Optical characterization of dissolved organic matter in tropical rivers of the Guayana Shield, Venezuela


  • Youhei Yamashita,

    1. Southeast Environmental Research Center, Florida International University, Miami, Florida, USA
    2. Department of Chemistry and Biochemistry, Florida International University, Miami, Florida, USA
    3. Now at Faculty of Environmental Earth Science, Hokkaido University, Sapporo, Japan.
    Search for more papers by this author
  • Nagamitsu Maie,

    1. Department of Environmental Biosciences, School of Veterinary Medicine, Kitasato University, Towada, Japan
    Search for more papers by this author
  • Henry Briceño,

    1. Southeast Environmental Research Center, Florida International University, Miami, Florida, USA
    Search for more papers by this author
  • Rudolf Jaffé

    1. Southeast Environmental Research Center, Florida International University, Miami, Florida, USA
    2. Department of Chemistry and Biochemistry, Florida International University, Miami, Florida, USA
    Search for more papers by this author


[1] Tropical rivers are an important source of dissolved organic matter (DOM) to coastal oceans. However, temporal and spatial variability of DOM composition and thus its quality in such rivers, on landscape and basin scales, have not been well documented. In this study, we present data on the spatial distribution of DOM quantity and quality based on source, molecular weight, and composition using optical properties including excitation emission matrix fluorescence with parallel factor analysis. We compared such DOM quantity and quality determinations in main river channels and their tributaries for three river systems of the Guayana Shield, Venezuela. Spatial variabilities of DOM parameters were strongly related to differences in the geological settings of the drainage basins and presumably their associated vegetation cover. Linear relationships between quantitative and qualitative DOM parameters were also evident, suggesting that high DOC concentration correlated with chromophoric dissolved organic matter (CDOM) characteristics of higher molecular weight associated with terrestrial sources, while low DOC concentrations correlated with CDOM characteristics of lower molecular weight associated primarily with microbial sources. Such relationships seem to imply that DOM concentrations and their sources/characteristics may be coupled in the studied fluvial systems. In addition, shifts in DOM compositions between terrestrial and microbial signals were observed with changes in water discharge and in watersheds disturbed by gold mining activities. The observed linkages between, and the changes among DOM quantity and quality, suggest that the biogeochemistry of DOM in tropical rivers may be quite sensitive to climatic and land use change.

1. Introduction

[2] Large world rivers convey a substantial amount of terrestrial organic matter to the ocean (0.4 GtC/yr) [Hedges et al., 1997] primarily derived from net uptake of carbon dioxide by terrestrial ecosystems (2.2 GtC/yr [Denman et al., 2007]). Thus, riverine transport of terrestrial organic matter is known to be an important process in the global carbon cycles, and ultimately integrates all sources and processes within the watersheds [Degens et al., 1991]. Quantitative and qualitative changes in DOM in riverine systems, and on watershed scales have been reported [Lara et al., 1998; Lobbes et al., 2000; Duan et al., 2007; Eckard et al., 2007; Cory et al., 2007], and its microbial and photochemical processing was suggested to be affected on landscape scales [Battin et al., 2008]. Since climate change as well as human impact such as changes in land use directly influence riverine scale DOM characteristics quantitatively and qualitatively [Monteith et al., 2007; Hood and Scout, 2008; Wilson and Xenopoulos, 2009], a better understanding of DOM variability in aquatic systems is necessary.

[3] Water discharge and associated riverine dissolved organic carbon (DOC) export to the coastal zone in the wet tropics accounts for ca. 51.3% and 65.5% loadings world wide, respectively [Spitzy and Leenheer, 1991]. In this respect, the spatial and temporal variability of organic matter in terms of quantity and quality have been surveyed in the Amazon River with the objective of determining its flux and fate [Richey et al., 1990; Hedges et al., 1994, 2000a; Mounier et al., 1999; Moreira-Turcq et al., 2003a, 2003b; Johnson et al., 2006; Saunders et al., 2006; Aufdenkampe et al., 2007]. The spatial variability of DOC concentration on riverine scales has been reported for some tropical rivers [Guyot and Wasson, 1994; Battin, 1998; Moreira-Turcq et al., 2003a]. In terms of chemical characteristics of organic matter, these studies found that particulate organic matter (POM) is less degraded than the corresponding DOM [Hedges et al., 1994; Aufdenkampe et al., 2001, 2007] which showed an enrichment in terms of its humic substances content [Ertel et al., 1986; Hedges et al., 1994]. Both the DOC concentration and the relative abundance of DOM in total riverine organic matter increased downstream [Hedges et al., 2000a]. Thus, highly degraded DOM seems to play an important biogeochemical role in tropical river systems as it is transported throughout their watershed and then discharged to coastal environments. It is well known that highly degraded DOM is optically active [Hedges et al., 2000b], and thus, studying chromophoric dissolved organic matter (CDOM) is ideal for a better understanding of DOM dynamics in rivers and coastal areas [Jaffé et al., 2008; Gonsior et al., 2008; Yamashita et al., 2008].

[4] CDOM is known to occur in a wide range in aquatic environments [Blough and Del Vecchio, 2002; Nelson and Siegel, 2002; Jaffé et al., 2008; Yamashita and Tanoue, 2008] and plays an important role as a factor determining light penetration [Frenette et al., 2003; Hayakawa and Sugiyama, 2008], as the substrate for photoproducts [Mopper et al., 1991; Moran and Zepp, 1997; Mopper and Kieber, 2002], as a pH buffer [Ceppi et al., 1999; García-Gil et al., 2004], and as a medium for the complexation of trace metals [Lu and Jaffé, 2001; Tosiani et al., 2004; Yamashita and Jaffé, 2008]. Thus, the chemical characteristics and temporal and seasonal variation of CDOM have extensive implications in aquatic ecology. As a result, several optical parameters have been reported to qualitatively characterize CDOM, such as the spectral slope (S) of the UV absorbance [Green and Blough, 1994; Battin, 1998], the maximum emission wavelength (λmax) [de Souza Sierra et al., 1997; Jaffé et al., 2004], the ratio of two regions in the S as an index of molecular weight (SR) [Helms et al., 2008], and the fluorescence index as a DOM source indicator [McKnight et al., 2001; Cory and McKnight, 2005]. Such indexes have been successfully applied in CDOM studies of a variety of aquatic environments [e.g., Jaffé et al., 2008; Helms et al., 2008]. More recently, the combined techniques of excitation-emission matrix (EEM) fluorescence with parallel factor analysis (PARAFAC) have successfully evaluated the environmental dynamics (i.e., source and fate) of fluorescent DOM components in diverse aquatic ecosystems [Stedmon et al., 2003; Cory and McKnight, 2005; Stedmon and Markager, 2005; Jaffé et al., 2008; Yamashita et al., 2008]. This technique provides higher resolution on EEM fluorescence components in DOM, and thus, may be ideally suited to detect small, but potentially significant variations in DOM composition in apparently similar aquatic environments.

