Site description

The Ruki River is a large, left-bank, blackwater tributary that joins the Congo River at the city of Mbandaka, approximately 645 km upstream from Kinshasa-Brazzaville (Fig. 1A). The Ruki Basin spans 188,800 km², is relatively flat (min/max: 250/750 masl), and has an average elevation of ~ 400 m asl. It receives an average rainfall of ~ 2000 mm annually (derived from rainfall data detailed below). The land cover within the Ruki Basin is composed of mostly lowland evergreen broadleaf and swamp forests (together comprising 93.6% of the Basin area) and contains no urbanized areas and only minimal deforested and agriculturally converted areas (Fig. 1B). The lowland forests are dominated by stands of Gilbertodendron dewevrei, with other areas populated by semi-deciduous Staudtia stipitata, Polyalthia suavaeoleus, Scorodophloeus zenkeri, Anonidium mannii, and Parinari glaberrimum. The seasonally flooded swamp forests contain Guibourtia demeusei, Mitragyna spp., Symphona globulifera, Entandrophragma palustre, Uapaca heudelotii, Sterculia subviolacea, and Alstonia congensis while the permanently flood swamp forests contain monospecific stands of Raffia palm. The soils of the lowland evergreen and swamp forests are predominantly Ferralsols and Gleysols, respectively (Jones et al. 2013).

Owing to its low gradient, the Ruki River and its network of tributaries are slowly moving and placid. Like much of the interior Congo Basin, these river networks are the primary means of transportation and commerce among the inhabitants. The small upriver towns of Boende and Ingende are serviced by diesel boats, although there is one unmaintained road that traverses the basin.

Discharge and precipitation

Instantaneous river discharge (Q) was measured fortnightly during one hydrologic year (21 November 2019 to 30 November 2020; n = 29) at the narrowest section of the Ruki (~ 430 m width) at the Jardin Botanique d'Eala (0.0698 N, 18.3167E) using a Sontek RiverSurveyor M5 Acoustic Doppler Current Profiler mounted to a pirogue. Reported discharge values represent the average of four transects, the coefficient of variation of which never exceeded 1.3%. Daily water height measurements were measured at noon using a staff gauge installed at the Jardin Botanique d'Eala. A standard power function (Q = a × H^b, Q = discharge, H = stage height, and a and b are constants) was fitted to the stage-discharge data and the resulting rating curve (see results below) was used to interpolate the fortnightly instantaneous measurements and generate a year-long discharge record for the Ruki.

Monthly accumulated precipitation data (half-hourly multi-satellite precipitation estimate with gauge calibration - GPM_3IMERGHH_v06) for the study period was downloaded from the NASA Global Precipitation Measurement repository on the Giovanni hub (giovanni.gsfc.nasa.gov/) and clipped to the Ruki catchment area and averaged using QGIS raster tools.

Sample collection and processing

On the same days and at the same location that discharge was measured, river samples were collected from the center of the channel directly offshore from the Jardin Botanique d'Eala using a motorized pirogue. At each sampling event, water for dissolved constituents was pumped from a depth of 0.5 m through a pre-combusted (450°C, >5 h) 0.7 μm glass fiber filter (Whatman GF/F) using a peristaltic pump into opaque acid-leached HDPE bottles and four 20 mL pre-combusted (450°C, >5 h) glass vials. Using a weighted intake tube and the peristaltic pump, unfiltered water for particulate constituents was pumped from approximately two-thirds depth of the center of the channel (as determined by the ADCP measurement, section 1.3) into two prerinsed 10 L Nalgene bottles. DIC, carbon dioxide (CO₂) and methane (CH₄) were sampled by injecting 6 mL of bubble-free river water taken 0.5 m below the river surface with a syringe into six N₂ flushed, gas-tight exetainers (12 mL, Labco, UK). Prior to sampling and N₂ flushing, three of the exetainers were dosed with 10 μL of 12 M HCl (to shift the pH < 2 and convert all carbonate species into CO₂ for DIC analysis) and all six exetainers were treated with 5 μL of 50% ZnCl₂⁻ (to suppress microbial activity). River water isotopes were sampled with 3 mL Exetainer vials filled to the brim using a syringe. Water pH was measured with a precalibrated portable pH meter (VWR pH 20) at a depth of 30 cm in the river.

