The Basins:

• Amazon
• Southeast Asia
• Puget Sound

Links:

• CENA/USP
• SEA START
• UW Oceanography

The River Story:

Carbon in Rivers: a Global Perspective

(drawn primarily from Richey 2004a,b)

As the main pathway for the ultimate preservation of terrigenous production in modern environments, the transfer of organic matter from the land to the oceans via fluvial systems is a key link in the global carbon cycle (Ittekot and Haake 1990; Degens et al. 1991; Hedges et al 1992). Hence, the “role” of rivers in the global carbon cycle is most typically expressed as the fluvial export of total organic and dissolved inorganic carbon from land to the ocean (e.g., Likens et al. 1981).

Within the overall  “role”  of river systems, an emerging issue is the importance of dissolved CO₂ (Richey et al 2002). The partial pressure of carbon dioxide dissolved in river water (pCO₂) represents a deceptively simple expression of the coupling of the water and carbon cycles between terrestrial and fluvial environments. The distribution of pCO₂ across a river basin is a function of a long sequence of complex biological and weathering processes and interactions, reflecting both internal carbon dynamics and external biogeochemical processes in upstream terrestrial ecosystems. The downstream expression of this coupling is the amount of organic matter and dissolved inorganic carbon mobilized to and through a river system, augmented by in-stream or riparian primary production and respiration. Perhaps the most evocative aspect of pCO₂   is that it is almost always present at concentrations much greater than the atmosphere (that is, it is supersaturated). The question is why, and what are the implications?

Dynamics of Fluvial Systems

Fluvial systems integrate hydrological and biogeochemical cycles, over scales from small streams to regional and ultimately to continental basins. Briefly, there are three primary forms of carbon of atmospheric origin that are transported through fluvial systems. Particulate organic carbon (POC) enters rivers from the erosion of soils (typically older materials) and as leaf litter (typically newly-produced). Dissolved organic carbon (DOC) is produced through solubilization of soil organic and enters streams via groundwater.  Total dissolved inorganic carbon (DIC) is produced via weathering, as the dissolution of carbonate and silicate rocks. This process sequesters atmospheric CO₂, establishes the alkalinity and influences the pH of water, which governs the subsequent partitioning of DIC between pCO₂, bicarbonate, and carbonate ions.  The dynamics of carbon in fluvial systems is not defined solely by the export fluxes of bulk C. Rather, it is defined as a complex interplay of multiple C fractions; each exhibits distinct dynamics and compositional traits that hold over very broad ranges of geological, hydrological and climatic conditions, (Hedges et al. 1994).

What are the sources of pCO₂, both direct and indirect? Total dissolved inorganic carbon (DIC) is produced via weathering, as the dissolution of primarily carbonate rocks. This process establishes the alkalinity and influences the pH of water, which governs the subsequent partitioning of DIC between pCO₂, bicarbonate, and carbonate ions. The DIC in groundwater is enriched many-fold by the CO₂ produced by the decomposition of organic matter in soils (in productive environments, soil CO₂ may be hundreds of times supersaturated relative to the atmosphere). Hence the DIC entering a stream has both an inorganic weathering component, and an organically-produced respiration component. The land also exports organic matter as dissolved organic carbon (DOC) in groundwater. The DOC available for export to rivers represents a balance between production of “fresh” DOC via the solubilization of soil organic matter and the adsorption to mineral particles. Particulate organic carbon (POC) enters rivers from the erosion of soils (typically older materials) and as leaf litter (typically newly-produced). Both DOC and POC may be mineralized within rivers, producing pCO₂. These are all considered as external, or allochthonous, sources. The in situ (autochthonous) production and respiration of organic matter (by plankton and attached aquatic plants) can both consume and produce pCO₂. The relative balance of autochthonous relative allochthonous sources and sinks for pCO₂   indicates what processes are dominant. The only way that pCO₂ can exist at supersaturated conditions is if allochthonous sources dominant, and the waters are net heterotrophic, fueled by carbon from land.

