fi" Acknowledaments QRSToIV! Fonds Docomentaire Regional pattern of riverine dissolved organic carbon in the Amazon drainage basin of Bolivia

Transcription

1 Limnol. ceanogr., 9(), 994, , by the American Society of Limnology and ceanography, Inc. Regional pattern of riverine dissolved organic carbon in the Amazon drainage basin of Bolivia )' Abstract-Dissolved organic carbon (DC) was analyzed at hydrometric stations and 4 Andean and lowland sites in the upper Rio Madeira drainage basin where anthropogenic alteration is low. The two regions, Andes and Llanos, behave very differently. Mean DC is. mg liter-' (range,.4. mg liter-i) in the Andes and.7 mg liter-i (range, mg liter-') in the Llanos. DC enrichment may occur rapidly as Andean rivers enter the Llanos. The specific DC export is correlated with the watershed forested area in the Andes but not in the Llanos. Specific DC yield for a given water runoff is times fi" Acknowledaments We ackruowledge the staff of the SENAMHI and IIQ- UMSA for field assistance in Bolivia. We are indebted to H. Sanejouand and P. Namour of the CEMAGREF (Lyon, France) for organic carbon analysis and to M. Meybeck for comments. higher in the Llanos. Regionalization seems important for predicting DC dynamics in such a large basin. The biogeochemistry of organic C in the Amazon River has been extensively investigated with regard to origin, processes, and budgets (e.g. Richey et al. 990). Richey et al. sampled Amazon tributaries near their confluence to determine their contribution to mainstem functioning. However, little is known about the origin of organic C in the upper drainage basins of the Amazon. The relative contribution of the different regions and landscapes to riverine carbon yields remains poorly documented. Although increasing colonization is leading to rapid landscape alterations in some QRSToIV! Fonds Docomentaire

2 Notes 4 o - 00 km Fig.. Map of the Amazon drainage basin of Bolivia: sampling points-.; PHICAB gauging stations-*, major cities-.. Letters refer to the five regions:.a-upper Rio ; B-Rio Grande; C-Chapare, for the Andes; D-Rio ; E-Rio, for the Llanos. Stations codes given in Table. regions, the global impact of such development on the organic C fluxes in these rivers is hardly predictable. The drainage basin of the Rio Madeira upstream of the Brazilian frontier extends over 900,000 km and supplies on average 8,000 m of water s-l. Andean tributaries are known to provide most of the yields for both dissolved and sediment loads (Roche and Fernandez 988; Guyot et al. 988), but nothing is known about organic C. This large, highly contrasted, and still relatively pristine tropical basin now constitutes a study area of high interest, inasmuch as it is covered by the PHICAB hydrological network which has been providing daily data on discharge and total dissolved and suspended solids (TDS and TSS) at gauging stations since 98. ur study was designed to investigate the spatial pattern of riverine dissolved organic C (DC), to identify possible causes of variation, and to propose a framework for the regionalization of the DC dynamics. The Bolivian portion of the upper Rio Ma-

