SCOPE 42 - Biogeochemistry of Major World Rivers

 3.1 INTRODUCTION

The great tropical floodplain rivers, such as the Amazon, Zaïre and Orinoco represent the downstream end of the continuum of rivers; they drain vast and important rain forest and savannah environments, and account for over 30% of the fresh water discharge to the oceans. Although their floodplains are extremely important for fisheries production (Welcomme 1985) and have been postulated to be sources of organic matter and nutrients to the main stem (Sioli 1975; Forsberg et al. 1988), there is little evidence that details linkages between main stem, floodplain and off-channel processes. Further , land use practices in these basins are changing rapidly. Regional impacts include potential changes in hydrology and sediment and nutrient transport. Because of the possible release of carbon dioxide and because of the role of evapotranspiration in the precipitation regime, long-range climate impact is possible.

From both the perspective of basic ecological theory and of understanding potential change, the biogeochemistry of these river systems must be understood. In this chapter, the spatial and temporal distributions and compositions of particulate and dissolved organic and inorganic carbon and nutrients in the Amazon main stem and tributaries are described. The actual distributions of materials may be unique to the Amazon, but the understanding derived should serve as a template to evaluate the other river systems described in this volume.

The presentation here represents a synthesis from CAMREX (Carbon in the Amazon River Experiment), a cooperative research project of the University of Washington (Seattle, USA), the Instituto Nacional de Pesquisas da Amazônia (Manaus, Brazil), and the Centro de Energia Nuclear na Agricultura (Piracicaba, Brazil).

3.2 THE BIOGEOCHEMISTRY OF LARGE RIVER SYSTEMS

Carbon and nutrient fluxes in large floodplain rivers are presumably reflections of their watersheds and floodplains, and some extension of the properties of the smaller rivers that form them (Welcomme 1979; Vannote et al. 1980; Richey et al. 1980). The main channel of large rivers acts as an 'integrator' of basin-wide properties. For example, the carbon measured in the main channel is a mixture of carbon originating from sources thousands of kilometers away in upland regions (terra firme), as well as carbon introduced continuously (spiralled) from the adjacent floodplain. Organic matter of both sources has been subjected to within-channel transport and reactive processes.

The sheer physical size and logistics of problems posed by very large rivers presents challenges in the effort to determine overall dynamics. If the distributions of the riverborne materials could be systematically measured, and both the origins and the processes producing the observed distributions and storage patterns could be determined, the dynamics of large river systems would be much better understood. A fundamental issue in defining carbon cycling in a river system is establishing the hierarchy of time and space scales at which the controlling processes operate. There are differences in spatial and temporal scales from the site a process originates to where its signal may be detected. Rapid cycling of materials may occur locally, but the integrated effect of this cycling may or may not influence cycling at greater temporal or spatial scales.

The problem, then, is how to determine material routing properties of a river system when the component parameters are spatially and temporally variable and not well characterized at a smaller scale. It is important to identify those mass routing properties that functionally integrate over time and space and that can be measured with the required precision to constrain processes. Richey (1983) outlined an analytical mass balance approach to studying the biogeochemistry of large rivers, and Richey (1982) discussed its application to the Amazon. The outcome of the application of this model approach is reviewed below.

3.2.1 THE AMAZON RIVER SYSTEM

The Amazon is a classic river basin, with a vast central plain bordered by highlands and a terrestrial drainage network within which the main channel and its extensive floodplain (várzea) receives inputs from a series of different-sized tributaries (Figure 3.1). A mean precipitation of 2500 mm/year over the 6 x 10⁶ km² basin results in nearly 20% of the world's runoff to the oceans.

The main channel of the Amazon (technically the Rio Solimões above the confluence with the Rio Negro) at São Paulo de Olivença is composed primarily of Andean water. Of the north-draining tributaries, the Rios Içá and Japurá have Andean origins, but are mostly lowland drainages. The Rio Negro drains primarily the caatinga forest on the Guyana Shield, though its major tributary, the Rio Branco, drains a drier savannah region. Of the south-draining tributaries, the Rios Jutaí, Juruá and Purús drain the sediments of the subandean trough and of the central plain, while the Rio Madeira begins in the Bolivian Andes, and passes across the Brazilian Shield and the central Amazon plain. The tributaries of the lower course of the Amazon, the Rios Trombetas and Uatumä, are shield-draining rivers and are characterized by having large 'mouthbays', where sediments are deposited. The main channel also receives input from smaller, ungauged tributaries and non-channelized floodplain (várzea) areas.