Uncertainties in the Current “Global Fluvial Systems Model”

We will now evaluate the schematic model of Figure 1 in terms of what is known globally.  The key issue here is the quantity of carbon removed from the atmosphere relative to the amount returned to the atmosphere, along the overall fluvial pathways. This sequence of processes can be summarized in the form of a box model (Fig. 2).  Evaluating this model is a challenge. The dynamics are complex, and multiple time constants are involved.  Data are scarce, particularly in many of the most anthropogenically-impacted systems. The distribution of the constituent processes varies dramatically across the face of the globe (with some of the most important regions being the least measured).  

The most common literature estimations of the magnitude of global river carbon fluxes are 0.4 Pg C y^-1 for total organic carbon (evenly divided between particulate and dissolved organic phases), and 0.4 Pg C yr^-1 for dissolved inorganic carbon. While these bulk fluxes are small components of the global C cycle, they are significant compared to the net oceanic uptake of anthropogenic CO₂ (Sarmiento and Sundquist 1992), and to the inter-hemispheric transport of carbon in the oceans (Aumont et al. 2001).

There are inconsistencies in this conventional model. The first is in the estimates of the flux quantities themselves, and in the relative influence of natural and anthropogenic processes in determining these fluxes. While much river research has emphasized concentrations of carbon, sediments, and/or nutrients, with a focus on export to the oceans (Degens et al 1991; Meybeck 1982, 1991; Milliman and Syvitski 1992), considerable uncertainties remain. The second problem is that the role of fluvial systems may not be limited to fluvial exports to the coastal zone.  Continental sedimentation may sequester large amounts of carbon in lower depressions and wetlands (Stallard 1998; Smith et al. 2001). More recent estimates indicate that CO₂ outgassing to the atmosphere from river systems may be an important pathway (Cole and Caraco 2001, Richey et al 2002).

Mobilization from Land to Water and Riparian Zones 

The fluxes from land to rivers are generally inferred directly from the fluxes out of a basin, especially at a global scale.  While there is considerable truth to this for dissolved species (especially conservative ones), it is less true for particulate species, especially with human intervention.

The modern terrestrial sediment cycle is not in equilibrium (Stallard 1998).  Meade et al (1990) estimated that agricultural land use typically accelerates erosion 10- to 100-fold, via both fluvial and Aeolian processes. Multiple other reports in the literature support this conclusion. With the maturation of farmlands worldwide, and with the development of better soil conservation practices, it is probable that the human-induced erosion is less than it was several decades ago. Overall, however, there has been a significant anthropogenic increase in the mobilization of sediments (and associated POC) through fluvial processes. The global estimates of the quantities, however, vary dramatically. Stallard (1998) poses a range of scenarios, from 24 to 64 Pg yr^-1 of bulk sediments (from 0.4 to 1.2 Pg y^-1 of POC). Smith et al. (2001) estimate that as much as 200 Pg y^-1 of sediment is moving, resulting in about 1.4 Pg y^-1 (using a lower %C than Stallard 1998).

Where does this material go? Does it all go downstream via big rivers, ultimately to the ocean, or is it stored inland? Stallard (1998) argues that between 0 and 40 Pg yr^-1 of sediments (0 to 0.8 Pg C y^-1 POC) is stored as colluvium and alluvium, and never makes it downstream. Smith et al. (2001), using a different approach, estimate that about 1 Pg C y^-1 of POC is stored this way. If this movement is merely transferring POC from one reservoir to another, with the same residence times, there is no net change in the C cycle. Then the issue is, to what degree can the (remaining) soils sequester carbon by sorption to the newly exposed mineral soils? Both Stallard (1998) and Smith et al. (2001) argue that carbon is removed from the upper portion of the soil horizon, where turnover times are relatively rapid (decades, or shorter), into either of two classes of environments with longer turnover times; wetlands and smaller, deeper depositional zones, coupled with new carbon accumulation at either erosional or depositional sites. Both assume that oxidation of organic C in transit is minimal, and both use quite conservative values for total suspended sediment (TSS) export. If true, this sequence of processes would result in a significant C sink, on the order of 1 Pg C y^-1. 

Within-River Transport and Reaction Processes (with an Emphasis on pCO₂)

Within river transport processes carry these eroded materials downstream through the river network. Transport is not passive; significant transformations occur along the way. Rivers exchange with their floodplains (depending on how canalized and diked a river is). The movement of POC is, of course, directly linked to the movement of suspended sediments. Sediments are deposited and remobilized multiple times and over long time scales. In the Amazon, for example, Dunne et al. (1998) computed that as much sediment was being recycled within a reach as was leaving it. Presumably, a significant amount of the erosion-excess sediment discussed in the previous section makes it some distance downstream, but is then slowed and retained within the alluvial floodplains.