In a clean room at the Woodwell Climate Research Center in Mbandaka, filtered samples DOC, DOM, and isotope analyses were acidified to a pH of 2 with 12 M HCl and stored in acid-leached HDPE bottles in a dark, cool place prior to transport to Zurich, Switzerland. Given the high DOC concentrations in the Ruki, samples were kept unfrozen to reduce the loss of DOC via solute exclusion and precipitation during ice formation (Fellman et al. 2008). Unfiltered water samples were agitated to re-suspend sediments and then filtered onto precombusted (450°C, >5 h) and preweighed 0.7 μm Whatmann GF/F filters. The amount of water passed through each filter was determined using a graduated cylinder and recorded. Filters were then sealed in plastic cases and frozen prior to transport. After transport to Zurich, the filters were thawed and dried at 60°C for 48 h.

Carbon, nitrogen, and water analyses

Concentrations of DOC and total dissolved nitrogen (TDN) were measured on the filtered water samples via high-temperature catalytic oxidation on a Shimadzu TOC-L total organic carbon analyzer equipped with a TNM-1 module using established methodology (Drake et al. 2018a). For this study, we report DOM C : N ratios as DOC : TDN, since dissolved organic nitrogen was not derived for all samples but determined to consistently comprise an average of 90% of TDN in preliminary analyses. Thus, our reported C : N ratios are slightly lower than actual DOM C : N ratios but still reflective of relative shifts in OM sources. Stable (δ¹³C) and radiocarbon (¹⁴C contents) compositions of OC were determined via isotope ratio mass spectrometry (0.1‰ precision) and accelerator mass spectrometry, respectively, at the Laboratory of Ion Beam Physics (LIP) using established procedures (Mcintyre et al. 2017). For DOC preparation, 150 mL of acidified DOC samples were concentrated using a freeze drier. Concentrated DOC samples (50–100 μg C) were converted to CO₂ using wet chemical oxidation. Sodium persulfate was then added to each of two 6 mL aliquots of the concentrated water in 12 mL gas-tight exetainer Labco vials. Prior to oxidation, inorganic CO₂ was removed by sparging the vials with He gas. The vials were then heated to 100°C for 1 h to convert the DOC to CO₂. The CO₂ from one exetainer was then analyzed for δ¹³C on a Thermo Scientific Gas Bench II at ETH Zurich. We report ¹³C/¹²C in traditional “delta” notation as δ¹³C = (¹³R_sample/¹³R_VPDB − 1) × 1000‰, where ¹³R is the ¹³C/¹²C ratio, VPDB is the Vienna Pee Dee Belemnite reference standard, and “‰” indicates parts per thousand. The formed CO₂ in the headspace of the second aliquot was analyzed for ¹⁴C activity using a mini carbon dating accelerator mass spectrometer (MICADAS) system equipped with a gas-accepting ion source at the LIP at ETH Zurich (Wacker et al. 2010). The samples have been corrected for constant contamination method (Haghipour et al. 2019) using processed blanks and standards (sucrose: Sigma, δ¹³C = −12.4‰, F_m = 1.053 ± 0.003 and phthalic acid: Sigma, δ¹³C = −33.6‰, F_m < 0.0025). Measured ¹⁴C/¹²C ratios are reported as fraction modern ¹⁴C activity (F_m, as described in Stuiver and Polach 1977). ¹⁴C analyses were prioritized when there was only one aliquot of CO₂ derived from the freeze-dried DOC, hence only five samples were analyzed for ¹³C-DOC.

