variations in dissolved greenhouse gases (co in the congo

Biogeosciences 16 3801ndash3834 2019httpsdoiorg105194bg-16-3801-2019copy Author(s) 2019 This work is distributed underthe Creative Commons Attribution 40 License

Variations in dissolved greenhouse gases (CO2 CH4 N2O)in the Congo River network overwhelmingly driven byfluvial-wetland connectivityAlberto V Borges1 Franccedilois Darchambeau1a Thibault Lambert1b Ceacutedric Morana2 George H Allen3Ernest Tambwe4 Alfred Toengaho Sembaito4 Taylor Mambo4 Joseacute Nlandu Wabakhangazi5 Jean-Pierre Descy1Cristian R Teodoru2c and Steven Bouillon2

1Chemical Oceanography Unit University of Liegravege Liegravege Belgium2Department of Earth and Environmental Sciences KU Leuven Leuven Belgium3Department of Geography Texas AampM University College Station Texas USA4Universiteacute de Kisangani Centre de Surveillance de la Biodiversiteacute Kisangani Democratic Republic of the Congo5Congo Atomic Energy Commission Kinshasa Democratic Republic of the Congoapresent address Direction Geacuteneacuterale Opeacuterationnelle Agriculture Ressources Naturelles et EnvironnementService Publique de Wallonie Jambes Belgiumbpresent address University of Lausanne Institute of Earth Surface Dynamics Lausanne Switzerlandcpresent address Eidgenoumlssische Technische Hochschule Zuumlrich Zuumlrich Switzerland

Correspondence Alberto V Borges (albertoborgesuliegebe)

Received 26 February 2019 ndash Discussion started 4 March 2019Revised 30 August 2019 ndash Accepted 15 September 2019 ndash Published 7 October 2019

Abstract We carried out 10 field expeditions between 2010and 2015 in the lowland part of the Congo River networkin the eastern part of the basin (Democratic Republic ofthe Congo) to describe the spatial variations in fluvial dis-solved carbon dioxide (CO2) methane (CH4) and nitrous ox-ide (N2O) concentrations We investigate the possible driversof the spatial variations in dissolved CO2 CH4 and N2O con-centrations by analyzing covariations with several other bio-geochemical variables aquatic metabolic processes (primaryproduction and respiration) catchment characteristics (landcover) and wetland spatial distributions We test the hypoth-esis that spatial patterns of CO2 CH4 and N2O are partlydue to the connectivity with wetlands in particular with agiant wetland of flooded forest in the core of the Congobasin the ldquoCuvette Centrale Congolaiserdquo (CCC) Two tran-sects of 1650 km were carried out from the city of Kisan-gani to the city of Kinshasa along the longest possible nav-igable section of the river and corresponding to 41 of thetotal length of the main stem Additionally three time se-ries of CH4 and N2O were obtained at fixed points in themain stem of the middle Congo (2013ndash2018 biweekly sam-pling) in the main stem of the lower Kasaiuml (2015ndash2017

monthly sampling) and in the main stem of the middleOubangui (2010ndash2012 biweekly sampling) The variationsin dissolved N2O concentrations were modest with valuesoscillating around the concentration corresponding to satura-tion with the atmosphere with N2O saturation level (N2Owhere atmospheric equilibrium corresponds to 100 ) rang-ing between 0 and 561 (average 142 ) The relativelynarrow range of N2O variations was consistent with lowNH+4 (23plusmn13 micromol Lminus1) and NOminus3 (56plusmn51 micromol Lminus1) lev-els in these near pristine rivers and streams with low agri-culture pressure on the catchment (croplands correspond to01 of catchment land cover of sampled rivers) domi-nated by forests (sim 70 of land cover) The covariations inN2O NH+4 NOminus3 and dissolved oxygen saturation level(O2) indicate N2O removal by soil or sedimentary den-itrification in low O2 high NH+4 and low NOminus3 environ-ments (typically small and organic matter rich streams) andN2O production by nitrification in high O2 low NH+4 andhigh NOminus3 (typical of larger rivers that are poor in organicmatter) Surface waters were very strongly oversaturatedin CO2 and CH4 with respect to atmospheric equilibriumwith values of the partial pressure of CO2 (pCO2) rang-

Published by Copernicus Publications on behalf of the European Geosciences Union

3802 A V Borges et al Dissolved greenhouse gases in the Congo River

ing between 1087 and 22 899 ppm (equilibrium sim 400 ppm)and dissolved CH4 concentrations ranging between 22 and71 428 nmol Lminus1 (equilibrium sim 2 nmol Lminus1) Spatial vari-ations were overwhelmingly more important than seasonalvariations for pCO2 CH4 and N2O as well as dayndashnightvariations for pCO2 The wide range of pCO2 and CH4 vari-ations was consistent with the equally wide range of O2(03 ndash1228 ) and of dissolved organic carbon (DOC)(18ndash678 mg Lminus1) indicative of generation of these twogreenhouse gases from intense processing of organic mat-ter either in ldquoterra firmerdquo soils wetlands or in-stream How-ever the emission rate of CO2 to the atmosphere from river-ine surface waters was on average about 10 times higherthan the flux of CO2 produced by aquatic net heterotrophy(as evaluated from measurements of pelagic respiration andprimary production) This indicates that the CO2 emissionsfrom the river network were sustained by lateral inputs ofCO2 (either from terra firme or from wetlands) The pCO2and CH4 values decreased and O2 increased with increas-ing Strahler order showing that stream size explains part ofthe spatial variability of these quantities In addition severallines of evidence indicate that lateral inputs of carbon fromwetlands (flooded forest and aquatic macrophytes) were ofparamount importance in sustaining high CO2 and CH4 con-centrations in the Congo river network as well as drivingspatial variations the rivers draining the CCC were charac-terized by significantly higher pCO2 and CH4 and signifi-cantly lower O2 and N2O values than those not drain-ing the CCC pCO2 and O2 values were correlated tothe coverage of flooded forest on the catchment The fluxof greenhouse gases (GHGs) between rivers and the atmo-sphere averaged 2469 mmol mminus2 dminus1 for CO2 (range 86 and7110 mmol mminus2 dminus1) 12 553 micromol mminus2 dminus1 for CH4 (range65 and 597 260 micromol mminus2 dminus1) and 22 micromol mminus2 dminus1 forN2O (rangeminus52 and 319 micromol mminus2 dminus1) The estimate of in-tegrated CO2 emission from the Congo River network (251plusmn46 TgC (1012 gC) yrminus1) corresponding to nearly half theCO2 emissions from tropical oceans globally (565 TgC yrminus1)and was nearly 2 times the CO2 emissions from the tropi-cal Atlantic Ocean (137 TgC yrminus1) Moreover the integratedCO2 emission from the Congo River network is more than3 times higher than the estimate of terrestrial net ecosys-tem exchange (NEE) on the whole catchment (77 TgC yrminus1)This shows that it is unlikely that the CO2 emissions from theriver network were sustained by the hydrological carbon ex-port from terra firme soils (typically very small compared toterrestrial NEE) but most likely to a large extent they weresustained by wetlands (with a much higher hydrological con-nectivity with rivers and streams)

1 Introduction

Emissions to the atmosphere of greenhouse gases (GHGs)such as carbon dioxide (CO2) methane (CH4) and nitrousoxide (N2O) from inland waters (rivers lakes and reser-voirs) might be quantitatively important for global budgets(Seitzinger and Kroeze 1998 Cole et al 2007 Bastviken etal 2011) Yet there are very large uncertainties in the esti-mates of GHGs emission to the atmosphere from rivers asreflected in the wide range of reported values between 02and 18 PgC (1015 gC) yrminus1 for CO2 (Cole et al 2007 Ray-mond et al 2013) 2 and 27 TgCH4 (1012 gCH4) yrminus1 forCH4 (Bastviken et al 2011 Stanley et al 2016) and 32and 2100 GgN2O-N (109 gN2O-N) yrminus1 for N2O (Kroeze etal 2010 Hu et al 2016) This uncertainty is mainly due tothe scarcity of data in the tropics that account for the major-ity of riverine GHG emissions (sim 80 for CO2 Raymondet al 2013 Borges et al 2015a b Lauerwald et al 201579 for N2O Hu et al 2016 70 for CH4 Sawakushiet al 2014) but also to scaling procedures of varying com-plexity that use different input data (Raymond et al 2013Borges et al 2015b Lauerwald et al 2015) in addition touncertainty in the estimate of surface area of rivers (Down-ing et al 2012 Raymond et al 2013 Allen and Pavelsky2018) and the parameterization of the gas transfer velocity(k) (Raymond et al 2012 Huotari et al 20013 Maurice etal 2017 Kokic et al 2018 McDowell and Johnson 2018Ulseth et al 2019) The exchange of CO2 between rivers andthe atmosphere is in most cases computed from the airndashwatergradient of the CO2 concentration and k parameterized as afunction of stream morphology (eg slope or depth) and wa-ter flow (or discharge) However there can be large errors as-sociated with the computation of the dissolved CO2 concen-tration from pH and total alkalinity (TA) for low alkalinitywaters in particular so that high-quality direct measurementsof dissolved CO2 concentration are preferred (Abril et al2015) although scarce In the tropics research on GHGs inrivers has mainly focussed on South American rivers and onthe central Amazon in particular (Richey et al 1988 2002Melack et al 2004 Abril et al 2014 Barbosa et al 2016Scofield et al 2016) while until recently African rivers werenearly uncharted with a few exceptions (Koneacute et al 20092010)

There is also a lack of understanding of the drivers ofthe fluvial concentrations of GHGs hence ultimately ofthe drivers of their exchange with the atmosphere It is un-clear to what extent CO2 emissions from rivers are sus-tained by in situ net heterotrophy andor by lateral inputsof CO2 The global organic carbon degradation by net het-erotrophy of rivers and streams given by Battin et al (2008)of 02 PgC yrminus1 is insufficient to sustain global riverineCO2 emissions given by the most recent estimates of 07ndash18 PgC yrminus1 (Raymond et al 2013 Lauerwald et al 2015)suggesting an important role of lateral CO2 inputs in sustain-ing emissions to the atmosphere from rivers In a regional

Biogeosciences 16 3801ndash3834 2019 wwwbiogeosciencesnet1638012019

A V Borges et al Dissolved greenhouse gases in the Congo River 3803

study in the US Hotchkiss et al (2015) estimated that in-stream organic matter degradation could only sustain 14 and 39 of CO2 riverine emissions in small and large sys-tems respectively It is also unclear to what extent CO2 andCH4 emissions from rivers are sustained by carbon inputs ei-ther from the terrestrial biome (terra firme) or from wetlands(flooded forests and macrophytes) (Abril and Borges 2019)This difference has large implications for our fundamentalunderstanding of carbon cycling in rivers and their connectiv-ity with respective catchments but also consequently for ourcapacity to predict how GHG emissions from rivers might bemodified in response to climate change (warming and modifi-cation of the hydrological cycle) water diversion (dammingwater abstraction) or land use change on the catchment (egKlaus et al 2018) In upland areas low-order stream CO2emissions are undoubtedly related to soil-water and ground-water CO2 inputs although CO2 degassing takes place oververy short distances from point sources (le 200 m) is highlyvariable in space and seasonally and mainly occurs duringshort-lived high-flow events that promote shallow flow paths(eg Duvert et al 2018) Low-order streams (1ndash3) mightaccount for about one-third of the global riverine CO2 emis-sions (Marx et al 2017) Recently acquired datasets allowedus to show that inputs from riparian wetlands seem to be ofparamount importance in sustaining CO2 and CH4 emissionsto the atmosphere from large tropical lowland rivers (Abril etal 2014 Borges et al 2015a b) About half of the globalsurface area of wetlands is located in the tropics and subtrop-ics (33 Nndash33 S) the rest are in the Northern Hemisphere(Fluet-Chouinard et al 2015) and more than half of riversurface area is located in the tropics and subtropics (Allenand Pavelsky 2018)

The Congo River (sim 4400 km length freshwater dischargesim 44000 m3 sminus1) has a large drainage basin (37times 106 km2)covered by evergreen forest (dense and mosaic) (sim 67 ) andsavannah (shrubland and grassland) (sim 30 ) owing to thetropical climate (annual average precipitation of 1530 mmand air temperature of 237 C) The Congo basin accountsfor 89 of African rainforests These rainforests are spreadbetween the Democratic Republic of the Congo (54 )Gabon (11 ) Cameroon (10 ) and the Republic of theCongo (10 ) the remaining 15 being shared by severalother countries The mean above-ground biomass of the rain-forests in Central Africa (43 kg dry biomass (db) mminus2) ismuch higher than the mean in Amazonia (29 kgdb mminus2) andnearly equals the mean in the notorious high biomass forestsof Borneo (44 kgdb mminus2) (Malhi et al 2013) Current es-timates of carbon transport from the Congo River close tothe mouth rank it as the Earthrsquos second-largest supplier oforganic carbon to the oceans (Coynel et al 2005) Despiteits overwhelming importance our knowledge of carbon andnutrient cycling in the Congo river basin is limited to sometransport flux data from the 1980s reviewed by Laraque etal (2009) and a number of more recent small-scale studies(Bouillon et al 2012 2014 Spencer et al 2012 Lambert

et al 2016) in sharp contrast to the extensive and sustainedwork that has been done on the Amazon river basin (Alsdorfet al 2016) The Congo basin has a wide range of contrastingtributaries (differing in lithology soil characteristics vegeta-tion rainfall patterns etc) and extensive flooded forests inits central region the ldquoCuvette Centrale Congolaiserdquo (CCC)with an estimated flooded cover of 360 000 km2 for a totalsurface area of 1 760 000 km2 (Bwangoy et al 2010) Exten-sive peat deposits are present beneath the swamp forest ofthe CCC that store below-ground 31 PgC of organic carbona quantity similar to the above-ground carbon stock of theforests of the entire Congo basin (Dargie et al 2017) Thetributaries partly drain semi-humid catchments with alternat-ing dry and wet seasons on both sides of the Equator result-ing in a relatively constant discharge and water level for themain stem Congo River (Runge 2008) Hence the CCC isan extended zone of year-round inundation (Bwangoy et al2010) in sharp contrast with other large tropical rivers suchas the Amazon where floodplain inundation shows clear sea-sonality (Hamilton et al 2002)

Data of dissolved GHGs (CH4 N2O and CO2) have beenreported in rivers in the western part of the Congo basin infour major river basins in the Republic of the Congo (Al-ima Lefini Sangha Likouala-Mossaka) (Mann et al 2014Upstill-Goddard et al 2017) and we previously reportedGHGs data collected in the eastern part of the basin in theDemocratic Republic of the Congo in the framework of abroad synthesis of riverine GHGs at the scale of the Africancontinent (Borges et al 2015a) and a general comparisonof the Congo and the Amazon rivers (Borges et al 2015b)Here we describe in more detail the variability of GHGsbased on a dataset collected during 10 field expeditions from2010 to 2015 (Figs 1 and 2) (Borges and Bouillon 2019)in particular with regards to spatial and seasonal patternsas well as with regards to the origin of fluvial CO2 by in-tegrating metabolic measurements (primary production andrespiration) stable isotope ratios of dissolved inorganic car-bon (DIC) and characteristics of the catchments with regardsto the cover of wetlands Comparison of data obtained instreams within and outside the giant wetland area of the CCCallows a natural large-scale test of the influence of the con-nectivity of wetlands on CO2 and CH4 dynamics in lowlandtropical rivers

2 Material and methods

21 Field expeditions and fixed monitoring

Samples were collected during a total of 10 field expedi-tions (Figs 1 and 2) Three were done from a mediumsized boat (22 m long) on which we deployed the equip-ment for continuous measurements of the partial pressureof CO2 (pCO2) (total n= 30490) as well as the field lab-oratory for conditioning water samples Sampling in the

wwwbiogeosciencesnet1638012019 Biogeosciences 16 3801ndash3834 2019


Figure 1 Freshwater discharge (m3 sminus1) of the Congo River at Kin-shasa from 2010 to 2015 with an indication of field expedition du-ration (thick lines)

main stem and large tributaries was made from the mediumsized boat while sampling in smaller tributaries was madewith a pirogue These large-scale field expeditions cov-ered the KisanganindashKinshasa transect twice (3ndash19 Decem-ber 2013 and 10ndash30 June 2014) and the Kwa river up to Ilebo(16 Aprilndash6 May 2015) During the other cruises the fieldlaboratory was installed in a base camp (typically in a vil-lage along the river) and traveling and sampling was madewith small pirogues on which it was not possible to deploythe apparatus for continuous measurements of pCO2 Threecruises covered the section from Kisangani to the mouth ofthe Lomami River (20 Novemberndash8 December 2012 17ndash26 September 2013 13mdash21 March 2014) one cruise cov-ered the section from Kisangani to the mouth of the ItimbiriRiver (8 Mayndash6 June 2010) and three cruises (previously re-ported by Bouillon et al 2012 2014) covered the Oubanguiriver network around the city of Bangui (21ndash23 March 201020ndash23 March 2011 and 20ndash24 November 2012)

Fixed sampling was carried out in the Congo main stemin proximity of the city of Kisangani (10 December 2012ndash16 April 2018) the Oubangui main stem in proximity ofthe city of Bangui (20 March 2010ndash31 March 2012) and theKasaiuml main stem in proximity of the village of Dima closeto the city of Bandundu (14 April 2015ndash15 May 2017) Sam-pling was carried out at regular intervals every 15 d in theCongo and Oubangui main stems and every month in theKasaiuml main stem Data from the Oubangui catchment werepreviously reported by Bouillon et al (2012 2014)

22 Continuous measurements

Continuous measurements (1 min interval) of pCO2 weremade with an equilibrator designed for turbid waters(Frankignoulle et al 2001) coupled to a nondispersive infra-red gas analyzer (IRGA) (Li-Cor 840) The equilibrator con-sisted of a Plexiglas cylinder (height of 70 cm and internaldiameter of 7 cm) filled with glass marbles pumped wa-ter flowed from the top to the bottom of the equilibrator

at a rate of about 3 L minminus1 water residence time withinthe equilibrator was about 10 s while 99 of equilibrationwas achieved in less than 120 s (Frankignoulle et al 2001)This type of equilibrator system was shown to be the fastestamong commonly used equilibration designs (Santos et al2012) In parallel to the pCO2 measurements water temper-ature specific conductivity pH dissolved oxygen saturationlevel (O2) and turbidity were measured with an YSI mul-tiparameter probe (6600) and position was measured witha Garmin geographical position system (Map 60S) portableprobe Water was pumped to the equilibrator and the multi-parameter probe (on deck) with a 12V powered water pump(LVM105) attached on a wooden pole to the side of the boatat about 1 m depth We did not observe bubble entrainmentin water circuit to the equilibrator as the sailing speed waslow (maximum speed 12 km hminus1) Instrumentation was pow-ered by 12 V batteries that were recharged in the evening withpower generators

23 Discrete sampling

In smaller streams sampling was done from the side of apirogue with a 17 L Niskin bottle (General Oceanics) forgases (CO2 CH4 N2O) and a 5 L polyethylene water con-tainer for other variables Water temperature specific con-ductivity pH and O2 were measured in situ with a YSImultiparameter probe (ProPlus) pCO2 was measured with aLi-Cor Li-840 IRGA based on the headspace technique withfour polypropylene syringes (Abril et al 2015) During twocruises (20 Novemberndash8 December 2012 17ndash26 Septem-ber 2013 n= 38) a PP Systems EGM-4 was used as anIRGA instead of the Li-Cor Li-840 During one of the cruises(13ndash21 March 2014 n= 20) the equilibrated headspace wasstored in pre-evacuated 12 mL Exetainer (Labco) vials andanalyzed in our home laboratory with a gas chromatograph(GC) (see below) Similarly the equilibrated headspace wasstored in pre-evacuated 12 mL Exetainer (Labco) vials forpCO2 analysis from the fixed sampling in Kisangani Thisapproach was preferred to the analysis of pCO2 from thesamples for CH4 and N2O analysis as the addition of HgCl2to preserve the water sample from biological alteration ledto an artificial increase in CO2 concentrations most probablyrelated to the precipitation of HgCO3 (Supplement Fig S1)

Both YSI multiparameter probes were calibrated accord-ing to manufacturerrsquos specifications in air for O2 and withstandard solutions for other variables commercial pH buffers(400 and 700) a 1000 microS cmminus1 standard for conductivityand a 124 nephelometric turbidity unit (NTU) standard forturbidity The turbidity data from the YSI 6600 compared sat-isfactorily with discrete total suspended matter (TSM) mea-surements (Fig S2) so hereafter we will refer to TSM forboth discrete samples and sensor data The Li-Cor 840 IR-GAs were calibrated before and after each cruise with ultra-pure N2 and a suite of gas standards (Air Liquide Belgium)