[5] Consequently, optical parameters such as those described above could provide additional insights into the biogeochemical dynamics of DOM in tropical rivers. However, only few studies have, to the best of our knowledge, been carried out on the combined quantitative and qualitative characterization of CDOM in tropical rivers and their watersheds [Battin, 1998]. Therefore, the main objective of this study is the assessment of the spatial distribution of CDOM in several remote tropical rivers of the Guayana Shield, Venezuela. In this study, special attention was paid to assess quantitative and qualitative similarities/differences in CDOM between three different, remote tropical river systems, between the mainstreams and associated tributaries for each of these systems, and to clarify the relationships between quantitative and qualitative parameters in tropical rivers using a variety of optical parameters including EEM-PARAFAC.

2. Methods

2.1. Study Sites

[6] Three tropical river systems located in southeast Venezuela, namely the upper Orinoco/Ventuari, the upper Paragua/Karun, and the Cuyuni/Uey rivers were chosen for this study (Figure 1). The Guayana region (400,000 km2), located in southeastern Venezuela, is characterized by its dense forest and patchy grassy-savanna cover, high precipitation rate (>3,000 mm) with maximum rainfall from May to October, and warm climate (24–26°C annual average). It features a series of large tropical river systems, including the upper Orinoco, and scarce population allows for mostly pristine environmental conditions. The overall region has been subjected to continuous tropical weathering processes and tectonic stability since Cretaceous times (some 60 million years), and conditions of quasi-equilibrium have been reached rendering landforms which are mostly the result of differential erosion [Briceño and Schubert, 1990]. Higher elevations and abrupt relief are the domain of resistant quartzites of the Roraima Group rocks (Proterozoic age), while extensive, undulated, low relief areas occur on the igneo-metamorphic terrain of the Imataca, Pastora and Cuchivero provinces (Archean to Lower Proterozoic age). Consequently, regoliths are thick, acidic (especially on Roraima-derived sandy soils) and extremely nutrient poor.

Figure 1.

Map of sampling locations. (a) Orinoco/Ventuari system; (b) Paragua/Karun system; (c) Cuyuni/Uey system. For each system, the two main river channels and their corresponding tributaries are labeled with symbols used throughout the manuscript. Specific sites with special labeling are (in Figure 1b) Ichun River (open diamond) and lower Paragua flooding lagoons (cross), and (in Figure 1c) tributaries with mining impact (open circle). Note: General water flow direction is indicated by dotted arrows.

[7] Despite the large distances between the studied sites (Figure 1), there are common geological features that result in similar landform, soil type, vegetation patterns and water quality in their drainages. Sedimentary rocks occur in the upper reaches of the Uey (not shown in Figure 1c), the whole Ichun basin, the upper Paragua and the right margin of the Orinoco river, upstream from its confluence with the Ventuari River. These rocks generate sandy, extremely lixiviated, acidic and permeable soils, covered with abundant organic detritus, whose acid drainages contain high DOC concentrations, of mostly soil-derived origin. On the other hand, igneo-metamorphic rocks, especially granites and gneisses, occur along the left margin of the Orinoco, the Ventuari, the Karun–lower Paragua, and also sporadically in the Cuyuni and lower Uey rivers. Intermediate-to-felsic metavolcanics rocks occur in the Paragua-Karun basin and prevail in the Cuyuni and lower Uey rivers. Extreme weathering of these igneo-metamorphic units develop thick, kaolinite-rich, clayey, poorly drained and locally lateritic soils, which occasionally are slightly richer in exchangeable Ca, Mg and K than those on sedimentary rocks. Waters draining these soils are less acidic, less colored (less DOC) and locally transparent and of greenish color (especially in the Ventuari River).

[8] In general, regional vegetation patterns and composition follow geomorphology, flooding magnitude and hydroperiod, and soil type [Huber, 1985, 1995; Huber and Alarcon, 1988]. Riverine forests are very diverse, productive and locally, as in the Orinoco/Ventuari, several kilometers wide. They are seasonally inundated as is the associated floodplain and large amounts of decomposing plant-derived debris are observed on local soils [Jaffé et al., 1996]. Also, vegetation on quartzites of Roraima and on sandy, quartz-rich soil has a different floristic composition and is usually less diverse than that on clay-rich soils on crystalline rocks. Grassy savannas (on granites) prevail on the left margin of the Orinoco bordering the riverine forests [Huber, 1995; Rodriguez et al., 2006], and as patches mixed with shrublands on thin and sandy soils over quartzites outcropping along its right margin. As presented below, these geologically derived characters of the landscape and their connections to resulting soil type and vegetation cover bear a relevant influence on the characters, quality and quantity of DOM in their waters.

2.2. Water Sample Collection

[9] The sampling at the Orinoco/Ventuari was carried out on May 2006 and May 2007, when for May 2006 (beginning of wet season) water levels were at least 2 m higher as compared to those of May 2007. Two meter is about 20% of the stage variation in annual hydrogram of the upper Orinoco River [Meade et al., 1990], and thus represent a significant difference between the two sampling events. Surface water samples (ca. 25 cm below the water surface) were obtained during November–December 2005 and January 2008 at the Paragua/Karun and Cuyuni/Uey, respectively. Immediately after sampling, the samples were filtered through precombusted GF/F filters, the filtrate were placed into brown high-density polyethylene bottles and transported on ice to the laboratory for further analysis. The brown high-density polyethylene bottles were precleaned with 0.5 M HCl followed by 0.1 M NaOH, and then, were extensively rinsed with Milli-Q water. All samples were analyzed within 2 weeks of sampling. Sample reanalyses after 4 weeks showed no detectable differences in the optical properties. Only the filtrated samples from the Cuyuni/Uey river system were kept frozen until analyzed.

2.3. Laboratory Analysis

[10] Water samples were allowed to stand until reaching room temperature prior to fluorescence and absorption analysis. Absorbance spectra were determined between 240 nm to 800 nm at 1 nm intervals using a dual-beam spectrophotometer (Cary 50 Bio, Varian) equipped with 1 cm quartz-windowed cells. A blank scan (Milli-Q) was subtracted from each spectrum. The blank corrected sample spectra were baseline corrected by subtracting average values ranging from 700 nm to 800 nm from the entire spectrum, and then converted to spectral absorption coefficients, a (λ, m−1) [Green and Blough, 1994].

