DOC and DIC in flowpaths of Amazonian headwater catchments

13 Pages·2006·0.28 MB·English

Biogeochemistry(2006)81:45–57
DOI10.1007/s10533-006-9029-3
ORIGINAL PAPER
DOC and DIC in flowpaths of Amazonian headwater
catchments with hydrologically contrasting soils
Mark S. Johnson Æ Johannes Lehmann Æ
Eduardo Guimara˜es Couto
Joa˜o Paulo Nova˜es Filho Æ Susan J. Riha
Received:21July2005/Accepted:25April2006/Publishedonline:6July2006
(cid:1)SpringerScience+BusinessMediaB.V.2006
Abstract Organic and inorganic carbon (C) These different hydrologic responses are attrib-
fluxes transported by water were evaluated for uted primarily to large differences in saturated
dominant hydrologic flowpaths on two adjacent hydraulic conductivity (K). Overland flow was
s
headwater catchments in the Brazilian Amazon found to be an important feature on both water-
with distinct soils and hydrologic responses from sheds. This was evidenced by the response rates
September 2003 through April 2005. The Ultisol- of overland flow detectors (OFDs) during the
dominated catchment produced 30% greater rainy season, with overland flow intercepted by
volumeofstorm-relatedquickflow(overlandflow 54 ± 0.5%and65 ± 0.5%ofOFDsfortheOxisol
and shallow subsurface flow) compared to the and Ultisol watersheds respectively during bi-
Oxisol-dominated catchment. Quickflow fluxes weekly periods. Small volumes of quickflow cor-
were equivalent to 3.2 ± 0.2% of event precipi- respond to large fluxes of dissolved organic C
tation for the Ultisol catchment, compared to (DOC); DOC concentrations of the hydrologic
2.5 ± 0.3% for the Oxisol-dominated watershed flowpathsthatcomprisequickflowareanorderof
(mean response ±1 SE, n = 27 storms for each magnitude higher than groundwater flowpaths
watershed).Hydrologicresponseswerealsofaster fueling base flow (19.6 ± 1.7 mg l–1 DOC for
on the Ultisol watershed, with time to peak flow overland flow and 8.8 ± 0.7 mg l–1 DOC for
occurring 10 min earlier on average as compared shallow subsurface flow versus 0.50 ± 0.04,mg l–1
to the runoff response on the Oxisol watershed. DOC in emergent groundwater). Concentrations
ofdissolved inorganicC(DIC, asdissolvedCO –
2
C plus HCO––C) in groundwater were found to
3
be an order of magnitude greater than quickflow
M.S.Johnson(&)ÆJ.Lehmann
DICconcentrations(21.5 mg l–1DICinemergent
DepartmentofCropandSoilSciences,Cornell
University,918BradfieldHall,Ithaca,NY14853, groundwater versus 1.1 mg l–1 DIC in overland
USA flow). The importance of deeper flowpaths in the
e-mail:[email protected]
transportofinorganicCtostreamsisindicatedby
the 40:1 ratio of DIC:DOC for emergent
E.G.CoutoÆJ.P.N.Filho
DepartmentofSoilScience,FederalUniversityof groundwater. Dissolved CO2–C represented 92%
MatoGrosso,78060-900Cuiaba´,MT,Brazil of DIC in emergent groundwater. Results from
this study illustrate a highly dynamic and tightly
S.J.Riha
coupled linkage between the C cycle and the
DepartmentofEarthandAtmosphericSciences,
CornellUniversity,Ithaca,NY14853,USA hydrologic cycle for both Ultisol and Oxisol
123
46 Biogeochemistry(2006)81:45–57
landscapes: organic C fluxes strongly tied to and even the ejection of soil solutes due to rain-
flowpaths associated with quickflow, and inor- drop impact (Gao et al. 2005). However, surficial
ganic C (particularly dissolved CO ) transported and near-surface flowpaths strongly contrast bio-
2
via deeper flowpaths. geochemically with hydrologic flowpaths that
interact with deeper soil horizons as a result of
Keywords Dissolved carbon dioxide Æ Dissolved adsorptionandmineralizationoforganicnutrients
organic carbon Æ Groundwater Æ Overland flow Æ (Quallset al.2002).
Quickflow Æ Stormflow Thequantificationofcarbonfluxestransported
bywaterfromterrestrialtoaquaticenvironments
is fundamental to resolving the C balance at
Introduction scalesrangingfromcatchment(Billettet al.2004)
to continental (Siemens 2003). Headwater catch-
Biogeochemical cycling within terrestrial ecosys- ments provide the scale at which stream water
temsandacrosstheterrestrial–aquaticinterfaceis exhibits the strongest connection with terrestrial
dynamicallylinkedwiththewatercycle.Notonly flowpaths (Hope et al. 2004). Because much of
is the movement of carbon (C) and nutrients what is transported by storm-event driven quick-
controlled in large part by the movement of flow is not captured in weekly streamwater sam-
water, but also processes of transformations pling strategies, a detailed consideration of C
between biogeochemical forms (e.g. inorganic fluxes of hydrologic flowpaths is needed to refine
andorganic)arestronglyinfluencedbytherateat determinations of terrestrial C transport to
which water cycles through the landscape (McC- streams.