A significant process within flowing water significantly affects organic matter (OM) – the mineralization to pCO₂. In fact, pCO₂ is present at elevated levels in most rivers of the world, from small streams to large. Kempe (1982) called early attention to the elevated levels of pCO₂ in many rivers, and that this was a sensitive indicator of the sources for river respiration. Jones and Mulholland (1998) analyzed a time series of elevated pCO₂ in a small temperate stream. Cole and Caraco (2001) computed that the average pCO₂ concentration in 47 rivers averaged 3230 µatm, or nearly 10 times saturation. Similar conclusions can be drawn from a wide survey of the literature. There are two important consequences of this. The first is that by far the majority of this CO₂ must be derived from the respiration of organic matter of terrestrial origin (allochthonous production). If the pCO₂ were derived from primary production within the water (autochthonous production), the pCO₂ would be near or below equilibrium (which certainly happens in localized environments).  The second consequence is that according to the rules of gas exchange, this CO₂ is outgassed (evaded) back to the atmosphere (that is, it becomes a source of CO₂ to the atmosphere). In toto, the export of CO₂ and organic matter from land to rivers constitutes a significant sink of terrestrial net ecosystem production.

How large is the return flux (outgassing) of CO₂ to the atmosphere? Telmer and Vuizer (1998) computed that outgassing was about 30% of the DIC export in the Ottawa River. Applying that ratio to the global export of DIC to the ocean, they computed that the flux of CO₂ to the atmosphere from rivers would be .13 Pg yr^-1, or about an order of magnitude higher than early estimates (e.g. Kempe 1982). Cole and Caraco (2001, using a gas exchange coefficient from the Hudson River, 47-river average as representative of flowing waters in general, and assuming that rivers cover ~0.5% of land surface area, computed a global outgassing of  ~0.3 Pg C year^–1.    More recently, Richey et al (2002) computed that outgassing of CO₂ from rivers and wetlands of the central Amazonian basin was about 1.2 Mg C ha^-1 y^-1; an amount comparable to conservative estimates of carbon storage in the Amazon (i.e., an equivalent partitioning of net ecosystem production). Extrapolated across the entire basin, this would produce a flux of about 0.5 Pg yr^-1  from the Amazon alone. This is an order of magnitude greater than the fluvial export of organic carbon and DIC from the Amazon to the ocean. In contrast to other studies, this calculation emphasized the full drainage network, from first-order streams to the river mainstem and floodplains, and was done for a specific region of the humid tropics. Assuming that the humid tropics behaves uniformly, then the total outgassing from the tropics would be about 0.9 Pg yr^-1.  If we then add the estimates of the tropics to the estimates for more temperate systems, total outgassing likely exceeds 1 Pg yr^-1.

The outgassing of CO₂ must be supported by extensive oxidation of organic matter of terrestrial origin within the river systems. This raises a very interesting question. What is the source of the organic matter being respired? Is it labile contemporary organic matter, recently fixed in the water by plankton or nearshore vegetation, or is it some fraction of the allochthonous (terrestrial) matter in transport? The prevailing wisdom is that riverborne organic matter is already very refractory and not subject to oxidation (after centuries on land). The “age” of riverine organic matter yields some important insights.

Measurements of the ¹⁴C ages of organic matter and CO₂ in river water are very few, but the results are intriguing (Hedges et al 1986). Cole and Caraco (2001) found that the POC (and to a lesser degree dissolved organic carbon (DOC)) entering the Hudson is greatly depleted in ¹⁴C, suggesting that the particulate material was originally formed on average ~5000 years ago. Their analysis of δ¹³C suggests that this material is of terrestrial origin, and unlikely to be ancient marine sedimentary rocks. Furthermore, they found that organic matter (OM) pools became selectively enriched in ¹⁴C downstream. Based on an inverse modeling approach, they hypothesized that this enrichment is due to utilization of old organic carbon, with dilution by recent primary production. That is, organic C that had resided in soils for centuries to millennia, without decomposing, is then decomposed in a matter of a few weeks in the riverine environment (how this happens is an intriguing question in its own right). Their results are not unique to the Hudson. Raymond and Bauer (2001) found that four rivers draining into the Atlantic are sources of old (¹⁴C-depleted) and young (¹⁴C-enriched) terrestrial dissolved organic carbon, and of predominantly old terrestrial particulate organic carbon. Much of the younger (relatively speaking) DOC  can be selectively degraded over the residence times of river and coastal waters, leaving an even older and more refractory component for oceanic export. Soemwhat in contrast, Mayorga et al (in review) argue that while much of the organic matter in transport is “old,” the material actually being respired is essentially contemporary.