3 44 Notes Discharge (m s-) - 4 u u U u, DC (mg liter-) Fig.. Mean daily discharge of Rio at Angosto del Bala (A4) and Cachuela Esperanza (D), Numbers on upper axis (-4) indicate sampling periods at the PHICAB stations. deira drainage basin (74% of the total basin area) is clearly divided into two geographic zones: the eastern slope of the Andes, ranging from 6, to 0 m, and the plain ( Llanos ), extending from the piedmont of the Andes to the Brazilian shield (Fig. ). Individual watersheds in the Andes are heavily contrasted; they range from semiarid basins on the Altiplano quaternary sediments to densely vegetated hyper-humid basins on Paleozoic rocks. That large zone has been subdivided into three regions. The upper Rio drainage basin (A on Fig. ) is made up of deeply incised, humid (rainfall,,-,000 mm yr-i), forested valleys: the Yungas and the Alto- foothills; it also includes some semiarid (rainfall, < 800 mm yr-l), highly erodible basins (Rios La Paz- Boopi and Consata). The Rio Grande drainage basin (B), which culminates at,00 m, is entirely semiarid. The Chapare region (C) is made up of densely forested mountains and reaches 4, m; in this region, as in other peripheral parts of the Amazon basin, precipitation is maximal (,000-7,000 mm yr-l) at very low elevations in the contact zone between the Andes and the Llanos and decreases to,000 mm yr-l with increasing altitude. The extreme variability of mechanical erosion rates in adjacent basins must be emphasized (,000-0,000 t kmf yrl, Guyot et al. 988). Highly turbid rivers flowing out from the Andes (Madre de Dios,, and ) correspond to white-water rivers. The Llanos is a nearly flat plain of recent alluvial sediments whose level is controlled by the Precambrian Brazilian Shield outcrop 0 J W ooo 7 wo Discharge (m s-) Mar 88 Fig.. DC vs. discharge on the Rio at Angosto del Bala (A4). near Cachuela Esperanza (D) and Guayaramerin (El ). About 0,000 km are flooded annually in this region, inducing the development of savannah patches. Annual rainfall is -,000 mm in the Rio catchment (D), which is covered mainly by tropical evergreen to seasonal forest. In region E, rainfall ranges from,000 mm yr-i in the north to, 0 mm yr- in the southeast where a tropical seasonal forest progressively gives place to tropical savannah. Human densities are low, -7 inhabitants kmw, including major cities, and - inhabitant km- in the Llanos. Accessibility to most rivers is difficult. Pollution is restricted to some large cities. Hydroelectric reservoirs are small and located at high altitude; channelization is unknown. Agriculture is traditional with many sites of active rural colonization in the Chapare, Alto-, and Llanos. The PHICAB gauging stations located in the Andean piedmont and the Llanos were sampled -4 times, corresponding to different 00 AHitude(m) I boo A p ICGQ - MM A P P AP %I A.E DC (mg liter-i) n AA Fig. 4. Mean DC value vs. sampling point altitude. Points referenced by region (Table, Fig. ); P indicates a polluted river. B C AD

4 Notes 4 Table. Data for rivers of the Amazonian drainage basin of Bolivia. River sample points, characteristics, and mean results. Alt.-sample point altitude (m); A-drainage basin area (km*); gs-specific discharge or runoff (liter s-* krr~-~); F-forested area (%) in the watershed ([.] = To of forested area for Bolivian area only); n-number of organic C samples; DC-average values (mg liter- ); gs x DC-DC-specific yields (g mwz yr-*). Code River Alt.. A qs F n DC qsxdc Al A A A4 A A6 A7 A8 A9 A0 Al A A A4 Al A6 A7 A8 A9 A0 A A A A4t B B B B4t c c c c4 c C6 c7 C8 Dlt Dt D D4t Dt El Et E E4 E E6t E7 E8 E9t Elt El lt San Cristobal Consata Mapiri Tipuani Challana Zongo San Juan Coroico Coroico Kaka Unduavi Taquesi Tamampaya La Paz* La Paz* Luribay La Paz* Miguillas La Paz Boopi Boopi Cotacajes Alto Rocha* Tapacari Arque Grande Esperitu Santos San Mateo Chimore Sajta Ichoa Ichilo Yapacani Piray* Madre de Dios rthon Grande Ichilo Ichilo- Secure Tijamuchi Apere Itenez-Guapore * Polluted river. t PHICAB gauging station. 900,00, ,00,00,000,0 soo soo ,40, ,00,400, ,400 8,800 60,00 9,00 80,00,000 7,400 0,900, 6,000 9,00 67,, ,0 9,800,0,900,00 6,00 6,900 4,00 9,000 4,00 4,000,0 8, 67,000 7,600 4,000 7,000 44,000 9,00 7,600 7,000 6,00 4,0 99, ? s ~ o o o ;