[11] EEM fluorescence spectra were obtained using a Horiba Jovin Yvon SPEX Fluoromax-3 spectrofluorometer. Analytical procedures to determine EEMs have been reported elsewhere [Maie et al., 2006; Yamashita and Jaffé, 2008] and are only briefly described here. The emission spectra in the range of excitation wavelength +10 nm to +250 nm were obtained every 2 nm intervals at excitation wavelengths between 240 nm and 455 nm at 5 nm intervals. Several postacquisition steps were involved in the correction of fluorescence spectra as follows: (1) the absorption spectra were used for inner filter corrections according to McKnight et al. [2001]. (2) The EEM of Milli-Q water was subtracted from the sample EEM to remove most of the effects due to Raman scattering. (3) The excitation and emission correction files supplied by the manufactures were applied for the correction of specific instrument's components. (4) Fluorescence intensities were also corrected to the area under the water Raman peak (excitation = 350 nm) analyzed daily [Cory and McKnight, 2005]. (5) Fluorescence intensities were converted to quinine sulfate unit (QSU) using a calibration with quinine sulfate monohydrate [Yamashita and Tanoue, 2003]. DOC concentration was analyzed using high-temperature combustion on a Shimadzu TOC-VCSH analyzer.

2.4. CDOM Characterization

[12] Several parameters were applied to characterize the CDOM in this study. The absorption coefficient at 350 nm (a350, m−1) was reported as a quantitative parameter in this study. The SR value, recently introduced by Helms et al. [2008] as the ratio of the spectral slope (S) obtained from two regions, i.e., 275 nm to 295 nm (S275–295) and 350 to 400 nm (S350–400), was calculated using linear regression of the log-transformed spectral ranges, respectively. The fluorescence index was calculated as the ratio of fluorescence intensities at 470 nm and 520 nm emission at 370 nm excitation according to Cory and McKnight [2005] and Maie et al. [2006]. The a350* were estimated as the absorption coefficients at 350 nm (m−1) divided by DOC (mgC L−1) according to Blough and Del Vecchio [2002].

[13] Parallel factor analysis (PARAFAC) was applied to enhance the resolution of the EEM data. PARAFAC statistically decomposes the EEMs into their individual components, and after pioneering work of Stedmon et al. [2003], has been successfully applied to characterize CDOM extracted from soils [Ohno and Bro, 2006], in waters of diverse terrestrial environments [Cory and McKnight, 2005], in coastal environments [Stedmon and Markager, 2005; Yamashita et al., 2008], and in the open ocean [Wedborg et al., 2007; Murphy et al., 2008].

[14] The approach of PARAFAC modeling to EEMs has been described in detail elsewhere [Stedmon et al., 2003; Ohno and Bro, 2006]. The PARAFAC modeling was carried out in MATLAB (Mathworks, Natick, MA) with the DOMFluor toolbox [Stedmon and Bro, 2008]. The data set used for this purpose was composed of 58 samples from the Orinoco/Ventuari and Paragua/Karun and wavelength ranges used were 260–455 nm and 290–500 nm for excitation and emission, respectively. The determination of the correct number of components was primarily achieved by the split half analysis and random initialization [Stedmon and Bro, 2008]. The Cuyuni/Uey samples were not used for the EEM-PARAFAC analysis (see below).

[15] Because sample sizes were small, differences in DOM characteristics among water types as well as between sampling periods were assessed only for water types containing more than three samples by performing nonparametric Mann–Whitney U test (StatView 5.0; SAS Institute Inc). Principal component analysis (PCA) using the relative abundance of PARAFAC components from the Orinoco/Ventuari and Paragua/Karun watersheds was carried out by JMP 5.0.1 software (SAS Institute Inc). The first and second principal component explained 65% and 31% of the variability, respectively.

2.5. Sample Preservation Effects

[16] Analytical artifacts for the characterization of DOM upon sample freezing have been reported in the literature [Fellman et al., 2008]. These authors reported a significant decrease in DOC and changes in the specific UV absorbance (SUVA) for frozen samples compared to those stored under refrigeration for samples which show high DOC concentration and/or high SUVA values. To determine possible effects of preservation in frozen compared to refrigerated samples on each of the parameters we used three samples from the Orinoco/Ventuari watershed for comparative purposes. The differences before and after freezing (n = 3) were relatively minor at 2.5 ± 6.9%, −0.4 ± 1.5%, and 2.8 ± 2.5% for a350, SR, and fluorescence index, respectively. In contrast, after freezing, significant increases/decreases in the fluorescence intensity of different components in the EEM-PARAFAC were found and such variations were variable among the three samples. Therefore, while EEM-PARAFAC analyses were not carried out on the frozen samples from the Cuyuni/Uey watershed, all other optical parameters on these samples were included in the study.

3. Results

3.1. DOC Concentration and CDOM Levels

[17] DOC concentrations in the three tropical river systems studied here were largely variable and ranged from 108 to 2130 μMC (Table 1 and Figure 2). DOC concentrations in the main river channels were less variable, but were different among different river systems. On the other hand, DOC concentrations in tributaries were highly variable. The range of DOC found in this study was similar to those reported for the Caura River, [Lewis et al., 1986] and the Orinoco River and its tributaries in Venezuela [Battin, 1998], as well as for the Amazon River and its tributaries in Brazil [Richey et al., 1990; Guyot and Wasson, 1994; Hedges et al., 1994; Moreira-Turcq et al., 2003a; Aufdenkampe et al., 2007]. Quantitative parameters of CDOM (a350) were also considerably different among the main river channels and between main river channels and their tributaries, but their spatial distributions were similar to those of DOC concentration. The observed a350 in this study ranged from 2 to 110 m−1.

Figure 2.

Box and whisker plots of DOC, CDOM quantitative parameters (a350), and CDOM qualitative parameters (a350*, SR and fluorescence index) at Orinoco/Ventuari, Paragua/Karun, and Cuyuni/Uey watersheds.

Table 1. Rock Type, Vegetation Cover, pH, Turbidity, DOC Concentration, and Optical Properties in Three Tropical Watersheds
WatershedWater TypenRock TypeaVegetation CoverbpHTurbidity (NTU)DOC (μMC)a350 (m−1)a350* (L mgC−1 m−1)SRFluorescence IndexComponent 1 (%)Component 2 (%)Component 3 (%)Component 4 (%)Component 5 (%)
  • a

    Here I and S mean igneo-metamorphic and sedimentary rocks, respectively.

  • b

    Here F, Sa, B, and BS mean forest, savanna, bush, and bare soil, respectively.

  • c

    Igneo-metamorphic rocks are evident after junction with Ventuari River.