Thus, pre-aging and degradation may alter significantly the structure, distribution, and quantity of terrestrial organic matter before its delivery to the oceans.  As noted by Ludwig (2001), the OM that runs from rivers into the sea is not necessarily identical to the OM upstream in river catchments. Cole and Caraco (2001) observe that the apparent high rate of decomposition of terrestrial organic matter in rivers may resolve the enigma of why OM that leaves the land doesn’t accumulate in the ocean (sensu Hedges et al. 1997). Overall, this sequence of processes suggests that the OM that is being respired is translocated in space and time from its points of origin, such that, over long times and large spatial scales, the modern aquatic environment may be connected with terrestrial conditions of another time.

Input to Reservoirs 

Just because dissolved and particulate materials enter a river doesn’t mean that they reach the ocean; modern reservoirs have had a tremendous impact on the hydrologic cycle. Starting about 50 years ago, large dams were seen as a solution to water resource issues, including flood control, hydroelectric power generation, and irrigation. Now there are more than 40,000 large dams worldwide (World Commission on Dams, 2000). This has resulted in a substantial distortion of freshwater runoff from the continents, raising the “age” of discharge through channels from a mean between 16-26 days to nearly 60 days (Vörösmarty et al. 1997). While erosion has clearly increased the mobilization of sediment off the land, the proliferation of dams has acted to retain those sediments. Vorosmarty et al (in press) estimates that the aggregate impact of all registered impoundments is on the order of 4 to 5 Pg y^-1 of suspended sediments (of the 15-20 Pg y^-1 total that he references). Stallard (1998) extrapolates from a more detailed analysis of the coterminous US to an estimate of about 10 Pg y^-1 worldwide (versus 13 Pg y^-1 efflux to the oceans), for a storage of about 0.2 Pg C y^-1 (which he includes as part of his overall calculation of continental sedimentation.

Export to the Coastal Zone 

The conventional wisdom is that the flux of POC and DOC are each about 0.2 Pg C y^-1, and DIC is 0.4 Pg C y^-1 (e.g., Schlesinger and Melack 1981; Degens, 1982, Meybeck 1982, 1994; Ittekot 1988; Ittekkot and Laane, 1991; Ludwig et al. 1996; Ver at al. 1999). That these analyses converge is not terribly surprising. They are all based on much of the same (very sparse) field data, and use variations of the same statistically-based interpolation schemes. Let us evaluate these numbers.

 Because direct measurements are few, POC flux estimations are typically a product of the flux of total suspended sediments (TSS) and the (estimated) weight-percent organic C (w%C) associated with the sediment (because the bulk of POC is organic C sorbed to mineral grains). The first problem is an adequate resolution of the TSS flux. Data on TSS are frequently poor and unknown quality. Many reported data are surface samples, and the depth integrations necessary to accurately characterize sediment flux are on the order of 2-3x higher. Additionally, much sediment moves during episodic storm events, when measurements are almost never made.

As summarized by Vorosmarty et al. (in press), estimates of TSS transport to the oceans have ranged from 9 Pg y^-1 to more than 58 Pg y^-1, with more recent studies converging around 15 to 20 Pg y^-1. These estimates are generally based on extrapolations of existing data, which are weighted to the large rivers of passive margins and temperate regions. Milliman and Syvitski (1992) called attention to the much higher yield rates from steep mountainous environments (without directly computing a global total). More recently, Milliman et al (1999) estimated that the total sediment flux from the East Indies alone (the islands of Sumatra, Borneo, New Guinea, Java, Sulawesi, and Timor), representing about 2% of the global land mass, is about 4 Pg y^-1, or 20-25% of the current global values. This type of environment (steep relief, draining directly to the oceans) is found elsewhere in the world, so the results are not likely to be unique. New data from Taiwan support these high levels, with isotopic analyses of the C showing that a significant part of the flux is human-driven (Kao and Liu 2000). 