Figure 3.1 The Amazon drainage basin (adapted from Richey et al. 1989). CAMREX main stem sampling stations include Vargem Grande (V), Santo Antônio do Içá (S), Xibeco (X), Tupé (T), Jutica (J), Itapeua (I), Anori (A), Manacapurú (M), São José do Amatari (D), Paura (P) and Óbidos (O); stations on tributaries are indicated by dot. DNAEE gauging stations include São Paulo de Olivença (C), Santo Antônio do Içá, Itapeua, Manacapurú and Óbidos

The river system can be divided into several geomorphically distinct reaches (Mertes et al. 1989). Upstream reaches (Vargem Grande to Itapeua) are characterized by deposition of sand in the main channel and floodplain channels, with subsequent rapid migration of these channels to produce an intricate scroll-bar topography with hundreds of long, narrow lakes. The middle reaches of the river (Itapeua to São José do Amatari) are controlled by structural features that tend to constrain the river and allow almost no morphological change. The floodplain is narrow and lakes are less numerous. Finally, in the downstream reaches (São José do Amatari to Óbidos), an incomplete levee system provides free access for overbank flows to a wide floodplain of relatively flat-lying topography and a patchwork of wide, shallow lakes. Differences in the degree of annual sediment deposition in the downstream reaches seem to control the location of the large lakes.

3.2.2 EXPERIMENTAL DESIGN

In accordance with our conceptual model, we developed a strategy for measuring the concurrent distributions and fluxes of water and the particulate and dissolved material fractions, at the space and time scales we anticipated to be dominant and in a logistically feasible manner (Richey et al. 1986; Hedges et al. 1986a,b; Richey et al. 1989, 1990). To highlight some of the problems in the quantitative measurement of very large rivers, we briefly review the sampling strategy.

The main stem was divided into a series of nine reaches, defined by 18 sampling stations: 11 stations located in the main stem just above primary tributary confluences and 7 tributary stations (Figure 3.1) .The data collection was done primarily on a series of nine cruises, each at a different stage of the hydrograph, on the research vessel LM Amanai, of the Instituto Nacional de Pesquisas da Amazônia (INPA). On each cruise we sampled from the uppermost station, Vargem Grande, down to the last station, Óbidos, in a quasi-Lagrangian mode (current velocity and effective boat speed are roughly comparable, 1-2 m/s).

The fundamental problem in sampling a large river is overcoming the non-uniform distributions of water velocity and constituent concentration, especially in the particulate phase. We developed a sampling methodology to provide horizontally and vertically integrated measurements of overall material transport (Richey et al. 1986). Briefly, sampling stations were selected from radar imagery (Radambrasil 1972) to have shores parallel to the flow and not be immediately downriver of major bends or islands, to minimize local irregularities in the channel and flow. On site, a modified equal-width- increment procedure (Nordin et al. 1983) was used to sample simultaneously current velocity and particulate and dissolved chemistry. A variable-speed hydraulic winch deployed an instrument array of a 130-kg sounding weight, Price AA current meter, and collapsible-bag sampler by lowering it from the surface to the bottom and back at a constant velocity (<20% mean current velocity). The bag sampler, which obtains a sediment or chemistry sample, fills at a rate proportional to the current velocity and the area of the nozzle. Eighteen verticals were taken at main stem stations (fewer on tributaries) an equidistance apart, with positioning determined by shipboard observation with a sextant monitoring angles from a three-marker base line on shore.

Samples for total suspended sediments (TSS) and chemistry were obtained on alternating verticals. For sediment, the contents of the bag sampler from each vertical were passed through a 0.063-mm sieve into a splitter. After the entire section was sampled, triplicate aliquots were withdrawn from the splitter and filtered through double, pre-weighed Millipore filters, and dried in a shipboard oven, while the contents of the sieve (the coarse fraction) were bottled. Samples for chemistry were processed slightly differently. The contents of the bag sampler were put directly into the splitter without sieving. After the entire section, triplicate 60 ml aliquots for dissolved organic carbon (DOC) analyses were withdrawn from the splitter, filtered through pre- combusted GFA filters, and preserved with 0.1 ml of 10% HgCl₂. After these aliquots were removed, the entire contents of the splitter (generally 20-30 liters) were passed through a 0.063 mm sieve into a reservoir and then on through a continuous flow centrifuge. The coarse (from the sieve) and fine (from the centrifuge) fractions were dried at 50 °C and returned to the laboratory for analysis.

Fine suspended sediments (FSS) were weighed and determined by difference, while the coarse fraction (CSS) was rewashed, dried and weighed, DOC samples were analysed in triplicate by ultraviolet and persulfate oxidation. Weight percentages of fine and coarse particulate organic carbon (FPOC, CPOC) were determined in duplicate on a CHN elemental analyser, after spot checks indicated no measurable inorganic carbon. Concentrations of POC were then calculated as the product of the respective weight-percent and sediment concentration. Nutrients were determined by colorimetric techniques.