Orinoco/VentuariOrinoco main4I ≫ SF ≫ Sa5.6 ± 0.917.2 ± 1.9564 ± 2331.2 ± 1.34.6 ± 0.30.72 ± 0.001.29 ± 0.0125.1 ± 0.638.7 ± 0.313.4 ± 0.113.8 ± 0.18.9 ± 0.5
   2006Ventuari main3I ≫ SF ≫ Sa4.8 ± 0.425.3 ± 4.6405 ± 2018.9 ± 0.53.9 ± 0.20.73 ± 0.001.30 ± 0.0022.9 ± 0.238.0 ± 0.314.9 ± 0.113.4 ± 0.210.9 ± 0.3
 Orinoco tributaries5ScSa > B > F5.9 ± 0.36.8 ± 0.6697 ± 25633.3 ± 17.73.8 ± 0.80.72 ± 0.041.22 ± 0.0731.7 ± 4.740.3 ± 1.58.8 ± 3.013.1 ± 1.46.1 ± 0.7
 Ventuari tributaries9I > SSa, F4.3 ± 0.210.4 ± 4.1326 ± 15410.2 ± 6.32.4 ± 0.60.76 ± 0.031.37 ± 0.0320.6 ± 2.938.0 ± 1.618.1 ± 2.314.5 ± 0.98.8 ± 2.8
Orinoco/VentuariOrinoco main3I ≫SF ≫ Sa4.4 ± 0.117.5 ± 3.1547 ± 2128.3 ± 0.54.3 ± 0.20.75 ± 0.011.26 ± 0.0123.4 ± 1.038.4 ± 0.313.9 ± 1.012.7 ± 0.811.6 ± 1.7
   2007Ventuari main2I > > SF ≫ Sa5.2 ± 1.07.6 ± 0.0327 ± 1415.0 ± 2.23.8 ± 0.40.76 ± 0.001.29 ± 0.0020.4 ± 1.036.0 ± 0.416.4 ± 0.512.7 ± 0.314.5 ± 2.2
 Orinoco tributaries5ScSa > B > F3.7 ± 0.61.3 ± 0.5891 ± 37042.5 ± 22.53.8 ± 0.50.71 ± 0.001.20 ± 0.1031.6 ± 5.541.1 ± 1.78.1 ± 4.512.9 ± 1.66.3 ± 0.8
 Ventuari tributaries9I > SSa, F3.9 ± 1.13.4 ± 1.4319 ± 1159.9 ± 3.72.8 ± 1.20.80 ± 0.041.38 ± 0.0718.2 ± 4.336.9 ± 3.318.5 ± 3.014.4 ± 1.612.0 ± 5.5
Paragua/KarunParagua main3S, IF5.0 ± 0.38.7 ± 3.7562 ± 17830.1 ± 10.44.5 ± 0.60.72 ± 0.031.22 ± 0.0326.1 ± 1.640.6 ± 2.111.7 ± 2.212.8 ± 0.18.7 ± 1.5
 Karun main2I ≫ SF ≫ Sa5.2 ± 0.020.8 ± 15.2195 ± 408.9 ± 1.33.9 ± 0.20.84 ± 0.011.35 ± 0.0318.3 ± 3.333.0 ± 2.217.3 ± 1.612.3 ± 0.919.1 ± 4.8
 Paragua tributaries2S, IF5.1 ± 0.85.2 ± 0.7362 ± 12918.0 ± 7.74.1 ± 0.30.75 ± 0.041.25 ± 0.0623.5 ± 2.140.2 ± 1.714.0 ± 2.813.2 ± 0.29.1 ± 0.8
 Karun tributaries3I ≫ SF ≫ Sa4.5 ± 0.523.0 ± 29.9142 ± 376.1 ± 1.83.6 ± 0.10.94 ± 0.061.34 ± 0.0214.8 ± 2.929.7 ± 1.818.8 ± 0.810.6 ± 0.426.1 ± 3.4
 lower Paragua small tributaries4IF4.6 ± 0.46.5 ± 2.2144 ± 286.5 ± 1.33.8 ± 0.40.92 ± 0.111.37 ± 0.0216.6 ± 2.232.4 ± 3.017.7 ± 1.712.5 ± 1.320.8 ± 6.8
 flooding lagoons2IF5.7 ± 0.02.5 ± 1.3445 ± 311.5 ± 2.02.1 ± 0.40.85 ± 0.051.39 ± 0.0121.0 ± 1.834.1 ± 0.321.2 ± 1.414.1 ± 0.29.6 ± 0.0
Cuyuni/UeyCuyuni main5I > SF6.1 ± 0.110.4 ± 4.7398 ± 1714.4 ± 2.23.0 ± 0.40.75 ± 0.061.21 ± 0.01-----
 Uey main2S, IF6.2 ± 0.01.3 ± 0.0594 ± 14322.7 ± 5.73.2 ± 0.00.70 ± 0.021.19 ± 0.04-----
 Cuyuni tributaries3IF5.9 ± 0.32.5 ± 1.4390 ± 849.0 ± 2.61.9 ± 0.20.77 ± 0.031.27 ± 0.02-----
 Uey tributaries7IF5.7 ± 0.213.3 ± 26.0508 ± 19118.8 ± 14.32.8 ± 1.10.71 ± 0.031.24 ± 0.08-----
 Cuyuni impacted tributaries3ISa, BS, F6.4 ± 0.4155.3 ± 152.0283 ± 293.5 ± 1.31.0 ± 0.40.93 ± 0.201.41 ± 0.08-----

[18] In the Orinoco/Ventuari watershed, levels of DOC and CDOM were higher in the Orinoco main river channel compared to the Ventuari main river channel for both 2006 and 2007 (Table 1 and Figure 2, p < 0.05 for 2006). In analogy with main river channels, the levels in Orinoco tributaries were significantly higher than those in Paragua tributaries for both 2006 and 2007 (p < 0.05). The variability in DOM quantity was larger in tributaries than in mainstreams of the Orinoco and Ventuari rivers for both 2006 and 2007. Similar characteristics have been reported for the Amazon main channel and its major tributaries [Hedges et al., 1994] using chemical characterizations of DOC.

[19] The highest DOC level was observed at the Ichun River in the Paragua/Karun watershed. The levels of DOM in the Paragua main channel and tributaries were generally higher than those in Karun main channel, Karun tributaries, and lower Paragua small tributaries (Table 1 and Figure 2). In the Cuyuni/Uey watersheds, levels of DOC and CDOM in the Uey main channel were higher than those for the Cuyuni main channel (Table 1 and Figure 2), while values for the tributaries were similar between the Uey and Cuyuni system (p > 0.05).

3.2. Qualitative Parameters of DOM and CDOM

[20] The SR parameter has recently been proposed as a proxy for DOM molecular weight, i.e., high and low SR values represent low and high molecular weight DOM, respectively [Helms et al., 2008], and has been successively applied to assess the seasonal changes in CDOM character in an Arctic river [Spencer et al., 2009]. The observed ranges in the SR parameter were 0.61 to 1.06 in this study (Table 1 and Figure 2), and were found to be similar in range to those reported for the CDOM-rich waters from the Great Dismal Swamp [Helms et al., 2008]. These values correspond to DOC as well as CDOM enriched in the high molecular weight (HMW) fraction (>1000 Da) compared to the low molecular weight (LMW) fraction (<1000 Da) [Helms et al., 2008]. Such a molecular weight distribution is similar to that reported for the Amazon River, where 60–85% of DOM was in the high molecular weight fraction [Amon and Benner, 1996].