Figure 2 Map showing the sampling stations of the 10 field expeditions in the Congo River network

with CO2 mixing ratios of 388 813 3788 and 8300 ppm Theoverall precision of pCO2 measurements was plusmn20

24 Metabolism measurements

Primary production (PP) was measured during 2 h incuba-tions along a gradient of light intensity using 13C-HCOminus3 asa tracer as described in detail by Descy et al (2017) Datawere integrated vertically with PAR profiles made with a Li-Cor Li-193 underwater spherical sensor and at daily scalewith surface PAR data measured during the cruises with a Li-190R quantum sensor In order to extend the number of PPdata points we developed a very simple model as a functionof chlorophyll a concentration (Chl a) and of Secchi depth(Sd)

PP=minus44+ 0166times Sd+ 3751timesChl a (1)

where PP is in mmol mminus2 dminus1 Sd in cm and Chl a in microg Lminus1

This approach is inspired from empirical models such asthe one developed by Cole and Cloern (1987) that accountsfor phytoplankton biomass given by Chl a light extinctiongiven by the photic depth and daily surface irradiance Weuse Sd as a proxy for photic depth and in order to simplify

the computations we did not include in the model daily sur-face irradiance since it is nearly constant year round in ourstudy region close to the Equator The model satisfactorilypredicts the PP (Fig S3) except at very low Chl a values atwhich the model overestimates PP (due to the Sd term) In or-der to overcome this we assumed a zero PP value for Chl aconcentrations lt 03 microg Lminus1

Pelagic community respiration (CR) was determined fromthe decrease in O2 in 60 mL biological oxygen demand bot-tles (Wheaton) over sim 24 h incubation periods The bottleswere kept in the dark and close to in situ temperature in acooler filled with in situ water The O2 decrease was deter-mined from triplicate measurements at the start and the endof the incubation with an optical O2 probe (YSI ProODO)At the end of incubation the sealed samples were homog-enized with a magnetic bar prior to the measurement of O2Depth integration was made by multiplying the CR in surfacewaters by the depth measured at the station with a portabledepth meter (Plastimo Echotest-II)

For methane oxidation measurements seven 60 mLborosilicate serum bottles (Wheaton) were filled sequentiallyfrom the Niskin bottle and immediately sealed with butyl



stoppers and crimped with aluminum caps The butyl stop-pers were cleaned of leachable chemicals by boiling in deion-ized water during 15 min in our home laboratory The firstand the last bottle to be filled were then poisoned with a sat-urated solution of HgCl2 (100 microL) injected through the butylstopper with a polypropylene syringe and a steel needle cor-responding to the initial concentration of the incubation (T0)The other bottles were stored in a cooler full of in situ wa-ter (to keep samples close to in situ temperature and in thedark) and were poisoned approximately 12 h after T0 (T1)and then approximately every 24 h after T0 (T2 T3 T4 andT5) The difference between the two T0 samples was close tothe typical precision of measurements showing that the wa-ter was homogeneous within the Niskin bottle with regardsto CH4 concentration and no measurable loss of CH4 oc-curred when filling the seven vials Methane oxidation wascomputed from the linear decrease in CH4 concentration withtime

25 Sample conditioning and laboratory analysis

Samples for CH4 and N2O were collected from the Niskinbottle with a silicone tube in 60 mL borosilicate serum bot-tles (Wheaton) poisoned with 200 microL of a saturated solu-tion of HgCl2 and sealed with a butyl stopper and crimpedwith aluminum cap Measurements were made with theheadspace technique (Weiss 1981) and a GC (SRI 8610C)with a flame ionization detector for CH4 (with a methanizerfor CO2) and electron capture detector for N2O calibratedwith CO2 CH4 N2O N2 gas mixtures (Air Liquide Bel-gium) with mixing ratios of 1 10 and 30 ppm for CH4404 1018 and 3961 ppm for CO2 and 02 20 and 60 ppmfor N2O The precision of measurements based on dupli-cate samples was plusmn39 for CH4 and plusmn32 for N2OThe CO2 concentration is expressed as partial pressure inparts per million (ppm) and CH4 as dissolved concentra-tion (nmol Lminus1) in accordance with convention in exist-ing topical literature and because both quantities were sys-tematically and distinctly above saturation level (400 ppmand 2ndash3 nmol Lminus1 respectively) Variations in N2O weremodest and concentrations fluctuated around atmosphericequilibrium so data are presented as percent of saturationlevel (N2O where atmospheric equilibrium corresponds to100 ) computed from the global mean N2O air mixing ra-tios given by the Global Monitoring Division (GMD) of theEarth System Research Laboratory (ESRL) of the NationalOceanic and Atmospheric Administration (NOAA) (httpswwwesrlnoaagovgmdhatscombinedN2Ohtml last ac-cess 15 August 2018) and using the Henryrsquos constant givenby Weiss and Price (1980)

The flux (F ) of CO2 (FCO2) CH4 (FCH4) and N2O(FN2O) between surface waters and the atmosphere wascomputed according to Liss and Slater (1974)

F = k1G (2)

where k is the gas transfer velocity (cm hminus1) and 1G is theairndashwater gradient of a given gas

Atmospheric pCO2 data from Mount Kenya station(NOAA ESRL GMD) and a constant atmospheric CH4 par-tial pressure of 2 ppm were used Atmospheric mixing ratiosgiven in dry air were converted to water-vapor-saturated airusing the water vapor formulation of Weiss and Price (1982)as a function of salinity and water temperature The k nor-malized to a Schmidt number of 600 (k600 in cm hminus1)were derived from the parameterization as a function slopeand stream water velocity given by Eq 5 of Raymond etal (2012)

k600 = 842+ 11838timesV S (3)

where V is stream velocity (m sminus1) and S is slope (unitless)We chose this parameterization because it is based on the

most exhaustive compilation to date of k values in streamsderived from tracer experiments and was used in the globalriverine CO2 efflux estimates of Raymond et al (2013) andLauerwald et al (2015) Values of V and S were derivedfrom a geographical information system (GIS) as describedin Sect 26

During the June 2014 field expedition samples for the sta-ble isotope composition of CH4 (δ13C-CH4) were collectedand preserved similarly to those described above for the CH4concentration The δ13C-CH4 was determined with a customdeveloped interface whereby a 20 mL He headspace was firstcreated and CH4 was flushed out through a double-hole nee-dle non-CH4 volatile organic compounds were trapped inliquid N2 CO2 was removed with a soda lime trap H2Owas removed with a magnesium perchlorate trap and theCH4 was quantitatively oxidized to CO2 in an online com-bustion column similar to that of an elemental analyzerThe resulting CO2 was subsequently pre-concentrated byimmersion of a stainless steel loop in liquid N2 passedthrough a micropacked GC column (Restek HayeSep Q 2 mlength 075 mm internal diameter) and finally measured ona Thermo DeltaV Advantage isotope ratio mass spectrome-ter (IRMS) Calibration was performed with CO2 generatedfrom certified reference standards (IAEA-CO-1 or NBS-19and LSVEC) and injected in the line after the CO2 trap Re-producibility of measurements based on duplicate injectionsof samples was typically better than plusmn05 permil

The fraction of CH4 removed by methane oxidation (Fox)was calculated with a closed-system Rayleigh fractionationmodel (Liptay et al 1998) according to

ln(1minusFox)=[ln(δ13CminusCH4_initial+ 1000

)minus ln

(δ13CminusCH4+ 1000

)][αminus 1] (4)

where δ13CminusCH4_initial is the signature of dissolved CH4 asproduced by methanogenesis in sediments δ13CminusCH4 is thesignature of dissolved CH4 in situ and α is the fractionationfactor



We used a value of 102 for α based on field measurementsin a tropical lake (Morana et al 2015) For δ13CminusCH4_initialwe used a value of minus602 permil which we measured in bubblesfrom the sediment trapped with a funnel on one occasion ina river dominated by Vossia cuspidata wetlands (16 Aprilndash6 May 2015) The model we used to compute Fox applies forclosed systems implying that CH4 is assumed not to be ex-changed with surroundings in contrast to open-system mod-els Running river water corresponds to a system that is in-termediate between closed and open chemical systems sinceit is open to the atmosphere and the sediments but on theother hand the water parcel can be partly viewed as a closedsystem being transported downstream with the flow As suchthe water parcel receives a certain amount of CH4 from sedi-ments and then is transported downstream away from the ini-tial input of CH4 We also applied two common open-systemmodels to estimate Fox given by Happel et al (1994) and byTyler et al (1997) that have also been applied in lake systems(Bastviken et al 2002) However both open-system modelsgave Fox values gt 1 in many cases (data not shown) since thedifference between δ13C of the CH4 source and measuredδ13C in dissolved CH4 was often much higher than expectedfrom the assumed isotopic fractionation (102) The same ob-servation (Fox gt 1) was also reported with open-system mod-els in the lakes studied by Bastviken et al (2002) Since Foxvalues gt 1 are not conceptually possible we preferred to usethe results from the closed-system model instead althoughwe acknowledge that flowing waters are in fact intermediarysystems between closed and open and that consequently thecomputed Fox values are likely underestimated

Samples for the stable isotope composition of DIC (δ13C-DIC) were collected from the Niskin bottle with a siliconetube in 12 mL Exetainer vials (Labco) and poisoned with50 microL of a saturated solution of HgCl2 Prior to the analy-sis of δ13C-DIC a 2 mL helium headspace was created and100 microL of phosphoric acid (H3PO4 99 ) was added in thevial in order to convert CO2minus

3 and HCOminus3 to CO2 Afterovernight equilibration up to 1 mL of the headspace was in-jected with a gastight syringe into a coupled elemental an-alyzer ndash IRMS (EA-IRMS Thermo FlashHT or Carlo ErbaEA1110 with DeltaV Advantage) The obtained data werecorrected for isotopic equilibration between dissolved andgaseous CO2 as described by Gillikin and Bouillon (2007)Calibration was performed with certified standards (NBS-19or IAEA-CO-1 and LSVEC) Reproducibility of measure-ments based on duplicate injections of samples was typicallybetter than plusmn02 permil

Water was filtered on Whatman glass fiber filters (GFFgrade 07 microm porosity) for TSM (47 mm diameter) partic-ulate organic carbon (POC) and particulate nitrogen (PN)(25 mm diameter) (precombusted at 450 C for 5 h) and Chl a(47 mm diameter) Filters for TSM and POC were stored dryand filters for Chl a were stored frozen at minus20 C Filtersfor POC analysis were decarbonated with HCl fumes for 4 hand dried before encapsulation into silver cups POC and PN

concentration and carbon stable isotope composition (δ13C-POC) were analyzed on an EA-IRMS (Thermo FlashHTwith DeltaV Advantage) with a reproducibility better thanplusmn02 permil for stable isotopic composition and better thanplusmn5 for bulk concentration of POC and PN Data were calibratedwith certified (IAEA-600 caffeine) and in-house standards(leucine and muscle tissue of Pacific tuna) that were previ-ously calibrated versus certified standards The Chl a sam-ples were analyzed by high-performance liquid chromatog-raphy according to Descy et al (2005) with a reproducibil-ity ofplusmn05 and a detection limit of 001 microg Lminus1 Part of theChl a data were previously reported by Descy et al (2017)

The water filtered through GFF Whatman glass mi-crofiber filters was collected and further filtered throughpolyethersulfone syringe encapsulated filters (02 microm poros-ity) for stable isotope composition of O of H2O (δ18O-H2O)TA dissolved organic carbon (DOC) major elements (Na+Mg2+ Ca2+ K+ and dissolved silicate Si) nitrate (NOminus3 )nitrite (NOminus2 ) and ammonium (NH+4 ) Samples for δ18O-H2O were stored at ambient temperature in polypropylene8 mL vials and measurements were carried out at the Interna-tional Atomic Energy Agency (IAEA Vienna) where watersamples were pipetted into 2 mL vials and measured twiceon different laser water isotope analyzers (Los Gatos Re-search or Picarro) Isotopic values were determined by aver-aging isotopic values from the last four out of nine injectionsalong with memory and drift corrections with final nor-malization to the VSMOWSLAP scales by using two-pointlab standard calibrations as fully described in Wassenaar etal (2014) and Coplen and Wassenaar (2015) The long-termuncertainty for standard δ18O values was plusmn01 permil Samplesfor TA were stored at ambient temperature in polyethylene55 mL vials and measurements were carried out by open-cell titration with HCl 01 mol Lminus1 according to Gran (1952)and data quality was checked with certified reference mate-rial obtained from Andrew Dickson (Scripps Institution ofOceanography University of California San Diego USA)with a typical reproducibility better thanplusmn3 micromol kgminus1 DICwas computed from TA and pCO2 measurements usingthe carbonic acid dissociation constants for freshwater ofMillero (1979) using the ldquoCO2sysrdquo package Samples to de-termine DOC were stored at ambient temperature and inthe dark in 40 mL brown borosilicate vials with polytetraflu-oroethylene (PTFE)-coated septa and poisoned with 50 microLof H3PO4 (85 ) and DOC concentration was determinedwith a wet oxidation total organic carbon analyzer (IO An-alytical Aurora 1030W) with a typical reproducibility bet-ter than plusmn5 Part of the DOC data were previously re-ported by Lambert et al (2016) Samples for major ele-ments were stored in 20 mL scintillation vials and preservedwith 50 microL of HNO3 (65 ) Major elements were mea-sured with inductively coupled plasma MS (ICP-MS Agilent7700x) calibrated with the following standards SRM1640afrom National Institute of Standards and Technology TM-273 (lot 0412) and TMRain-04 (lot 0913) from Environ-



ment Canada and SPS-SW2 Batch 130 from SpectrapureStandard Limit of quantification was 05 micromol Lminus1 for Na+Mg2+ and Ca2+ 10 micromol Lminus1 for K+ and 8 micromol Lminus1 forSi Samples were collected during four cruises (8 Mayndash6 June 2010 20 Novemberndash8 December 2012 3ndash19 Decem-ber 2013 and 16 Aprilndash6 May 2015) for NOminus3 NOminus2 andNH+4 and were stored frozen (minus20 C) in 50 mL polypropy-lene vials NOminus3 and NOminus2 were determined with the sulfanil-amide colorimetric was determined with the vanadium re-duction method (APHA 1998) and NH+4 was determinedwith the dichloroisocyanurate-salicylate-nitroprussiate col-orimetric method (SCA 1981) Detection limits were 03001 and 015 micromol Lminus1 for NH+4 NOminus2 and NOminus3 respec-tively Precisions were plusmn002 plusmn002 and plusmn01 micromol Lminus1 forNH+4 NOminus2 and NOminus3 respectively

26 GIS

The limits of the river catchments and Strahler order ofrivers and streams were determined from the geospatial Hy-droSHEDS dataset (httpshydroshedscrusgsgov last ac-cess 27 September 2018) using ArcGISreg (1031) Landcover data were extracted from the global land cover(GLC) 2009 dataset (httpdueesrinesaintpage_globcoverphp last access 11 January 2019) from the European SpaceAgency GlobCover 2009 project for the following classescroplands mosaic croplandvegetation dense forest floodeddense forest open forestwoodland shrublands mosaic for-est or shrublandgrassland grasslands flooded grasslandand water bodies Shrubland and grassland classes were ag-gregated for the estimate of savannah Theoretical contribu-tion of C4 vegetation were extracted based on the vegeta-tion δ13C isoscape for Africa from Still and Powell (2010)but corrected for agro-ecosystems according to the methodof Powell et al (2012)

The geospatial and statistical methods to compute riverwidth length Strahler stream order surface area slope flowvelocity and discharge throughout the Congo River networkare described in detail in the Supplement

27 Statistical analysis

Statistical tests were carried out using GraphPad Prismreg

software at 005 level and the normality of the distributionwas tested with the DrsquoAgostinondashPearson omnibus normalitytest

3 Results and discussion

This section starts with the description of the spatial varia-tions in general limnological variables as well as GHGs alongthe two main transects (KinsanganindashKinshasa and Kwa) andas a function of stream order and the presence of the CCCThe following parts of this section deal with the seasonalvariability and the drivers of CO2 dynamics based on one

hand on the metabolic measurements and on the other handon stable isotopic composition of DIC The final part of thissection deals with fluxes of GHGs between the river and theatmosphere that are presented and discussed firstly as arealfluxes and finally as integrated fluxes at the scale of the basinand at global scale

31 Spatial variations along the KisanganindashKinshasatransect of general limnological variables anddissolved GHGs

Figures 3 and 4 show the spatial distribution of variables insurface waters of the main stem Congo River and confluencewith tributaries along the KisanganindashKinshasa transect dur-ing high water (HW December 2013) and falling water (FWJune 2014) periods (Figs 1 and 2) Numerous variables showa regular pattern in the main stem (increase or decrease) dueto the gradual inputs from tributaries with a different (higheror lower) value than the main stem Specific conductivityTA δ13C-DIC pH O2 N2O TSM pH and δ18O-H2Odecreased while pCO2 and DOC increased in the main stemalong the KisanganindashKinshasa transect Numerous tributarieshad black-water characteristics (low conductivity TA pHO2 TSM and high DOC) while the main stem had gener-ally white-water characteristics The black-water tributarieswere mainly found in the region of the CCC while tribu-taries upstream or downstream of the CCC had in generalmore white-water characteristics The differences betweenblack-waters and white-waters were apparent in the patternsof continuous measurements of pCO2 showing a negativerelationship with O2 TSM pH and specific conductivity(Fig S4)

Specific conductivity in the main stem Congo River de-creased from 483 (HW) and 789 (FW) microS cmminus1 in Kisan-gani to 265 (HW) and 327 (FW) microS cmminus1 in Kinshasa(Figs 3 and 4) This decreasing pattern was related to a grad-ual dilution with tributary water with lower conductivitieson average 276plusmn 99 (HW) and 319plusmn 178 (FW) microS cmminus1The lowest specific conductivity was measured in the LefiniRiver that is part of the Teacutekeacute Plateau where rainwater infil-trates into deep aquifers across thick sandy horizons lead-ing to water with a low mineralization (Laraque et al1998) TA in the main stem decreased from 344 (HW)and 697 (FW) micromol kgminus1 in Kisangani to 195 (HW) and269 (FW) micromol kgminus1 in Kinshasa with an average in tribu-taries of 141plusmn 119 (HW) and 136plusmn 141 (FW) micromol kgminus1δ13C-DIC roughly followed the patterns of TA decreas-ing in the main stem from minus79 (HW) and minus138 (FW) permilin Kisangani to minus118 (HW) and minus178 (FW) permil in Kin-shasa with an average in tributaries of minus217plusmn 32 (HW)and minus209plusmn 43 (FW) permil TSM in the main stem decreasedfrom 929 (HW) and 232 (FW) mg Lminus1 in Kisangani to 454(HW) and 182 (FW) mg Lminus1 in Kinshasa with an average intributaries of 93plusmn 134 (HW) and 85plusmn 93 (FW) mg Lminus1The highest TSM values were recorded in the Kwa (444



(HW) and 158 (FW) mg Lminus1) and in the Nsele a smallstream-draining savannah and flowing into pool Malebo(714 [HW] and 348 [FW] mg Lminus1) pH in the main stemdecreased from 673 (HW) and 738 (FW) in Kisangani to611 (HW) and 629 (FW) in Kinshasa with an average intributaries of 544plusmn 088 (HW) and 491plusmn 105 (FW) O2values in the main stem decreased from 892 (HW) and 926(FW) in Kisangani to 578 (HW) and 797 (FW) in Kin-shasa with an average in tributaries of 367plusmn 333 (HW)and 437plusmn 358 (LW) DOC increased from 59 (HW)and 51 (FW) mg Lminus1 in Kisangani to 119 (HW) and 94(FW) mg Lminus1 in Kinshasa with an average in tributaries of161plusmn 106 (HW) and 179plusmn 152 (FW) mg Lminus1 Extremelylow pH values were recorded in rivers draining the CCCwith values as low as 36 coinciding with nearly anoxicconditions (O2 down to 03 ) in surface waters and veryhigh DOC content (up to 678 mg Lminus1) δ18O-H2O in themain stem decreased from minus22 (HW) and minus16 (FW) permilin Kisangani to minus29 (HW) and minus22 (FW) permil in Kinshasawith an average in the tributaries of minus27plusmn 09 (HW) andminus22plusmn08 permil (FW) Temperature increased in the main stemfrom 260 (HW) and 271 (FW) C in Kisangani to 278(HW) and 273 (FW) C in Kinshasa due to exposure in theuncovered main stem to solar radiation as temperature waslower in tributaries with an average of 260plusmn 11 (HW) and261plusmn 17 (FW) C due to more shaded conditions (forestcover)

