Measurements of carbon dioxide flux over undisturbed tropical rain forest in Brazil for 55 days in the wet and dry seasons of 1992 to 1993 show that this ecosystem is a net absorber of carbon dioxide. Photosynthetic gains of carbon dioxide exceeded respiratory losses irrespective of the season. These gains cannot be attributed to measurement error, nor to loss of carbon dioxide by drainage of cold air at night. A process-based model, fitted to the data, enabled estimation of the carbon absorbed by the ecosystem over the year as 8.5 ± 2.0 moles per square meter per year.

We present the energy and mass balance of cerrado sensu stricto (a Brazilian form of savanna), in which a mixture of shrubs, trees and grasses forms a vegetation with a leaf area index of 1-0 in the wet season and 0-4 in the dry season. In the wet season the available energy was equally dissipated between sensible heat and evaporation, but in the dry season at high irradiance the sensible heat greatly exceeded evaporation. Ecosystem surface conductance gî n the wet season rose abruptly to 0-3 mol m"^ s"^ and fell gradually as the day progressed. Much of the total variation in gs was associated with variation in the leaf-to-air vapour pressure deficit of water and the solar irradiance. In the dry season the maximal g^ values were only 0-1 mol m"^ s~\ Maximal net ecosystem fluxes of CO2 in the wet and dry season were -10 and -15 /Jmol CO2 m'^ s~', respectively (sign convention: negative denotes fluxes from atmosphere to vegetation). The canopy was well coupled to the atmosphere, and there was rarely a significant build-up of respiratory CO2 during the night. For observations in the wet season, the vegetation was a carbon dioxide sink, of maximal strength 0*15 mol m"^ d~\ However, it was a source of carbon dioxide for a brief period at the height of the dry season. Leaf carbon isotopic composition showed all the grasses except for one species to be C4, and all the palms and woody plants to be C3. The CO2 coming from the soil had an isotopic composition that suggested 40% of it was of C4 origin.

A simple 'big leaf ecosystem gas exchange model was developed, using eddy covariance data collected at an undisturbed tropical rainforest in south-western Amazonia (Brazil). The model used mechanistic equations of canopy biochemistry combined with an empirical stomatal model describing responses to light, temperature and humidity. After calibration, the model was driven using hourly data from a weather station at the top of the tower at the measurement site, yielding an estimate of gross primary productivity (annual photosynthesis) in 1992/1993 of about 200 mol C m~^ year"\ Although incoming photon flux density emerged as the major control on photosynthesis in this forest, at a given PAR CO2 assimilation rates were higher in the mornings than in the afternoons. This was attributable to stomatal closure in the afternoon in response to increasing canopy-to-air vapour pressure differences. Although most morning gas exchange was clearly limited by the rate of electron transport, afternoon gas exchange was generaWy observed to be very nearly co-limited by both Rubisco activity (V^^^) and electron transport rate. The sensitivity of the model to changes in nitrogen allocation showed that the modelled ratio of V^^^^ to electron transport (•/max) served nearly to maximize the annual carbon gain, and indeed, would have resulted in almost maximum annual carbon gain at the pre-industrial revolution atmospheric CO2 concentration of 27 Pa. Modelled gross primary productivity (GPP) was somewhat lower at 27 Pa, being about 160mol C m~^ year"'. The model suggests that, in the absence of any negative feedbacks on GPP, future higher concentrations of atmospheric CO2 will continue to increase the GPP of this rainforest, up to about 230 mol C m"^ year' at 70 Pa.

1 Carbon dioxide and water vapour fluxes were measured for 55 days by eddy covariance over an undisturbed tropical rain forest in Rondonia, Brazil. Profiles of CO2 inside the canopy were also measured. 2 During the night, CO2 concentration frequently built up to 500 ppm throughout the canopy as a result of low rates of exchange with the atmosphere. In the early morning hours, ventilation of the canopy occurred. 3 Ecosystem gas exchange was calculated from a knowledge of fluxes above the canopy and changes of CO2 stored inside the canopy. Typically, uptake by the canopy was 15 |imol m"^ s"^ in bright sunlight and dark respiration was 6-7 pmol m^^ s"\ The quantum requirement at low irradiance was: 40 mol photons per mol of CO2. 4 Bulk stomatal conductance of the ecosystem was maximal in the early moming (0.4-1.0 mol m^^ s'^) and declined over the course of the day as leaf-to-air vapour pressure difference increased.

1 Eddy covariance measurements of CO2 flux, based on four and six week campaigns in Ronddnia, Brazil, have been used in conjunction with a model to scale up data to a whole year, and thus estimate the carbon balance of the tropical forest ecosystem, and the changes in carbon balance expected from small interannual variations in cUmatological conditions. 2 One possible source of error in this estimation arises from the difficulty in measuring fluxes under stably stratified meteorological conditions, such as occur frequently at night. Flux may be 'lost' because of low velocity advection, caused by nocturnal radiative cooling at sites on raised ground. Such effects may be detected by plotting the net ecosystem flux of CO2, Fgco/ 3S a function of wind speed. If flux is 'lost' then F^^^ is expected to decline with wind speed. In the present data set, this did not occur, and Fgco was similar to the nocturnal flux estimated independently from chamber measurements. 3 The model suggests that in 1992/3, the Gross Primary Productivity (GPP) was 203.3 mol C m"^ y"^ and ecosystem respiration was 194.8 mol C m"^ y"^, giving an ecosystem carbon balance of 8.5 mol C m"^ y"^, equivalent to a sink of 1.0 ton C ha"^ y"^. However, the sign and magnitude of this figure is very sensitive to temperature, because of the strong influence of temperature on respiration. 4 The model also suggests that the effect of temperature on the net carbon balance is strongly dependent on the partial pressure of CO2.