance estimates.
Methods
Instruments
In the EUROFLUX network, the same CO 2/H2O infrared gas analyser (LI-6262, Licor Inc.) and sonic anemometer (Solent, Gill Inst.) are used. All the analysers are calibrated against the same CO2 standard (NOAA Climatic Monitoring and Diagnostics Laboratory, Boulder, Colorado). The software for eddy covariance data acquisition and calculation have been extensively tested and compared against reference data sets, resulting in a maximum variation of calculated fluxes of less than 1% (ref. 15).
Data treatment
The collected data are quality controlled, corrected for frequency losses and sensor separation and, when needed, corrected for night-time fluxes under stable conditions with low wind speed or friction velocity typically less than 0.2 ms-1 (refs 14, 15).
NEE values for the entire year are obtained by summation of fluxes measured on a 30-min time step. The average coverage of directly measured data for all of the sites was more than 70% of the annual half-hour periods (60–95%). Fluxes in stable conditions and data gaps have been filled through site-based functional relationships using meteorological variables, such as radiation during the day and temperature during the night. Small gaps (a few half-hours) during single days were filled by simple interpolation. For data gap filling procedure see Supplementary Information.
GPP values have been obtained by summing annual NEE and ecosystem respiration (RE). RE comes from the summation of night-time fluxes (all sites), whole-day fluxes for leafless periods (deciduous forests) and of the day-time respiration.
Night-time and leafless-period fluxes are obtained by summation of the fluxes measured by eddy covariance on a 30-min time step, including the CO 2 storage component. Daytime respiration has been obtained by extrapolating the night-time fluxes to the rest of the day, using functional relationships with soil or air temperature.
A multivariate statistical analysis, based on different procedures, namely the forward and stepwise selections and the maximum R-square improvement, was used to test the controlling factors on NEE28.