The pCO2 values in the main stem increased from 2424(HW) and 1670 (FW) ppm in Kisangani to 5343 (HW) and2896 (FW) ppm in Kinshasa with an average in tributariesof 8306plusmn 4089 (HW) and 8039plusmn 5311 (FW) ppm pCO2 intributaries was in general higher than in the main stem with afew exceptions namely in rivers close to Kinshasa (1582 to1903 [HW] and 1087 to 2483 [FW] ppm) due to degassingat waterfalls upstream of the sampling stations as the terrainis more mountainous in this area than in more gently slop-ing catchments upstream and also possibly due to a largercontribution of savannah to the catchment cover The highestpCO2 values (up to 16 942 ppm) were observed in streamsdraining the CCC CH4 in the main stem decreased from 85(HW) and 63 (FW) nmol Lminus1 in Kisangani to 24 (HW) and22 (FW) nmol Lminus1 just before Kinshasa and then increasedagain in the Malebo pool (82 [HW] and 78 [FW] nmol Lminus1)possibly related to the shallowness (sim 3 m in Malebo poolversus sim 30 m depth at station just upstream) The generaldecreasing pattern of CH4 in the main stem resulted fromCH4 oxidation as indicated by 13C-enriched values in themain stem (minus381 permil to minus494 permil) with regards to sedimentCH4 (minus602 permil) and the increasing 13C enrichment along thetransect (Fig 5) The calculated fraction of oxidized CH4ranged between 043 and 068 in the main stem and also in-creased downstream along the transect (Fig S5) CH4 in thetributaries showed a very large range of CH4 concentration(68 to 51 839 nmol Lminus1 (HW) and 102 to 56 236 nmol Lminus1

(FW)) CH4 in the tributaries showed a variable degree of 13C

enrichment compared to sediment CH4 (δ13C-CH4 betweenminus193 permil and minus563 permil Fig 5) and the calculated fractionof oxidized CH4 ranged between 018 and 088 (Fig S5)Unlike the large rivers of the Amazon where a loose neg-ative relation has been reported (Sawakuchi et al 2016)the δ13C-CH4 values in the Congo River were unrelated todissolved CH4 concentrations and a relatively high 13C en-richment (δ13C-CH4 up to minus193 permil) was observed even athigh CH4 concentrations (correspondingly 3118 nmol Lminus1)(Fig 5) This lack of correlation between concentration andisotope composition of CH4 was probably related to spatialheterogeneity of sedimentary (and corresponding water col-umn) CH4 content over a very large and heterogeneous sam-pling area The highest CH4 concentrations were observedat the mouths of small rivers in the CCC At the confluencewith the Congo main stem the flow is slowed down leadingto the development of shallow delta-type systems with verydense coverage of aquatic macrophytes (in majority Vossiacuspidata with a variable contribution of Eichhornia cras-sipes but also other species Fig S6) Such sites are favorablefor intense sediment methanogenesis but due to the stableenvironment related to near stagnant waters also very favor-able for the establishment of a stable methanotrophic bac-terial community in the water column and associated withthe root system of floating macrophytes that sustain intensemethane oxidation (Yoshida et al 2014 Kosten et al 2016)Indeed we found a very strong relation between CH4 oxi-dation and CH4 concentration on a limited number of incu-bations carried out in the Kwa river network in April 2015(Fig S7) Such conditions can explain the apparently para-doxical combination of high CH4 concentrations in somecases associated with a high degree of CH4 oxidation Micro-bial oxidation of CH4 might also explain the occurrence ofsamples with extremely 13C depleted POC (δ13C-POC downto minus390 permil) observed in low O2 and high CH4 environ-ments (Fig 6) located in small streams of the CCC that werealso characterized by low POC and TSM values (not shown)and high POC Chl a ratios (excluding the possibility thatin situ PP would be at the basis of the 13C depletion) Thissuggests that in these environments poor in particles (typ-ical of black-water streams) but with high CH4 concentra-tions methanotrophic bacteria that are able to incorporate13C-depleted CH4 into their biomass contribute substantiallyto POC While such patterns have been reported at the oxicndashanoxic transition zone of lakes with high hypolimnetic CH4concentrations such as Lake Kivu (Morana et al 2015) thishas never been reported in surface waters of rivers

N2O decreased from 198 (HW) and 139 (FW) inKisangani to 168 (HW) and 153 (FW) in Kinshasa andin most cases N2O was lower in tributaries on average114plusmn 73 (HW) and 120plusmn 69 (FW) The undersaturationin N2O was observed in streams with high DOC low O2and relatively low NOminus3 and is most probably related to den-itrification of N2O as also reported in the Amazon river(Richey et al 1988) and in temperate rivers (Baulch et al



Figure 3 Variation in surface waters of the specific conductivity (microS cmminus1) water temperature (C) oxygen stable isotope composition ofH2O (δ18O-H2O in permil) total suspended matter (TSM in mg Lminus1) total alkalinity (TA in micromol kgminus1) dissolved organic carbon (DOC inmg Lminus1) carbon stable isotope composition of dissolved inorganic carbon (δ13C-DIC in permil) pH partial pressure of CO2 (pCO2 in ppm)dissolved O2 saturation level (O2 in ) dissolved CH4 concentration (nmol Lminus1) and dissolved N2O saturation level (N2O in ) as afunction of the distance upstream of Kinshasa along a transect along the Congo River from Kisangani (3ndash19 December 2013 n= 10505)Grey and black symbols indicate samples from the main stem and green samples from tributaries

2011) Denitrification could have occurred in river sedimentsor soils although the lowest N2O (and O2) occurredin the CCC dominated by flooded soils in the flooded for-est Indeed there was a general negative relationship be-tween N2O and O2 (Fig 7) with an average N2O

of 785plusmn 593 for O2 lt 25 and an average N2O of1557plusmn577 for O2 gt 25 (MannndashWhitney plt00001)The decreasing pattern of NH+4 DIN and increasing patternof NOminus3 DIN with O2 indicated the occurrence of nitrifi-cation in oxygenated (typical of high Strahler stream order)



Figure 4 Variation in surface waters of the specific conductivity (microS cmminus1) water temperature (C) oxygen stable isotope composition ofH2O (δ18O-H2O in permil) total suspended matter (TSM in mg Lminus1) total alkalinity (TA in micromol kgminus1) dissolved organic carbon (DOC inmg Lminus1) carbon stable isotope composition of dissolved inorganic carbon (δ13C-DIC in permil) pH partial pressure of CO2 (pCO2 in ppm)dissolved O2 saturation level (O2 in ) dissolved CH4 concentration (nmol Lminus1) and dissolved N2O saturation level (N2O in ) as afunction of the distance upstream of Kinshasa along a transect along the Congo River from Kisangani (10ndash30 June 2014 n= 12968) Greyand black symbols indicate samples from the main stem and green samples from tributaries

rivers and prevalence of NH+4 in the more reducing and loweroxygenated (typical of low Strahler order) streams drainingthe CCC in particular where NOminus3 was probably also re-moved from the water by sedimentary or soil denitrification(Fig 7) In addition in black-water rivers and streams low

pH (down to 4) might have led to the inhibition of nitrifica-tion (Le et al 2019) and also contributed higher NH+4 DINvalues Furthermore the positive relation between N2Oand NOminus3 DIN and the negative relation between N2Oand NH+4 DIN (Fig 8) support the hypothesis of N2O re-



Figure 5 Carbon stable isotope composition of CH4 (δ13C-CH4 inpermil) in surface waters of the Congo River main stem (black symbols)and tributaries (green symbols) as a function of dissolved CH4 con-centration (nmol Lminus1) and as a function of the distance upstream ofKinshasa obtained along a longitudinal transect along the CongoRiver from Kisangani (10ndash30 June 2014) The dotted line indicateslinear regression

Figure 6 Carbon stable isotope composition of particulate organiccarbon (δ13C-POC in permil) in surface waters of the Congo River net-work as a function of dissolved O2 saturation level (O2 in )and dissolved CH4 concentration (nmol Lminus1) Grey lines indicateaverage (full line) and standard deviation (dotted lines) of aver-age soil organic carbon stable isotope composition with a domi-nance of C4 (minus155plusmn 08 permil) and C3 plants (minus284plusmn 07 permil) (Birdand Pousai 1997) Data with δ13C-POC ltminus30 permil were associatedwith significantly lower O2 and higher CH4 than data with δ13C-POC gtminus30 permil (MannndashWhitney plt00001 for both tests)

moval by denitrification in O2-depleted environments (ei-ther in stream sediments or soils) while in more oxygenatedrivers N2O was produced by nitrification In the main stemof the Congo there might in addition be a loop of nitrogenrecycling (ammonification-nitrification) sustained by phyto-plankton growth and decay that contributed to maintain over-saturation of N2O with respect to atmospheric equilibriumas phytoplankton growth was only observed in the main stem(Descy et al 2017) The generally low N2O values in theCongo River network were probably due to the near pristinenature of these systems with low NH+4 (23plusmn 13 micromol Lminus1)and NOminus3 (56plusmn 51 micromol Lminus1) levels typical of rivers andstreams draining a large fraction (sim 70 ) of forests Crop-lands only represented at most 17 on average of the land

cover of the studied river catchments where traditional agri-culture is practised with little use of artificial fertilizers Thiscorresponds to an upper bound of cropland surface area sinceit was estimated aggregating the ldquocroplandrdquo and ldquomosaiccroplandvegetationrdquo GLC 2009 categories the latter corre-sponding to mixed surfaces with lt 50 of cropland Theldquocroplandrdquo GLC 2009 category only accounts for 01 onaverage of the land cover of the studied river catchments Ni-trogen inputs from waste water can also sustain N2O produc-tion in impacted rivers and streams (Marwick et al 2014)but the largest cities along the Congo River main stem areof relatively modest size such as Kisangani (1 600 000 habi-tants) and Mbandaka (350 000 habitants) especially consid-ering the large dilution due to the massive discharge of themain stem (sampling was done upstream of the influence ofthe megacity of Kinshasa with 11 900 000 habitants)

The input of the Kwa led to distinct changes of TA δ18O-H2O (decrease) and TSM (increase) of main stem values(comparing values upstream and downstream of the Kwamouth) (Figs 3ndash4) In the main stem continuous measure-ments of pCO2 pH O2 and conductivity showed morevariability compared to discrete samples acquired in the mid-dle of channel in particular in the region of CCC (Figs 3ndash4)These patterns were related to gradients across the section ofchannel as the boat sailed either along the mid-channel orcloser to shore The water from the tributaries flowed alongthe riverbanks and did not mix with main stem middle chan-nel waters as visible in natural color remotely sensed im-ages (Fig S8) leading to strong gradients across the sectionof main stem channel During the June 2014 field expeditionthis was investigated in more detail by a series of six transectsperpendicular to the river main stem channel (Fig 9) In theupper part (1590 km from Kinshasa) the variables showedlittle cross section gradients except for a decrease in conduc-tivity towards the right bank due to inputs from the LindiRiver that had distinctly lower specific conductivities (272[HW] and 351 [FW] microS cmminus1) than the main stem (483[HW] and 789 [FW] microS cmminus1) At 795 km from Kinshasamarked gradients appeared in all variables with the presenceof black-water characteristics close to the right bank (higherpCO2 and lower O2 pH specific conductivity tempera-ture and TSM values) This feature was related to inputsfrom large right-bank tributaries such as the Aruwimi andthe Itimbiri (Supplement Table S1) Upstream of this sectionthe only major left-bank tributary is the Lomami which hadwhite-water characteristics relatively similar to those of themain stem (Figs 3ndash4) The presence of black-water charac-teristics became apparent also on the left bank from crosssections at 307 and 254 km upstream of Kinshasa wherethe river is particularly wide (gt 6 km wide) and received theinputs from the Ruki the second-largest left-bank tributary(Table S1) with black-water characteristics The cross sec-tion gradients became less marked at 203 km upstream fromKinshasa as in this region the river becomes more narrow(2 km wide) leading to increased currents and more lateral



Figure 7 Dissolved N2O saturation level (N2O in ) ratio ofNH+4 to dissolved inorganic nitrogen (DIN=NH+4 +NOminus2 +NOminus3 )(micromol micromol) and ratio of NOminus3 to DIN (micromol micromol) in surface wa-ters of the Congo River network as a function of dissolved O2 sat-uration level (O2 in ) Outliers (red dots) were identified witha Cookrsquos distance procedure and removed prior to linear regressionanalysis (solid line)

mixing The cross section gradients nearly disappeared at158 km upstream of Kinshasa (and 30 km downstream ofthe Kwa mouth) due to homogenization by the large Kwainputs (nearly 20 of total freshwater discharge from theCongo River Table S1) into a relatively narrow river section(sim 1 km)

32 Spatial variations along the Kwa transect

Spatial features of biogeochemical variables along the tran-sect in the Kwa River network (Fig 10) showed somesimilarities with two KisanganindashKinshasa transects alongthe Congo main stem (Figs 3ndash4) The main stem Kwa

Figure 8 Dissolved N2O saturation level (N2O in ) as afunction of the ratio of NH+4 to dissolved inorganic nitrogen(DIN=NH+4 +NOminus2 +NOminus3 ) (micromol micromol) and of the ratio of NOminus3to DIN (micromol micromol) in surface waters of the Congo River networkOutliers (red dots) were identified with a Cookrsquos distance procedureand removed prior to linear regression analysis (solid line)

had a higher specific conductivity than the tributaries(251plusmn42 versus 212plusmn115 microS cmminus1) higher TA (281plusmn64versus 119plusmn 118 micromol kgminus1) higher temperatures (277plusmn07 versus 263plusmn 22 C) higher TSM (401plusmn 89 versus152plusmn 351 mg Lminus1) higher pH (61plusmn 02 versus 45plusmn 07)higher O2 (670plusmn 70 versus 379plusmn 268 ) higher δ13C-DIC (minus165plusmn 12 versus minus226plusmn 34 permil) and lower DOC(52plusmn 14 versus 130plusmn 85 mg Lminus1) Unlike the KisanganindashKinshasa transects along the Congo main stem the δ18O-H2O was lower in the main stem Kwa (minus43plusmn 03 permil) thanin the tributaries (minus33plusmn06 permil) Note that both the main up-stream branches of the Kwa (Kasai and Sankuru) had lowδ18O-H2O values The δ18O-H2O values of the upper Ka-sai and Sankuru (minus44 permil and minus47 permil respectively) werelower than those of the Lualaba (Congo at Kisangani) (minus22[HW] and minus16 [FW] permil) This difference can be in part ex-plained by the spatial patterns of δ18O-H2O in rainwater asthe annual averages over of river catchments areminus31 permil andminus42 permil for the Lualaba and the Kasai respectively basedon the global grids of the O isotope composition of precip-itation given by Bowen et al (2005) The river δ18O-H2Owas close to those of rain for the Kasai but less depleted in18O for the Lualaba This was probably related to the lowerevapotranspiration over the catchments of the upper Kasaiand Sankuru than those of Lualaba (Bultot 1972) leading tomore 18O-depleted water (eg Simpson and Herczeg 1991)Additionally the catchment of Kwa has a high fraction ofunconsolidated sedimentary (414 ) and siliciclastic sedi-mentary (443 ) rocks than the catchment of the Lualabathat is dominated by metamorphic rocks (684 ) Uncon-solidated sedimentary and siliciclastic sedimentary rocks aremore favorable to the infiltration of water and development ofaquifers that will minimize evaporation and 18O enrichmentunlike catchments dominated by metamorphic rocks Notethat the tributaries with the lowest δ18O-H2O values weresituated downstream of the Kwa and upstream of Kinshasa



Figure 9 Variations in surface water of the partial pressure of CO2 (pCO2 in ppm) dissolved oxygen saturation level (O2 in ) pHspecific conductivity (microS cmminus1) water temperature (C) and total suspended matter (TSM in mg Lminus1) along perpendicular transects to themain stem Congo River as a function of distance from the left bank (m) and at a variable distance from Kinshasa (10ndash30 June 2014)

(Figs 3ndash4) and are part of the Teacutekeacute Plateau These rivers arefed by deep aquifers derived from infiltration of rain throughsandy soils (Laraque et al 1998)

Another difference with the KisanganindashKinshasa transectsalong the Congo main stem relates to N2O values that werecloser to saturation in the Kwa main stem (1101plusmn 88 )while surface waters oscillated from undersaturation to over-

saturation in the tributaries (1224plusmn 595 ) This differ-ence could be due to variations in biogeochemical cyclingor in physical settings leading to changes in k The latterseems more likely due to the strong flow in the Kwa thatprobably led to high gas transfer velocities and strong de-gassing of N2O to the atmosphere In the Kwa main stempCO2 (3473plusmn974 ppm) and CH4 (255plusmn150 nmol Lminus1) val-



Figure 10 Variation in surface waters of the specific conductivity (microS cmminus1) water temperature (C) oxygen stable isotope compositionof H2O (δ18O-H2O in permil) total suspended matter (TSM in mg Lminus1) total alkalinity (TA in micromol kgminus1) dissolved organic carbon (DOC inmg Lminus1) carbon stable isotope composition of dissolved inorganic carbon (δ13C-DIC in permil) pH partial pressure of CO2 (pCO2 in ppm)dissolved O2 saturation level (O2 in ) dissolved CH4 concentration (nmol Lminus1) and dissolved N2O saturation level (N2O in ) as afunction of the distance upstream of the Kwa mouth along a transect along the Kwa River (16 Aprilndash6 May 2015 n= 7017) Grey and blacksymbols indicate samples from the main stem and green samples from tributaries

ues were lower than in tributaries (8804plusmn5108 ppm 6783plusmn16479 nmol Lminus1) The highest pCO2 and CH4 values of theentire dataset in the Congo River network were observed in atributary of the Fimi (22 899 ppm and 71 428 nmol Lminus1) thatis bordered by very extensive meadows of the aquatic macro-

phyte Vossia cuspidata and unrelated to inputs from the shal-low Lake Mai Ndombeacute that showed lower pCO2 and CH4values (3143 ppm and 250 nmol Lminus1 respectively)



33 Spatial variations as a function of stream order andas a function of the influence of the CCC

The large differences in pCO2 CH4 O2 and N2O(Figs 3 4 10) among the various sampled tributaries of theCongo River can be analyzed in terms of size classes as givenby Strahler order (Fig 11) There were distinct patterns inCH4 and pCO2 versus Strahler order with a decrease in thecentral value (median and average) for both quantities as afunction of Strahler order in streams draining and not drain-ing the CCC (Fig 11) For nearly all the stream orders thestreams draining the CCC had significantly higher pCO2 andCH4 values than streams not draining the CCC (Fig 11) TheO2 values per Strahler order did not show any distinct pat-tern (increase or decrease) in the streams not draining theCCC but the streams draining the CCC showed an increas-ing pattern as a function of Strahler order The O2 valueswere significantly lower in the streams draining the CCC thanthose not draining the CCC (Fig 11) For N2O the ten-dency of the central value (median and average) as a func-tion of Strahler order did not show a clear pattern for streamsnot draining the CCC however there was a clear increasingpattern with Strahler order for streams draining the CCC Inaddition the N2O values were significantly lower for halfof the cases in streams draining the CCC compared to thosenot draining it (Fig 11)

In US rivers a decreasing pattern as a function of Strahlerorder has previously been reported for pCO2 (Butman andRaymond 2001 Liu and Raymond 2018) This has been in-terpreted as reflecting inputs of soil-water enriched in ter-restrial respired CO2 that have a stronger impact in smallerand lower Strahler order systems in particular headwaterstreams followed by degassing of CO2 in higher Strahler or-der rivers (Hotchkiss et al 2015) although soil-water CO2inputs in headwater streams are seasonally variable and spa-tially heterogeneous (Duvert et al 2018) Nevertheless all ofthe low-Strahler-order streams we sampled were in lowlandsso the decreasing pattern of pCO2 as a function of Strahlerorder could alternatively reflect the stronger influence of ri-parian wetlands on smaller streams rather than larger sys-tems The mechanism remains the same a high ratio of lat-eral inputs to water volume in small streams that is related tosoil-water in temperate streams such as in the US but relatedin addition to riparian wetlands in tropical systems such asthose sampled in the Congo River network

The influence of riparian wetlands on stream pCO2 CH4and N2O can be also highlighted when data were separatedinto rivers draining or not draining the CCC but aggregatinginto systems smaller or larger than Strahler order 5 to ac-count simultaneously for the effect of stream size (Fig 12)The pCO2 values were statistically higher in rivers drain-ing the CCC than those not draining it with median valuesmore than 2-fold higher in both small and large rivers Con-versely O2 levels were statistically lower in rivers drain-ing the CCC than those not draining it with median val-

ues 11-fold and 2-fold lower in small and large rivers re-spectively Additional evidence on the influence of the con-nectivity of wetlands with rivers in sustaining high pCO2 andlow O2 values was provided by the positive relationshipbetween pCO2 and flooded dense forest cover and the con-verse negative relationship between O2 and flooded denseforest cover (Fig 13) These patterns were also consistentwith the positive relation between DOC concentration andflooded dense forest reported by Lambert et al (2016) Notethat aquatic macrophytes (Vossia cuspidata) also most prob-ably strongly contributed in addition to flooded forest tohigh pCO2 and low O2 levels based on visual observationsof dense coverage (Fig S9) although macrophytes have notbeen systematically mapped and GIS data are unavailable (asfor flooded dense forest)