5 46 Notes Table. Minimum, average, and maximum values of mean DC contents (mg liter-') in unpolluted stations of the five regions. (Regions identified in Fig..) No. -number of stations by region. DC (mg liter-i) Domain Region No. Min Av8 Max ~~ Andes A..9.8 B.0.4. C I Llanos D E hydrological phases (Fig. ): low flow (July- August 987), rising levels (November 987, December 988), and high flows (March 988). Major rivers in the Andes were sampled once in March, April, or June 988, and a -d flood episode was studied at one station. In the turbulent Andean rivers, samples were taken at the river's edge. In the Llanos, surface samples were taken in the middle of the river. Measurements made with depth-integrating samplers (USD 49) at gauging stations exhibited very good homogeneity of TDS over the whole cross-section, the C.V. being 0.0 for 9% of the trials. Samples were passed, on location, through a glass-fiber filter (GFK) previously heated at 0 C. The samples were then acidified (HP04, ph ), stored in the dark, and sent to the CEMAGREF laboratory in Lyon (France). DC determinations were performed with a nondispersive infrared analyzer (Dhormann DCS). Ninety-three samples from sites were processed. Percentage of forested area was calculated for each station watershed with a vegetation map of Bolivia (Brockmann 978), assuming that no significant large-scale alteration of the vegetative cover has occurred in the past yr. Five stations contaminated by the cities of La Paz (A4, A, A7), Cochabamba (Bl), and Santa Cruz (C8) were not used. DC was measured daily during an annual flood episode at the Angosto del Bala gauging station (A4) which controls the Rio output from the Andes. DC contents varied between.7 and. l mgliter-', but with no direct relation with either discharge (Fig. ) or TDS during such events. The flushing effect (i.e. increased DC with increased discharge) observed by various investigators therefore seems limited. Still, the lowest value for this station (. mg liter-', 8 August 987) corresponds RioGrande / 000 loo0 Distance from river confluence (h) Fig.. Downstream evolution of mean DC content along the Rio and its Andean tributaries. Stations identified in Table. to the lowest flow situation observed in the period. Thus, as previously observed in tropical South America (Lewis et al. 986; Richey et al. 990) and in the Gambia River (Lesack et al. 984), DC values for a particular river seem to have limited variations (--fold) on a broad temporal scale. Consequently, we used average values to characterize spatial variations. Detailed results are given in Table. By plotting DC values vs. altitude (Fig. 4), the contrast between the Andes (A, B, C) and Llanos (Dy E) is immediately apparent. Mean DC content is. mg liter-' for unpolluted Andean stations and.7 mg liter-' for the Llanos below 00 m. Data for each region are summarized in Table. The ranges for Andes and Llanos do not overlap and correspond to the world DC distribution proposed by Meybeck ( 98). Extreme values measured in the Llanos are.6 and 9.4 mg liter-'. These values are in the same range as the data from Amazonian rivers in Brazil (Richey et al. 990) and from the rinoco basin (Lewis et al. 986; Depetris and Paolini 99 ). But the salient feature of this pattern is the evidence of a DC enrichment of the whitewater rivers (Rios and ) as they enter the Llanos. This transition can be fast, as illustrated by the evolution of DC along the Rio, coming from the B and C regions (Fig. ). The Rio Ichilo, sampled at km from the Chapare foothills (E) has the highest DC content (7.4 mg liter) while the mean DC value of its four tributaries (C- C6) is only. mg liter-'. The same phenom-

