Net Exchange of CO2 in a Mid-Latitude Forest

The eddy correlation method was used to measure the net ecosystem exchange of carbon dioxide continuously from April 1990 to December 1991 in a deciduous forest in central Massachusetts. The annual net uptake was 3.7 � 0.7 metric tons of carbon per hectare per year. Ecosystem respiration, calculated from the relation between nighttime exchange and soil temperature, was 7.4 metric tons of carbon per hectare per year, implying gross ecosystem production of 11.1 metric tons of carbon per hectare per year. The observed rate of accumulation of carbon reflects recovery from agricultural development in the 1800s. Carbon uptake rates were notably larger than those assumed for temperate forests in global carbon studies. Carbon storage in temperate forests can play an important role in determining future concentrations of atmospheric carbon dioxide.

. The spectrograph collects the radiation from an area of 3 am in diameter of the sample through a pinhole with 50 gm in diameter. The temperature gradient over this area is only a few degrees. The error bars in Fig. 2 reflect the laser power fluctuation (root mean square is less than <0.5%). We followed the method suggested by Boehler et al. (2) in using a ruby crystal on the top of the sample as the isolator. The use of crystals eliminates the possibility of any significant reaction with the sample. In particular at the subsolidus temperature range of our study, reaction between the iron and the ruby was not a problem. A circular area with diameter of 3 gm was sampled by the spectrograph to get one pressure value; this method reduces errors in pressure determination resulting from the presence of a pressure gradient in the chamber. The pressure was determined at room temperature. Following the discussion by D. L. Heinz [Geophys. Res. Lett. 17, 1161Lett. 17, (1990], we added 7% to the observed value to account for the thermal pressure. Before measurement, we heated a spot with the laser to relax the sample assemblage mechanically. The pressure usually dropped by up to 16% of the initial unrelaxed value. The pressure drop was almost linear and could be stabilized by repeated heating. The errors in pressure shown in Fig. 2 are based on a value of 7% and thus accounts for not only point to point variation within the heated area of the sample (within 1 %) and any misjudgment of the thermal pressure. 19 (3), with equilibrium carbon stocks estimated on the basis of allometric relations derived from a small number of destructive harvests (9-1 1) and with limited information on belowground components (10). Estimates of rates for carbon uptake by forests can be significantly refined if the eddy correlation technique is used to determine the net ecosystem flux (12), for time scales from hours to years. This method provides a direct measurement of annual net uptake as well as information on underlying ecosystem processes. However, eddy correlation studies have not been carried out for periods long enough to define seasonal or annual carbon exchange in forests (13). In this report we present nearly continuous measurements of net ecosystem exchange (NEE) for CO2 for 2 years in a regenerating temperate forest using the eddy correlation method. We determined hourly, daily, and seasonal net fluxes for the ecosystem for 1990 and 1991. Our study site was located at Harvard Forest, Petersham, Massachusetts (42.540N, 72.18'W; elevation, 340 m), in a 50to 70-year-old mixed deciduous forest (red oak, red maple, and white and red pine, with scattered individuals of yellow and white birch, beech, ash, sugar maple, and hemlock). The terrain was moderately hilly, -95% forested, with the nearest paved roads > 1 km away and small towns > 10 km away.
We instrumented a tower to measure the exchange of CO2 by eddy correlation at an altitude of 30 m, 6 m above the canopy (14). We computed the vertical flux from the covariance of fluctuations of vertical wind speed with CO2 concentrations, averaged over 30 min (14,15), with special consideration of nighttime data (16). We obtained companion measurements of CO2 concentrations sequentially (two cycles per hour) at 29, 24, 18, 12, 6, 3, 1, and 0.05 m, using an independent gas analyzer. The net exchange of CO2 between the atmosphere and the vegetation plus the soil, the NEE (in kilograms of carbon per hectare per hour) is defined by Combustion of fossil fuel releases -5.5 Gt of carbon per year (1 Gt = 109 metric tons) to the atmosphere (1), with an additional 1 to 2 Gt year-l released from tropical deforestation (2,3). About 2 Gt yearis removed by the ocean (4, 5) and 3 Gt year-l accumulates in the atmosphere (6). The balance, 1 to 2 Gt year-l, is often presumed (6, 7) to be stored by aggrading temperate 1314 forests. However, available estimates (3,8) indicate that the rates of net carbon uptake by temperate forests are insufficient to balance the global carbon budget.
Mean rates for carbon uptake by aggrad- is the rate of change of CO2 content between 0 and 30 m. The NEE of CO2 was positive (respiration exceeded assimilation) from fall through spring, except for warm intervals when photosynthesis by conifers (-25% of the canopy) and understory plants was observed. Carbon uptake increased dramatically in June after leaves emerged, peaked in July and August, and then declined in late September during leaf senescence (Fig. 1).
Chamber measurements at Harvard Forest (14) indicated that belowground respiration accounted for >80% of R, with much of the balance due to stem respiration. Values for R are similar to soil fluxes ob--0 U ul z served in a variety of temperate forests (17,18). The annual integral for R, based on the use of hourly soil temperatures in Eq. 2, was 7.4 metric tons of carbon per hectare in 1991, slightly larger than obtained from a regression of soil flux against mean annual air temperature for a variety of forests (18) (6.5 metric tons per hectare per year at 7.50C, the annual mean air temperature for Harvard Forest).
Carbon uptake increased systematically with incident photosynthetically active radiation (PAR) (Fig. 3A). Uptake was slightly less after noon than in the morning for a given PAR, indicating modest effects of water stress, carbohydrate status, elevated air or soil temperatures, or lower ambient CO2. We fit hourly mean data to a2PAR  (4) as the number of CO2 molecules fixed per day by the canopy per incident photon.