Wetlands coverage had also a major importance on CH4distribution as the CH4 values were statistically higher inrivers draining the CCC than those not draining it (Fig 12)with median values 10-fold and 2-fold higher in small andlarge rivers respectively N2O was also statistically lowerin rivers draining the CCC than those not draining it withmedian values 27-fold lower in small rivers but 14-foldhigher in large rivers The pattern of N2O followed the oneof O2 and the very low to null N2O values were observedin the systems draining the CCC where the lowest O2 val-ues (close to 0) were also observed due to sedimentary or soilorganic matter degradation leading to a decrease in O2 andN2O in surface waters (consumption of N2O by denitrifica-tion)

The CH4 CO2 molar ratio ranged between 00001 and01215 with a mean of 00097plusmn0018 Such ratios were dis-tinctly higher than those typically observed in marine waters(00005) and in the atmosphere (0005) The CH4 CO2 mo-lar ratio strongly increased with the decrease in O2 andwas significantly higher in small rivers draining the CCC(Fig 14) These patterns were probably related to inputsof organic matter from wetlands and in particular aquaticmacrophytes that lead to important organic matter transferto sediments and high sedimentary degradation of organicmatter This led to O2 decrease in surface waters and alarge fraction of organic matter degradation by anaerobicprocesses compared to aerobic degradation leading to an in-crease in the CH4 CO2 ratio The decrease in O2 and in-crease in CO2 in the water in presence of floating macro-phytes was probably in part also related to autotrophic rootrespiration and not fully related to microbial heterotrophicrespiration

34 Drivers of CO2 dynamics ndash metabolicmeasurements and daily variations in CO2

We compared the balance of depth-integrated planktonic PPto water column CR and compared it to FCO2 to test ifin situ net heterotrophy was sufficient to sustain the emis-sions of CO2 to the atmosphere (Fig 15) or if alterna-



Figure 11 Box plot as a function of Strahler stream order of the partial pressure of CO2 (pCO2 in ppm) dissolved CH4 concentration(nmol Lminus1) dissolved N2O saturation level (N2O in ) and dissolved O2 saturation level ( O2 in ) for rivers and streams of the CongoRiver network draining and not draining the Cuvette Centrale Congolaise The box represents the first and third quartile the horizontalline corresponds to the median the cross corresponds to the average the error bars correspond to the maximum and minimum and thesymbols show all data points A MannndashWhitney test was used to test statistical differences ns represents not significant lowastlowastlowastlowast=plt00001lowastlowastlowast=plt0001 lowastlowast=plt001 lowast=plt005

tively fluvial CO2 had a lateral origin (soils or wetlands)A detailed description of spatial and seasonal variations inPP as well as main phytoplankton communities is given byDescy et al (2017) In brief phytoplankton biomass wasmainly confined to the main stem and was low in most trib-utaries The PP values in the Congo River ranged between00 and 575 mmol mminus2 dminus1 and were higher than previouslyreported in tropical river channels whereas in other tropi-cal rivers phytoplankton production mainly occurred in thefloodplain lakes This is due to generally lower TSM valuesin the Congo and to its relative shallowness that allows netphytoplankton growth in the mainstream unlike other deeperand more turbid tropical rivers such as the Amazon Mea-sured CR was lower than PP 3 out of 49 times and on aver-age the PP CR ratio was 028 Volumetric rates of CR rangedbetween 07 and 466 mmol mminus3 dminus1 while integrated ratesof CR ranged between 31 and 7904 mmol mminus2 dminus1 CR wasunrelated to TSM POC NH+4 and Chl a but showed a pos-itive relation with DOC after binning the data (Fig S10)The same pattern emerged when using modeled PP to ex-tend the number of data points with PP higher than CR 2 outof 169 times and a PP CR ratio of 015 on average Thisindicates that a generalized and strongly net heterotrophicmetabolism was encountered in the sampled sites Yet in 174out of 187 cases FCO2 was higher than CR and in 162 outof 169 cases FCO2 was higher than net community produc-tion (NCP) CR averaged 81 mmol mminus2 dminus1 NCP averaged

minus75 mmol mminus2 dminus1 and the corresponding average FCO2was 740 mmol mminus2 dminus1 The FCO2 CR ratio was higher inlower-order streams than in higher-order streams with me-dian values ranging between 21 and 139 in stream orders 2ndash5and between 3 and 17 in stream orders 6ndash10 (Fig 16) Thisindicates a prevalence of lateral CO2 inputs either from soil-water or riparian wetlands in sustaining FCO2 in lower-orderstreams over higher-order streams where in-stream CO2 pro-duction from net heterotrophy is more important These pat-terns are in general agreement with the conceptual frame-work developed by Hotchkiss et al (2015) although lateralCO2 inputs were exclusively attributed by these authors tosoil-water or ground-water inputs and riparian wetlands werenot considered These patterns are also in agreement with theresults reported by Ward et al (2018) who show that in largehigh-order rivers of the lower Amazon in-stream productionof CO2 from respiration is sufficient to sustain CO2 emis-sions to the atmosphere

CR was estimated from measurements of O2 concentra-tion decrease in bottles that were not rotated and this hasbeen shown to lead to an underestimation of CR up to a fac-tor of 2 (Richardson et al 2013 Ward et al 2018) The un-derestimation of our CR measurements due to the absenceof rotation is most likely not as severe as in the Richardsonet al (2013) and Ward et al (2018) studies as the organicmatter in our samples was mostly in dissolved form (me-dian DOC of 86 mg Lminus1) with a low particulate load (me-



Figure 12 Box plot of the partial pressure of CO2 (pCO2 in ppm) dissolved O2 saturation level (O2 in ) dissolved CH4 concen-tration (nmol Lminus1) and dissolved N2O saturation level (N2O in ) in surface waters of rivers and streams of the Congo river networkdraining and not draining the Cuvette Centrale Congolaise for small and large systems (Strahler stream order 5le and gt 5 respectively)(3ndash19 December 2013 10ndash30 June 2014) The box represents the first and third quartile the horizontal line corresponds to the medianthe cross corresponds to the average the error bars correspond to the maximum and minimum and the symbols show all data points AMannndashWhitney test was used to test statistical differences lowastlowastlowastlowast=plt00001 lowastlowastlowast=plt0001 lowastlowast=plt001

dian TSM of 14 mg Lminus1 and POC of 13 mg Lminus1) while themedian of TSM at the sites studied by Ward et al (2018)in the Amazon was higher (285 mg Lminus1 based on data re-ported by Ward et al 2015) We acknowledge that our CRmeasurements might be underestimated due to bottle effectsand lack of rotation up to a factor of 2 based on the studies ofRichardson et al (2013) and Ward et al (2018) neverthelessit seems unrealistic to envisage an underestimation of CR bya factor of 10 that would allow reconciling the CR (and NCP)estimates with those of FCO2 Although we did not mea-sure sediment respiration the average value reported by Car-doso et al (2014) of 21 mmol mminus2 dminus1 for tropical rivers andstreams does not allow accounting for the imbalance betweenFCO2 and NCP This then suggests that the emission of CO2from the Congo lowland river network is to a large extentsustained by lateral inputs rather than by in-stream produc-tion of CO2 by net heterotrophy It remains to be determinedto what extent this lateral input of CO2 is sustained by ripar-ian wetlands or soil-groundwater from terra firme

The low PP CR ratio of 015 to 028 on average and gen-erally low PP values (on average 12 mmol mminus2 dminus1) were

also reflected in the low diurnal variations in pCO2 We didnot carry out dedicated 24 h cycles to look at the dayndashnightvariability of pCO2 due to lack of opportunity given theimportant navigation time to cover large distances but wecompared the data acquired at the anchoring site on shore(typically around 1700 universal time UT just before dusk)with the data on the same spot the next day (typically around0430 UT just after dawn) (Fig S11) Unsurprisingly wa-ter temperature measured just before dusk was significantlyhigher than just after dawn (on average 05 C higher) whilespecific conductivity was not significantly different indicat-ing that the same water mass was sampled at dusk and dawn(Fig S11) The pCO2 and O2 concentration measured justbefore dusk were not significantly different than just afterdawn showing that daily variability in these variables waslow (Fig S11) The difference between pCO2 at dusk anddawn ranged between minus2307 and 1186 ppm and averaged39 ppm (n= 39) The wide range of values of the differenceof pCO2 at dusk and at dawn might reflect occasional small-scale variability of pCO2 as the boat anchored for the nightclose to shore frequently in close proximity to riparian veg-



Figure 13 Partial pressure of CO2 (pCO2 in ppm) and dissolvedO2 saturation level (O2 in ) in surface waters of rivers andstreams of the Congo River network as a function of the floodeddense forest over the respective catchment (GLC 2009) Grey opendots are individual data points and black full dots are binned aver-ages (plusmn standard deviation) by intervals of 20

etation Nevertheless the average difference is not of the ex-pected sign (in the case of a strong diurnal change of pCO2due to PP and CR pCO2 should have been lower at dusk thandawn so the difference should have been negative) This dif-ference was also very small compared to the overall range ofspatial variations in pCO2 (1087 to 22 899 ppm) Dayndashnightvariations in pCO2 have been reported in temperate headwa-ter and low-order streams and in one lowland river with anamplitude fromsim 50 tosim 700 ppm (Lynch et al 2010 Dins-more et al 2013 Peter et al 2014 Crawford et al 2017Reiman and Xu 2019) although daily signals of pCO2 werenot systematically observed and were absent for instance instreams covered by forest canopy (Crawford et al 2017)In a low turbidity and very shallow low-order stream of theTana River network Tamooh et al (2013) reported on oneoccasion dayndashnight variation in pCO2 with an amplitude ofsim 400 ppm and in the Zambezi river during the dry sea-son corresponding to very low TSM values (lt 10 mg Lminus1)Teodoru et al (2015) reported dayndashnight pCO2 variation inthe range of 475 ppm In both cases the dayndashnight varia-tions were also small compared to spatial variations of 300to 5204 ppm in the Tana and 300 to 14 004 ppm in the Zam-bezi In floodplain lakes of the Amazon daily variations inpCO2 can be intense (with an amplitude up to sim 2000 ppm)during cyanobacterial blooms (Abril et al 2013 Amaral et

al 2018) but have not been documented in the river chan-nels of the Amazon Our data show that in nutrient-poorand light-limited lowland tropical rivers such as the CongoRiver where pelagic PP is low dayndashnight variations in pCO2were negligible compared to spatial variations in pCO2 Weconclude that accounting for dayndashnight variations in pCO2should not lead to a dramatic revision of global CO2 emis-sions unlike the recent claim based on data from a lowlandtemperate river by Reiman and Xu (2019) given that tropicalrivers account for 80 of CO2 emissions (Raymond et al2013 Borges et al 2015a b Lauerwald et al 2015)

35 Drivers of CO2 dynamics ndash stable isotopecomposition of DIC

The stable isotope composition of DIC can provide informa-tion on the origin of CO2 although the signal depends onthe combination of the biological processes that remove oradd CO2 to the water column (CR and PP) rock weatheringthat adds HCOminus3 and outgassing which removes CO2 that is13C-depleted relative to HCOminus3 The variable contribution toDIC in time and in space of CO2 relative to HCOminus3 compli-cates the interpretation of δ13C-DIC data The stable isotopecomposition of DIC due to rock weathering will depend onthe type of rock (silicate or carbonate) and on the origin ofthe CO2 involved in rock dissolution (either CO2 from theatmosphere or from respiration of soil organic matter) Thestable isotopic composition of DIC due to the degradationof organic matter will depend in part on whether the organicmatter is derived from terrestrial vegetation following the C3photosynthetic pathway (woody plants and trees temperategrasses δ13C simminus27 permil) compared to the less fractionatingC4 photosynthetic pathway (largely tropical and subtropicalgrasses δ13C simminus13 permil) (Hedges et al 1986 Bird et al1994) Spatial and temporal changes of δ13C are related tothe change of the relative abundance of HCOminus3 over CO2The degassing of CO2 to the atmosphere and the addition ofHCOminus3 from rock weathering lead to δ13C-DIC values be-coming dominated by those related to HCOminus3 Since CO2is isotopically depleted relative to HCOminus3 CO2 degassingleads to a gradual enrichment of the remaining DIC pool(eg Doctor et al 2008 Deirmendjian and Abril 2018) Thecombination of these processes can lead to spatial changes ofδ13C-DIC that covary with CO2 and HCOminus3 concentrationsFor instance in the Tana River network an altitudinal gra-dient of δ13C-DIC was attributed to a downstream accumu-lation of HCOminus3 due to rock weathering combined to CO2degassing (Bouillon et al 2009 Tamooh et al 2003) Fig-ure 17 shows δ13C-DIC as a function of the TA DIC ratio inthe rivers and streams of the Congo with a general increasein δ13C-DIC from TA DIC= 0 (all of the DIC is in the formof CO2) towards TA DIC= 1 (nearly all of the DIC is in theform of HCOminus3 ) This pattern reflects the mixing of two dis-tinct types of rivers and streams the lowland systems drain-ing the CCC with low rock weathering due to dominance of



Figure 14 Ratio of dissolved CH4 and CO2 concentration (micromolmicromol) plotted as a function of O2 saturation level (O2 in ) and plottedin box plots draining and not draining the Cuvette Centrale Congolaise for small and large systems (Strahler stream order 5le and gt 5respectively) (3ndash19 December 2013 10ndash30 June 2014) The box represents the first and third quartile the horizontal line corresponds to themedian the cross corresponds to the average the error bars correspond to the maximum and minimum and the symbols show all data pointsA MannndashWhitney test was used to test statistical differences ns=not significant lowastlowastlowastlowast=plt00001

deep organic soils (low HCOminus3 ) and high CO2 from respira-tion leading to TA DIC close to 0 with low δ13C-DIC andthe systems draining highland regions (Lualaba and Kasaiuml)with high rock weathering and lower generation of CO2 fromrespiration andor higher CO2 degassing leading to TA DICclose to 1 with high δ13C-DIC The plots of TA Na+ andMg2+

Na+ versus Ca2+ Na+ (Fig 18) showed aggrega-

tion of data close to what would be expected for silicate rockweathering based on the average values proposed by Gail-lardet et al (1999) This is in agreement with the dominanceof silicate rocks over carbonate rocks in the Congo basin(Nkounkou and Probst 1987) Note that TA values from theCongo are generally very low compared to other large riversglobally (Meybeck 1987) due to the large proportion ofrelatively insoluble rocks on the catchment (70 of meta-morphic rocks) and a small proportion of low soluble rockssuch as siliciclastic sedimentary rocks (10 mainly as sandstone) unconsolidated sediments (17 as sand and clays)and a very small proportion of highly soluble volcanic rocks(1 ) The high TA values in the Lualaba are due to a largerproportion of volcanic rocks in high-altitude areas such asthe Virunga region which is rich in volcanic rocks (includ-ing basalts) and has been shown to be a hotspot of chemicalweathering (Balagizi et al 2015)

At TA DIC= 0 the δ13C-DIC values are exclusively re-lated to those of CO2 and might be indicative of the sourceof mineralized organic matter The δ13C-DIC values rangedbetween minus282 permil and minus159 permil and averaged minus241 permil(Fig 17) indicating that CO2 was produced from the degra-dation of mixture of organic matter from C3 and C4 ori-

gin Furthermore δ13C-DIC was positively related to pCO2(Fig 17) indicating that in the streams with high pCO2 val-ues the CO2 was generated from the degradation organicmatter with a higher contribution from C4 plants The δ13C-DIC values were unrelated to the contribution of C4 veg-etation on the catchment (terra firme) as modeled by Stilland Powell (2010) and the cover by savannah on the catch-ment given by GLC2009 (Fig S12) Further the most en-riched δ13C-DIC values (gtminus21 permil) corresponded on aver-age to a low contribution of C4 vegetation on the catchment(74 ) and low contribution of savannah cover on the catch-ment (08 ) These patterns are inconsistent with an originfrom terra firme of the C4 material that led to high pCO2 instreams (Fig 17) but is rather more consistent with a largercontribution of degradation of C4 aquatic macrophyte mate-rial in high CO2 streams

36 Seasonal variations

The difference between the HW and FW in the main stemalong the KisanganindashKinshasa transects (Figs 3 and 4)were relatively modest with higher specific conductivity TAδ13C-DIC and TSM (at Kisangani) biogeochemical sig-natures of higher surface run-off during the HW sampling(December 2013) and water flows from deeper soil hori-zons (or ground-water) during the FW sampling (June 2014)The comparison of tributaries that were sampled during bothHW and FW periods along the KisanganindashKinshasa tran-sects shows that pCO2 was significantly higher (Wilcoxonmatch-pairs signed rank test p = 0078) during HW for largesystems (freshwater discharge ge 300 m3 sminus1 Table S1) but



Figure 15 Primary production (measured and modeled) (mmol mminus2 dminus1) as a function of community respiration (mmol mminus2 dminus1) andairndashwater CO2 fluxes (FCO2 in mmol mminus2 dminus1) as a function of community respiration (mmol mminus2 dminus1) and net community production(mmol mminus2 dminus1) in surface waters of rivers and streams of the Congo River network Insets show the data on a linear scale (instead of alogndashlog scale)

not significantly different for small systems (freshwater dis-charge lt 300 m3 sminus1) irrespective of whether it was from theleft or right bank (Fig 19) However no significant differ-ences among the HW and FW periods occurred in O2 CH4and N2O for either small of large systems irrespective ofwhether they were taken from the left or right bank (Fig 19)This indicates that across the basin spatial differences amongtributaries are more important than seasonal variations withina given tributary The average difference for large rivers be-tween HW and FW was on average only 1745 ppm for pCO2when the respective range of variation for the whole datasetwas from 1087 to 22 899 ppm

Yearly cycles of CH4 and N2O were established on theLualaba (6 years) the Oubangui (2 years) and the Kasaiuml(2 years) (Figs 20 and 21) while pCO2 is only availableduring 2 years in the Lualaba In the Oubangui and the KasaiumlCH4 concentration peaked with the onset of rising water anddecreased as water level continued to increase The decreasein CH4 as discharge peaked in the Oubangui and the Kasaiumlis most probably related to dilution by surface runoff as alsotestified by the decrease in specific conductivity and TA (notshown) The increase at the onset of rising water could be

related to initial flushing of soil atmosphere enriched in CH4as rain penetrates superficial layers of soils In the Lualabathe CH4 seasonal variations seem to follow more closelythose of freshwater discharge (for instance in 2014) Unlikethe Kasaiuml and the Oubangui the Lualaba has very extensivepermanent marshes and swamps such as the Upemba wet-land system (inundated area of 18 000 km2) and the Luamaswamps (inundated area of 6000 km2) (Hughes and Hughes1992) Yet it remains to be determined if the seasonality ofCH4 observed in Lualaba can be attributed to higher inputsduring high water from these major wetlands located respec-tively 1000 and 600 km upstream of Kisangani Even sothere are numerous more modest marshes and swamps thatborder upstream tributaries of the Lualaba closer to Kisan-gani (Hughes and Hughes 1992) The seasonal amplitude ofdissolved CH4 in the Oubangui was about 10 times higherthan in the Kasaiuml and the Oubangui and the seasonal ampli-tude of CH4 in these three rivers seemed to be related to therelative seasonal amplitude of freshwater discharge as indi-cated by the positive relation with the ratio of maximum andminimum of freshwater discharge (Fig 22)



Figure 16 Ratio of airndashwater CO2 fluxes (FCO2 inmmol mminus2 dminus1) and community respiration (mmol mminus2 dminus1)as a function of Strahler stream order for rivers and streams ofthe Congo River network The box represents the first and thirdquartile the horizontal line corresponds to the median the crosscorresponds to the average the error bars correspond to the max-imum and minimum and the symbols show all data points Insetshows on a log scale the median of FCO2 CR and an exponentialfit (r2

= 088)