Then to obtain POC flux estimates, these values (and their uncertainties) must be multiplied by w%C. Meybeck (1991) divided particulate carbon into inorganic (PIC) and organic (POC) phases, and assumed that high sediment rivers have very low carbon fractions (0.5 w%C), representative of shale; he essentially does not consider the latter to be "atmospherically derived" and hence discounts it from estimates of fluxes to the ocean. More recent values for w%C tend to be in the 1-2% range, and higher for organic-rich systems (Richey et al. 1990, Stallard et al 1998; Gao et al. 2002). 

To account for this range, POC flux  can be computed as an ensemble based on different combinations of w%C and TSS fluxes, resulting in a range of 0.3 Pg C y^-1 to 0.8 Pg C y^-1 , with a “more likely” of about 0.5 Pg C y^-1 (depending on assumptions used). Therefore, it is possible that the common estimate of 0.2 Pg C y^-1 is low, and that the overall value lies in the range of 0.2-0.5 Pg C y^-1.  The common estimate for PIC of 0.2 Pg C y^-1 (Meybeck 1991, Ver et al. 1999) may also be underestimated if sediment fluxes are higher.

The value of 0.2 Pg C y^-1 for DOC export may also be low. DOC is also subject to sparse (and questionable) measurements, without the availability of a proxy like TSS for POC. Aitkenhead and McDowell (2000) developed a model of riverine DOC flux as a function of soil C:N. Using this model, they computed a global flux of 0.4 Pg C y^-1, or twice the common estimate. That is, the total organic C output from fluvial systems may well be approximately double the original estimates, in the ~0.8 Pg C y^-1 range.

Marine Fate

Long-term preservation of terrestrially-derived organic matter in the oceans occurs largely within sediments that accumulate along continental margins. Organic carbon within these sediments is thought to be preserved largely because it is adsorbed to mineral grains (Keil et al., 1994, Mayer 1994; Bishop et al., 1992). Hedges and Keil (1995) estimated that carbon preservation along continental margins over the Holocene was split roughly evenly between  sediments accumulating within the delta or sedimentary plume of rivers and non-deltaic sediments accumulating outside the direct influence of major rivers (but within range of multiple smaller systems).

The efficiency of storage between deltaic and non-deltaic systems is different.  The amount of organic carbon in non‑deltaic continental shelf sediments falls in a narrow range (0.5‑1.1 mg C m^‑2 of mineral surface), and typically >90% of the preserved organic matter is adsorbed to mineral surfaces.  Deltaic sediments are distinctly different, containing only a fraction of the organic carbon (by weight) found in other margin sediments. Suspended sediments from the Amazon River, for example, have loadings (~0.67 mg C m^‑2) that are 3 times higher than the corresponding deltaic sediments (Keil et al. 1997), so over 2/3 of the terrestrial particulate organic load delivered to the Amazon delta is lost from the mineral matrix and is not preserved.  The Mississippi, Yellow and other river/delta systems also show extensive loss of terrestrial organic matter. Thus many deltaic systems bury only a small fraction of the potential organic load normally sorbed to mineral particles, with the balance presumably desorbed or mineralized (and entering the DIC pool).

The organic matter lost by mineralization and not buried is one of the factors in maintaining the historical perspective that marginal seas are net heterotrophic (Chen 2003). But Chen (2003) reviews more recent evidence, based on direct measurements of pCO₂ (again, showing the critical importance of actual field measurements of key parameters!) and comes to the conclusion that these seas are net autotrophic, driven primarily by nutrients delivered via upwelling (with enhanced nutrients delivered by rivers leading to eutrophication constituting only a minor source), and net consumers of atmospheric CO₂. The overall implication of this sequence of processes is that much of the anthropogenically-mobilized riverborne OM (and perhaps the naturally mobilized OM) is liable to remain in the marine environment over time scales longer than the current increase of atmospheric CO₂.