Because of the slight turbulence generated by the filling of the collapsible- bag sampler, samples for dissolved gases were taken at five profiles at a 3 m depth with a Niskin bottle, after tests showed that the water column was well mixed with regard to gases. Samples for pH determination were analysed immediately after collection in a closed container. Alkalinity was determined after sampling by micro-Gran titration. Total dissolved inorganic carbon (DIC) and free dissolved carbon dioxide (pCO₂)were calculated from pH and alkalinity. Oxygen concentrations were determined using an oxygen meter . Due to the non-uniform vertical distribution of particulate matter, respiration rates were determined on subsamples of the composite by changes in oxygen concentration in samples incubated in the dark (ambient conditions) at in situ temperatures and were calibrated against calculated oxygen invasion rates (Devol et al. 1987; Richey et al. 1990).

3.3 DISTRIBUTIONS OVER TIME AND SPACE

3.3.1 DISCHARGE REGIME

The hydrology of the river over this time period has been described in detail by Richey et al. (1990). The most striking features of the hydrograph of the main stem of the Amazon River, in comparison to other rivers, are its relative uniformity and small difference between minimum and maximum discharges (Figure 3.2). Over a 15-year period, the average minimum and maximum discharges were about 20 000 m³/s and 60 000 m³/s upstream at Vargem Grande and 100 000 m³/s and 220 000 m³/s downriver at Óbidos.

The damped hydrograph of the main stream reflects in part the offset input from tributaries. The peak flows from the north- and south-draining tributaries are typically three months out of phase as a result of seasonal differences in precipitation. The hydrographs of the northern tributaries are relatively synchronous, with low water in November and high water in May or June.

The damped hydrograph of the main stem also reflects the storage of water on the floodplain and inputs from local channels (paranas) and small, ungaged tributaries. The discharge of the Parana Copeá, for example, ranged from about 3000 m³/s to about 6000 m³/s, which is comparable to the Rios Jutaí and Juruá, and is 5-10% of the discharge of the main stem at Itapeua. Further down the river, the discharges of the Paranas Autaz, Urucara, Uraria and Nhamundá and Rio Uatumá are comparable to the Parana Copeá, and cumulatively represent about 10% of the discharge at Óbidos.

Net lateral exchange of water between the floodplain and main stem in the upriver sections was greatest in the area of the Rio Japurá, with net flow from the floodplain to the main stem ranging up to 20% of the total flow. Net exchanges were generally lower in the reach between Anori and São José do Amatari where the area of the floodplain is relatively small. Downstream of São José do Amatari net exchanges increase again to a maximum of 5% of total flow.

Figure 3.2 Discharge (10³ m³/s), along ordinate, calculated from DNAEE stage records for main stem (Santo Antônio do Içá and Itapeua are not shown) and tributary gauging stations ( adapted from Richey et al. 1989) .Years indicated by abscissa scale are from 1 September through 31 August (the water year for São Paulo de Olivença)

3.3.2 DISTRIBUTIONS OF PARTICULATE AND DISSOLVED MATERIALS

The distributions measured on the CAMREX cruises represent nine synoptic 'snapshots' of the spatial and temporal variability of the respective constituents. Discharge and concentrations of DIC, DOC, TSS, FSS, CSS, FPOC, CPOC, NO₃ and PO₄ are shown in Figure 3.3 (concentrations of NH₄, dissolved and particulate organic N and P are referred to in the text, but are not shown).

3.3.2.1 Sediments, POC and DOC

Over 95% of the particulate material transported by the Amazon River is carried in suspension, with computed bedload transport rates only 1-2% of the total sediment load (Richey et al. 1986; Dunne et al. 1989). Most of the suspended fraction is carried as fine sediment (FSS). Concentrations of TSS generally decrease downstream due to dilution of Andean-derived materials with relatively sediment-poor tributary waters of the lower basin. The highest concentration of TSS (600 mg/l) was measured at Vargem Grande on mid- rising water, with the lowest of 93 mg/l at Óbidos at low water.

Particulate organic carbon concentrations closely track the respective sediment distributions, with the highest concentrations during early rising water and the lowest during falling water. Thus, FPOC was greatest upriver at Vargem Grande, with concentrations ranging from 2.5 mg/l to 5.0 mg/l, and decreased down river to 0.8 mg/l to 2.5 mg/l at São José do Amatari, then increased slightly to 1.2 mg/l to 3.3 mg/l at Óbidos. Coarse POC exhibited similar distribution patterns, however concentrations were considerably lower, averaging less than 20-30% of the fines. Concentrations were greatest upstream (between 0.6 mg/l and 1.1 mg/l), and were lowest downriver at Óbidos (0.05 mg/l to 0.4 mg/l). Except for the Rio Madeira, tributary concentrations were lower and more variable than in the main stem.