The seasonal cycles of N2O show patterns that were con-sistent with those CH4 with a loose parallelism of N2O anddischarge in the Lualaba an inverse relation in the Kasaiuml andpeak of N2O during rising waters (although delayed withrespect to the CH4 peak) in the Oubangui In addition to theseasonal cycle there seemed to be a longer-term decrease inthe annual average of N2O in the Lualaba (Fig S13) Therewere no significant long-term changes of the annual aver-age of other variables such as POC and PN (not shown) Sothe observed decrease in annual N2O is probably unrelated tochanges in terrestrial productivity (Zhou et al 2014) or nitro-gen content of terrestrial vegetation (Craine et al 2018) thatact at longer timescales (several decades) and seem to be re-lated to long-term changes in climate (precipitation) Fresh-water discharge showed an increasing pattern during the timeperiod (Fig S13) An increasing discharge could lead to in-creased gas transfer velocities and a loss of N2O to the at-mosphere Water temperature in the Congo River is alreadyclose to the optimum of denitrification (sim 27 C Canion etal 2014) so an enhancement of denitrification with increasein temperature is unlikely Water temperature did not show aclear pattern but freshwater discharge increased during the2013ndash2018 period excluding the year 2017 It is likely thatthe decreasing trend in N2O during the 2013ndash2018 period isa transient feature in response to inter-annual fluctuations inhydrology that led to a period of sustained increase in fresh-water discharge

The pCO2 time series in the Lualaba at Kisangani isshorter than for CH4 but a general positive relationship be-tween pCO2 and discharge was observed (Fig S14) A sim-ilar positive relationship between pCO2 and discharge wasalso observed in the Oubangui (Bouillon et al 2012) and theMadeira River (Almeida et al 2017) as well in several riversin the Amazon (Richey et al 2002) based on pCO2 calcu-lated from pH and TA Such pCO2 discharge patterns wereinterpreted as resulting from higher connectivity during highwater between the river main stem and the floodplains andwetlands In large temperate rivers a negative relationshipbetween pCO2 and discharge was observed in the Meuse(Borges et al 2018) and in more than half of the US riversanalyzed by Liu and Raymond (2018) The difference in therelationship of pCO2 versus discharge between tropical andtemperate large rivers might be related to lower interactionsbetween river and floodplains in temperate rivers particu-larly in highly human-impacted and channelized rivers suchas the Meuse This is in agreement with the analysis of Ahoand Raymond (2019) who reported in the Salmon River net-work positive relationships between pCO2 and flow in wa-ter watersheds with a high presence of wetlands and nega-tive relationships between pCO2 and flow in watersheds withlow presence of wetlands Also in temperate rivers temper-ature covaries strongly with discharge in temperate rivers sothat the warmer months that promote biological productionof CO2 are also characterized by lower discharge (Borges etal 2018)

The seasonal amplitude of CH4 (sim 50 nmol Lminus1 in theKasaiuml and the Congo versus 200ndash400 nmol Lminus1 in theOubangui) and N2O (20 ndash90 in the 3 rivers) was over-all much lower than the spatial gradients across the basin of22 and 71 428 nmol Lminus1 for CH4 and 0 and 561 for N2O

37 Variability of GHG fluxes in the Congo rivernetwork

Since pCO2 and CH4 and N2O concentrations followed dis-tinct patterns as a function of Strahler stream order (Fig 11)we used k600 and stream surface area as a function of Strahlerstream order (Fig S15) to compute the airndashwater GHG fluxesand to integrate them at basin scale This was done sepa-rating data for streams draining and not draining the CCCsince pCO2 and CH4 and N2O patterns were very differ-ent (Figs 11 and 12) The stream surface area decreasedregularly with increasing Strahler stream order but showeda large increase for Strahler stream order 9 (Fig S15)The latter mainly corresponds to the Congo main stem thatdownstream of Kisangani is characterized by anastomosingriver channels with extended sand bars and numerous is-lands (Runge 2008 OrsquoLoughlin et al 2013) In particu-lar along the about 500 km long section between the Mban-daka and Kwa mouths the main stem river channel un-dergoes a general expansion and width increases from sim 4to sim 10 km This corresponds to the area of CCC depres-



Figure 17 Carbon stable isotope composition of dissolved inorganic carbon (DIC) (δ13C-DIC in permil) as a function of the total alkalinity(TA) to DIC ratio (micromol micromol) and as a function of the partial pressure of CO2 (pCO2 in ppm) for a TADIC ratio equal to zero in surfacewaters of rivers and streams of the Congo River network Open dots indicate individual data points full dots indicate binned averages (plusmnstand deviation) (bins lt 5000 5000ndash10 000 and gt 10 000 ppm) Horizontal dotted lines indicate the δ13C values of atmospheric CO2 and ofaverage soil organic carbon stable isotope composition with a dominance of C4 (minus155plusmn 08 permil) and C3 plants (minus284plusmn 07 permil) (Bird andPousai 1997) Dotted red line provides polynomial fit (δ13C-DIC=minus225+ 2137timesTA DICminus697times (TA DIC)2 r2

= 076)

Figure 18 Total alkalinity (TA) and Mg2+ as a function of Ca2+ of the surface waters of rivers and streams of the Congo River network inNa+ normalized plots (micromol micromol) showing the composition fields for rivers draining different lithologies from a global compilation of the60 largest rivers in the world (Gaillardet et al 1999)

sion with a corresponding decrease in the slope from 6 to3 cm kmminus1 (OrsquoLoughlin et al 2013) The calculated k600 de-creased regularly with increasing Strahler stream order aspreviously reported (Butman and Raymond 2011 Raymondet al 2012 Deirmendjian and Abril 2018 Liu and Ray-mond 2018) due to higher turbulence in low-order streamsassociated with higher stream flow due to steeper slopes Hy-droSHEDS stream order classification is missing at least 1stream order because small streams are not correctly repre-sented (Benstead and Leigh 2012 Raymond et al 2013)Hence to correct this bias we added 1 to stream orders de-termined by HydroSHEDS meaning that the lowest streamorder to which GHG were attributed was 2 We then extrap-olated pCO2 and CH4 and N2O concentrations to stream or-

der 1 separating streams draining and not draining the CCCusing a linear regression with higher orders or by using thesame value as for order 2 (Fig S16)

An error analysis on the GHG flux computation and up-scaling was carried out by error propagation of the GHGconcentration measurements the k value estimates and theestimate of surface areas of river channels to scale the arealfluxes using a Monte Carlo simulation with 1000 iterationsThe uncertainty in the GHG concentrations led to an uncer-tainty of areal fluxes of plusmn12 plusmn23 and plusmn71 forCO2 CH4 and N2O respectively The uncertainty in k de-rived from tracer experiments is typically plusmn300 (Ulsethet al 2019) This leads to a cumulated uncertainty of arealfluxes of plusmn176 plusmn179 and plusmn190 for CO2 CH4 and



Figure 19 Comparison of the partial pressure of CO2 (pCO2 in ppm) dissolved CH4 concentration (nmol Lminus1) dissolved O2 saturationlevel (O2 in ) and dissolved N2O saturation level (N2O in ) in surface waters of the Congo River tributaries sampled during bothhigh water (3ndash19 December 2013) and falling water periods (10ndash30 June 2014) Tributaries were separated into left and right bank as wellas into large and small systems with a freshwater discharge (Q) ltand ge 300 m3 sminus1 respectively

N2O respectively The uncertainty of the river and streamsurface areas based on GIS analysis of Allen and Pavel-sky (2018) is estimated toplusmn10 leading to an overall uncer-tainty of integrated fluxes ofplusmn183 plusmn183 andplusmn196 for CO2 CH4 and N2O respectively

The calculated FCO2 ranged between 86 and7110 mmol mminus2 dminus1 averaging 2469plusmn 435 mmol mminus2 dminus1

(weighted by surface area of Strahler stream order) encom-passing the range FCO2 reported by Mann et al (2014)(312 to 1429 mmol mminus2 dminus1) in 25 sites during a singleperiod (November 2010) from four major river basins in theRepublic of the Congo (Alima Lefini Sangha Likouala-Mossaka) The pCO2 values ranged between 1087 and22 899 ppm also encompassing the values reported by Mannet al (2014) (2600 to 15 802 ppm) that were not measureddirectly but computed from pH and DIC measurementsalthough pH measurements in black-water rivers can bebiased by the presence of humic-dissolved organic matter(Abril et al 2015) and the addition of HgCl2 seems to alterthe CO2 content of samples (Fig S1) The calculated FCH4

ranged between 65 and 597 260 micromol mminus2 dminus1 averaging12553plusmn 2247 micromol mminus2 dminus1 (weighted by surface area ofStrahler order) encompassing the range FCH4 reported byUpstill-Goddard et al (2017) (33 to 48 705 micromol mminus2 dminus1)in 41 sites draining the Congo basin in the Republic of theCongo (November 2010 and August 2011) in the same fourriver basins sampled by Mann et al (2014) The CH4 valuesranged between 22 and 71 428 nmol Lminus1 also encompassingthe values reported by Upstill-Goddard et al (2017) (11to 9553 nmol Lminus1) The calculated FN2O ranged betweenminus52 and 319 micromol mminus2 dminus1 averaging 22plusmn4 micromol mminus2 dminus1

(weighted by surface area of Strahler order) encompassingthe range FN2O reported by Upstill-Goddard et al (2017)(minus19 to 67 micromol mminus2 dminus1) The N2O values ranged be-tween 0 and 561 also encompassing the values reportedby Upstill-Goddard et al (2017) (6 to 266 ) The widerranges of GHGs and respective fluxes we report comparedto those of Mann et al (2014) and Upstill-Goddard etal (2017) reflect the larger number of river systems sampledover a wider geographical area (n= 25 versus n= 278 for



Figure 20 Time series of dissolved CH4 concentration (nmol Lminus1) in surface waters and freshwater discharge (grey line) in the Congo(at Kisangani 2013ndash2018) the Oubangui (at Bangui 2010ndash2012) and the Kasaiuml (at Dima 2015ndash2017) rivers The black line shows afive sample running average Time series of the partial pressure of CO2 (pCO2 in ppm) in surface waters was also obtained in the CongoRiver (at Kisangani 2017ndash2018) Data in the Oubangui were previously reported by Bouillon et al (2012 2013)

pCO2 and n= 41 versus n= 367 for CH4 and N2O) hencerepresenting a wider range of river types morphologiescatchment characteristics and wetland densities

Information on the seasonal variability of concurrentFCO2 FCH4 and FN2O values was only available inthe Lualaba (at Kisangani) where the three GHGs weremeasured simultaneously (Fig S17) FCO2 FCH4 andFN2O were loosely positively correlated with freshwaterdischarge as the seasonal variations in k600 were small(ranging between 234 and 303 cm hminus1) and althoughpCO2 was correlated to freshwater discharge (Fig S14)this was not the case for CH4 and N2O concentra-tions (Figs 20 and 21) The range of seasonal varia-tions at Kisangani of FCO2 (234 and 948 mmol mminus2 dminus1)FCH4 (116 and 876 micromol mminus2 dminus1) and FN2O (minus2 and40 micromol mminus2 dminus1) was small compared to the range ofspatial variations in FCO2 (86 and 7110 mmol mminus2 dminus1)FCH4 (65 and 597 260 micromol mminus2 dminus1) and FN2O (minus52 and319 micromol mminus2 dminus1)

38 Significance of integrated GHG fluxes at basin andglobal scales

The FCO2 and FCH4 decreased with increasing Strahler or-der as given by surface area and also when integrated bysurface area of the streams (Table 1) Strahler orders 1-2 ac-counted for nearly 80 of the integrated FCO2 (796 )and FCH4 (770 ) while Strahler orders 1ndash4 accounted forgt 90 of the integrated FCO2 (907 ) and FCH4 (919 )Strahler orders 5ndash10 only accounted for 93 of integratedFCO2 and 81 of integrated FCH4 The rivers draining theCCC contributed to 6 of the basin-wide emissions for CO2and 22 for CH4 although the contribution in stream sur-face area was only 11 The low contribution of FCO2 fromthe CCC to the basin-wide emissions was due to the lowerk600 values although pCO2 values were higher than the restof the basin (Table 1) In the case of FCH4 the much higherCH4 concentrations in the CCC overcome the lower k600 val-ues FN2O per surface area in rivers and streams outside theCCC were relatively similar for Strahler orders 10 to 3 andincreased for Strahler orders 1 and 2 (Table 1) FN2O persurface area in rivers and streams draining the CCC steadily



Figure 21 Time series of dissolved N2O saturation level (N2O in ) in surface waters and freshwater discharge (grey line) in the Congo (atKisangani 2013ndash2018) the Oubangui (at Bangui 2010ndash2012) and the Kasaiuml (at Dima 2015ndash2017) rivers The black line shows a five samplerunning average Data in the Oubangui were previously reported by Bouillon et al (2012 2013)

Figure 22 Seasonal amplitude of dissolved CH4 concentration(1CH4 in nmol Lminus1) in surface waters as a function of the ra-tio of seasonal maximum and minimum of freshwater discharge(Qmax Qmin) in the Congo (at Kisangani 2013ndash2018) the Ouban-gui (at Bangui 2010ndash2012) and the Kasaiuml (at Dima 2015ndash2017)rivers (Fig 20)

decreased from Strahler order 8 to 1 with rivers of orders 53 2 and 1 acting as sinks for N2O Consequently the relativecontribution per Strahler order of integrated FN2O was lessskewed than for integrated FCO2 and FCH4 Strahler or-ders 1ndash4 contributed 699 of integrated FN2O comparedto gt 90 for FCO2 and FCH4

The rivers and streams draining the CCC were a very smallsink of atmospheric N2O (minus001 GgN2O-N yrminus1) while therivers and streams outside the CCC were a source of N2O(51 GgN2O-N yrminus1) The integrated FN2O for the CongoRiver network was 51plusmn10 GgN2O-N yrminus1 corresponded to14 ndash17 of total riverine emissions of N2O reported byHu et al (2016) Note that the N2O riverine emissions com-puted by Hu et al (2016) were indirectly computed from dataon global nitrogen deposition on catchments and on emis-sion factors rather than derived from direct measurements ofdissolved N2O concentrations The estimates given by Hu etal (2016) are more conservative than older estimates (egKroeze et al 2010) because they are based on revised emis-sion factors and converge with a similar more recent studyby Maavara et al (2018) Our estimate of the integratedFN2O is consistent with the range of N2O emissions of 38



Tabl

e1

Part

ial

pres

sure

ofC

O2

(pC

O2

inpp

m)

diss

olve

dC

H4

conc

entr

atio

n(n

mol

Lminus

1 )

diss

olve

dN

2Osa

tura

tion

leve

l(

N2O

in

)ga

str

ansf

erve

loci

ty(k

600

incm

hminus1 )

ai

rndashw

ater

fluxe

sof

CO

2(F

CO

2in

mm

olmminus

2dminus

1an

din

TgC

yrminus

1 )C

H4

(FC

H4

inmicrom

olmminus

2dminus

1an

din

TgC

H4

yrminus

1 )a

ndN

2O(F

N2O

inmicrom

olmminus

2dminus

1an

din

GgN

2O-N

yrminus

1 )

stre

amsu

rfac

ear

ea(S

Ai

nkm

2 )a

ndas

afu

nctio

nof

Stra

hler

stre

amor

der

(SO

)in

the

Con

goR

iver

netw

ork

for

river

san

dst

ream

sdr

aini

ngan

dno

tdra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

Val

ues

inita

licw

ere

extr

apol

ated

tost

ream

orde

r1u

sing

alin

earr

egre

ssio

nw

ithhi

gher

orde

rsor

byus

ing

the

sam

eva

lue

asfo

rstr

eam

orde

r2(F

igS

16)

SO

p

CO

2C

H4

N2O

N

2Ok

600

Tem

pF

CO

2F

CH

4F

N2O

SA

F

CO

2F

CH

4F

N2O

ppm

nmol

Lminus

1nm

olLminus

1

cmhminus

1C

mm

olm

2dminus

1microm

olm

2dminus

1microm

olm

2dminus

1km

2T

gCyrminus

1T

gCH

4yrminus

1G

gN2O

-Nyrminus

1

Dra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

112

304

8349

05

73

264

270

2892

6173

2minus

431

350

44

012

6minus

015

42

1230

48

349

27

441

161

270

1766

3774

7minus

159

335

26

007

4minus

005

43

1192

374

114

977

513

026

113

7625

765

minus5

232

42

00

049

minus0

017

412

065

9409

65

104

512

326

713

2333

431

09

291

17

005

70

003

513

705

9533

20

322

125

270

1525

3325

5minus

148

278

19

005

4minus

004

26

7830

532

132

214

712

227

583

418

5524

748

01

80

005

012

17

6643

503

113

188

811

727

567

015

8218

216

30

50

002

003

08

6977

401

129

219

813

628

782

415

8128

636

61

30

003

010

7

Not

drai

ning

the

Cuv

ette

Cen

tral

eC

ongo

lais

e

110

719

1584

94

136

616

16

241

1541

766

695

107

222

3515

09

087

12

449

289

2112

759

413

58

565

241

4450

1917

136

421

4341

80

240

079

83

7225

1185

93

133

431

124

019

7295

5418

918

8316

30

105

036

44

5766

726

85

126

321

524

310

6540

9810

117

528

20

042

018

15

4110

538

86

131

417

225

558

924

529

616

884

40

024

016

56

1929

299

93

143

416

825

623

913

2812

917

721

90

014

023

37

2667

220

93

153

616

327

534

310

3015

421

683

30

013

034

08

3010

226

67

109

212

927

731

083

72

116

962

30

008

003

69

2521

170

87

145

710

927

721

753

38

846

394

40

014

041

910

3445

339

415

31

204

277

574

178

192

646

16

000

10

127

Tota

l25

11

75

1



to 43 GgN2O-N yrminus1 given by Maavara et al (2018) for theCongo river network also based on an indirect calculationbased on nitrogen deposition and emission factors

The integrated FCH4 for the Congo River network of17plusmn 03 TgCH4 yrminus1 is nearly 2 times higher than the es-timate given by Bastviken et al (2011) for all tropical riversand corresponds to 6 of the global emission of CH4 fromrivers given by Stanley et al (2016) while the surface areaof the Congo River network corresponds to a lower propor-tion (3 ) of the global riverine surface area (773 000 km2Allen and Pavelsky) Note that the meta-analysis of Stanleyet al (2016) includes part of our dataset from the CongoRiver as published by Borges et al (2015a) The inte-grated FCH4 we report for the Congo River network is alsomore than 3 times higher than the estimate of CH4 emis-sions for Amazonian large rivers reported by Sawakuchi etal (2014) (049 TgCH4 yrminus1) The integrated FCH4 we re-port for the Congo River network corresponds to 7 ofthe emission from the Congo wetlands inferred from re-motely sensed atmospheric CH4 data (Bloom et al 2010)(257 TgCH4 yrminus1) However the top-down estimate givenby Bloom et al (2011) includes the CH4 emission from allecosystems over the Congo basin and should also includethe fluvial emissions Note that the CH4 emissions we re-port for the Congo River basin only include the diffusiveflux component when the ebullitive CH4 emission com-ponent represents the majority of CH4 emissions from in-land waters (Bastviken et al 2011) which would be consis-tent with the gap between our emission estimates and thosefrom atmospheric CH4 inventories Finally note that the CH4emissions were only calculated and integrated for the riversand streams draining the CCC but not for the actual wet-land flooded area of the CCC The emission of CH4 fromactual wetland flooded area of the CCC can be estimatedto a massive 51 TgCH4 yrminus1 by extrapolating the area av-eraged FCH4 from the streams (24 468 micromol mminus2 dminus1 Ta-ble 1) to the flooded extent (360 000 km2 Bwangoy et al2010) This corresponds to about 29 of the CH4 emis-sions from natural wetlands (sim 180 TgCH4 yrminus1 Saunois etal 2010) and would in this case be higher than the estimateof CH4 emissions from the Congo wetlands inferred fromremotely sensed atmospheric CH4 data (Bloom et al 2010)