[21] The fluorescence index has been proposed as a proxy to assess the plankton/microbial versus higher plant/terrigenous contribution to the fulvic acid pool in DOM [McKnight et al., 2001]. High and low values indicate enrichment in microbial and terrestrial origin, respectively. The values of fluorescence index found in this study ranged from 1.06 to 1.50 and generally varied between 1.2 and 1.4 (Table 1 and Figure 2), suggesting a predominantly terrestrial CDOM signal with some contributions of microbial CDOM [McKnight et al., 2001; Jaffé et al., 2008]. A similar range of fluorescence index was reported for the Orinoco river and its tributaries [Battin, 1998].

[22] The a350* indicates the contribution of CDOM to bulk DOM. In the Orinoco/Ventuari watershed, the values of a350* were higher in the Orinoco main channel as compared to the Ventuari main channel (Table 1 and Figure 2, p < 0.05 for 2006). The values in SR and fluorescence index fell within a relatively narrow range, but a350 values were different between main channel of the Orinoco and Ventuari rivers, both for the 2006 and 2007 sampling periods. Such CDOM characteristics suggest that while the abundance of CDOM was significantly different, the chemical characteristics and thus the sources of CDOM, were similar for the main Orinoco and Ventuari rivers. The variability of SR, the fluorescence index, as well as the levels of CDOM in the Orinoco and Ventuari tributaries, however, were larger than those observed for the main river channels. Such larger difference suggests that CDOM in tributaries were more variable compared to main rivers in terms of quality and quantity. However, SR and fluorescence index values tended to be high but a350* to be low in Ventuari tributaries compared to the Orinoco tributaries for 2006 and 2007 (p < 0.05, excluding SR for 2006 (p = 0.10)).

[23] In the Paragua/Karun watershed, the spatial distribution of DOC concentration was similar to that of CDOM. However, a350* values in the flooding lagoons were remarkably lower compared to other water types (Table 1 and Figure 2). The ranges of CDOM level, SR, and fluorescence index did not overlap between the Paragua and Karun main river channels or for their tributaries (Table 1 and Figure 2), pointing to both quantitative and qualitative differences in CDOM between these water types. Excluding Cuyuni polluted tributaries, the highest SR values were found in the Karun tributaries, in small tributaries of the lower Paragua and in flooding lagoons in the Paragua/Karun watershed (Figure 2). The fluorescence index values at those sites were also highest, although similar values were also found in Ventuari tributaries. Lowest fluorescence index and lowest SR values were for the Ichun and Perro de Agua of the Orinoco tributaries, indicating a strong terrestrial source of CDOM.

[24] The ranges of SR and fluorescence index in the Cuyuni/Uey watershed were similar to those found in Orinoco/Ventuari watershed (Table 1 and Figure 2). It should be noted that, in the human impacted Cuyuni tributaries, a350* and fluorescence index values were significantly different (lower and higher, respectively; p < 0.05) and SR values were higher (p > 0.05) compared to those in undisturbed tributaries. Quantitative parameters (DOC and CDOM) were statistically lower (p < 0.05) in impacted areas.

[25] Among the three tropical river systems, a strong linear relationships between DOC and a350 (Figure 3a; R2 = 0.91, p < 0.001, n = 77) was evident. The spatial distributions of SR and fluorescence index, were similar to each other, but were opposite to those of quantitative parameter of DOM, i.e., a350 and DOC (Figure 2). As expected, a positive correlation between SR and fluorescence index (Figure 3g; R2 = 0.40, p < 0.001, n = 77), and a negative correlation between the fluorescence index and DOC (Figure 3c; R2 = 0.40 without Ichun, p < 0.001, n = 76) and SR and DOC (Figure 3d; R2 = 0.35 without Ichun, p < 0.001, n = 76) was evident. The Ichun sample deviated from the relationships between SR and DOC as well as fluorescence index and DOC, but not from between SR and fluorescence index (Figures 3c, 3d, and 3g). Thus, including the Ichun sample, fluorescence index and SR were also positively correlated with DOC on natural log scales (R2 = 0.44, p < 0.001, n = 77 for fluorescence index; R2 = 0.45, p < 0.001, n = 77 for SR). This is likely due to the fact that the Ichun sample has an extremely high DOC value, derived from waters draining thick peat deposits, and is consequently influenced by a different end-member characterized by higher aromaticity and SR values (Table 1 and Figure 2).

Figure 3.

Relationships among DOC and CDOM quantitative and qualitative parameters. (a) DOC versus a350; (b) DOC versus a350*; (c) DOC versus fluorescence index; (d) DOC versus SR; (e) fluorescence index versus a350*; (f) SR versus a350*; (g) fluorescence index versus SR; dashed lines show the 95% predicted intervals.

[26] The distributional pattern of a350* was similar to those of DOC and a350 but opposite to those of fluorescence index and SR among three river systems (Table 1). Linear relationships, however, were not evident between a350* and DOC (Figure 3b; R2 = 0.13, p < 0.001, n = 77; R2 = 0.12 without Ichun, p < 0.001, n = 76) as well as a350* and SR (Figure 3f; R2 = 0.10, p < 0.001, n = 77). On the other hand, a weak negative linear relationship was found between a350* and fluorescence index (Figure 3e; R2 = 0.25, p < 0.001, n = 77).

3.3. PARAFAC Components

[27] In the present study, five fluorescent components were identified by PARAFAC (Figure 4). Spectral characteristic of the components identified in these tropical rivers were similar to those of DOM previously reported in other aquatic environments [Cory and McKnight, 2005; Stedmon and Markager, 2005; Yamashita and Jaffé, 2008; Yamashita et al., 2008].

Figure 4.

Contour plots of five components obtained by PARAFAC. Two gray lines in excitation and emission spectra shows results of split half validation. The source of each component was assigned by comparison of spectral shape with previous studies (i.e., component 1, fulvic acid-type; component 2, humic acid-type; component 3, microbial humic-like; component 4, microbial humic-like; component 5, protein-like).

[28] There were two excitation maxima (<260 nm and 315 nm) observed in the EEM of component 1. This component had emission maximum at 447 nm and was categorized as mixture of the traditional terrestrial humic-like peaks A and C [Coble, 1996]. The spectral features were also similar to reported terrestrial components (component C10 [Cory and McKnight, 2005]; component 1 [Stedmon and Markager, 2005]). Component 2 was also composed of two peaks with excitation maxima at <260 nm and 370 nm at >500 nm emission. This peak has not been traditionally defined [Coble, 1996], but was similar to a terrestrial reduced quinone-like component (component SQ1 [Cory and McKnight, 2005]) as well as a terrestrial humic-like component found in the Florida coastal Everglades (component 2 [Yamashita and Jaffé, 2008]). Components with similar spectral features to components 1 and 2 in the present study have been reported as the most abundant components in fulvic and humic acid fractions extracted from sediments/soils, respectively [Santin et al., 2009].