Anthropogenic Transient

To what degree have these fluxes been influenced or impacted by human activities; i.e., how much of this carbon is an anthropogenic transient? The consensus is that human-induced erosion has dramatically accelerated the movement of sediments (and POC). While some of this material “hangs up” on land (sedimentation, reservoirs), some of it likely escapes to the sea (some of the regions with highest sediment yields have very few dams). There is evidence that anthropogenic processes have an affect on DIC. Raymond and Cole (2003) report an increase in the alkalinity of the Mississippi, which implies an increase in the consumption of atmospheric CO₂ through weathering. However, Jones et al (in press) report a systematic decrease in pCO₂ in rivers across the U.S., which they attribute to large-scale declines in terrestrial CO2 production and import into aquatic ecosystems, and not to terrestrial weathering or in-stream processes.

 In the case of DOC, there is simply not enough information available to make conclusions. Clair et al. (1999) suggested that DOC export from basins in Canada might increase by 14% with a doubling in atmospheric CO₂. An additional factor rarely addressed in rivers is direct loading from urban and industrial sources (Ver et al 1999; Abril et al. 2002). In evaluating the consequences of continental sedimentation and the potentially higher fluxes of POC, a net transient exported from land of roughly 1 Pg C y^-1 is possible, with perhaps half of that going to the sea, and the other half divided equally between outgassing and sedimentation.  

Overall Fluvial System – Atmosphere Exchange: Alternative Scenarios

Considerable uncertainty remains in the assessment of the carbon cycle of fluvial systems, including the actual magnitude of fluxes, how to include processes not previously considered, and delineation of anthropogenic and natural processes; all with an explicit recognition of geography. To assess the implications of each, the above discussions are summarized here via five “scenarios” (Fig. 3). 

The bulk transfer of atmospheric C through the land to fluvial systems (assuming a steady-state summation of downstream processes, in a non-steady state environment) ranges from about 0.6 Pg C y^-1 (CW) to 2.6 Pg C y^-1 (+Outgassing). Continental sedimentation results in a significant sink, but that sink is reduced with CO₂ outgassing (because of the way the sedimentation was computed).  The inclusion of continental sedimentation, and then the larger export of OM to the sea (about twice conventional assumptions, under +POC, DOC), yields net sinks of atmospheric CO₂ of up to 1.6 Pg C y^-1. However, if outgassing is included, then the fluvial net sink is reduced to 0.2 Pg C y^-1. While partitioning the total fluvial fluxes into natural conditions and anthropogenic transients is problematic at best, there is substantial evidence that there has been a dramatic increase in the mobilization of sediments. While much of this material is captured in reservoirs, it is reasonable to expect that a considerable amount escapes to the sea (especially in non-deltaic regions with steep slopes and few dams.   

Summarizing all the components of the riverine carbon cycle, several images emerge. As a global steady-state aggregate, there appears to be a sink (between continental sedimentation and marine sedimentation and dissolution) on the order of 1 to 1.5 Pg C yr^-1, with a significant anthropogenically-enhanced component. A return flux to the atmosphere, on the order of 1 Pg C y^-1, reduces the net sink to about .2 or .3 Pg C yr^-1. However, there are certainly disjunctures in space and time in this view. Because the organic matter in transport appears to be “old,” the subsequent mobilization and oxidation/outgassing would essentially be mining old C. There are significant regional implications in this analysis. With its preponderance of land mass, extensive reservoirs, and agriculture, the bulk of continental sedimentation (and its implications for C sink), is focused in the Northern Hemisphere, between 30^o-50^o N. C sequestration in paddy lands would be closer to the equator. The greatest amount of sediment flux to the ocean (and the greatest uncertainty) is in South and Southeast Asia and Oceania.  Outgassing is function of both pCO₂ concentrations (driven by in situ oxidation) and surface area of water. It is likely most significant in the humid tropics, particularly during the peak of the wet seasons. The highly canalized temperate areas have less area available. The northern latitudes, particularly with warming, are liable to have significant fluxes.

Overall, more carbon is moving through the river system than previously assumed; however, the exact magnitudes remain uncertain. The degree to which fluvial systems constitute a net source/sink relative to the atmosphere is essentially governed by the interplay between mobilization of materials off the landscape and oxidation of those materials back to the atmosphere.  Additional data will be necessary to constrain these magnitudes, and it is necessary to partition the model by geographic zone.

Click here for References