Dissolved organic matter (DOM) in river samples was isolated for Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) analysis via solid-phase extraction with 100 mg Bond Elut (Agilent Technologies) styrene-divinylbenzene copolymer (PPL) cartridges. Using a sample's DOC concentration, the volume of water used for extraction was adjusted to target 50 μg of C eluted with 1 mL of HPLC-grade methanol into a precombusted glass vial. The molecular composition of acidic species from PPL-extracted DOM was determined by negative-ion micro-electrospray ionization coupled to a 21 Tesla FT-ICR-MS at the National Magnetic Field Laboratory at Florida State University, USA (Smith et al. 2018; Hendrickson et al. 2015). Typical conditions for negative ion formation were: emitter voltage of − 2.4–2.9 kV, S-lens RF level at 45%, and a heated metal capillary temperature of 350°C with a 500 nL min⁻¹ flow rate. Time-domain transients of 3.1 s were conditionally co-added and acquired with the Predator data station, with 100 time-domain acquisitions averaged for all experiments, which were initiated by a TTL trigger from the commercial Thermo data station (Blakney et al. 2011). Mass spectra were phase-corrected and internally calibrated with 10–15 highly abundant homologous series that span the entire molecular weight distribution based on the “walking” calibration method (Xian et al. 2010; Savory et al. 2011). Peaks with signal magnitude greater than six times the baseline root-mean-square (rms) noise at m/z 500 were exported to peak lists and elemental compositions were assigned based on constraints published (Koch et al. 2007; Stubbins et al. 2010) on all signals > 6σ root mean square baseline noise with PetroOrg© software (Corilo 2014). Molecular formulae with elemental combinations of C_4-50H_8-100 N_0-4O_1-25S_0-2 and mass errors < 200 ppb were assigned. The modified aromaticity index (AI_mod) for each neutral elemental composition was calculated following the procedure of Koch and Dittmar (2006, 2016). Formulae were classified into compound classes by their elemental stoichiometries and AI_mod (Šantl-Temkiv et al. 2013). AI_mod values of 0.5 to 0.67 and > 0.67 were classified as aromatic and condensed aromatic structures, respectively. Further compound classes were defined as follows: condensed aromatic = AI_mod > 0.67, polyphenolic = 0.67 ≥ AI_mod > 0.5, unsaturated low oxygen = AI_mod < 0.5, H/C < 1.5, O/C < 0.5; unsaturated high oxygen = AI_mod < 0.5, H/C < 1.5, O/C > 0.5; aliphatics = H/C 1.5–2.0, O/C < 0.9, N = 0; and peptides = H/C 1.5–2.0, O/C < 0.9, N > 0.

Total suspended solids (TSS) concentrations were determined by dividing the weight difference between the pre-weighed filters and the dried filter after filtration by the amount of water filtered. POC and PON concentrations were determined by feeding the whole filter to an elemental analyzer (Automated Nitrogen Carbon Analyzer, SerCon; Crewe, UK).

Mole fractions of DIC, dissolved CO₂, and dissolved CH₄ were measured on equilibrated gas headspaces of the exetainers via gas chromatography (456-GC, Scion Instruments, UK) using a suite of standards that spanned the expected range of concentrations. Dissolved concentrations were calculated using headspace mole fractions and exetainer headspace/water volume ratios according to Henry's law.

All water samples were analyzed for δD and δ¹⁸O values using the high-temperature carbon reduction method by coupling a high-temperature elemental analyzer (TC/EA; Finnigan MAT, Bremen, Germany) to a Delta^plusXP isotope ratio mass spectrometer via a ConFlo III interface (Finnigan MAT; Werner et al. 1999). Post-run off-line calculations like offset-, memory effect- and drift corrections for assigning the final δD and δ¹⁸O-values on the V-SMOW scale were performed according to Werner and Brand (2001). The long-term precision (3 yr) of the quality control lab standard water (Isolab 1, 4 replicates in a measurement sequence) for δD and δ¹⁸O was 0.57‰ and 0.22‰, respectively.

Flux estimates

Daily carbon fluxes (kg C d⁻¹) for all dissolved C species were modeled using instantaneous discharge measurements (m³ s⁻¹) and concentration data input into the FORTRAN Load Estimator (LOADEST) program (Runkel et al. 2004). Daily fluxes were then summed for the study period to generate an annual flux estimate. For each C species, the LOADEST calibration equation was determined using the adjusted maximum likelihood estimate (AMLE) with the regression model number set to default (MODNO = 0), allowing the model to determine the best fitting regression using Akaike Information Criteria. The model output includes the standard error and the standard error of the prediction of the fluxes in addition to the r² of the AMLE, residuals, and the serial correlation of the residuals. These statistics allow for model validation and for confirmation that the residuals are normally distributed according to established protocols (Dornblaser and Striegl 2009).

CO₂ outgassing from the Ruki River water surface to the atmosphere was estimated from fortnightly pCO₂ concentrations. The flux to the atmosphere was calculated as F = k_CO2 ✕ ΔpCO₂, where k_CO2 is the gas-transfer velocity of CO₂ and ΔpCO₂ is the difference between the partial pressure of CO₂ in the water and the atmosphere (ΔpCO₂ = CO_2-river − CO_2-atmosphere). Monthly global average atmospheric pCO₂ concentrations for the study period were obtained from the National Oceanic and Atmospheric Administration's Global Monitoring Program (https://gml.noaa.gov/ccgg/trends/gl_data.html). A commonly used literature-derived standardized k₆₀₀ value of 4.13 m d⁻¹ for rivers > 100 m in width (Mann et al. 2014; Borges et al. 2015) was converted to a gas- and temperature-specific gas transfer velocity (k_CO2) using temperature-dependent Schmidt numbers for CO₂ (Raymond et al. 2012). pCO₂ concentrations measured during midday when in-stream photosynthesis is at its potential maximum and daily pCO₂ concentrations are at their potential minimum have the potential to underestimate daily average concentration (Gómez-Gener et al. 2021), however, previous work in Congo blackwater systems has shown that diel variations in pCO₂ are undetectable because phytoplankton primary production is light limited from high DOC concentrations (Borges et al. 2019, Descy et al. 2017).