The integrated FCO2 for the Congo River network is251plusmn 46 TgC yrminus1 and is equivalent to the CO2 emissionvalue for rivers globally given by Cole et al (2007) and to14 and 39 of the CO2 emission value for rivers glob-ally given by Raymond et al (2013) (18 000 TgC yrminus1) andLauerwald et al (2015) (650 TgC yrminus1) respectively The in-tegrated CO2 emission from the Congo River network corre-sponds to 44 of the CO2 emissions from tropical (24 Nndash24 S) oceans globally (565 TgC yrminus1) and 183 of CO2emissions from the tropical Atlantic Ocean (137 TgC yrminus1)based on the Takahashi et al (2002) FCO2 climatology Theterrestrial net ecosystem exchange (NEE) of the watershedof the Congo River can be estimated based on the NEE

estimate of 23 gC mminus2 dminus1 for savannahs given by Ciais etal (2011) and of 20 gC mminus2 dminus1 for forests given by Fisheret al (2013) and based on the respective land cover fromGLC2009 (30 savannah and 70 forest) The correspond-ing terrestrial NEE of 77 TgC yrminus1 is more than 3 times lowerthan the riverine CO2 emission from the Congo River This isextremely surprising since the hydrological export from terrafirme forests of DOC and DIC that are assumed to sustain flu-vial emissions are typically 2 ndash3 compared to terrestrialNEE (Kindler et al 2011 Deirmendjian et al 2018) Hy-drological carbon export is higher compared to NEE in Eu-ropean grasslands (on average 22 ) (Kindler et al 2011)We ignore if this is transposable to tropical grasslands suchas savannahs although they only occupy 30 of the Congocatchment surface Accordingly the CO2 emission from theCongo River network should have been an order of magni-tude lower than the estimates of terrestrial NEE from terrafirme biomes rather than more than 3 times higher Howeverin wetlands such as peatlands in Europe the hydrological ex-port of DOC and DIC represent 109 of NEE and is enoughto sustain the riverine CO2 emission that represents 17 ofNEE (Billett et al 2004) Indeed the carbon export fromflooded forests to riverine waters of the Congo basin can beroughly estimated to 396 TgC yrminus1 and is in excess of the in-tegrated FCO2 (calculated from the export per surface areaof flooded forest of 1100 gC mminus2 yrminus1 reported by Abril etal 2014 for the central Amazon and surface area of floodedforest of the CCC 360 000 kmminus2 Bwangoy et al 2010) Al-together this would then strongly suggest that the CO2 emis-sion from the lowland Congo River network is to a large ex-tent sustained by another source of carbon than from the ter-restrial terra firme biome The most likely alternative sourcewould be wetlands (flooded forest and aquatic macrophytes)in agreement with the analysis in the central Amazon Riverby Abril et al (2013)

4 Conclusions

Net heterotrophy in rivers and lakes sustained by inputs oforganic matter from the terrestrial vegetation on the catch-ments is the prevailing paradigm to explain oversaturation ofCO2 in inland surface waters and corresponding emissionsto the atmosphere based on process studies the earliest inthe Amazon (Wissmar et al 1981) and boreal systems (DelGiorgio et al 1999 Prairie et al 2002) and then general-ized at global scales for lakes (Cole et al 1994) and rivers(Cole and Caraco 2001) Yet the comparison of 169 mea-surements of aquatic NCP and FCO2 estimates in the centralCongo River network covering a wide range of size and typeof rivers and streams shows that the aquatic NCP cannot ac-count for fluvial FCO2 This implies that lateral inputs ofCO2 sustain a large part of the CO2 emissions from riversand streams in the Congo River network



The comparison of the integrated CO2 emission from theCongo River network with terrestrial terra firme NEE showsthat it unlikely that fluvial CO2 emissions are sustained bylateral hydrological transfer of carbon from terra firme In-deed integrated FCO2 from the river network was morethan 3 times higher than terrestrial NEE when forests typ-ically only export a very small fraction of NEE as carbonto rivers that can sustain fluvial CO2 emissions It is thenlikely that the fluvial CO2 emissions from the central CongoRiver are sustained by organic matter inputs as well as directCO2 inputs from extensive riparian wetlands (flooded forestand aquatic floating macrophytes) This is consistent with thestable isotopic signature of DIC the differences in the spatialdistribution of dissolved CO2 O2 and CH4 between riversand streams draining or not draining the CCC a large wet-land region in the core of the basin and based on the correla-tion between pCO2 levels and the cover of flooded forest inthe catchment Indeed the calculated export of carbon fromthe CCC to the riverine network is sufficient to sustain thefluvial CO2 emission from the Congo

The fact that fluvial CO2 emissions in lowland rivers areto a large extent sustained by carbon inputs from wetlandsin addition to those from terra firme has consequences forthe conceptualization of statistical and mechanistic modelsof carbon cycling in river networks While progress has beenmade in integrating wetland connectivity in mechanistic re-gional models (Lauerwald et al 2017) this has not been thecase so far for statistical global models that rely on terres-trial (terra firme) productivity (Lauerwald et al 2015) Thecomparison of the output of such a statistical model for theCongo River with observational data (Fig S18) shows thatthe model fails to represent spatial gradients and the higherpCO2 values of streams and rivers draining the CCC in par-ticular This illustrates how ignoring the riverndashwetland con-nectivity can lead to the misrepresentation of pCO2 dynam-ics in river networks in particular tropical ones that accountfor the vast majority (80 ) of global riverine CO2 emissions

Data availability Full dataset is available at httpszenodoorgrecord3413449 (Borges and Bouillon 2019)

Supplement The supplement related to this article is available on-line at httpsdoiorg105194bg-16-3801-2019-supplement

Author contributions AVB and SB conceived the study AVB FDTL ET ATS TB JB CRT and SB collected field samples AVBFD TL CM CRT JPD and SB made laboratory analysis GHA andTL carried out GIS analysis AVB drafted the manuscript with sub-stantial inputs from SB All authors contributed to the manuscript

Competing interests The authors declare that they have no conflictof interest

Acknowledgements We thank the Isotope Hydrology Laboratoryof IAEA for analyses of water stable isotope ratios which con-tribute to the Coordinated Research Project CRPF33021 (Applica-tion and development of isotope techniques to evaluate human im-pacts on water balance and nutrient dynamics of large river basins)implemented by the IAEA Jonathan Richir (University of Liegravege)for advice about the statistical analysis Sandro Petrovic and Marc-Vincent Commarieu (University of Liegravege) Bruno Leporcq (UN-amur) and Zita Kelemen (KU Leuven) for analytical assistancethe Reacutegie des Voies Fluviales (RVF Kinshasa) for providing waterheight data and the two anonymous reviewers for their commentson the previous version of the paper Publication costs were partlycovered by European Geosciences Union as part of the 2018 Out-standing Reviewer Award to Alberto V Borges Alberto V Borgesis a senior research associate at the FNRS

Financial support This research has been supported by the Euro-pean Research Council (grant no 240002) the Fonds National dela Recherche Scientifique (grant no 14711103) the Belgian Fed-eral Science Policy (BELSPO grant no COBAFISH SDAR05A)the Research Foundation Flanders (FWO-Vlaanderen) the Re-search Council of the KU Leuven and the Fonds Leopold-III pourlrsquoExploration et Conservation de la Nature The Boyekoli-Ebale-Congo 2010 Expedition was funded by the Belgian DevelopmentCooperation BELSPO and Belgian National Lottery

Review statement This paper was edited by Ji-Hyung Park and re-viewed by two anonymous referees

References

Abril G and Borges A V Carbon leaks from flooded land do weneed to re-plumb the inland water active pipe Biogeosciences16 769ndash784 httpsdoiorg105194bg-16-769-2019 2019

Abril G Martinez J-M Artigas L F Moreira-Turcq PBenedetti M F Vidal L Meziane T Kim J-H BernardesM C Savoye N Deborde J Albeacuteric P Souza M FL Souza E L and Roland F Amazon river carbon diox-ide outgassing fuelled by wetlands Nature 505 395ndash398httpsdoiorg101038nature12797 2014

Abril G Bouillon S Darchambeau F Teodoru C R Mar-wick T R Tamooh F Omengo F O Geeraert N Deir-mendjian L Polsenaere P and Borges A V Technical noteLarge overestimation of pCO2 calculated from pH and alkalinityin acidic organic-rich freshwaters Biogeosciences 12 67ndash78httpsdoiorg105194bg-12-67-2015 2015

Aho K S and Raymond P A Differential responseof greenhouse gas evasion to storms in forested andwetland streams J Geophys Res 124 649ndash662httpsdoiorg1010292018JG004750 2019



Allen G H and Pavelsky T M Global ex-tent of rivers and streams Science 28 eaat0636httpsdoiorg101126scienceaat0636 2018

Almeida R M Pacheco F S Barros N Rosi E andRoland F Extreme floods increase CO2 outgassing froma large Amazonian river Limnol Oceanogr 62 989ndash999httpsdoiorg101002lno10480 2017

Alsdorf D Beighley E Laraque A Lee H TshimangaR OrsquoLoughlin F Maheacute G Dinga B Moukandi Gand Spencer R G M Opportunities for hydrologic re-search in the Congo Basin Rev Geophys 54 378ndash409httpsdoiorg1010022016RG000517 2016

Amaral J H F Borges A V Melack J M Sarmento HBarbosa P M Kasper D Melo M L de Fex Wolf Dda Silva J S and Forsberg B R Influence of plank-ton metabolism and mixing depth on CO2 dynamics in anAmazon floodplain lake Sci Total Environ 630 1381ndash1393httpsdoiorg101016jscitotenv201802331 2018

APHA Standard methods for the examination of water and wastew-ater American Public Health Association 1325 pp 1998

Balagizi C M Darchambeau F Bouillon S Yalire M M Lam-bert T and Borges A V River geochemistry chemical weath-ering and atmospheric CO2 consumption rates in the VirungaVolcanic Province (East Africa) Geochem Geophy Geosy 162637ndash2660 httpsdoiorg1010022015GC005999 2015

Barbosa P M Melack J M Farjalla V F Amaral J H FScofield V and Forsberg B R Diffusive methane fluxesfrom Negro Solimotildees and Madeira rivers and fringing lakesin the Amazon basin Limnol Oceanogr 61 S221ndashS237httpsdoiorg101002lno10358 2016

Bastviken D Ejlertsson J and Tranvik L Measure-ment of methane oxidation in lakes A comparisonof methods Environ Sci Technol 36 3354ndash3361httpsdoiorg101021es010311p 2002

Bastviken D Tranvik L J Downing J A Crill PM and Enrich-Prast A Freshwater methane emissionsoffset the continental carbon sink Science 331 p 50httpsdoiorg101126science1196808 2011

Battin T J Kaplan L A Findlay S Hopkinson C S Marti EPackman A I Newbold J D and Sabater F Biophysical con-trols on organic carbon fluxes in fluvial networks Nat Geosci1 95ndash100 httpsdoiorg101038ngeo101 2008

Baulch H M Schiff S L Maranger R and Dillon PJ Nitrogen enrichment and the emission of nitrous ox-ide from streams Global Biogeochem Cy 25 GB4013httpsdoiorg1010292011GB004047 2011

Benstead J P and Leigh D S An expandedrole for river networks Nat Geosci 5 678ndash679httpsdoiorg101038ngeo1593 2012

Billett M F Palmer S M Hope D Deacon C Storeton-West R Hargreaves K J Flechard C and FowlerD Linking land-atmosphere-stream carbon fluxes in a low-land peatland system Global Biogeochem Cy 18 GB1024httpsdoiorg1010292003GB002058 2004

Bird M I and Pousai P Variations of δ13C in the surfacesoil organic carbon pool Global Biogeoch Cy 11 313ndash322httpsdoiorg10102997GB01197 1997

Bird M I Giresse P and Chivas A R Effect of forest and sa-vanna vegetation on the carbon-isotope composition from the

Sanaga River Cameroon Limnol Oceanogr 39 1845ndash1854httpsdoiorg104319lo19943981845 1994

Bowen G J Wassenaar L I and Hobson K A Global appli-cation of stable hydrogen and oxygen isotopes to wildlife foren-sics Oecologia 143 337ndash348 httpsdoiorg101007s00442-004-1813-y 2005

Bloom A A Palmer P I Fraser A Reay D S and Franken-berg C Large-scale controls of methanogenesis inferred frommethane and gravity spaceborne data Science 327 322ndash325httpsdoiorg101126science1175176 2010

Borges A V Darchambeau F Teodoru C R Marwick T RTamooh F Geeraert N Omengo F O Gueacuterin F LambertT Morana C Okuku E and Bouillon S Globally signifi-cant greenhouse gas emissions from African inland waters NatGeosci 8 637ndash642 httpsdoiorg101038NGEO2486 2015a

Borges A V Abril G Darchambeau F Teodoru C R DebordeJ Vidal L O Lambert T and Bouillon S Divergent biophys-ical controls of aquatic CO2 and CH4 in the Worldrsquos two largestrivers Sci Rep 5 15614 httpsdoiorg101038srep156142015b

Borges A V Darchambeau F Lambert T Bouillon SMorana C Brouyegravere S Hakoun V Jurado A Tseng H-C Descy J-P and Roland F A E Effects of agricul-tural land use on fluvial carbon dioxide methane and ni-trous oxide concentrations in a large European river theMeuse (Belgium) Sci Total Environ 610611 342ndash355httpsdoiorg101016jscitotenv201708047 2018

Borges A V and Bouillon S Data-base of CO2 CH4 N2O andancillary data in the Congo River available at httpszenodoorgrecord3413449XYm2eUYzaUk last access 24 Septem-ber 2019

Bouillon S Abril G Borges A V Dehairs F Govers GHughes H J Merckx R Meysman F J R Nyunja J Os-burn C and Middelburg J J Distribution origin and cyclingof carbon in the Tana River (Kenya) a dry season basin-scale sur-vey from headwaters to the delta Biogeosciences 6 2475ndash2493httpsdoiorg105194bg-6-2475-2009 2009

Bouillon S Yambeacuteleacute A Spencer R G M Gillikin D PHernes P J Six J Merckx R and Borges A V Organic mat-ter sources fluxes and greenhouse gas exchange in the Ouban-gui River (Congo River basin) Biogeosciences 9 2045ndash2062httpsdoiorg105194bg-9-2045-2012 2012

Bouillon S Yambeacuteleacute A Gillikin D P Teodoru C Dar-chambeau F Lambert T and Borges A V Contrastingbiogeochemical characteristics of right-bank tributaries anda comparison with the mainstem Oubangui River CentralAfrican Republic (Congo River basin) Sci Rep 4 5402httpsdoiorg101038srep05402 2014

Bultot F Atlas Climatique du Bassin Congolais Publica-tions de LrsquoInstitut National pour LrsquoEtude Agronomique duCongo (INEAC) Troisieme Partie Temperature et Humiditede LrsquoAir Rosee Temperature du Sol 253 pp 1972

Butman D and Raymond P A Significant efflux of carbon diox-ide from streams and rivers in the United States Nat Geosci 4839ndash842 httpsdoiorg101038NGEO1294 2011

Bwangoy J-R B Hansen M C Roy D P De Grandi Gand Justice C O Wetland mapping in the Congo Basinusing optical and radar remotely sensed data and derived



topographical indices Remote Sens Environ 114 73ndash86httpsdoiorg101016jrse200908004 2010

Canion A Overholt W A Kostka J E Huettel M Lavik Gand Kuypers M M M Temperature response of denitrificationand anaerobic ammonium oxidation rates and microbial commu-nity structure in Arctic fjord sediments Environ Microbiol 163331ndash3344 httpsdoiorg1011111462-292012593 2014

Cardoso S J Enrich-Prast A Pace M L and Roland FDo models of organic carbon mineralization extrapolate towarmer tropical sediments Limnol Oceanogr 59 48ndash54httpsdoiorg104319lo20145910048 2014

Ciais P Bombelli A Williams M Piao S L Chave J RyanC M Henry M Brender P and Valentini R The carbon bal-ance of Africa synthesis of recent research studies Philos T RSoc A 369 2038ndash2057 httpsdoiorg101098rsta201003282013

Coplen T B and Wassenaar L I LIMS for Lasers 2015for achieving long-term accuracy and precision of δ2Hδ17O and δ18O of waters using laser absorption spec-trometry Rapid Commun Mass Spectr 29 2122ndash2130httpsdoiorg101002rcm73722015

Cole B E and Cloern J E An empirical model for estimatingphytoplankton productivity in estuaries Mar Ecol Prog Ser36 299ndash305 httpsdoiorg103354meps036299 1987

Cole J J and Caraco N F Carbon in catchments connecting ter-restrial carbon losses with aquatic metabolism Mar Fresh Res52 101ndash110 httpsdoiorg101071MF00084 2001

Cole J J Caraco N F Kling G W and Kratz T K Carbondioxide supersaturation in the surface waters of lakes Science265 1568ndash1570 httpsdoiorg101126science265517815681994

Cole J J Prairie Y T Caraco N F McDowell W H TranvikL J Striegl R G Duarte C M Kortelainen P Downing JA Middelburg J J and Melack J Plumbing the global carboncycle Integrating inland waters into the terrestrial carbon bud-get Ecosystems 10 171ndash184 httpsdoiorg101007s10021-006-9013-8 2007

Coynel A Seyler P Etcheber H Meybeck M and Orange DSpatial and seasonal dynamics of total suspended sediment andorganic carbon species in the Congo River Global BiogeochemCy 19 GB4019 httpsdoiorg1010292004GB002335 2005

Craine J M Elmore A J Wang L Aranibar J Bauters MBoeckx P Crowley B E Dawes M A Delzon S FajardoA Fang Y Fujiyoshi L Gray A Guerrieri R GundaleM J Hawke DJ Hietz P Jonard M Kearsley E KenzoT Makarov M Marantildeoacuten-Jimeacutenez S McGlynn T P Mc-Neil B E Mosher S G Nelson D M Peri P L RoggyJ C Sanders-DeMott R Song M Szpak P Templer P HVan der Colff D Werner C Xu X Yang Y Yu G andZmudczynska-Skarbek K Isotopic evidence for oligotrophica-tion of terrestrial ecosystems Nat Ecol Evol 2 1735ndash1744httpsdoiorg101038s41559-018-0694-0 2018

Crawford J T Stanley E H Dornblaser M and StrieglR G CO2 time series patterns in contrasting headwa-ter streams of North America Aquat Sci 79 473ndash486httpsdoiorg101007s00027-016-0511-2 2017

Dargie G C Lewis S L Lawson I T Mitchard E T A PageS E Bocko Y E and Ifo S A Age extent and carbon storage

of the central Congo Basin peatland complex Nature 542 86ndash90 httpsdoiorg101038nature21048 2017

Deirmendjian L and Abril G Carbon dioxide degassing at thegroundwater-stream-atmosphere interface isotopic equilibrationand hydrological mass balance in a sandy watershed J Hydrol558 129ndash143 2018

Deirmendjian L Loustau D Augusto L Lafont S ChipeauxC Poirier D and Abril G Hydro-ecological controls on dis-solved carbon dynamics in groundwater and export to streamsin a temperate pine forest Biogeosciences 15 669ndash691httpsdoiorg105194bg-15-669-2018 2018

Dinsmore K J Wallin M B Johnson M S Billett M FBishop K Pumpanen J and Ojala A Contrasting CO2concentration discharge dynamics in headwater streams amulti-catchment comparison J Geophys Res 118 445ndash461httpsdoiorg101002jgrg20047 2013

Del Giorgio P A Cole J J Caraco N F and Pe-ters R H Linking planktonic biomass and metabolismto net gas fluxes in northern temperate lakes Ecol-ogy 80 1422ndash1431 httpsdoiorg1018900012-9658(1999)080[1422LPBAMT]20CO2 1999

Descy J-P Hardy M-A Steacutenuite S Pirlot S Lep-orcq B Kimirei I Sekadende B Mwaitega S R andSinyenza D Phytoplankton pigments and community com-position in Lake Tanganyika Freshwater Biol 50 668ndash684httpsdoiorg101111j1365-2427200501358x 2005

Descy J-P Darchambeau F Lambert T Stoyneva M PBouillon S and Borges A V Phytoplankton dynam-ics in the Congo River Freshwater Biol 62 87ndash101httpsdoiorg101111fwb12851 2017

Doctor D H Kendall C Sebestyen S D Shanley J B OhteN and Boyer E W Carbon isotope fractionation of dissolvedinorganic carbon (DIC) due to outgassing of carbon dioxidefrom a headwater stream Hydrol Process 22 2410ndash2423httpsdoiorg101002hyp6833 2008

Downing J A Cole J J Duarte C M Middelburg J JMelack J M Prairie Y T Kortelainen P Striegl R G Mc-Dowell W H and Tranvik L J Global abundance and sizedistribution of streams and rivers Inland Waters 2 229ndash236httpsdoiorg105268IW-24502 2012

Duvert C Butman D E Marx A Ribolzi O and Hut-ley L B CO2 evasion along streams driven by groundwa-ter inputs and geomorphic controls Nat Geosci 11 813ndash818httpsdoiorg101038s41561-018-0245-y 2018

Fisher J B Sikka M Sitch S Ciais P Poulter B GalbraithD Lee J-E Huntingford C Viovy N Zeng N AhlstroumlmA Lomas M R Levy P E Frankenberg C Saatchi S andMalhi Y African tropical rainforest net carbon dioxide fluxesin the twentieth century Philos T R Soc B 368 20120376httpsdoiorg101098rstb20120376 2013

Fluet-Chouinard E Lehner B Rebelo L-M Papa F andHamilton S K Development of a global inundation map athigh spatial resolution from topographic downscaling of coarse-scale remote sensing data Remote Sens Environ 158 348ndash361httpsdoiorg101016jrse201410015 2015

Frankignoulle M Borges A and Biondo R A new de-sign of equilibrator to monitor carbon dioxide in highly dy-namic and turbid environments Water Res 35 1344ndash1347httpsdoiorg101016S0043-1354(00)00369-9 2001