Dissolved organic carbon accounts for approximately 50% of the total organic matter transported by the Amazon, and is by far the predominant form of organic matter in most tributaries. Spatial and temporal variations were less than for POC. Main stem DOC ranged from 2.7 to 4.7 mg/l, and was highest during rising water. Some increases occur downstream, due to high DOC in blackwater tributaries, in particular the Rio Negro (6.5 to 10.6 mg/l).

3.3.2.2 Dissolved inorganic carbon

Approximately 80-90% of the DIC is bicarbonate, with the balance being pCO₂ (carbonate alkalinity is virtually zero). On all cruises DIC decreased downstream; low values were generally observed at low discharges and high values at high discharge. DIC ranged from 687 µmol/l to 1228 µmol/l at Vargem Grande to 485 µmol/l and 667 µmol/l at Óbidos. With the exception of the Rio Japurá, tributary concentrations of DIC were lower than those in the main stem.

Figure 3.3 Synoptic 'snapshots' of the distributions of the respective variables from Vargem Grande (upstream, U) to Óbidos (downstream, D) as a function of the hydrograph Q, on CAMREX cruises 1 (Cr 1) to 8 (Cr 8), 1982-84. Data are from Richey et al. (1990) and Devol et al. (1989) 

3.3.2.3 Nitrogen and phosphorus

Nutrient distributions varied systematically over time and space. Nitrate was the dominant form of combined N in the rivers, varying from 5 to 25 µmol/l, with NH₄less than 1-2 µmol/l. In general, NO₃ concentrations were highest upriver and decreased downriver. A distinct inverse relationship with discharge was seen, with the highest NO₃ concentrations observed at the lowest discharges. Both the downstream concentration decrease and dischargecentrations at lower discharges. Tributary DOP was comparable to or slightly stem waters with tributary waters of generally lower nitrate concentration. Phosphate distribution patterns were similar to those of NO₃, but not as well defined. Variations in the tributaries were less extreme than with NO₃, though the southern tributaries were more concentrated and variable than the northern ones.

Dissolved organic N (DON) concentrations were relatively constant along the main channel and about equal to NO₃. As with DOC, concentrations of DON were higher in the tributaries. Main stem dissolved organic P (DOP) values were somewhat lower than those of PO₄ (0.12 to 1.23 µmol/l) with most values clustered around 0.50 µmol/l. Variability in the data obscured any downstream trends. However, there was a tendency toward lower concentrations at lower discharges. Tributary DOP was comparable to or slightly higher than main stem DOP.

Particulate organic nitrogen (PON) and particulate phosphorus (PP) distributions followed the POC patterns. Nitrogen was rather evenly distributed between NO₃, DON and PON (overall means were 12.1, 14 and 23 µmol/l, for NO₃, DON and PON, respectively). In contrast, PP was by far the largest pool of phosphorus in transit in the river with values ranging from about 4 to 15 µmol/l (overall means of 0.79, 0.50 and 7.85 µmol/l for P04, DOP and PP, respectively). Virtually all of the particulate nitrogen is organic, while the bulk of the phosphorus is not organically bound; rather it is probably sorbed to suspended inorganic particles (Devol et al. 1990). The overall TN : TP ratio of the main stem was about 5, indicating that, if the main stem waters are the basis of várzea fertility, then the várzea should be severely N-limited.

3.3.2.4 Gas exchange and respiration

That pCO₂ is maintained considerably in excess of atmospheric equilibrium and O₂ is maintained at less than equilibrium implies that significant gas exchange processes are involved in producing observed distributions (Richey et al. 1980; Devol et al. 1987). To evaluate the distributions of pCO₂ and O₂ in the Amazon main stem and floodplain, Richey et al. (1988) posed the model:

where C is the average gas concentration in a water parcel, F_ex is exchange with the atmosphere, F_r is biological processes (production and respiration), F_t is net physical transport (advection and diffusion) from adjacent parcels or sediments, F_i represents ionic equilibrium reactions, and t is time. Richey et al. (1990) evaluated Equation (3.1). Reflecting the riverine distributions of pCO₂ and O₂, the downstream gradients in the evasion of CO₂ and the invasion of O₂ at high water were small, and their relative magnitudes were comparable, ranging from 2 µmol/m²/s to 6 µmol/m²/s. Richey et al. argued that the invasion rate of O₂ was essentially driven by in situ respiration, and, for methodological reasons, was a better estimator of respiration rates than direct incubation measurements.