Although the direct outgassing of greenhouse gasses from river surfaces to the atmosphere (i.e., vertical fluxes) has become increasingly recognized as an important and often dominant flux relative to lateral export, accurate quantification at the basin scale is still hindered by high spatio-temporal variability and the lack of spatially representative measurements. As a result, we only present local instantaneous fluxes from the water surface and refrain from offering a basin-wide estimate due to the inaccuracy of such estimates upscaled from a single location.

Statistical analyses

FT-ICR MS data analysis for compound class statistics, heteroatomic content, and Spearman's rank correlations were performed with the “fouriertransform” Python package (Hemingway 2017).

Ruki River hydrology

The average width of the Ruki at the study site was 490 m, the average velocity was 0.80 m s⁻¹, and the average depth was 9.5 m. Measured Ruki River water discharge (n = 29) ranged from 2348 to 5515 m³ s⁻¹ and averaged 3778 m³ s⁻¹ for the study period (Table 1). Daily stage height ranged from 0.47 to 3.51 m for the study period, with measured discharges spanning stage heights from 0.55 to 3.44 m (95% of the total range). Measured discharge related well to stage height when modeled with a power function (r² = 0.85, Supporting Information Fig. S1), likely due to slight hysteresis due to flood conditions near the confluence of Ruki and Congo Rivers. As a result, interpolated and modeled discharge values varied slightly from measured values and tended to be slightly lower during peak and low flows and slightly higher during moderate flows.

Total annual discharge for the Ruki was 115 km³ yr⁻¹, which represents 8.6% of the average annual water flux of the Congo River at Kinshasa-Brazzaville (1336 km³ yr⁻¹; Spencer et al. 2016). Like the Congo, the Ruki exhibited a bi-modal hydrograph with a dominant peak in December and a smaller second peak in June (Fig. 2). Intra-annual variability of the Ruki (measured daily max Q/min Q) was 2.35, slightly higher than the Congo average for 1977–2006 (1.94; Spencer et al. 2012). This relatively low seasonal variability in discharge arises due to the regular and high rainfall (1800–2200 mm) in the Central Basin, which dampens extremes in flow (Maurice 1947; Runge 2008).

Indeed, satellite-derived monthly average rainfall on the Ruki Basin area for the study period ranged from 98 to 239 mm, with an average of 162 mm. The total average rainfall for the study period was 1960 mm, which lies in the middle of the range for Central Congo. All values of monthly specific discharge (Q/drainage area) plotted against the corresponding monthly precipitation fell below the 1 : 1 line (Fig. 3), indicating high evapotranspiration rates in the watershed. Over the study period, the Ruki River discharged 31% the amount of water it received as rainfall over the Basin. Peak discharge (~ 75 mm) lagged approximately two and a half months behind peak rainfall (~ 230 mm), providing an average travel time for water in the Basin.

Stable isotopic values (δD and δ¹⁸O) of Ruki River water ranged from − 28.2 to − 0.5‰ and − 5.1 to − 2.1‰, respectively (Table 1). Water isotopic ratios exhibited a strong seasonality (Fig. 2) and varied inversely with water discharge (Supporting Information Fig. S2). The depletion of δD and δ¹⁸O during high flows corresponds to both the higher inputs of precipitation relative to baseflow and to the lower isotopic signature of precipitation during the wet season (basin-integrated average δD of rainfall is − 17.0‰ for October compared to 15.7‰ for July; Terzer-Wassmuth et al. 2021).

