Carbon dioxide goes with the flow

Measurements of the rate at which carbon dioxide is released from rivers running through tropical forests provide a surprise. They will help in developing an improved picture of the carbon cycle.

Rainforests contain not only trees but also lots of water, largely in the form of river systems. The Amazon is by far the largest such system in the world, contributing 20% of all water flowing from rivers to the ocean. But how this and other great rivers participate in the global carbon cycle is a puzzle — relatively small quantities of carbon are detected in the outflow, yet organic material from the adjacent forest is commonly observed as floating debris. On page 617 of this issue¹, Richey et al. provide insight into the watery fate of the organic matter produced by the forest, showing that much of it is released into the atmosphere as CO₂. There is some way to go, however, before we can balance the carbon-budget books.

The study¹ focuses on the central part of the Amazon basin (see the map on page 617). The authors show that the river water contains high concentrations of CO₂, which they infer is being released into the atmosphere at a surprisingly high rate per unit area — a rate that is comparable to that from respiration by soil organisms, and which contributes substantially to the 'decomposition flux' of CO₂ from the basin as a whole.

The recognition that the forest exports carbon, which decomposes in water and is released as CO₂, helps to reconcile some of the conflicting results from studies of the carbon cycle and carbon budget, both in the Amazon and on a global scale. The data concerned come from measurements by eddy covariance^2,3 (direct measurements of CO₂ flux made above the forest, using towers), analyses of forest inventories⁴ (calculations from repeated measurements of the number and size of trees in sample plots), and studies of the global atmosphere⁵ (calculations of the geographical distribution of CO₂ sources and sinks made from frequent and precise measurements of concentration in the Earth's atmosphere).

Several studies of eddy covariance suggest that about 5 × 10⁶ g of carbon accumulate per hectare per year in the dry-land — 'terra firme' — forests of the Amazon basin. This is a surprisingly large amount, but to our knowledge only two studies have produced much smaller estimates². Although it is expected that the 'fertilization' that arises from increasing levels of CO₂ in the atmosphere would enhance the uptake of carbon, with photosynthesis exceeding respiration, it seems from both calculations and experiments that this effect is rather small for mature forests⁶. If the high rate of carbon accumulation were to apply to the entire basin, we would have to conclude that the Amazon is a giant carbon sink, perhaps absorbing two-thirds of the world's fossil-fuel emissions. This seems unlikely, and moreover is not consistent with either forest-inventory analyses or global atmospheric studies, which suggest that the Amazonian sink is much smaller. Richey et al.¹ show that significant quantities of the carbon assimilated by the forest are almost certainly carried away, and decompose elsewhere. In other words, the eddy-covariance towers situated in the forests do not give a true picture of the overall decomposition.

These, however, are highly complicated issues, and inevitably there are questions to be asked about the new results. For instance, is the authors' method for calculating CO₂ flux reliable? The method is the same as that used by oceanographers, and is relatively simple: flux is calculated as the difference in CO₂ partial pressure between the water and the atmosphere (in this case, measured at 1,800 sites), multiplied by an exchange coefficient. As used by Richey et al.¹, the calculation can be expected to yield a conservative estimate of CO₂ flux under most conditions because the exchange coefficient is determined inside floating chambers, without normal air movement, so the true flux may be even higher.

To see the new work in a broader context, the central Amazon basin can be represented as a three-component system: the terra firme forest, the seasonally inundated forest and the permanent river. To try to identify the uncertainties in the carbon budget, we have combined Richey and colleagues' data on outgassing of CO₂ by the river with published carbon fluxes over terra firme forest⁷ and textbook data on fluxes over inundated forest⁸. The outcome is shown in Fig. 1. We calculate that the residual carbon flux from the river, which must balance the inputs and outputs, is 1.13 × 10¹⁴ g C yr⁻¹. This suggests either that the authors' estimate of CO₂ flux is indeed conservative, or that the calculated input of dissolved carbon from terra firme forest, estimated by Richey et al. as 40% of the total riverine CO₂ flux, is too high.

Figure 1: Estimates of carbon flows in the central Amazon basin.

As defined by Richey et al.¹, this is a region of 1.77 million km², and here it is shown divided into terra firme forest (T), seasonally inundated forest (S) and permanent river (R). All units are 10¹² g C yr⁻¹. The area of seasonally inundated forest (153,000 km²) represents the mean area that is flooded over the year. N, net primary productivity; R, heterotrophic respiration; VOC, flux of volatile organic compounds (methane, isoprene, monoterpenes); F_aq, flow of carbon from forest to river as dissolved CO₂ and dissolved organic carbon; ▵B, net carbon accumulation in biomass; ▵S, net carbon accumulation in soils; L_S, direct litter flow from flooded and riverside forests into water; E_R, net export of dissolved or inorganic carbon in the Amazon's main stream. N_T and R_T are derived from ref. 7. N_S per unit area is assumed to be the same as N_P (the net primary productivity of the terra firme forest⁸). ▵B and ▵S are derived from forest inventories and estimates of soil-carbon residence times⁹. E_R is scaled down from the estimate for the entire Amazon basin¹. X is a residual efflux from the river, which is required to account for the fact that the flux reported by Richey et al. is less than the carbon entering the river.

Even taking the export of carbon into account, there remains a problem in balancing the books for terra firme forests. Net carbon uptake is often measured by eddy covariance as 500 g C m⁻² yr⁻¹, of which about 70 g are accumulated in the biomass and 70 g in the soil carbon pool, 50 g are lost as volatile compounds, and 60 g are exported to the river. This leaves 250 g C m⁻² yr⁻¹ of carbon uptake as 'missing carbon', which must be accounted for. One possible explanation, which is currently being explored, is that, during the night, CO₂-rich air in the forest canopy drains to valley bottoms and towards rivers.

Finally, to interpret fully the results of Richey et al., we need to know more about the origins of the CO₂ that comes from the river. The sources are likely to be organic matter from terra firme forest, from inundated forest and from vegetation on the riverside and in the river itself. Richey et al. provide an estimate of the relative importance of these sources, but their figures are highly uncertain. We also need clarification on related questions, such as whether the carbon is young (from decaying litter) or old (from within the soil). Studies using chemical and isotopic tracers are the way forward in this respect.