3.3.3 MASS BALANCES

To determine the relative magnitude of processes operating externally to and within a channel section, a mass balance model can be used. The model considers the river as a series of linked reaches, with each homogeneous reach receiving inputs from upstream (F_in), its catchment via major tributary input (F_tr), and local channel and floodplain input (F_ot), and exporting to downriver (F_ot), and including gas exchange for the dissolved gaseous species:

The fluxes F_in, F_at and F_tr for each reach were calculated as the product of the respective concentrations for each species at each reach boundary and water discharge for that site (Richey et al. 1986; Richey et al. 1990). The floodplain input F_fp was calculated as the product of the mean concentration for that species on the floodplain in that reach (from the discrete floodplain sampling) and the floodplain discharge (Richey et al. 1990). Differences between the fluxes observed in a reach and the fluxes expected in a reach can provide insight into processes not accounted for in the initial mass balance.

Richey et al. (1990) report on detailed mass balances for each carbon species by reach; their approach is illustrated here for total organic carbon (Figure 3.4) and for nutrients. To consider the pool of the total organic carbon (TOC) in the river, the DOC, FPOC and CPOC fractions were aggregated. The total annual export of TOC from Óbidos was 30.9 Tg/year, of which 62% was DOC, 37% was FPOC and 5% was CPOC. Per unit of drainage basin area, TOC export ranged from 3.6 g/m²/year in the Rio Madeira to 12.6 g/m²/year in the Rio Negro; the total export from Óbidos was 8.5 g/m²/year. Including only advective fluxes, the TOC pool balances within less than 20% .However, if we assume that respiration is derived from the TOC, then the calculation of the expected TOC must include a loss term for respiration. In this case, there is a shortfall of 40% to over 200% , depending on river stage. That is, there is not nearly sufficient organic carbon observed to support in situ oxidation.

Figure 3.4 Total organic carbon mass balance, including inputs (In-sum of up-stream, tributary, and floodplain), output at Óbidos (at), respiration (Rs), and mass balance (D = Ot + Rs -In)

Mass balances for nitrogen and phosphorus can also be constructed (Devol et al. 1990). The variation in concentration of PO₄ and especially NO₃ is considerable downstream and with time. For both species, the flux in Vargem Grande plus the tributary inputs are always significantly (95% level) less than the outputs at Óbidos by a factor of 1.5 to 2. Anomalies occur primarily upstream between Xibeco and Itapeua and below Paura.

The same pattern of downstream gain is also seen for dissolved organic nitrogen (TDN) and total dissolved phosphorus (TDP), except that these anomalies are not statistically significant. Due to the close relationship between the particulate forms of the nutrients and suspended sediments, losses and gains of PP and PO_N mimic losses and gains in suspended sediments. Analysis of the budgets of TN and especially TP are complicated by the relative magnitude of particulate terms and their relationship to sediment erosion and deposition. Devol et al. (1990) showed the overwhelming dominance of PP in the TP budget, but that TN is not dominated by PON, indicating that factors other than sediment transport playa role in the overall nitrogen budget.

The overall carbon and nutrient budgets now indicate an important problem. There is a shortage of organic carbon in transit, downstream to support the observed levels of in situ oxidation by an amount roughly equal to the oxidation rate. Thus, we require a large additional-and as yet unidentified- source of labile organic substrates that can be rapidly oxidized. Similarly, additional NO₃ is required.

3.3.4 ORGANIC COMPOSITION AND SCALES OF PROCESS CONTROL

3.3.4.1 Basin level control of organic compositions

We need to know both the sources and condition of the organic matter in transport. The degree to which the fine and coarse particulate organic carbon (FPOC and CPOC) in the Amazon River system has been diagenetically altered is an important consideration both in understanding carbon cycling processes and identifying biological and geographic sources.

An outstanding characteristic of both the CPOC and FPOC in the Amazon main stem is their nearly constant composition and relative concentrations (Hedges et al. 1986b), Based on analyses of ¹³C and of lignin oxidation products characteristic of particular tissue types, the CPOC fraction in transport is predominantly a mix of leaf (about 80%) and wood (about 20% ) remains. Várzea grass remains (d¹³C -12 per thousand) are at most a minor <10%) component (Hedges et al. 1986b). Because leaves and grasses are produced and degraded rapidly in the Amazon region (Jordan 1982), such tissues should have a content closely matching that of the recent atmosphere; in fact, the CPOC fraction had essentially current atmospheric D¹⁴C levels ( +250 per thousand) (Hedges et al. 1986a).