[29] Two excitation maxima at <260 nm and <315 nm, at 421 nm emission were evident in component 3. These fluorescence characteristics could be categorized as the previously defined marine humic-like peak M [Coble, 1996]. This component was also similar to the microbial oxidized components (Q3) as defined by Cory and McKnight [2005]. The spectral characteristics of component 4 were characterized by peaks at 370 nm and <260 nm excitation with 440 nm emission wavelengths. This component has also not been traditionally defined [Coble, 1996], but partially resembled a reported microbial reduced quinone-like component (component SQ2 [Cory and McKnight, 2005]).

[30] In the present study, we obtained one component (component 5) in the excitation and emission wavelength range where protein-like fluorophores and polyphenols have been previously reported [Coble, 1996; Yamashita and Tanoue, 2003, 2004; Maie et al., 2007, 2008]. A component with similar fluorescence characteristics to component 5 was reported in estuarine areas and suggested to represent freshly produced tryptophan-like DOM (component 4 [Yamashita et al., 2008]).

3.4. Assessing CDOM Quality Through EEM-PARAFAC

[31] To better assess CDOM differences spatially, PCA on the EEM-PARAFAC data set was carried out using the relative abundance of the five PARAFAC components (Table 1 and Figure 5). Figure 5a shows the property-property plots between first and second factor loadings. The fulvic acid-type component (component 1) and humic acid-type component (component 2) concurrently showed positive first factor loadings, but showed negative and positive second factor loadings, respectively. The two microbial humic-like components (components 3 and 4) showed positive second factor loadings. Component 5, the protein-like component, showed negative first and second factor loadings. Thus, PARAFAC-PCA could separate the characteristics of CDOM quality, i.e., source strength of humic-like components versus the contributions of protein-like component in the total fluorescence analyses.

Figure 5.

Results of PARAFAC-PCA. (a) Property-property plots between first and second factor loadings; (b) property-property plots between first and second principal component scores at Orinoco/Ventuari watershed in 2006 (relatively high water revels); (c) property-property plots between first and second principal component scores at Orinoco/Ventuari watershed in 2007 (relatively low water revels); (d) property-property plots between first and second principal component scores at Paragua/Karun watershed.

[32] The first and second principal component scores of the Orinoco/Ventuari watershed samples from the 2006 and 2007 sampling were plotted in Figures 5b and 5c, respectively. Interestingly, the clusters for most sampling sites were quite similar between 2006 (relatively higher water levels) and 2007 (relatively lower water levels), although data for 2007 were more dispersed compared to 2006. In addition, the first and second principal component scores were similar between 2006 and 2007 for Orinoco and Ventuari tributaries (p > 0.05). Data for the main river samples, Orinoco tributaries, and Ventuari tributaries were roughly grouped together. The first and second principal component scores were significantly different between Orinoco and Ventuari tributaries for both 2006 and 2007 (p < 0.01). Samples for the Orinoco tributaries tended to cluster in the region with higher first (approximately 1 to 4) and lower second (−3 to 0.5) principal component scores. In contrast, the data corresponding to the Ventuari tributaries were usually located in the region with lower first (approximately −2 to 1) and higher second (0 to 2) principal component scores. Such distinct distribution patterns may be the results of higher contributions of terrestrial and microbial CDOM in the Orinoco versus Ventuari tributaries, respectively (Table 1 and Figure 5a). The data for the Orinoco and Ventuari main river channels were located in the middle range between their respective tributaries (Figures 5b and 5c), indicating that CDOM in the main rivers were the results of mixture of both sources. PARAFAC-PCA results regarding CDOM source coincided reasonably well with those of the fluorescence index (Figure 2).

[33] The data from various sites in the Paragua/Karun watersheds were also basically clustered together (Figure 5d). The first and second principal component score for the Ichun blackwaters was 3.9 and −2.6, respectively. This feature in the PARAFAC-PCA was likely the result of a particularly high contribution of terrestrial CDOM (Table 1 and Figures 5a and 5d) and is consistent with the lowest value of the fluorescence index (Table 1 and Figure 2). The CDOM characteristics of main Karun river channel, its tributaries and three of four lower Paragua small tributaries showed a high contribution of protein-like component compared to other samples from Paragua/Karun and Orinoco/Ventuari watersheds. In addition, higher contributions of protein-like component in three Karun tributaries compared to the Karun main river channel were evident. The data for the Paragua flooding lagoons display a high contribution of microbial humic-like components. The samples from the main Paragua river channel and its tributaries were clustered between the terrestrial humic-like components and microbial humic-like components, suggesting a mixture of these two CDOM sources.

4. Discussion

4.1. General Characteristics of CDOM in the Tropical Rivers of the Guayana Shield

[34] A strong linear relationship between DOC and a350 among the three tropical river systems (Figure 3a) indicates that the dynamics of CDOM are coupled to those of DOC. On the other hand, the DOC concentration was negatively correlated to fluorescence index and SR but not to a350* (Figures 3b3d). In a previous study, no relationship between DOM quantity and quality (e.g., DOC versus fluorescence index) was observed for a study comparing DOM characteristics of different aquatic ecosystems ranging from tropical to subarctic regions [Jaffé et al., 2008]. This suggests that variable sources and alteration processes might have obscured any broad relationships between quality and quantity of DOM in a data set corresponding to a wide range of the aquatic environments. In the present study, however, the negative relationship between DOM quantity and quality suggests that two end-members primarily control the DOM sources and quantity in tropical river ecosystems, where high DOC levels are coupled to soil/degraded littler layers derived DOM, while lower DOC levels are produced by planktonic sources. This seems to be confirmed by the observed correlation between high values of SR and fluorescence index, which are indicative of a linkage between low molecular weight and microbially derived CDOM (Figures 3c and 3d). Although on a different geological scale, an increase in DOC concentration accompanied with increase in HMW-DOM from high latitude tributaries down to the Amazon River has been reported [Hedges et al., 2000a].

[35] Among the qualitative parameters, the a350* was weakly but linearly correlated to fluorescence index (Figure 3e) but not to SR (Figure 3f). Similarly, a weak linear relationship between SUVA (corresponding to a254*) and fluorescence index was also evident for samples from a wide range of terrestrial aquatic environments [Jaffé et al., 2008], suggesting that negative relationships between a350* and fluorescence index may be ubiquitous in such systems.