6 00 DC (g m- yr-i) Notes 47 m A A - B o c 0. A D M E 0..I Forested arca in the tasin (%) Fig. 6. Mean DC-specific yield vs. forested area in the drainage basin. For Andean rivers (A. B, C) n =, r = 0.664, P < 0.00, qs x DC = exp(0.08 x F - 0.7). D, E-Llanos rivers. Points referenced by region (Table, Fig. ) Runoff Qiten s-i h ) Fig. 7. Mean DC-specific yield vs. runoff. For Andean rivers (A, B, C) n =, r = 0.4, P < 0.00, qs x DC = x qs For Llanos rivers (D, E) n = 6, rz = 0.9, P < 0.00, qs x DC = 0.8 x qs Points referenced by region (Table, Fig. ). enon is observed along the Rio Grande between stations B4 and El with DC contents of. and 6. mg liter. In both cases, this enrichment occurs in a short distance at the beginning of the floodplain and can be related to the presence along the piedmont of extensive swamps and riparian forests in zones of active fluvial dynamics that maximize the connectivity between rivers and wetlands. At the Chapare (C), this phenomenon may be intensified by huge local rainfalls. The DC-specific export rate (mean DC x specific discharge) has been plotted vs. total forested area for each watershed. These two parameters exhibit a good correlation, but that is exclusively due to the Andean stations. The correlation is significant for regions A, By and C but not for the Llanos (Fig. 6). Thus, the role of forested areas in river DC production seems to be different depending on the geographical context. The relationship between forest area and DC production is poorly documented and principally in relation to forestry practices (Collier et al. 989). ur results indicate that in the deeply incised and humid Andean valleys (Yungas and Chapare), the valleyside forests of the watershed are involved in DC production. Conversely, the importance of carbon inputs from the floodplain has been emphasized (Ittekkot and Arain 986; Richey et al. 990). In the Bolivian Llanos, high DC production may be achieved in the floodplain, and the contribution of the wholebasin forest seems to be of minor importance. As observed by Spitzy and Leenheer ( 99 l), DC production increases linearly with increasing runoff. DC specific yield is obviously correlated with specific discharge, as DC contents exhibit limited kariations during the hydrological cycle. But the salient feature of the Bolivian data (Fig. 7) is once more the clear separation between the Andes and the Llanos, the regression coefficients for both zones being significantly different. The DC production for a given runoff is - times higher in the Llanos than in the Andes, suggesting that different processes prevail in each region. Despite the relatively small number of samples, the first insight into the DC spatial pattern in the Amazonian basin of Bolivia has shown a clear distinction and a rapid transition between two domains: the Andes and the Llanos. This difference in DC content seems to correspond to a distinct origin of DC production: the valleyside forests in the Andes and, presumably, the floodplain wetlands in the Llanos. As a consequence, deforestation should not have the same impact on DC dynamics in these two domains. Jean Loup Guyot RSTM C.P Brasilia DF, Brazil Jean Gabriel Wasson CEhGREF, Div. B.E.A. Lab. Hydroécologie Quantitative B.P Lyon Cedex 09, France

7 48 Notes References solutes in the Gambia River, West Africa. Limnol. ceanogr. 9: BRCKMA", C. E Mapa de cobertura y uso actual LEWIS, W. M., JR., J. F. SALJNDERS III, S. N. LEVINE, AND de la Tierra, Bolivia. Geobol, La Paz. F. H. WEIBEZAHN rganic carbon in the Caura CLLIER, K. J., R. J. JACKSN, AND M. J. WINTERBURN. River, Venezuela. Limnol. ceanogr. : Dissolved organic carbon dynamics of devel- MEYBECK, M. 98. Carbon, nitrogen, and phosphorus,i oped and undeveloped wetland catchments in West- transport by world rivers. Am. J. Sci. 8: land, New Zealand. Arch. Hydrobiol. 7: -8. RICHFI, J. E., AND THERS Biogeochemistry of! DEPETRIS, P. P., AND J. E. PALINI. 99. Biogeochemical carbon in the Amazon River. Limnol. ceanogr. : Ir aspects of South American rivers: The Paraná and the -7. I rinoco, p. 0-. In E. T. Degens et al. [eds.], RCHE, M. A., AND C. FEWANDEZ Water re- Biogeochemistry of major world rivers. SCPE 4. sources, salinity and salt exportations of the rivers of Wiley. the Bolivian Amazon. J. Hydrol. 0: -. GUYT, J. L., AND THERS Exportation de ma- SPITZY, A., AND J. LEENHEER. 99. Dissolved organic tihres en suspension des Andes vers l'amazonie par carbon in rivers, p. -. In E. T. Degens et al. le Rio, Bolivie, p In M. P. Bordas and [eds.], Biogeochemistry of major world rivers. SCPE D. E. Walling [eds.], Sediment budgets. IAHS Publ. 4. Wiley. 74. ITTEICKT, V., AND R. AFMN Nature ofparticulate organic matter in the River Indus, Pakistan. Geochim. Cosmochim. Acta 0: LESACK, L. F. W., R. E. HECKY, AND J. M. MELACK Transport of carbon, nitrogen, phosphorus, and major Submitted: January 99 Accepted: June 99 Amended: August 99