The coexisting FPOC is compositionally distinct from the CPOC and is characterized by a higher N content and low concentrations of lignin that have been extensively degraded by microorganisms. These compositional features and the near-conservative behavior of the FPOC suggest that the bulk of this organic matter is soil derived and refractory. This inference is supported by the radiocarbon content of the FPOC fraction, which contains a large component of prebomb carbon and thus is retained within the drainage basin a longer period of time than the CPOC or DOC before export by the river.

Total dissolved humic substances average 60% of the total riverine DOC, with slightly higher values for blackwater tributaries (Ertel et al, 1986). Fulvic to humic acid carbon ratios (FA: HA) were about 5 from Vargem Grande to Manacapurú, and 3 downriver, due to the large input of humic-acid-rich water from the Rio Negro (FA: HA = 1.6). The lignin composition of the dissolved humic substances was relatively uniform in the main stem, and had undergone a much greater degree of oxidative degradation than the particulate fractions. These data indicate that the diagenetic transformation of the dissolved humic substances also has occurred predominantly in the terrestrial environment prior to river introduction. Mass balances indicated that most of the dissolved humic and fulvic acids mix conservatively in the upper river . However, there is a substantial loss (40% ) of dissolved humic acids down- stream of the confluence of the Rio Negro, which results from selective adsorption of blackwater material on to fine particles carried by the main stem.

Dissolved fulvic acids from the Amazon River at Óbidos and the Rio Negro have D¹⁴C values (+ 300 to + 336 per thousand) which significantly exceed present-day atmospheric D¹⁴C levels in the tropical atmosphere. This indicates that at least 38% of the component carbon was photosynthetically fixed over the time period 1963-83 (maximum average residence time 100 years). In contrast, the coexisting humic acids have significantly lower D¹⁴C values (+ 144 to + 185 per thousand) than either the CPOC or fulvic acids, which corresponds to a maximum average residence time of less than 150 years. The residence time differences between the FA and HA indicate that Amazon humic and fulvic acids are not only chemically distinct, but also have different dynamics within the drainage system. These distinct residence times may be related to the relative affinity of the various organic matter fractions for soil minerals.

These results suggest that the composition of the bulk organic matter in transport is controlled on a 'slow-dynamic' basis by basin-wide processes. The observed compositions and dynamics of dissolved humic substances and soil organic matter in the Amazon drainage basin are all consistent with a system controlled by selective partitioning of solubilized vascular plant degradation products between water and soil minerals in upland forests. One implication of this model is that most of the FPOC within the river may be nitrogen-rich humic substances strongly adsorbed to the coexisting clay minerals and/or their metal oxide coatings. Thus adsorption may be a key physical process controlling the forms and transport of organic substances within the entire drainage basin.

3.3.4.2 Main stem-floodplain exchange

A central problem of river carbon flow models is estimating the degree to which organic matter 'spirals' downriver. Although basin-derived 'slow-dynamic' processes control the bulk chemical composition of the main stem, linkages between the main stem and floodplain appear to influence the dynamics of organic matter and nutrients on 'intermediate-dynamic' weekly to seasonal scales over tens to hundreds of kilometers.

The compositional indices show that the bulk of the particulate organic matter in transport is long-lived, recalcitrant material of terrestrial origin. Does that mean it is derived from far up in the drainage basin, or can organic matter be exchanged laterally?

Quay et al. (1989) showed that the O¹³ of the main stem FPOC averages -26 per thousand at Vargem Grande, then decreases downstream to -27 per thousand during rising water cruises and -28 per thousand during falling water cruises. The downstream depletion occurs mainly in the geomorphologically active Rio Japurá area between Vargem Grande and Itapeua. The d¹³C of FPOC at Vargem Grande and in the Rio Madeira is similar to 'global biospheric material' at -26 per thousand. These data suggest that there is 'spiralling' between the material in transport and the floodplain. That is, there may be a replacement of the riverborne POC derived from the Andean region with compositionally similar, but isotopically distinct material of the Amazon lowlands. This distinct material could have originated (i.e. was photosynthetically fixed) in the lowlands, or be upland material that has been diagenetically altered during residence in the floodplain. The greatest exchange appears to be in areas where the geomorphological characteristics imply the greatest river-floodplain interaction, especially in the reach between Vargem Grande and Itapeua. This possible input of lowland-basin-derived material is maximized during falling water when localized input from the floodplain is greatest.