[36] On the other hand, a positive linear relationship was evident between fluorescence index and SR (Figure 3g). SR values may also change due to photodegradation and microbial reworking of CDOM [Helms et al., 2008]. These authors reported SR values to increase and decrease thorough photodegradation and microbial reworking, respectively. The increases in SR values during photodegradation have recently been confirmed by Zhang et al. [2009]. In contrast, the fluorescence index reflects CDOM sources [McKnight et al., 2001] and decreases with photodegradation [Cory et al., 2007]. Thus, property-property plots between SR and fluorescence index (Figure 3g) might provide information on both sources and alteration processes of DOM in tropical rivers. The linear relationship (Figure 3g) was composed of samples from the main river channels and their tributaries where residence times (e.g., exposure to sunlight) might be largely different. Thus, if microbial reworking of CDOM decreases and increases the SR and fluorescence index, respectively, only minor effects of microbial reworking on CDOM in these systems would be expected in order to maintain the observed linear relationship. Similarly, if photodegradation was a major controlling factor for determining the CDOM character in these river systems, linear relationships would be less likely to be observed because the photodegradation of CDOM increases and decreases the SR and fluorescence index, respectively. The deep (∼30 m) main river channels and the relatively high CDOM levels suggest limited light penetration and thus reduced photodegradation of the CDOM in the main river channels. The limited light penetration to smaller tributaries due to developed rain forest canopy may also limit the photodegradation of DOM in tributaries. Although correlations between DOM quantity and quality as well as among different quality parameters were observed, these were relatively weak (R2 = 0.35–0.45), suggesting that the linkages between concentration and composition are controlled by a combination of environmental drivers which include diagenetic processing (photodegradation and biodegradation), biological productivity (planktonic primary productivity) and hydrology (end-member mixing).

[37] Several samples with high values of SR and fluorescence index, however, deviated significantly from the 95% predicted intervals of the linear relationship (Figure 3g), and might represent exceptional cases with more significant degradation effects and/or different sources. Two of these samples were from human impacted Cuyuni tributaries, and some from the lower Paragua small tributaries and Karun tributaries which are potentially impacted by illegal gold mining operations. The PARAFAC-PCA for the Paragua/Karun system showed that samples from three of four lower Paragua small tributaries and the Karun tributaries were strongly influenced by protein-like component (Figure 5) possibly as a result of enhanced nutrient inputs and associated plankton productivity. It has been suggested that enrichment of bio-components in CDOM may result in larger spectral slopes at wavelengths shorter than 300 nm [Yamashita and Tanoue, 2009] and protein molecules have a peak around 275 nm in their absorption spectrum [Sarpal et al., 1995]. This suggests that a large contribution of bio-components in CDOM such as protein-like component would enhance the spectral slope at 275–295 nm, and thus, enhance the SR value. Therefore, deviation of these samples might be influenced by higher contributions of protein-like component.

4.2. Potential Effects of Geology and Associated Land Cover on DOM Quantity and Quality

[38] Even though some general trends regarding DOM quality and quantity parameters were determined for several remote tropical rivers and their watersheds (Figure 3), significant variability, especially among tributaries was observed (Table 1 and Figures 2 and 5). Similar observations have previously been reported based on chemical characterizations of DOM [Hedges et al., 1994, 2000a]. For example, the levels of DOC and a350 in Orinoco tributaries were high compared to those of Ventuari tributaries. Qualitative optical parameters (fluorescence index and SR) as well as EEM-PARAFAC characteristics showed varying contributions of terrestrial versus microbial DOM sources for Orinoco and Ventuari tributaries, respectively, suggesting that differences in watershed characteristics control DOM dynamics. As such, Orinoco tributaries are mainly located on sedimentary rocks while the Ventuari tributaries are located on igneo-metamorphic rocks. Similar trends on differences in DOM quantity and quality with such lithological differences were also observed between the Paragua (high levels of DOM with terrestrial characteristics) and the Karun and lower Paragua tributaries (low levels of DOM with enhanced microbial characteristics). While the watershed of the Karun and lower Paragua tributaries are overlaying igneo-metamorphic rocks, the upper Paragua tributaries sampled are on intermediate-to-felsic metavolcanics substrates, which upstream overlay on sedimentary rocks.

[39] The Ichun is the best example of waters derived almost exclusively from sedimentary rocks (Roraima) and associated peat bogs. Other sites in the Orinoco and Paragua tributaries are characterized by their thick detritus layers developed on sedimentary rocks [Huber, 1992]. Thus, the high levels of DOM associated with terrestrial characteristics (Orinoco and Paragua tributaries) are mainly derived from such organic detritus rich layers. On the other hand, microbial characteristics found in Ventuari, Karun, and lower Paragua tributaries might be results of high microbial activity. The soil pH on igneo-metamorphic rock is usually close to neutral value compared to those on sedimentary rocks [Huber, 1992]. Such differences in pH might affect the water quality and microbial activity, and thus, result in the observed differences in DOM quality in streams associated with sedimentary and igneo-metamorphic rocks, respectively.

[40] Samples from Karun tributaries and lower Paragua small tributaries showed considerably high SR values compared to Ventuari tributaries, possibly resulting from high contributions of protein-like component as mentioned above, and may represent a seasonal phenomenon.

[41] It is interesting to note that the values of fluorescence index in the flooding lagoon samples were highest in the Paragua/Karun watersheds, but SR values were relatively low compared to Karun and lower Paragua tributaries (Figure 2). PCA results using EEM-PARAFAC showed that a significant contribution of microbial humic-like components, but not of protein-like component for two flooding lagoon samples (Table 1 and Figure 5). This might be due to coupling of high production and degradation of labile protein-like components in these shallow water environments because of extended water residence times, particularly after the lagoon have been isolated from the river system, as was the case in this study. The low values of a350* in flooding lagoons (Figure 2) also suggest autochthonous production of noncolored DOM in the flooding lagoons.

[42] The Cuyuni/Uey watersheds sampled were associated with a geological setting of intermediate-to-felsic metavolcanics with sporadical granites. Interestingly, the ranges of DOM quantity and quality were intermediate between tributaries on sedimentary rock and igneo-metamorphic rock geology. It should be noted that the Uey main channel showed higher DOC levels enriched in terrestrial humic components compared to those in the Cuyuni main channel (Figure 2). Such differences in DOM quality and quantity may be the result of the upstream geology, i.e., sedimentary rocks and associated vegetation cover of the Uey river. In contrast, DOM quantity and quality for human impacted Cuyuni tributaries showed low levels of DOM with enriched microbial characteristics (Figure 2). These tributaries are affected by gold mining activities which disturb the soil and forest in the watershed and introduce miscellaneous wastewaters into the tributaries. Thus, low level of DOM with microbial characteristics might be due to combination of low inputs of DOM from organic rich soils/sediments and higher microbial activity as a result of wastewater inputs.

