Concerns about the effects of increasing levels of atmospheric carbon dioxide on climate and ocean acidity have intensified the study of how carbon is exchanged between the atmosphere, land and oceans — especially how CO2 is removed from the atmosphere and transferred into long-term storage (sinks). The rates of removal into terrestrial and ocean sinks seem to be slowing down1, highlighting the need for improved data on the individual sinks, their controls and likely future behaviour, and their relationship to other carbon fluxes.

In the global carbon cycle, rivers are a fundamental, multi-faceted link between atmosphere, land and ocean2,3. Five years ago, a study3 of the Mississippi River in the United States found a large increase in the export of dissolved inorganic carbon to the ocean during the past 50 years, a result of the enhanced dissolution (weathering) of rock minerals. Links to changing agricultural practices and crop management, as well as to increased precipitation, were found, but the mechanisms remained unclear. On page 449 of this issue, Raymond et al.4 revisit this topic using a remarkable, century-long data set and more robust approaches. They conclude that direct effects from changes in land use outweigh the indirect influence of climate change.

The chemical weathering of rock material in soils typically consumes CO2 of atmospheric origin through reactions with carbonate and silicate minerals. These reactions are controlled by factors including climate, mineral type, the availability of acids (particularly carbonic acid) and physical erosion5. The resulting bicarbonate ion (HCO3) is the dominant form of dissolved inorganic carbon in most rivers, and does not exchange readily with atmospheric CO2. Instead, it is transported by water run-off to rivers and, ultimately, to the ocean, where it contributes to the ocean's alkalinity — and so its capacity to absorb CO2 and buffer changes in acidity.

The weathering of silicate rocks such as basalt and granite results in complete conversion of consumed CO2 to bicarbonate in soils and groundwater, and the weathering of carbonates such as limestone converts one molecule of mineral CO2 to bicarbonate for each molecule of atmospheric CO2 consumed5. Overall, chemical weathering is a sink of CO2 on land, and represents a transfer of atmospheric carbon and buffering capacity to the ocean at timescales relevant to human activity (hundreds to thousands of years). But in all this, human alterations of weathering fluxes through land management had typically been neglected in large-scale assessments.

Raymond et al.4 exploit alkalinity measurements taken at the water-treatment plants that supply the city of New Orleans with drinking water drawn from the Mississippi. Combined with information about river discharge from the US Geological Survey, these data provide an extraordinary 100-year record of monthly variability in water and bicarbonate export from North America's largest river. They show little change before the 1950s, but confirm the approximately 40% increase in bicarbonate export over the past 50 years previously assessed using an independent record3.

To investigate the factors driving this increased annual flux, Raymond et al. separate out two possible mechanisms. First, there are increases in water discharge that increase total amounts of bicarbonate export, but imply no change in bicarbonate concentration. Second, there are increases in bicarbonate flux at equivalent discharge values, representing net increases in mean bicarbonate concentration. The authors assess this second effect by normalizing annual bicarbonate flux to the long-term mean annual water discharge. They conclude that more than 60% of the increased bicarbonate flux results from increased concentrations, and not mainly from increased water discharge as was previously thought3.

Alterations in how the watersheds of the Mississippi basin work seem to have driven the increased bicarbonate discharges of the past 50 years. This period coincides with agricultural intensification, and changes in crop type and land-management practices in the basin. One must bear in mind, however, that the data from New Orleans, near the mouths of the Mississippi, provide only a highly aggregated picture. So Raymond et al.4 turn to an extensive sub-watershed data set of bicarbonate discharge, land use and precipitation distribution to probe the role of cropland extent, changes in hydrology and increased precipitation.

They find that increases in discharge in agricultural — but not grassland or forested — watersheds outpaced increases in precipitation. This conclusion agrees with recent hydrological studies6 that found that agricultural changes in the Mississippi basin, especially the expansion of soya-bean row crops, might be reducing evapotranspiration and increasing the fraction of precipitation that reaches the groundwater and flows to streams (Fig. 1). Together with practices such as liming soil to neutralize acidity, this enhanced flow through deeper mineral-rich soils results in increased bicarbonate exports from agricultural areas. The authors suggest that more than 50% of the total increase in bicarbonate export in the past 50 years results from the direct effect of land-use changes. The influence of precipitation increase and CO2 fertilization from atmospheric CO2 increase is comparatively small. Agriculturally driven hydrological alterations will, however, magnify the effect of precipitation increases.

Figure 1: Carbon exchange and land-use change.
figure 1

Even with similar soil types and well-buffered (mineral-rich) bedrock material, water and CO2 fluxes through soils in cropland (a) and non-agricultural ecosystems (b) can be very different. For the same precipitation rate, certain agricultural conditions (including drainage systems, crop type and tillage practice) result in a greater fraction of annual water flows occurring as groundwater run-off and a lower fraction occurring as evapotranspiration. The increase in groundwater run-off results in larger export to rivers of bicarbonate (HCO3) derived from the consumption of atmospheric CO2 during chemical weathering, as demonstrated by Raymond and colleagues' 100-year Mississippi time series4. Agricultural practices such as lime application might also increase the concentration of bicarbonate in soils and groundwater.

Does enhanced weathering tip croplands into becoming a net sink of atmospheric CO2 in comparison with the situation in pristine ecosystems such as forests or grasslands? The answer is complex. First, the acceleration of weathering reported by Raymond et al.4 is probably limited to mineral-rich soils and particular agricultural practices. More importantly, other direct human pressures must be considered. Besides the removal of biomass during conversion from pristine ecosystems, agricultural activities commonly lead to enhanced erosion of soils and loss of associated organic carbon, providing a potential source of CO2 to the atmosphere through oxidation downstream2. Sedimentation in the reservoirs that are often built to support irrigation can, on the other hand, serve as a long-term sink for eroded-soil organic carbon2. Consequently, the net global effect of agricultural soil erosion on atmospheric CO2 remains contentious2,7. Land-management effects on river-carbon exports should be considered in parallel with the more extensively studied effects on nutrient exports. The Mississippi basin represents a notable example of both8.

An intensification of groundwater flow and bicarbonate flux analogous to, but of smaller magnitude than, that highlighted by Raymond et al. seems to be occurring in some Arctic basins. Here, it is an effect of climate change on freeze–thaw cycles and permafrost coverage9. Thus, changes in chemical weathering caused both by direct human alterations and by the indirect effects of climate change seem to be a more significant component of regional and global perturbations of the carbon cycle than was thought.

This latest study4 adds to mounting evidence that the levels of carbon export from developed watersheds today might not be representative of conditions before pressures such as intensive agriculture and acid deposition from industrial emissions became widespread10. Long biogeochemical time-series such as that from the New Orleans water-treatment stations are invaluable tools for discerning what the dominant controls on river-carbon exports are, and how they respond to large-scale human activities.