Gaillardet J Dupreacute B Louvat P and Allegravegre C J Globalsilicate weathering and CO2 consumption rates deduced fromthe chemistry of large rivers Chem Geol 159 3ndash30httpsdoiorg101016S0009-2541(99)00031-5 1999

Gillikin D P and Bouillon S Determination of δ18O of water andδ13C of dissolved inorganic carbon using a simple modificationof an elemental analyzer ndash isotope ratio mass spectrometer (EA-IRMS) an evaluation Rapid Commun Mass Spectr 21 1475ndash1478 httpsdoiorg101002rcm2968 2007

Gran G Determination of the equivalence point in po-tentiometric titrations Part II The Analyst 77 661ndash671httpsdoiorg101039AN9527700661 1952

Hamilton S K Sippel S J and Melack J M Com-parison of inundation patterns among major South Ameri-can floodplains J Geophys Res 107 LBA 5-1-LBA 5-14httpsdoiorg1010292000JD000306 2002

Happell J Chanton J P and Showers W The influence ofmethane oxidation on the stable isotopic composition of methaneemitted from Florida Swamp forests Geochim Cosmochim Ac58 4377ndash4388 httpsdoiorg1010160016-7037(94)90341-71994

Hedges J I Clark W A Quay P D Richey J E Devol AH and de M Santos U Compositions and fluxes of particulateorganic material in the Amazon River Limnol Oceanogr 31717ndash738 httpsdoiorg104319lo19863140717 1986

Hotchkiss E R Hall Jr R O Sponseller R A ButmanD Klaminder J Laudon H Rosvall M and Karlsson JSources of and processes controlling CO2 emissions changewith the size of streams and rivers Nat Geosci 8 696ndash699httpsdoiorg101038ngeo2507 2015

Hu M Chen D and Dahlgren R A Modeling nitrous oxideemission from rivers a global Assessment Glob Change Biol22 3566ndash3582 httpsdoiorg101111gcb13351 2016

Hughes R H and Hughes J S A directory of African wetlandsIUCN ISBN 2-88032-949-3 820 pp 1992

Huotari J Haapanala S Pumpanen J Vesala T andOjala A Efficient gas exchange between a boreal riverand the atmosphere Geophys Res Lett 40 5683ndash5686httpsdoiorg1010022013GL057705 2013

Kindler R Siemens J Kaiser K Walmsley D C BernhoferC Buchmann N Cellier P Eugster W Gleixner G Grun-wald T Heim A Ibrom A Jones S K Jones M KlumppK Kutsch W Steenberg Larsen K Lehuger S Loubet BMcKenzie R Moors E Osborne B Pilegaard K Reb-mann C Saunders M Schmidt M W I Schrumpf M Seyf-ferth J Skiba U Soussana J-F Sutton M A Tefs CVowinckel B Zeeman M J and Kaupenjohann M Dis-solved carbon leaching from soil is a crucial component of netecosystem carbon balance Glob Change Biol 17 1167ndash1185httpsdoiorg101111j1365-2486201002282x 2011

Klaus M Geibrink E Jonsson A Bergstroumlm A-K BastvikenD Laudon H Klaminder J and Karlsson J Green-house gas emissions from boreal inland waters unchangedafter forest harvesting Biogeosciences 15 5575ndash5594httpsdoiorg105194bg-15-5575-2018 2018

Kokic J Sahleacutee E Sobek S Vachon D and Wallin M BHigh spatial variability of gas transfer velocity in streams re-vealed by turbulence measurements Inland Waters 8 461ndash473httpsdoiorg1010802044204120181500228 2018

Koneacute Y J M Abril G Kouadio K N Delille B and BorgesA V Seasonal variability of carbon dioxide in the rivers andlagoons of Ivory Coast (West Africa) Estuar Coast 32 246ndash260 httpsdoiorg101007s12237-008-9121-0 2009

Koneacute Y J M Abril G Delille B and Borges A VSeasonal variability of methane in the rivers and lagoonsof Ivory Coast (West Africa) Biogeochemistry 100 21ndash37httpsdoiorg101007s10533-009-9402-0 2010

Kosten S Pintildeeiro M de Goede E de Klein J Lamers L P Mand Ettwig K Fate of methane in aquatic systems dominated byfree-floating plants Water Res 104 200ndash207 2016

Kroeze C Dumont E and Seitzinger S P Future trends in emis-sions of N2O from rivers and estuaries J Integr Environ Sc 771ndash78 httpsdoiorg1010801943815X2010496789 2010

Lambert T Bouillon S Darchambeau F Massicotte P andBorges AV Shift in the chemical composition of dissolved or-ganic matter in the Congo River network Biogeosciences 135405ndash5420 httpsdoiorg105194bg-13-5405-2016 2016

Laraque A Mietton M Olivry J C and Pandi A Impact oflithological and vegetal covers on flow discharge and water qual-ity of Congolese tributaries from the Congo river Rev Sci Eau11 209ndash224 1998

Laraque A Bricquet J P Pandi A and Olivry J C A reviewof material transport by the Congo River and its tributaries Hy-drol Process 23 3216ndash3224 httpsdoiorg101002hyp73952009

Lauerwald R Laruelle G G Hartmann J Ciais P andRegnier P A G Spatial patterns in CO2 evasion from theglobal river network Global Biogeochem Cy 29 534ndash554httpsdoiorg1010022014GB004941 2015

Lauerwald R Regnier P Camino-Serrano M Guenet B Guim-berteau M Ducharne A Polcher J and Ciais P OR-CHILEAK (revision 3875) a new model branch to simu-late carbon transfers along the terrestrialndashaquatic continuumof the Amazon basin Geosci Model Dev 10 3821ndash3859httpsdoiorg105194gmd-10-3821-2017 2017

Le T T H Fettig J and Meon G Kinetics and simulation ofnitrification at various pH values of a polluted river in the tropicsEcohydrol Hydrobiol 19 54ndash65 2019

Liptay K Chanton J Czepiel P and Mosher B Useof stable isotopes to determine methane oxidation inlandfill cover soils J Geophys Res 103 8243ndash8250httpsdoiorg10102997JD02630 1998

Liss P S and Slater P G Flux of gases across the air sea interfaceNature 247 181ndash184 httpsdoiorg101038247181a0 1974

Liu S and Raymond P A Hydrologic controls on pCO2 and CO2efflux in US streams and rivers Limnol Oceanogr Lett 3 428ndash435 httpsdoiorg101002lol210095 2018

Lynch J K Beatty C M Seidel M P Jungst L J andDeGrandpre M D Controls of riverine CO2 over an an-nual cycle determined using direct high temporal resolu-tion pCO2 measurements J Geophys Res 115 G03016httpsdoiorg1010292009JG001132 2010

Maavara T Lauerwald R Laruelle GG Akbarzadeh ZBouskill N J Van Cappellen P and Regnier P Ni-trous oxide emissions from inland waters Are IPCCestimates too high Glob Change Biol 25 473ndash488httpsdoiorg101111gcb14504 2018



Malhi Y Adu-Bredu S Asare R A Lewis S L and MayauxP African rainforests past present and future Philos T R SocB 368 20120312 httpsdoiorg101098rstb20120312 2013

Mann P J Spencer R G M Dinga B J Poulsen J RHernes P J Fiske G Salter M E Wang Z A Ho-ering K A Six J and Holmes R M The biogeochem-istry of carbon across a gradient of streams and rivers withinthe Congo Basin J Geophys Res-Biogeo 119 687ndash702httpsdoiorg1010022013JG002442 2014

Maurice L Rawlins B G Farr G Bell R and GooddyD C The influence of flow and bed slope on gastransfer in steep streams and their implications for eva-sion of CO2 J Geophys Res-Biogeo 122 2862ndash2875httpsdoiorg1010022017JG004045 2017

Marwick T R Tamooh F Ogwoka B Teodoru C Borges AV Darchambeau F and Bouillon S Dynamic seasonal nitro-gen cycling in response to anthropogenic N loading in a tropicalcatchment AthindashGalanandashSabaki River Kenya Biogeosciences11 1ndash18 httpsdoiorg105194bg-11-1-2014 2014

Marx A Dusek J Jankovec J Sanda M Vogel T vanGeldern R Hartmann J and Barth J A C A re-view of CO2 and associated carbon dynamics in headwaterstreams A global perspective Rev Geophys 55 560ndash585httpsdoiorg1010022016RG000547 2017

McDowell M J and Johnson M S Gas transfer velocitiesevaluated using carbon dioxide as a tracer show high stream-flow to be a major driver of total CO2 evasion flux for aheadwater stream J Geophys Res-Biogeo 123 2183ndash2197httpsdoiorg1010292018JG004388 2018

Melack J M Hess L L Gastil M Forsberg B R Hamil-ton S K Lima I B T and Novo E M L M Regional-ization of methane emissions in the Amazon Basin with mi-crowave remote sensing Glob Change Biol 10 530ndash544httpsdoiorg101111j1365-2486200400763x 2004

Meybeck M Global chemical weathering of surficial rocks esti-mated from river dissolved loads Am J Sci 287 401ndash428httpsdoiorg102475ajs2875401 1987

Millero F J The thermodynamics of the carbonate systemin seawater Geochem Cosmochem Ac 43 1651ndash1661httpsdoiorg1010160016-7037(79)90184-41979

Morana C Borges A V Roland F A E DarchambeauF Descy J-P and Bouillon S Methanotrophy withinthe water column of a large meromictic tropical lake(Lake Kivu East Africa) Biogeosciences 12 2077ndash2088httpsdoiorg105194bg-12-2077-2015 2015

Nkounkou R R and Probst J L Hydrology and geochemistryof the Congo river system Mitt GeolndashPalaont Inst Univ Ham-burg SCOPEUNEP 64 483ndash508 1987

OrsquoLoughlin F Trigg M A Schumann G J-P and BatesP D Hydraulic characterization of the middle reach ofthe Congo River Water Resour Res 49 5059ndash5070httpsdoiorg101002wrcr20398 2013

Peter H Singer G A Preiler C Chifflard P Steniczka G andBattin T J Scales and drivers of temporal pCO2 dynamics inan Alpine stream J Geophys Res-Biogeo 119 1078ndash1091httpsdoiorg1010022013JG002552 2014

Powell R L Yoo E-H and Still C J Vegetation and soilcarbon-13 isoscapes for South America integrating remote sens-

ing and ecosystem isotope measurements Ecosphere 3 1ndash25httpsdoiorg101890ES12-001621 2012

Prairie Y T Bird D F and Cole J J The sum-mer metabolic balance in the epilimnion of southeast-ern Quebec lakes Limnol Oceanogr 47 316ndash321httpsdoiorg104319lo20024710316 2002

Raymond P A Zappa C J Butman D Bott T L Potter CMulholland P Laursen A E McDowell W H and NewboldD Scaling the gas transfer velocity and hydraulic geometry instreams and small rivers Limnol Oceanogr Fluids Environ 241ndash53 httpsdoiorg10121521573689-1597669 2012

Raymond P A Hartmann J Lauerwald R Sobek S Mc-Donald C Hoover M Butman D Striegl R Mayorga EHumborg C Kortelainen P Duumlrr H Meybeck M Ciais Pand Guth P Global carbon dioxide emissions from inland wa-ters Nature 503 355ndash359 httpsdoiorg101038nature127602013

Reiman J H and Xu J Y Diel variability of pCO2 andCO2 outgassing from the Lower Mississippi River Implica-tions for riverine CO2 outgassing estimation Water 11 1ndash15httpsdoiorg103390w11010043 2019

Richardson D C Newbold J D Aufdenkampe A K Tay-lor P G and Kaplan L A Measuring heterotrophic res-piration rates of suspended particulate organic carbon fromstream ecosystems Limnol Oceanogr-Method 11 247ndash261httpsdoiorg104319lom201311247 2013

Richey J E Devol A H Wofy S C Victoria R and RiberioM N G Biogenic gases and the oxidation and reduction of car-bon in Amazon River and floodplain waters Limnol Oceanogr33 551ndash561 httpsdoiorg104319lo19883340551 1988

Richey J E Melack J M Aufdenkampe A K Ballester V Mand Hess L Outgassing from Amazonian rivers and wetlandsas a large tropical source of atmospheric CO2 Nature 416 617ndash620 httpsdoiorg101038416617a 2002

Runge J Large Rivers Geomorpholgy and Management editedby Gupta A John Wiley amp Sons 293ndash309 2008

Santos I R Maher D T and Eyre B D Couplingautomated radon and carbon dioxide measurements incoastal waters Environ Sci Technol 46 7685ndash7691httpsdoiorg101021es301961b 2012

Saunois M Bousquet P Poulter B Peregon A Ciais PCanadell J G Dlugokencky E J Etiope G Bastviken DHouweling S Janssens-Maenhout G Tubiello F N CastaldiS Jackson R B Alexe M Arora V K Beerling D J Berga-maschi P Blake D R Brailsford G Brovkin V BruhwilerL Crevoisier C Crill P Kovey K Curry C Frankenberg CGedney N Houmlglund-Isaksson L Ishizawa M Ito A Joos FKim H-S Kleinen T Krummel P Lamarque J-F Langen-felds R Locatelli R Machida T Maksyutov S McDonaldK C Marshall J Melton J R Morino I Naik V OrsquoDohertyS Parmentier F-J W Patra P K Peng C Peng S PetersG Pison I Prigent C Prinn R Ramonet M Riley W JSaito M Sanyini M Schroeder R Simpson I J SpahniR Steele P Takizawa A Thornton B F Tian H TohjimaY Viovy N Voulgarakis A van Weele M van der Werf GWeiss R Wiedinmyer C Wilton D J Wiltshire A WorthyD Wunch D B Xu X Yoshida Y Zhang B Zhang Z andZhu Q The global methane budget Earth Syst Sci Data 8697ndash751 httpsdoiorg105194essd-8-697-2016 2016



Sawakuchi H O Bastviken D Sawakuchi A O KruscheA V Ballester M V R and Richey J E Methane emis-sions from Amazonian Rivers and their contribution to theglobal methane budget Glob Change Biol 20 2829ndash2840httpsdoiorg101111gcb12646 2014

Sawakuchi H O Bastviken D Sawakuchi A O Ward N DBorges C D Tsai S M Richey J E Ballester M V R andKrusche A V Oxidative mitigation of aquatic methane emis-sions in large Amazonian rivers Glob Change Biol 22 1075ndash1085 httpsdoiorg101111gcb13169 2016

Scofield V Melack J M Barbosa P M Amaral J H F Fors-berg B R and Farjalla V F Carbon dioxide outgassing fromAmazonian aquatic ecosystems in the Negro River basin Bio-geochemistry 129 77ndash91 httpsdoiorg101007s10533-016-0220-x 2016

Seitzinger S P and Kroeze C Global distribution of ni-trous oxide production and N inputs in freshwater and coastalmarine ecosystems Global Biogeochem Cy 12 93ndash113httpsdoiorg10102997GB03657 1998

Simpson H and Herczeg A Stable isotopes as an indicator ofevaporation in the River Murray Australia Water Resour Res27 1925ndash1935 httpsdoiorg10102991WR00941 1991

Spencer R G M Hernes P J Aufdenkampe A K Baker AGulliver P Stubbins A Aiken G R Dyda R Y Butler KD Mwamba V L Mangangu A M Wabakanghanzi J Nand Six J An initial investigation into the organic matter bio-geochemistry of the Congo River Geochim Cosmochim Ac84 614ndash627 httpsdoiorg101016jgca201201013 2012

SCA (Standing committee of Analysts) Ammonia in waters Meth-ods for the examination of waters and associated materials 16pp 1981

Stanley E H Casson N J Christel S T Crawford J T LokenL C and Oliver S K The ecology of methane in streams andrivers patterns controls and global significance Ecol Monogr86 146ndash171 httpsdoiorg10189015-1027 2016

Still C J and Powell R L Continental-scale distributions ofvegetation stable carbon isotope ratios edited by West J BBowen G J Dawson T E Tu K P Isoscapes the Nether-lands Springer Netherlands 179ndash193 2010

Takahashi T Sutherland S C Wanninkhof R Sweeney CFeely R A Chipman D W Hales B Friederich G ChavezF Sabine C Watson A Bakker D C E Schuster U MetzlN Yoshikawa-Inoue HI Ishii M Midorikawa T Nojiri YKoumlrtzinger A Steinhoff T Hoppema M Olafsson J Arnar-son T S Tilbrook B Johannessen T Olsen A Bellerby RWong C S Delille B Bates N R and de Baar H J W Cli-matological mean and decadal change in surface ocean pCO2and net sea-air CO2 flux over the global oceans Deep-Sea ResPt II 56 554ndash577 httpsdoiorg101016jdsr22008120092009

Tamooh F Borges A V Meysman F J R Van Den Meer-sche K Dehairs F Merckx R and Bouillon SDynamicsof dissolved inorganic carbon and aquatic metabolism in theTana River basin Kenya Biogeosciences 10 6911ndash6928httpsdoiorg105194bg-10-6911-2013 2013

Teodoru C R Nyoni F C Borges A V Darachambeau FNyambe I and Bouillon S Spatial variability and temporal dy-namics of greenhouse gas (CO2 CH4 N2O) concentrations andfluxes along the Zambezi River mainstem and major tributaries

Biogeosciences 12 2431ndash2453 httpsdoiorg105194bg-12-2431-2015 2015

Tyler S C Bilek R S Sass R L and Fisher F MMethane oxidation and pathways of production in a Texaspaddy field deduced from measurements of flux δ13Cand δD of CH4 Global Biogeochem Cy 11 323ndash348httpsdoiorg10102997GB01624 1997

Ulseth A J Hall Jr R O Canadell M B Madinger H LNiayifar A and Battin T J Distinct airndashwater gas exchangeregimes in low- and high-energy streams Nat Geosci 12 259ndash263 httpsdoiorg101038s41561-019-0324-8 2019

Upstill-Goddard R C Salter M E Mann P J Barnes JPoulsen J Dinga B Fiske G J and Holmes R MThe riverine source of CH4 and N2O from the Republic ofCongo western Congo Basin Biogeosciences 14 2267ndash2281httpsdoiorg105194bg-14-2267-2017 2017

Ward N D Krusche A V Sawakuchi H O Brito D CCunha A C Sousa Moura J M da Silva R Yager P LKeil R G and Richey J E The compositional evolution ofdissolved and particulate organic matter along the lower Ama-zon RiverndashOacutebidos to the ocean Mar Chem 177 244ndash256httpsdoiorg101016jmarchem201506013 2015

Ward N D Sawakuchi H O Neu V Less D F S ValerioA M Cunha A C Kampel M Bianchi T S Krusche AV Richey J E and Keil R G Velocity-amplified microbialrespiration rates in the lower Amazon River Limnol OceanogrLett 3 265ndash274 httpsdoiorg101002lol210062 2018

Wassenaar L I Coplen T B and Aggarwal P K Ap-proaches for achieving long-term accuracy and precisionof δ18O and δ2H for waters analyzed using laser absorp-tion spectrometers Environ Sci Technol 48 1123ndash1131httpsdoiorg101021es403354n 2014

Weiss R F Determinations of carbon dioxide and methane by dualcatalyst flame ionization chromatography and nitrous oxide byelectron capture chromatography J Chromatogr Sci 19 611ndash616 httpsdoiorg101093chromsci1912611 1981

Weiss R F and Price B A Nitrous oxide solubility in water andseawater Mar Chem 8 347ndash359 httpsdoiorg1010160304-4203(80)90024-9 1980

Wissmar R C Richey J E Stallard R F and Edmond J MPlankton metabolism and carbon processes in the Amazon riverits tributaries and floodplain waters Peru-Brazil MayndashJune1977 Ecology 62 1622ndash1633 httpsdoiorg10230719415171981

Yoshida N Iguchi H Yurimoto H Murakami A and Sakai YAquatic plant surface as a niche for methanotrophs Front Micro-biol 30 1ndash9 httpsdoiorg103389fmicb201400030 2014

Zhou L Tian Y Myneni R B Ciais P Saatchi SL Yi Y Shilong P Chen H Vermote E F SongC and Hwang T Widespread decline of Congo rain-forest greenness in the past decade Nature 509 86ndash90httpsdoiorg101038nature13265 2014


Abstract

Introduction

Material and methods

Field expeditions and fixed monitoring

Continuous measurements

Discrete sampling

Metabolism measurements

Sample conditioning and laboratory analysis

GIS

Statistical analysis

Results and discussion

Spatial variations along the Kisangani--Kinshasa transect of general limnological variables and dissolved GHGs

Spatial variations along the Kwa transect

Spatial variations as a function of stream order and as a function of the influence of the CCC