4.3. Hydrological Effects on Quality of DOM in the Tropical Rivers

[43] Increases in riverine DOC concentration with increasing water discharge have often been reported [Mayer and Tate, 1983; Boyer et al., 1997; Hood et al., 2006; Neff et al., 2006]. An increment in indices of aromaticity (SUVA) with high discharge was also observed in steams in Oregon and Siberian rivers [Hood et al., 2006; Neff et al., 2006]. Such quantitative and qualitative changes in DOM have been explained by varying hydrological contributions of interstitial and soil water DOC into the streams [Allan, 2007]. In the present study, however, the levels of DOC and a350 in the main Orinoco and Ventuari river channels and their tributaries were similar between periods of higher water (2006) and lower water levels (2007) (Table 1 and Figure 2, p > 0.05 for 2006). It needs to be stressed that the difference between the two sampling periods was, although noticeable (ca. 2 m), not nearly as large as what can be expected between the maximum and minimum of the observed hydrogram for the region [Depetris and Paolini, 1991]. An absence of a clear relationship between DOC concentration and water discharge was also observed for the Orinoco and Amazon rivers and their watersheds [Battin, 1998; Moreira-Turcq et al., 2003a]. Such limited variations in the DOC concentration in tropical rivers was suggested to be due to the buffer capacity of the mature tropical forest soil systems [Lewis et al., 1986]. Even though floodplains are considered to be source of DOC for tropical rivers [Degens et al., 1991], the floodplain drainage back into the Orinoco and Ventuari main river channels should occur during falling waters (ca. October/November) and not during the early part of the wet season (May). However, while initial washout of accumulated soil DOM could affect its dynamics during this time period [Depetris and Paolini, 1991] this phenomenon was not observed in this study.

[44] By contrast, SR values in main Orinoco and Ventuari river channels were larger for the period of lower water levels compared to the period of higher water levels (Table 1 and Figure 2, p < 0.05 for main Orinoco river channels). The increases in SR values due to photodegradation would be accompanied with decreasing of fluorescence index, as mentioned above. The fluorescence indices for the main Orinoco and Ventuari river channels during the period of low water level were lower than those during high water level (Table 1 and Figure 2, p < 0.05 for main Orinoco river channels), suggesting the possibility of higher photodegradation rates of DOM at main Orinoco and Ventuari river channels during the period of lower water levels due to relatively low turbidities (Table 1; 16.0–28.4 and 7.6–19.5 NTU for higher and lower water levels, respectively). However, these main channels are quite deep, and thus, only surface waters would be affected by this process and unlikely exert a significant overall change in the DOM quality.

[45] As mentioned above, SR values could increase potentially through the high contribution of protein-like components. The EEM-PARAFAC clearly shows differences in the compositions of fluorescent components at the Orinoco and Ventuari river systems between the two years (Figure 5). Thus the deviations for the period of lower water levels compared to the period of higher water levels could also result from higher contributions of protein-like component by enhanced plankton primary productivity during the low turbidity period (Table 1 and Figure 5).

[46] The values of SR and fluorescence index for the period of lower water level (2007) showed a wider range compared to those during higher water level (2006) for the Orinoco and Ventuari tributaries (Figure 2). It should be noted that the largest change in PARAFAC-PCA between the two periods was found for the Guapuchi station (Figures 5b and 5c). The fluorescence index and SR at Guapuchi changed from 1.39 to 1.38 and from 0.76 to 0.86 in periods of higher and lower water levels, respectively. Such increase in SR without change in fluorescence index and deviation of PARAFAC-PCA between two periods were consistent with those found in main Orinoco and Ventuari river channels.

[47] The PARAFAC-PCA plots of two other sites of the Orinoco/Ventuari system, namely Perro de Agua and Negro for the period of lower water level, fell into a region with more positive first and negative second principal component scores compared to the higher water level (Figure 5), suggesting relatively higher contributions of terrestrial humic-like components compared to microbial humic-like components during the period of lower water level (Figure 5). This is in agreement with other reports [Battin, 1998] who found lower S values in Orinoco tributaries compared to the main Orinoco River, suggesting a relatively strong contribution of terrestrial sources of CDOM through the tributaries.

[48] Thus, CDOM characteristics in streams could change to enhance either terrestrial or microbial signals by changing hydrological conditions. During the period of lower water levels, enhanced planktonic activity seems to more significantly affect the CDOM quality in main river channels through increased light penetration as a result of lower turbidity, while it enhances terrestrial/soil derived DOM inputs through the tributaries. DOM dynamics within fluvial networks has also been proposed to be strongly linked to residence time in aquatic environments [Battin et al., 2008]. Thus, the enhanced microbial signals found in CDOM in the main river channels during lower water levels compared to those for higher water levels also might be affected by longer residence time. It is important to point out that water discharge (and depth) differences between the highest and lowest point in the hygrogram of these streams is significantly more pronounced than for the two sampling events presented here. Under such extreme condition, DOM dynamics may be quite different to these discussed in this paper.

5. Conclusions

[49] The analytical approach used in the present study, namely a combination of various optical parameters of DOM as well as EEM-PARAFAC, clearly identified the degree of variability in the CDOM characteristics (i.e., composition) as well as in the levels of CDOM in three tropical river systems. Spatial variability was strongly related to the differences in the geological setting and presumably the associated vegetation cover. However, linear relationships between DOC and optical parameters (quantity versus quality) were also evident, indicating that DOM concentrations and their sources/characteristics may be coupled in tropical river systems. As such, (1) high DOC concentration correlated with CDOM characteristics of higher molecular weight associated with terrestrial sources, while low DOC concentrations correlated with CDOM characteristics of lower molecular weight associated with more microbial contributions. (2) Human disturbance of tributaries affected the DOM character shifting it to low concentrations enriched in microbial components. (3) Hydrological changes such as water discharge differences and likely associated floodplain extensions, seem to affect CDOM characteristics in tributaries to a larger extent than the main river channels.

[50] Climate change is expected to further modify the hydrological conditions, vegetation coverage, and biological activity in the tropics. The variability of CDOM characteristics in tropical river systems found in the present study strongly suggest that climatic change may influence the biogeochemistry of DOM in tropical river systems. Thus, the monitoring of DOM quantity and quality may be required to accurately assess changes in biogeochemistry of DOM associated with changes in climate and land use. This work demonstrates the potential of applying optical properties measurements (in contrast to chemical characterizations) as a simple, high-sample throughput monitoring tool to assess long-term, large-scale, biogeochemical changes in river systems.


[51] The authors thank the Cisneros Foundation for financial support, Conservation International, EDELCA, and the staff from the Manaka Camp for logistical support, and O. Pisani for assistance of optical analysis. We are indebted to Rose Cory for her assistance with the PARAFAC training course. Special thanks to Peter Tinoco for his continued support and flexibility throughout this study and two anonymous reviewers and the associate editor for helpful comments and suggestions that improve the quality of this manuscript. Y.Y. thanks the College of Arts and Science at Florida International University for financial support. SERC contribution 463.