Carbon concentration dynamics

DOC concentrations ranged from 12.7 to 30.1 mg L⁻¹ and averaged 21.3 mg L⁻¹ for the study period (Table 1). These concentrations are much higher than those observed in the Congo River (7.8–11.8 mg L⁻¹; Spencer et al. 2016) and other large tropical blackwater rivers such as the Rio Negro (7.1–10.8 mg L⁻¹; Richey et al. 1990; Pérez et al. 2011; Gonsior et al. 2016), Epulu River (5.2–9.0 mg L⁻¹; Spencer et al. 2010), or Caroní River (5.87 mg L⁻¹; Paolini et al. 1987). DOC concentrations exhibited a strong positive correlation with discharge, with a positive b-value (0.69) in the fitted power relationship indicating a mobilization or flushing response (Fig. 4A). No discernable hysteresis was observed in the DOC-discharge relationship, indicating a consistent supply of DOC sources available for mobilization throughout the year.

TSS and POC ranged from 0.68 to 6.89 and 0.39 to 1.61 mg L⁻¹ and averaged 3.68 and 0.88 mg L⁻¹, respectively (Table 1). These TSS concentrations are much lower than those observed in the Congo mainstem (15.5–27.2 mg L⁻¹; Spencer et al. 2016) and large rivers globally (7.0–26,900 mg L⁻¹; Meybeck and Ragu 1997), while the POC concentrations are only slightly lower than in the Congo River (1.01–1.97 mg L⁻¹; Spencer et al. 2016). Compared to the Rio Negro, Ruki TSS concentrations are generally lower while POC concentrations are similar (Negro TSS: 3.7–11 mg L⁻¹ and POC: 0.27–1.1 mg L⁻¹; Moreira-Turcq et al. 2003). The relative proportion of POC to TSS (%POC) in the Ruki was high, averaging 28% and generally exceeding the OC content of sediments within other tropical rivers (Fig. 5). The Ruki thus represents an end-member along the chemical weathering (low TSS and high POC) to mechanical erosion (high TSS and low POC) continuum of rivers, exemplifying a tropical basin with minimal mechanical erosion and very low sediment mobilization (Seyler et al. 2005). Indeed, TSS and POC concentrations were both negatively correlated with discharge (b-values of − 1.03 and − 0.85, respectively), indicating a dilution response to increased flow (Fig. 4B,C).

DIC concentrations ranged from 4.5 to 7.7 mg L⁻¹ and averaged 5.7 mg L⁻¹ for the study period (Table 1). This range of DIC concentrations is higher than average values for the Congo River (3.1 mg L⁻¹; Wang et al. 2013) and an order of magnitude higher than values for the Rio Negro (0.4 ± 0.2; de Fátima et al. 2013). However, like the Congo River, Ruki DIC did not exhibit a significant relationship with discharge; DIC increased slightly with moderate discharge but decreased again at high flows (Fig. 4D). This pattern could be due to a decrease in pH with discharge driven by the increasing concentration of DOC and associated organic acids that have a titrating effect on DIC (Fig. 4D; Wang et al. 2013). The potential effect of DOC on the carbonate equilibrium system is also supported by the strong negative relationship between DOC and pH (r² = 0.61; Supporting Information Fig. S3). At low pH, bicarbonate ions are protonated and the DIC system shifts toward carbonic acid, which equilibrates with pCO₂. Ruki pCO₂ concentrations ranged from 3069 to 9088 ppm by volume (1.34–3.95 mg C L⁻¹) and averaged 6366 ppm (2.77 mg L⁻¹; Table 1). These values compare well to two CO₂ concentrations, 8686 and 6326 ppm, measured previously on the Ruki during the larger (December) and smaller (June) discharge peaks, respectively (Borges et al. 2019). Ruki pCO₂ concentrations from this study are higher than those measured in the Congo River at Kinshasa-Brazzaville (2018–6853 ppm; Wang et al. 2013) and the Rio Negro (1724–4720 ppm; de Fátima et al. 2013) but within the range for tropical rivers of the Congo Basin (2000–16,000 ppm; Mann et al. 2014; Borges et al. 2015) and close to the average concentration reported in African rivers (6415 ppm; Borges et al. 2015). Like DOC, Ruki pCO₂ concentrations exhibited a strong positive relationship with discharge (Fig. 4E). Such a relationship could indicate the increased mobilization of soil or wetland-derived CO₂ to the Ruki with higher flows. It also could arise from the increase in DOC that occurs with discharge, since higher DOC concentrations might fuel higher rates of heterotrophic respiration. However, given the observed pattern in DIC, this strong correlation with discharge might further indicate that it was the DOC titration effect on DIC that increased pCO₂ concentrations at high discharge. Indeed, the coefficient of variation (CV) for DIC was only 16% while pCO₂ had a CV of 27%, slightly higher than that of DOC (21%). These results suggest that pCO₂ would most likely have shown less seasonal variation if DOC concentrations had remained static.