Drivers of CO2 dynamics -- metabolic measurements and daily variations in CO2

Drivers of CO2 dynamics -- stable isotope composition of DIC

Seasonal variations

Variability of GHG fluxes in the Congo river network

Significance of integrated GHG fluxes at basin and global scales

Conclusions

Data availability

Supplement

Author contributions

Competing interests

Acknowledgements

Financial support

Review statement

References


ing between 1087 and 22 899 ppm (equilibrium sim 400 ppm)and dissolved CH4 concentrations ranging between 22 and71 428 nmol Lminus1 (equilibrium sim 2 nmol Lminus1) Spatial vari-ations were overwhelmingly more important than seasonalvariations for pCO2 CH4 and N2O as well as dayndashnightvariations for pCO2 The wide range of pCO2 and CH4 vari-ations was consistent with the equally wide range of O2(03 ndash1228 ) and of dissolved organic carbon (DOC)(18ndash678 mg Lminus1) indicative of generation of these twogreenhouse gases from intense processing of organic mat-ter either in ldquoterra firmerdquo soils wetlands or in-stream How-ever the emission rate of CO2 to the atmosphere from river-ine surface waters was on average about 10 times higherthan the flux of CO2 produced by aquatic net heterotrophy(as evaluated from measurements of pelagic respiration andprimary production) This indicates that the CO2 emissionsfrom the river network were sustained by lateral inputs ofCO2 (either from terra firme or from wetlands) The pCO2and CH4 values decreased and O2 increased with increas-ing Strahler order showing that stream size explains part ofthe spatial variability of these quantities In addition severallines of evidence indicate that lateral inputs of carbon fromwetlands (flooded forest and aquatic macrophytes) were ofparamount importance in sustaining high CO2 and CH4 con-centrations in the Congo river network as well as drivingspatial variations the rivers draining the CCC were charac-terized by significantly higher pCO2 and CH4 and signifi-cantly lower O2 and N2O values than those not drain-ing the CCC pCO2 and O2 values were correlated tothe coverage of flooded forest on the catchment The fluxof greenhouse gases (GHGs) between rivers and the atmo-sphere averaged 2469 mmol mminus2 dminus1 for CO2 (range 86 and7110 mmol mminus2 dminus1) 12 553 micromol mminus2 dminus1 for CH4 (range65 and 597 260 micromol mminus2 dminus1) and 22 micromol mminus2 dminus1 forN2O (rangeminus52 and 319 micromol mminus2 dminus1) The estimate of in-tegrated CO2 emission from the Congo River network (251plusmn46 TgC (1012 gC) yrminus1) corresponding to nearly half theCO2 emissions from tropical oceans globally (565 TgC yrminus1)and was nearly 2 times the CO2 emissions from the tropi-cal Atlantic Ocean (137 TgC yrminus1) Moreover the integratedCO2 emission from the Congo River network is more than3 times higher than the estimate of terrestrial net ecosys-tem exchange (NEE) on the whole catchment (77 TgC yrminus1)This shows that it is unlikely that the CO2 emissions from theriver network were sustained by the hydrological carbon ex-port from terra firme soils (typically very small compared toterrestrial NEE) but most likely to a large extent they weresustained by wetlands (with a much higher hydrological con-nectivity with rivers and streams)

1 Introduction

Emissions to the atmosphere of greenhouse gases (GHGs)such as carbon dioxide (CO2) methane (CH4) and nitrousoxide (N2O) from inland waters (rivers lakes and reser-voirs) might be quantitatively important for global budgets(Seitzinger and Kroeze 1998 Cole et al 2007 Bastviken etal 2011) Yet there are very large uncertainties in the esti-mates of GHGs emission to the atmosphere from rivers asreflected in the wide range of reported values between 02and 18 PgC (1015 gC) yrminus1 for CO2 (Cole et al 2007 Ray-mond et al 2013) 2 and 27 TgCH4 (1012 gCH4) yrminus1 forCH4 (Bastviken et al 2011 Stanley et al 2016) and 32and 2100 GgN2O-N (109 gN2O-N) yrminus1 for N2O (Kroeze etal 2010 Hu et al 2016) This uncertainty is mainly due tothe scarcity of data in the tropics that account for the major-ity of riverine GHG emissions (sim 80 for CO2 Raymondet al 2013 Borges et al 2015a b Lauerwald et al 201579 for N2O Hu et al 2016 70 for CH4 Sawakushiet al 2014) but also to scaling procedures of varying com-plexity that use different input data (Raymond et al 2013Borges et al 2015b Lauerwald et al 2015) in addition touncertainty in the estimate of surface area of rivers (Down-ing et al 2012 Raymond et al 2013 Allen and Pavelsky2018) and the parameterization of the gas transfer velocity(k) (Raymond et al 2012 Huotari et al 20013 Maurice etal 2017 Kokic et al 2018 McDowell and Johnson 2018Ulseth et al 2019) The exchange of CO2 between rivers andthe atmosphere is in most cases computed from the airndashwatergradient of the CO2 concentration and k parameterized as afunction of stream morphology (eg slope or depth) and wa-ter flow (or discharge) However there can be large errors as-sociated with the computation of the dissolved CO2 concen-tration from pH and total alkalinity (TA) for low alkalinitywaters in particular so that high-quality direct measurementsof dissolved CO2 concentration are preferred (Abril et al2015) although scarce In the tropics research on GHGs inrivers has mainly focussed on South American rivers and onthe central Amazon in particular (Richey et al 1988 2002Melack et al 2004 Abril et al 2014 Barbosa et al 2016Scofield et al 2016) while until recently African rivers werenearly uncharted with a few exceptions (Koneacute et al 20092010)

There is also a lack of understanding of the drivers ofthe fluvial concentrations of GHGs hence ultimately ofthe drivers of their exchange with the atmosphere It is un-clear to what extent CO2 emissions from rivers are sus-tained by in situ net heterotrophy andor by lateral inputsof CO2 The global organic carbon degradation by net het-erotrophy of rivers and streams given by Battin et al (2008)of 02 PgC yrminus1 is insufficient to sustain global riverineCO2 emissions given by the most recent estimates of 07ndash18 PgC yrminus1 (Raymond et al 2013 Lauerwald et al 2015)suggesting an important role of lateral CO2 inputs in sustain-ing emissions to the atmosphere from rivers In a regional







































F = k1G (2)



k600 = 842+ 11838timesV S (3)






)minus ln


)][αminus 1] (4)













26 GIS








































































































= 088)









= 076)




























Tabl

e1

Part

ial

pres

sure

ofC

O2

(pC

O2

inpp

m)

diss

olve

dC

H4

conc

entr

atio

n(n

mol

Lminus

1 )

diss

olve

dN

2Osa

tura

tion

leve

l(

N2O

in

)ga

str

ansf

erve

loci

ty(k

600

incm

hminus1 )

ai

rndashw

ater

fluxe

sof

CO

2(F

CO

2in

mm

olmminus

2dminus

1an

din

TgC

yrminus

1 )C

H4

(FC

H4

inmicrom

olmminus

2dminus

1an

din

TgC

H4

yrminus

1 )a

ndN

2O(F

N2O

inmicrom

olmminus

2dminus

1an

din

GgN

2O-N

yrminus

1 )

stre

amsu

rfac

ear

ea(S

Ai

nkm

2 )a

ndas

afu

nctio

nof

Stra

hler

stre

amor

der

(SO

)in

the

Con

goR

iver

netw

ork

for

river

san

dst

ream

sdr

aini

ngan

dno

tdra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

Val

ues

inita

licw

ere

extr

apol

ated

tost

ream

orde

r1u

sing

alin

earr

egre

ssio

nw

ithhi

gher

orde

rsor

byus

ing

the

sam

eva

lue

asfo

rstr

eam

orde

r2(F

igS

16)

SO

p

CO

2C

H4

N2O

N

2Ok

600

Tem

pF

CO

2F

CH

4F

N2O

SA

F

CO

2F

CH

4F

N2O

ppm

nmol

Lminus

1nm

olLminus

1

cmhminus

1C

mm

olm

2dminus

1microm

olm

2dminus

1microm

olm

2dminus

1km

2T

gCyrminus

1T

gCH

4yrminus

1G

gN2O

-Nyrminus

1

Dra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

112

304

8349

05

73

264

270

2892

6173

2minus

431

350

44

012

6minus

015

42

1230

48

349

27

441

161

270

1766

3774

7minus

159

335

26

007

4minus

005

43

1192

374

114

977

513

026

113

7625

765

minus5

232

42

00

049

minus0

017

412

065

9409

65

104

512

326

713

2333

431

09

291

17

005

70

003

513

705

9533

20

322

125

270

1525

3325

5minus

148

278

19

005

4minus

004

26

7830

532

132

214

712

227

583

418

5524

748

01

80

005

012

17

6643

503

113

188

811

727

567

015

8218

216

30

50

002

003

08

6977

401

129

219

813

628

782

415

8128

636

61

30

003

010

7

Not

drai

ning

the

Cuv

ette

Cen

tral

eC

ongo

lais

e

110

719

1584

94

136

616

16

241

1541

766

695

107

222

3515

09

087

12

449

289

2112

759

413

58

565

241

4450

1917

136

421

4341

80

240

079

83

7225

1185

93

133

431

124

019

7295

5418

918

8316

30

105

036

44

5766

726

85

126

321

524

310

6540

9810

117

528

20

042

018

15

4110

538

86

131

417

225

558

924

529

616

884

40

024

016

56

1929

299

93

143

416

825

623

913

2812

917

721

90

014

023

37

2667

220

93

153

616

327

534

310

3015

421

683

30

013

034

08

3010

226

67

109

212

927

731

083

72

116

962

30

008

003

69

2521

170

87

145

710

927

721

753

38

846

394

40

014

041

910

3445

339

415

31

204

277

574

178

192

646

16

000

10

127

Tota

l25

11

75

1







4 Conclusions













References



















































































































































Abstract

Introduction




Discrete sampling



GIS








Seasonal variations



Conclusions

Data availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References






































F = k1G (2)



k600 = 842+ 11838timesV S (3)






)minus ln


)][αminus 1] (4)













26 GIS








































































































= 088)









= 076)




























Tabl

e1

Part

ial

pres

sure

ofC

O2

(pC

O2

inpp

m)

diss

olve

dC

H4

conc

entr

atio

n(n

mol

Lminus

1 )

diss

olve

dN

2Osa

tura

tion

leve

l(

N2O

in

)ga

str

ansf

erve

loci

ty(k

600

incm

hminus1 )

ai

rndashw

ater

fluxe

sof

CO

2(F

CO

2in

mm

olmminus

2dminus

1an

din

TgC

yrminus

1 )C

H4

(FC

H4

inmicrom

olmminus

2dminus

1an

din

TgC

H4

yrminus

1 )a

ndN

2O(F

N2O

inmicrom

olmminus

2dminus

1an

din

GgN

2O-N

yrminus

1 )

stre

amsu

rfac

ear

ea(S

Ai

nkm

2 )a

ndas

afu

nctio

nof

Stra

hler

stre

amor

der

(SO

)in

the

Con

goR

iver

netw

ork

for

river

san

dst

ream

sdr

aini

ngan

dno

tdra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

Val

ues

inita

licw

ere

extr

apol

ated

tost

ream

orde

r1u

sing

alin

earr

egre

ssio

nw

ithhi

gher

orde

rsor

byus

ing

the

sam

eva

lue

asfo

rstr

eam

orde

r2(F

igS

16)

SO

p

CO

2C

H4

N2O

N

2Ok

600

Tem

pF

CO

2F

CH

4F

N2O

SA

F

CO

2F

CH

4F

N2O

ppm

nmol

Lminus

1nm

olLminus

1

cmhminus

1C

mm

olm

2dminus

1microm

olm

2dminus

1microm

olm

2dminus

1km

2T

gCyrminus

1T

gCH

4yrminus

1G

gN2O

-Nyrminus

1

Dra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

112

304

8349

05

73

264

270

2892

6173

2minus

431

350

44

012

6minus

015

42

1230

48

349

27

441

161

270

1766

3774

7minus

159

335

26

007

4minus

005

43

1192

374

114

977

513

026

113

7625

765

minus5

232

42

00

049

minus0

017

412

065

9409

65

104

512

326

713

2333

431

09

291

17

005

70

003

513

705

9533

20

322

125

270

1525

3325

5minus

148

278

19

005

4minus

004

26

7830

532

132

214

712

227

583

418

5524

748

01

80

005

012

17

6643

503

113

188

811

727

567

015

8218

216

30

50

002

003

08

6977

401

129

219

813

628

782

415

8128

636

61

30

003

010

7

Not

drai

ning

the

Cuv

ette

Cen

tral

eC

ongo

lais

e

110

719

1584

94

136

616

16

241

1541

766

695

107

222

3515

09

087

12

449

289

2112

759

413

58

565

241

4450

1917

136

421

4341

80

240

079

83

7225

1185

93

133

431

124

019

7295

5418

918

8316

30

105

036

44

5766

726

85

126

321

524

310

6540

9810

117

528

20

042

018

15

4110

538

86

131

417

225

558

924

529

616

884

40

024

016

56

1929

299

93

143

416

825

623

913

2812

917

721

90

014

023

37

2667

220

93

153

616

327

534

310

3015

421

683

30

013

034

08

3010

226

67

109

212

927

731

083

72

116

962

30

008

003

69

2521

170

87

145

710

927

721

753

38

846

394

40

014

041

910

3445

339

415

31

204

277

574

178

192

646

16

000

10

127

Tota

l25

11

75

1







4 Conclusions













References



















































































































































Abstract

Introduction




Discrete sampling



GIS








Seasonal variations



Conclusions

Data availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References











26 GIS








































































































= 088)









= 076)




























Tabl

e1

Part

ial

pres

sure

ofC

O2

(pC

O2

inpp

m)

diss

olve

dC

H4

conc

entr

atio

n(n

mol

Lminus

1 )

diss

olve

dN

2Osa

tura

tion

leve

l(

N2O

in

)ga

str

ansf

erve

loci

ty(k

600

incm

hminus1 )

ai

rndashw

ater

fluxe

sof

CO

2(F

CO

2in

mm

olmminus

2dminus

1an

din

TgC

yrminus

1 )C

H4

(FC

H4

inmicrom

olmminus

2dminus

1an

din

TgC

H4

yrminus

1 )a

ndN

2O(F

N2O

inmicrom

olmminus

2dminus

1an

din

GgN

2O-N

yrminus

1 )

stre

amsu

rfac

ear

ea(S

Ai

nkm

2 )a

ndas

afu

nctio

nof

Stra

hler

stre

amor

der

(SO

)in

the

Con

goR

iver

netw

ork

for

river

san

dst

ream

sdr

aini

ngan

dno

tdra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

Val

ues

inita

licw

ere

extr

apol

ated

tost

ream

orde

r1u

sing

alin

earr

egre

ssio

nw

ithhi

gher

orde

rsor

byus

ing

the

sam

eva

lue

asfo

rstr

eam

orde

r2(F

igS

16)

SO

p

CO

2C

H4

N2O

N

2Ok

600

Tem

pF

CO

2F

CH

4F

N2O

SA

F

CO

2F

CH

4F

N2O

ppm

nmol

Lminus

1nm

olLminus

1

cmhminus

1C

mm

olm

2dminus

1microm

olm

2dminus

1microm

olm

2dminus

1km

2T

gCyrminus

1T

gCH

4yrminus

1G

gN2O

-Nyrminus

1

Dra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

112

304

8349

05

73

264

270

2892

6173

2minus

431

350

44

012

6minus

015

42

1230

48

349

27

441

161

270

1766

3774

7minus

159

335

26

007

4minus

005

43

1192

374

114

977

513

026

113

7625

765

minus5

232

42

00

049

minus0

017

412

065

9409

65

104

512

326

713

2333

431

09

291

17

005

70

003

513

705

9533

20

322

125

270

1525

3325

5minus

148

278

19

005

4minus

004

26

7830

532

132

214

712

227

583

418

5524

748

01

80

005

012

17

6643

503

113

188

811

727

567

015

8218

216

30

50

002

003

08

6977

401

129

219

813

628

782

415

8128

636

61

30

003

010

7

Not

drai

ning

the

Cuv

ette

Cen

tral

eC

ongo

lais

e

110

719

1584

94

136

616

16

241

1541

766

695

107

222

3515

09

087

12

449

289

2112

759

413

58

565

241

4450

1917

136

421

4341

80

240

079

83

7225

1185

93

133

431

124

019

7295

5418

918

8316

30

105

036

44

5766

726

85

126

321

524

310

6540

9810

117

528

20

042

018

15

4110

538

86

131

417

225

558

924

529

616

884

40

024

016

56

1929

299

93

143

416

825

623

913

2812

917

721

90

014

023

37

2667

220

93

153

616

327

534

310

3015

421

683

30

013

034

08

3010

226

67

109

212

927

731

083

72

116

962

30

008

003

69

2521

170

87

145

710

927

721

753

38

846

394

40

014

041

910

3445

339

415

31

204

277

574

178

192

646

16

000

10

127

Tota

l25

11

75

1







4 Conclusions













References



















































































































































Abstract

Introduction




Discrete sampling



GIS








Seasonal variations



Conclusions

Data availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References




























































































= 088)









= 076)




























Tabl

e1

Part

ial

pres

sure

ofC

O2

(pC

O2

inpp

m)

diss

olve

dC

H4

conc

entr

atio

n(n

mol

Lminus

1 )

diss

olve

dN

2Osa

tura

tion

leve

l(

N2O

in

)ga

str

ansf

erve

loci

ty(k

600

incm

hminus1 )

ai

rndashw

ater

fluxe

sof

CO

2(F

CO

2in

mm

olmminus

2dminus

1an

din

TgC

yrminus

1 )C

H4

(FC

H4

inmicrom

olmminus

2dminus

1an

din

TgC

H4

yrminus

1 )a

ndN

2O(F

N2O

inmicrom

olmminus

2dminus

1an

din

GgN

2O-N

yrminus

1 )

stre

amsu

rfac

ear

ea(S

Ai

nkm

2 )a

ndas

afu

nctio

nof

Stra

hler

stre

amor

der

(SO

)in

the

Con

goR

iver

netw

ork

for

river

san

dst

ream

sdr

aini

ngan

dno

tdra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

Val

ues

inita

licw

ere

extr

apol

ated

tost

ream

orde

r1u

sing

alin

earr

egre

ssio

nw

ithhi

gher

orde

rsor

byus

ing

the

sam

eva

lue

asfo

rstr

eam

orde

r2(F

igS

16)

SO

p

CO

2C

H4

N2O

N

2Ok

600

Tem

pF

CO

2F

CH

4F

N2O

SA

F

CO

2F

CH

4F

N2O

ppm

nmol

Lminus

1nm

olLminus

1

cmhminus

1C

mm

olm

2dminus

1microm

olm

2dminus

1microm

olm

2dminus

1km

2T

gCyrminus

1T

gCH

4yrminus

1G

gN2O

-Nyrminus

1

Dra

inin

gth

eC

uvet

teC

entr

ale

Con

gola

ise

112

304

8349

05

73

264

270

2892

6173

2minus

431

350

44

012

6minus

015

42

1230

48

349

27

441

161

270

1766

3774

7minus

159

335

26

007

4minus

005

43

1192

374

114

977

513

026

113

7625

765

minus5

232

42

00

049

minus0

017

412

065

9409

65

104

512

326

713

2333

431

09

291

17

005

70

003

513

705

9533

20

322

125

270

1525

3325

5minus

148

278

19

005

4minus

004

26

7830

532

132

214

712

227

583

418

5524

748

01

80

005

012

17

6643

503

113

188

811

727

567

015

8218

216

30

50

002

003

08

6977

401

129

219

813

628

782

415

8128

636

61

30

003

010

7

Not

drai

ning

the

Cuv

ette

Cen

tral

eC

ongo

lais

e

110

719

1584

94

136

616

16

241

1541

766

695

107

222

3515

09

087

12

449

289

2112

759

413

58

565

241

4450

1917

136

421

4341

80

240

079

83

7225

1185

93

133

431

124

019

7295

5418

918

8316

30

105

036

44

5766

726

85

126

321

524

310

6540

9810

117

528

20

042

018

15

4110

538

86

131

417

225

558

924

529

616

884

40

024

016

56

1929

299

93

143

416

825

623

913

2812

917

721

90

014

023

37

2667

220

93

153

616

327

534

310

3015

421

683

30

013

034

08

3010

226

67

109

212

927

731

083

72

116

962

30

008

003

69

2521

170

87

145

710

927

721

753

38

846

394

40

014

041

910

3445

339

415

31

204

277

574

178

192

646

16

000

10

127

Tota

l25

11

75

1







4 Conclusions













References



















































































































































Abstract

Introduction




Discrete sampling



GIS








Seasonal variations



Conclusions

Data availability

Supplement


Competing interests

Acknowledgements

Financial support

Review statement

References

variations in dissolved greenhouse gases (co in the congo

Documents