Values of (q,,ff) in midsummer averaged 0.02, 40 to 50% of a2. Climatic conditions had only slight influence on (qff) (for example, the modest decline during dry weather in July 1990), indicating that (q. is a well-defined property of the ecosystem. Eddy correlation measurements of NEE for selected intervals, along with measurements of incident PAR and T,, could provide sufficient data to parameterize (qeff) and R as functions of PAR and TS in mesic forests. Annual net carbon storage could then be computed from climatological observations of PAR and T, (21).
We measured the diurnal course of leaf CO2 exchange under ambient light at two levels in the canopy. The leaf data were scaled to ground area on the assumption that observations at upper and lower levels represented 1 and 2 m2 of leaf area for each square meter of ground, respectively (1 and 1.5 in September) (22); R, computed from Eq. 2, was added to the scaled leaf measurements for comparison with the tower data. The leaf measurements reproduced most features of the diurnal and seasonal changes observed in the tower data ( Fig. 1), except for a tendency to underestimate photosynthesis at midday.
Harvard Forest took up 6 metric tons of carbon per hectare in the growing season and released 2 metric tons per hectare in the dormant period, both in 1990 and in 1991 (Fig. 4), implying a net ecosystem production (NEP) of 4 metric tons of car-Ts (CC) Fig. 2. System respiration between 2300 and 0300 (EST) (R), and soil temperature (T, at a depth of 5 cm), averaged over 1 0-day periods, for 1990 (0) and 1991 (a). The linear leastsquares fit (Eq. 2) accounts for 70% of the range of 3.7 ± 0.7 metric tons per hectare per year. This value is consistent with recent allometric measurements of net growth at nearby plots (3.1 to 3.6 metric tons per hectare per year) (23), but larger than the 0 to 2.5 metric tons per hectare assumed in global carbon models (3,8) or the 0.5 metric ton per hectare inferred from silvicultural inventories of European forests (9). Gross ecosystem production, the annual assimilation of C02, was approximately 11.1 metric tons of carbon per hectare per year (the sum of NEP and heterotrophic and autotrophic respiration).
Forest age and historical land use are expected to be major factors that regulate the rate of CO2 uptake. Harvard Forest is heterogeneous, reflecting topography and past land use. Portions were cleared for agriculture between 1790 and 1830, then abandoned, reverting to forest by 1890 (24). Commercial logging in the 1920s and 1930s declined after a devastating hurricane in 1938. Eddy correlation measurements sample forest metabolism 100 to 500 m upwind of the tower during the day, with longer fetch at night (25). Our data show that R was highest when the fetch sampled poorly drained land northwest of the tower, an area used as a woodlot during the agricultural period, presently with patches of older hemlock (24). Canopy photosynthesis (NEE -R) did not show this asymmetry. We estimate that annual NEP for the former woodlot was about half of that for the more extensive area used previously for crops and pasture, indicating the importance of past land use, stand age, and soil characteristics in regulating net carbon storage.
Influx of nitrate and ammonium from the atmosphere, 10 kg of nitrogen per hectare per year (26), could stimulate growth, with associated uptake of as much as 2 metric tons of carbon per hectare per year if the additional biomass were primarily wood with a carbon/nitrogen ratio of -200. However, nitrogen amendments to a nearby plot (23) indicated efficient microbial immobilization of nitrogen in these low-nutrient soils, implying that less than 10% of the net annual carbon uptake may be attributed to current rates of nitrogen deposition.
Ecosystem respiration rates (R) and quantum yields at Harvard Forest are similar to those in other temperate forests (17)(18)(19)21). The potential global area of temperate deciduous forests is 1.32 x 109 ha,~50% cultivated at present (27). If the rate of carbon accumulation at Harvard Forest were representative, global uptake by temperate forests could exceed 2 Gt of carbon per year. Much of the currently forested area is managed for wood products or was formerly used for agriculture. If carefully managed, temperate forests could represent a significant global sink for CO2. Boreal forests may also take up significant quantities of carbon: we observed a net uptake of 0.6 metric ton of carbon per hectare in Quebec [500N, 720W (28)1 during July and August 1990, using the eddy correlation method.
The direct flux measurements presented here define NEE for time scales from 1 hour to several years, accounting fully for the storage of carbon above and below ground. The study demonstrates that eddy correlation observations may be carried out for long periods with sufficient resolution to integrate and partition carbon budgets for major ecosystems. We anticipate that similar investigations in a variety of ecosystems could advance our understanding of the global carbon cycle, by helping to develop and test mechanistic process models and remote-sensing algorithms and by providing data on the response of ecosystems to climatic variations.  In 1992 we measured belowground, stem, and leaf respiration using an array of ten chambers, continuously ventilated, sampled for several days at one site and moved weekly to cover ten locations around the tower. 15. B. B. Hicks and R. T. McMillen, Boundary Layer Meteorol. 42, 79 (1988). 16. Observations at night were analyzed for errors associated with stratification, which shifts transport toward higher frequency turbulence and decouples air flow at the sensor from the forest. We corrected lack of response to high-frequency CO2 fluctuations by computing, in each hourly interval, the error in the sensible heat flux (usually 4 to 10%) when the high-frequency temperature signal was filtered to simulate the effective bandpass of the CO2 analyzer (15). We assessed errors due to decoupling by comparing NEE in stratified periods with data from well-coupled (windy) nighttime intervals (momentum flux < -500 cm-1 s-2), and with the sum of measurements (14) for belowground, stem, and leaf respiration. It appears that unstable temperature gradients usually maintain turbulent exchange between 0 and 30 m, and errors associated with decoupling are small: nighttime NEE may be underestimated in midsummer by 0 to 0.5 kg of carbon per hectare per hour, corresponding to a possible overestimate of the annual net uptake of 0 to 0.5 metric ton per hectare.