pCH₄ concentrations ranged from 117 to 517 ppm, with an average of 251 ppm for the study period (Table 1). These concentrations compare well and envelope the two previous CH₄ concentrations, 346 and 254 ppm, measured during the larger (December) and smaller (June) discharge peaks, respectively (Borges et al. 2019). The seasonal range of pCH₄ concentrations in the Ruki River from this study corresponds well with the median ranges of pCH₄ measured within large and small rivers of the Congo Basin (~ 100–550 ppm; Borges et al. 2015). pCH₄ showed a moderate positive relationship with discharge (R² = 0.49) and suggests that the fluvial network becomes more hydrologically connected with wetlands at high flow.

Overall, C concentrations in the Ruki River were strongly controlled by hydrology. High flows mobilized plentifully available DOC from the watershed and allow for drainage of CH₄-rich wetlands. In turn, organic acids from these terrestrial or wetland inputs likely exerted a titration effect on DIC, shifting the carbonate equilibrium toward carbonic acid and strongly increasing the concentration of pCO₂. Moreover, these higher DOC inputs may have also fueled more CO₂ production at high discharge. In contrast to the generally high concentrations and enrichment of dissolved C with discharge, POC exhibited a dilution response and indicated the predominance of chemical weathering over physical erosion in this low-lying tropical basin.

Fluxes of carbon

Given the predominantly positive relationships between dissolved C concentrations and discharge, the resultant modeled fluxes tended to track discharge (Fig. 6). Thus, periods of high flow are disproportionately important periods for the export of dissolved C in the Ruki Basin due to high concentrations of C species concurrent with elevated discharge (Figs. 4 and 6 and Table 1). The annual total flux of DOC was 2.48 Tg C yr⁻¹, which equates to a yield of 13 g m⁻² yr⁻¹ (Table 2). This high DOC yield highlights the Ruki as one of the most efficient exporters of DOC globally, registering at 1.5 × the yield of the Rio Negro (8.7 g m⁻² yr⁻¹; Seyler et al. 2005), 1.5 × the Canoni River (9 g m⁻² yr⁻¹; Paolini et al. 1987), 3.9 × the Congo (3.4 g m⁻² yr⁻¹; Spencer et al. 2016), and almost 20 × the Mississippi River (0.7 g m⁻² yr⁻¹; Raymond and Spencer 2015). Moreover, the annual flux of DOC from the Ruki represents ~ 20% of the Congo Basin's total flux (12.48 Tg; Spencer et al. 2016), despite only comprising 5% of the Congo Basin by area. Similarly, DIC and lateral CO₂ fluxes (0.66 and 0.32 Tg, respectively; Table 2) comprised 19% and 16% of the Congo's annual DIC and lateral CO₂ fluxes (3.46 and 2.0 Tg; Wang et al. 2013), once again highlighting the C saturation of the Ruki relative to the Congo River. The downstream flux of CH₄ was considerably lower than CO₂, with only 0.52 Gg of C exported per year as CH₄ (Table 2).

Given the inverse relationship between POC concentration and discharge (Fig. 4C), POC fluxes were relatively invariant across the study year despite the strong fluctuations in discharge (Fig. 6). POC fluxes amounted to only 0.10 Tg yr⁻¹ and represented 5% of the Congo's annual POC flux (1.96 Tg yr⁻¹; Table 2). In summary, the total downstream C flux from the Ruki River for the study period was 3.25 Tg yr⁻¹, of which DOC, DIC, and POC comprised 76.6%, 20.3%, and 3.1%, respectively. These results highlight the dominance of dissolved C species in the Ruki, a typical characteristic of placid, low-elevation, blackwater rivers.

Instantaneous outgassing CO₂ fluxes from our study site on the Ruki ranged from 5.5 to 18.1 g C m⁻² d⁻¹ and averaged 12.4 g C m⁻² d⁻¹. Because CO₂ concentrations increased sharply with discharge, the CO₂ outgassing fluxes also exhibited a strong positive relationship with discharge (Supporting Information Fig. S4). The seasonal range of CO₂ outgassing fluxes for the Ruki were similar to average values of rivers throughout the Congo Basin (13.8–18.2 g C m⁻² d⁻¹; Borges et al. 2015) and forest-dominated rivers (> 100 m in width) in the Republic of Congo (10.3 g C m⁻² d⁻¹; Mann et al. 2014). Outgassing fluxes were higher in the Ruki compared to the Oubangui River (0.3–0.6 g m⁻² d⁻¹; Bouillon et al. 2012), a large right-bank tributary to the Congo that drains predominantly wooded savannah. This difference is likely the result of the Ruki's much higher pCO₂ concentrations that arise from a combination of higher inputs of both soil-derived pCO₂ and organic acidity from inundated lowland forests and greater overall mineralization within the river driven by the high DOC concentrations.