For there to be significant exchange of particulate organic matter between the main channel and floodplain, reasonable physical mechanisms must exist. There is a high correlation between the transport of the coarse and fine POC fractions and the transport of sediment in the Amazon River (Hedges et al. 1986b) .Therefore, it can be assumed that the distribution of the POC is tied to the processes that control sediment transport.

The dominant mechanisms of erosion include resuspension of sediments stored temporarily in the main channel and on the floodplain and bank erosion through lateral channel migration and island migration. The dominant mechanisms of deposition include seasonal storage in the main channel, long-term storage (> 100 years) on the bed and bars of the main channel (as sand bars accrete laterally and the channel migrates), and long-term storage in floodplain channels and on the floodplain surface (Meade et al. 1985; Dunne et al. 1989; Mertes et al. 1989). Hence, there is both a seasonal and long-term disequilibrium of sediment transport potentially causing accumulation of sediment in the floodplain.

3.3.4.3 Metabolic gradients and bioactive elements

So far we have dealt primarily with particulate and dissolved fractions controlled on slow- to intermediate-dynamic scales. Carbon oxidation rates, dissolved gas distributions and mass balances suggest that there is also a 'fast- dynamic' scale level of control of organic matter and nutrients.

The overall carbon and nutrient budgets indicated a shortage of organic carbon and nutrients in transit downstream to support the observed levels of in situ oxidation by an amount roughly equal to the oxidation rate. The compositional indices previously discussed indicate that most of the POC and DOC is refractory and behaves relatively conservatively. Thus, a large additional-and as yet unidentified-source of labile organic substrates that can be rapidly oxidized is required.

What is the potential of the system to provide labile organic matter and nutrients, and what might their form be? There are several potential sources of such material. There is an exchange of particulate material between the main stem and floodplain; such an exchange would likely mobilize labile POC, which could be a substrate for oxidation. Though the inventory of this material would be small relative to the bulk POC in transport, supply rates on a fast -dynamic scale would maintain a rapid turnover. Labile dissolved organics supplied laterally from the várzea are a likely substrate for in- channel respiration, which in turn would release nutrients. If so, there must be a gradient at some space scale in metabolic properties between the main stem and floodplain. Richey et al. (1988) showed that as a function of an oxidation-reduction sequence in combination with physical mixing between the floodplain and main stem, inputs from the more reducing várzea can impact the main channel.

Clearly, organic matter sources, of the magnitude required to sustain oxidation, do exist and could continuously supply carbon to the river by lateral exchange on time and space scales shorter than have been measured to date. The overall implication is that a small but highly labile carbon pool coexists with a pool of more refractory material that accounts for the bulk of organic matter in transport.

3.4 SUMMARY

A model of the biogeochemistry of the Amazon River has been outlined. Water and materials in the main stem carry signals of drainage basin processes operative at several different time and space scales. Variations in discharge and concentrations of dissolved and particulate materials in transport occur on a relatively damped and predictable basis, with changes occurring over several weeks and tens to hundreds of kilometers. This signal, however, also appears to be affected by processes operating on very different scales. The chemical compositions of the bulk organic matter in transport appear to be established on a decadal basis over large areas of the Andean and tributary basins (and perhaps on the floodplain). Conversely, the level of oxidation, the concentration of bioactive elements such as nitrogen, and certain compositional indices indicate that rapid processes occur, where the relevant scale parameters are hours to days and distances of several kilometers.

These results exemplify both the necessity of viewing organic carbon dynamics on a basin-wide perspective and the utility of using coupled chemical and isotopic analyses to deconvolve their dynamics. How will the distributions and processes and their respective time and space scales represented in the Amazon model differ for the other great rivers of the world represented in this volume? The in-depth study represented here should provide a guide for analysing the characteristics of other, less-well studied systems.

ACKNOWLEDGEMENTS

We gratefully acknowledge the contributions of our colleagues of the CAMREX project, and the intellectual spirit and camaraderie of Egon Degens and the SCOPE Biogeochemistry of Major World Rivers project. Supported by the National Science Foundation grant BSR-8107522. Contribution No.44 of the CAMREX project.

REFERENCES

Devol, A. H., Richey, J. E., Quay, P. and Martinelli, L. (1987) The role of gas exchange in the inorganic carbon, oxygen, and ²²²Rn budgets of the Amazon River . Limnol. Oceanogr. 32, 235-48.

Devol, A. H., Richey, J. E., Forsberg, B. F., Hedges, J. I. and dos Santos, U. (1990) The biogeochemistry of the Amazon River: the dynamics of N and P. Limnol. Oceanogr. (submitted).

Dunne, T., Meade, R. H., Richey, J. E. and Mertes, L. (1989) Sediment transport in the Amazon River: annual fluxes, aggradation, and hydraulics. (in prep.).