Collectively, these flux results demonstrate the overall importance of the Ruki River as a major transporter of dissolved C in-and-of itself and relative to its areal extent within the Congo Basin. Furthermore, these results highlight the dominant control of hydrology in dissolved C transport and how high flows may indirectly enhance the physical (mobilization), chemical (acid titration of carbonate system), and biological (heterotrophic respiration) transfer of CO₂ to the atmosphere in the Ruki Basin. Future work might endeavor to better partition the relationship between seasonal organic acid inputs and CO₂ outgassing in blackwater and other DOC-rich rivers.

Isotopic and molecular composition of DOC, POC, and DOM

Stable carbon isotopic ratios of DOC and POC ranged from − 31.0 to − 29.7‰ and − 36.3 to − 28.5‰, respectively (Table 1; Fig. 7A). In general, such depleted δ¹³C values indicate that OC in the Ruki River is ubiquitously sourced from biogenic C fixed via the C3 photosynthetic pathway. Given the dominance of C3 plants in the Basin, such a monodominant source of OC in the river is expected. Indeed, the δ¹³C-POC values measured in the Ruki River were generally more depleted than average values for the Congo (− 28 to − 26.4‰; Spencer et al. 2016) and Oubangui (− 30.6 to − 25.8‰; Bouillon et al. 2012), both rivers that contain large proportions of savannah woodlands and C4 vegetation. Despite the dominance of a C3 source, Ruki δ¹³C-POC ratios exhibited slight seasonality, becoming slightly more enriched as discharge increased during the rising limbs before shifting to a more depleted signature at peak flows (Fig. 7A). This pattern suggests that the stock of more processed material in the Ruki Basin, presumably sourced from litter or top soils, is exhausted and replaced by either fresher, less degraded material or autochthonous biomass sourced from lakes that become hydrologically connected to the river system during peak flows.

The radiocarbon contents of DOC in the Ruki River were predominantly modern (F_m > 1.00; Table 1; Fig. 7B), indicating that DOC is sourced from recently photosynthesized vegetation or litter. Similarly, POC radiocarbon values were mostly modern (Fig. 7B), with values indicating slightly older C ages than DOC (Avg. F_m of 0.95; Table 1). The observed pattern of older POC ages relative to DOC is consistent with OC ages in African freshwaters and rivers globally (Marwick et al. 2015). The Ruki's average ¹⁴C-POC content (F_m = 0.95) was similar to the averages from African rivers (F_m = 0.97; Marwick et al. 2015).

Interestingly, two timepoints on the falling limb of the most prominent discharge peak exhibited shifts toward older ¹⁴C contents for both DOC and POC, with F_m values of 0.98 and 0.86 for DOC and 0.94 and 0.88 for POC (Fig. 7B). For DOC, the timing of these older values on the falling limb coincides with the highest proportion of subsurface flow of a typical flood hydrograph (Hursh and Brater 1941). This transition from surface runoff to subsurface flow on the falling limb is also evidenced by the shift to more enriched water isotopic ratios during this period (Fig. 2). If the DOC and H₂O isotopic ratios are the result of aged DOC arriving within subsurface water, they may indicate the presence of aged soil OC or even peat-derived OC in the bulk DOC pool. The dip in POC age occurred 1 month later than DOC (Fig. 7B), coinciding with minimum discharge and thus highest TSS/POC concentrations (Fig. 4). If the similarity in age between the oldest DOC and POC values (F_m of 0.88 and 0.86, respectively) reflects a common source, then the offset between the oldest DOC and POC values is likely explained by the divergent mobilization responses observed between the dissolved and particulate phases (Fig. 4).