Ertel, J., Hedges, J. I., Richey, J. E., Devol, A. H. and dos Santos, U. (1986) Dissolved humic substances of the Amazon River System. Limnol. Oceanogr. 31, 739-54.

Forsberg, B. F., Devol, A. H., Richey, J. E., Martinelli, L. and dos Santos, U. (1988) Factors controlling nutrient distributions in Amazon lakes. Limnol. Oceanogr. 33, 41-56.

Hedges, J. I., Ertel, J. R., Quay, P. D., Grootes, P. M., Richey, J. E., Devol, A. H., Farwell, G. W., Schmidt, 
F. H. and Salati, E. (1986a) Organic carbon-14 composition in the Amazon River system. Science 231, 1129-31.

Hedges, J. I., Clark, W., Quay, P. D., Richey, J. E., Devol, A. H. and Ribeiro, N. (1986b) Composition and fluxes of organic matter in the Amazon River. Limnol. Oceanogr. 31, 717-38.

Jordan, C. F. (1982) The nutrient balance of an Amazonian rain forest. Ecology 63, 647-54.

Meade, R. H., Dunne, T., Richey, J. E., de M. Santos, U. and Salati, E. (1985) Storage and remobilization of sediment in the lower Amazon River of Brazil. Science 228, 488-90.

Mertes, L., Dunne, T. and Richey, J. E. (1989) Floodplain development and sediment transport of the Solimões-Amazon River. (in prep.).

Nordin, C. F., Jr, Cranston, C. C. and Mejia, B. A. (1983) New technology for measuring water and suspended-sediment discharge of large rivers. In: Proceedings of the 2nd International Symposium on River Sedimentation, Water Resources and Electric Power Press, Beijing, China, pp. 1145-58.

Quay, P., Stuiver, M., Hedges, J. I., Devol, A. H. and Richey, J. E. (1989) Carbon cycling in the Amazon River: Implications from the 13-C composition of particulate and dissolved carbon. Limnol. Oceanogr. (in review).

Radambrasil (1972) Mosaico semi-controlado de radar, escala 1: 250000. Ministério das Minas e Energia, Departamento Nacional de Produção Mineral, Rio de Janeiro.

Richey, J. E. (1982) The Amazon River system: A biogeochemical model. In: Degens, E. T. (Ed.) Transport of Carbon and Minerals in Major World Rivers, Pt. 1. Mitt. Geol-Paläont. Inst. Univ. Hamburg, SCOPE/UNEP Sonderbd. 58, pp. 365-78.

Richey, J. E. (1983) Interactions of C, N, P and S in river systems: a biogeochemical model. In: Bolin, B. and Cook, R. (Eds) The Interaction of Biogeochemical Cycles, John Wiley and Sons, New York, pp. 365-83.

Richey, J. E., Brock, J. T., Naiman, R. J., Wissmar, R. C. and Stallard, R. F. (1980) Organic carbon: oxidation amd transport in the Amazon River. Science 207, 1348- 51.

Richey, J. E., Meade, R. H., Salati, E., Devol, A. H., Nordin, C. F. and dos Santos, U. (1986) Water discharge and suspended sediment concentrations in the Amazon River: 1982-1984. Water Resources Res. 22, 756-64.

Richey, J. E., Devol, A. H., Victoria, R. and Wofsy, S. (1988) Biogenic gases and the oxidation and reduction of carbon in the Amazon River and floodplain waters. Limnol. Oceanogr. 33, 551-61.

Richey, J. E., Mertes, L. A., Victoria, R. L., Forsberg, B. R., Dunne, T., Oliveira, E. and Tancredi, A. (1989a) Sources and routing of the Amazon River fioodwave. Global Biogeochemical Cycles 3, 191-204.

Richey, J. E., Devol, A. H., Hedges, J. I., Forsberg, B. R., Victoria, R., Martinelli, L. A. and Ribeiro, N. (1989b) Distributions and fluxes of carbon in the Amazon River. Limnol. Oceanogr. 35, 352-71.

Sioli, H. (1975) Amazon tributaries arid drainage basins. Ecol. Stud. 10, 199-213. 

Vannote, R. L., Minshall, G. W., Cummins, K. W., Sedell, J. R. and Cushing, C. E.  (1980) The river continuum concept. Can. J. Fish. Aquat. Sci. 37, 130-7. 

Welcomme, R. L. (1979) Fisheries Ecology of Floodplain Rivers, Longman, London,  317pp.

Welcomme, R. L. (1985) River Fisheries, FAO Fish. Tech. Pap. 262, 330pp.