The slight increase in C : N ratio (Fig. 7C) of DOM along with the sharp decrease in δ¹³C-POC at this time (Fig. 7A) may suggest a unique source of OC material mixing into the Ruki during the falling limb. Given that the oldest DOC value exhibited a high C : N ratio of DOM (~ 38) and a more depleted δ¹³C signature (− 31.0‰) relative to the range of values (C : N = 25.4–39.0, δ¹³C = −30.7 to − 29.7‰), it is likely that this old OC was sourced from non-degraded peat rather than aged soil OC, which would have a lower C : N ratio as a result of microbial processing and necromass (Boström et al. 2007). Previous studies have also inferred the mobilization of peat-derived OC from the presence of aged material in water draining deeper subsurface flows, however, these studies took place in disturbed rather than pristine peatlands (Moore et al. 2013; Cook et al. 2018). Given the isolated nature of the aged signature in our dataset, the presence of peat OC in the Ruki River during the falling limb needs to be corroborated with additional high-resolution sampling. Nevertheless, the presence of aged material in the Ruki indicates some amount of mobilization of old C stores within the Basin, which is an interesting finding given the generally low level of human impact on an areal basis.

The composition of DOM determined via FT-ICR MS revealed that the proportions of broad compound classes were stable throughout the study period (Supporting Information Fig. S5). This finding indicates a high level of seasonal homogeneity and consistent source of DOM to the River. Ruki DOM was characterized by high proportions of unsaturated phenolic and polyphenolic compounds, which is typical of DOM in natural waters sourced predominantly from terrestrial vegetation (Wagner et al. 2015). Compared to the Congo River, the Ruki had higher oxygen-rich (O/C > 0.5) unsaturated phenolics (55.75% vs 38.4%), lower oxygen-poor (O/C < 0.5) unsaturated phenolics (21.8% vs. 38.4%), lower condensed aromatics (3.2% vs. 4.9%), and lower aliphatics (1.2% vs 3.7%; Kurek et al. 2022). The Ruki also exhibited higher relative abundances of CHON and CHOS (20.1% and 3.1%, respectively) than the Congo (13.9% and 2.4%, respectively; Kurek et al. 2022). These results suggest that DOM in the Ruki is more processed, oxidized, and heteroatomic, indicating higher relative inputs of degraded OM derived from decomposed litter and N- and S-rich soils. Meanwhile the Congo contains more combustion derived DOM, likely due to inputs from the savanna dominated Kasai River Basin (Drake et al. 2020b).

Despite the observed homogeneity of DOM based on first-order compound classes, it is well known that these broad classifications often fail to capture trends that may exist among individual molecular formulae or groups of formulae that either fall into less abundant compound classes or span multiple compound classes. Using Spearman-rank correlation between discharge (a proxy for seasonality) and the relative abundance of individual molecular formulae to delve beyond the broad compound classes, ~ 31% of all formulae (n = 7775) present in at least half of the samples were found to be significantly (p < 0.05) correlated with discharge (Fig. 8). Among these formulae, two distinct clusters composed primarily of CHO-containing formulae were correlated with high discharge: one low H/C, high O/C group (G1) and a second smaller group with high H/C and low O/C formulae (G2). G1 is indicative of aromatic vascular plant-derived DOM, which is generally low H/C, high O/C (Stubbins et al. 2010; Kellerman et al. 2018; Spencer et al. 2019). Formulae in the same region as G2 (H/C of 1.25–1.5 and O/C of 0.3) are often ascribed to lipid-like compounds (D'Andrilli et al. 2015), which can also be derived from terrestrial vegetation. Previous studies that examined DOM composition across multiple tropical catchments have shown similar groups of formulae to be correlated with the extent of forests within a catchment (Drake et al. 2019, 2020a; Spencer et al. 2019).

The spearman-rank correlation also revealed a heteroatomic (i.e., containing N or S) cluster with elevated H/C and moderate O/C ratios (G3, Fig. 8) that was negatively correlated with discharge (blue formulae). This negative correlation of N-containing formulae with discharge corresponds well with the observed decrease in C : N ratios of DOM at low discharge (Fig. 7C). Together, the low C : N ratios and relatively more aliphatic (high H/C) and heteroatomic composition indicates an increasing influence of soil OM, which tends to have low aromaticity and higher N content derived from microbial necromass and processed organic matter (Kögel-Knabner et al. 2000; Wilson and Xenopoulos 2008; Kaiser and Kalbitz 2012; Kellerman et al. 2018). The correlation of the “soil-like” G3 group with baseflow conditions is likely a result of a larger proportion of water flowing through subsurface soils during low flow periods compared to high flow periods. Overall, these spearman-rank correlation results indicate that despite the general seasonal homogeneity and predominance of forest vegetation as a source of DOM in the Ruki River, more nuanced compositional shifts from soil-derived DOM to vascular-plant derived DOM driven by hydrologic flowpaths are still evident.