lain and Elsenbeer 2001), exerting a primary Elsenbeer (2001) advanced the concept of
control on biotic factors controlling C minerali- hydrologic end-members for tropical soils as
zation and humification (Zech et al. 1997). comprised of Acrisols (Ultisols in the USDA
Recent advances in hydrologic research have classification), which are dominated by rapid and
refined the conceptualization of hydrologic laterally-oriented flowpaths, and Ferralsols (Ox-
flowpaths and their contributions to stream flow isols) that exhibit slower responses as a result of
(McDonnell 2003). Broadly, these include rapid more vertically-oriented flowpaths. Both Oxisols
flowpaths occurring at or near the soil surface, and Ultisols are highly weathered soils typical of
and slower flowpaths occurring deeper in the soil humid tropical regions, with Ultisols character-
profile. A useful corollary to the distinction of ized by a greater increase in clay content with
rapidity between different flowpaths is their depth. Here, we present results from our
temporal continuity: punctuated versus continu- Amazonian site with the useful features (from a
ous. The concept of quickflow versus deeper research perspective) of adjacent catchments
flowpaths encompasses this distinction. Quick- under the same climatic conditions and forested
flow consists of laterally-oriented overland flow ecosystem, but with hydrologically contrasting
and shallow subsurface storm flow, in addition to soils. The only other published study in which
direct precipitation of throughfall onto stream these end-member soil formations are found in
channels. Deeper flowpaths follow vertically- close proximity was conducted in Panama
oriented percolation in the upper soil horizons (Godsey et al. 2004).
prior to their routing through deeper soil hori- In this paper, we use paired-watersheds with
zons and emergence as groundwater-derived hydrologically contrasting soils to evaluate: (1)
base flow. the role of soil properties in controlling the acti-
The biogeochemical distinction between over- vation ofrapidversus slow flowpaths, and (2) the
land flow and shallow subsurface storm flow can relativeimportanceofthesehydrologicflowpaths
becomeblurredasaresultofexfiltrationofreturn on controlling C fluxes at the terrestrial–aquatic
flow (Walter et al. 2005), emergence of pipe-flow interface for upland forested catchments in the
on upland soils (Elsenbeer and Vertessy 2000), Amazon.
123
Biogeochemistry(2006)81:45–57 47
Methods 8841650
UUlltisol watershed (UW) 1.94 ha
Study site
8841600 287
ResearchwasconductedinthesouthernAmazon
near Juruena, Mato Grosso, Brazil (10(cid:2)28¢ S, 8841550
UW
58(cid:2)28¢ W) in a region characterized by rolling m)
topography and strong seasonality. The study g ( 8841500 281
catchments are located at about 250 m above sea hin 284 OW
lseevqeulenotinalttrhiebutBarraiezsiltiaonassthrieealmd,flaonwdingcoinmtoprtihsee Nort 8841450 278 272 273 281 284
Juruena River. Soils in the region overlie the 278 281
Precambrian gneisses of the Xingu Complex 8841400 278
(Ministry of Mines and Energy (Brazil) 1980), 275
and have pH values ca. 5 (Nova˜es Filho 2005), 272
8841350 OOxxiissooll wwaatteerrsshheedd ((OOWW)) 00..9955 ha
precluding the presence of carbonate minerals.
339250 339300 339350 339400 339450 339500
Mean annual temperature in the region is 24 (cid:2)C,
Easting (m)
with annual precipitation of 2200 mm distributed
Fig.1 LocationmapfortheOxisolwatershed(OW)and
in a unimodal pattern with a five month dry
theUltisolwatershed(UW)inUTMzone21Scoordinates,
season from May–September (Nunes 2003). and1mcontourintervalswith elevationinmetersabove
Two adjacent, forested headwater catchments sealevel.TheUltisolwatershedshowsahigherdegreeof
incisementthantheOxisolwatershed
with contrasting soil physical characteristics were
selected based on an initial recognizance field
campaign that used soil color and degree of in- watershed.Compositesamplesweretakenfrom5
cisement as distinguishing and readily observable sub-samplesobtainedwithin1 m2at10 cmdepth,
criteria. Each catchment consists of a topose- and a single sample was collected at 50 cm. The
quence of hillslope landscape positions and a sample design allowed a robust determination of
perennialfirst-orderstreamthatoriginatesfroma thespatialvariabilityofchangesinsoiltextureand
spring. The results presented in this paper com- other parameters across the landscape and with
prise monitoring and sampling conducted depth. Following the initial soil survey, a soil pit
between September 1, 2003 and April 1, 2005. wasdugforeachwatershedontherepresentative
soil immediately adjacent to the watersheds but
Topographic and soil characterization outside of areas contributing to the catchments.
Increases in clay content with depth across the
A topographic survey was conducted using stan- landscape were determined from comparison of
dard surveying techniques, with transects estab- 10 cmand50 cmdepthsateachsamplepoint,and
lished across each watershed at 20 m intervals were found to be greater for the Ultisol-domi-
perpendicular to the predominant hillslope nated watershed than for the Oxisol-dominated
direction. Elevation was determined every 5 m watershed (61 ± 2% increase versus 45 ± 2%,
along transects. GPS was used to adjust the field mean ± 1 SE, P < 0.001 for two-sample T-test).
grid to UTM coordinates. Elevation data from This feature coupled with more detailed investi-
ASTER (Advanced Spaceborne Thermal Emis- gation of soil diagnostic horizons shows the pre-
sion and Reflection Radiometer) on the Terra dominant soil of the Ultisol watershed to be a
satellite was used to offset the field datum to Plinthic Kandiustult in the USDA classification
meters above sea level (Fig. 1). (Soil Survey Staff 1999), and a Plinthic Acrisol in
Soilsampleswerecollectedattwodepthsevery the FAO classification (FAO-UNESCO 1987).
20 m along transects by auger, comprising 43 ThepredominantsoiloftheOxisolwatershedisa
samples at each depth for the Oxisol watershed Typic Haplustox in USDA classification (Rhodic
and 65 samples at each depth for the Ultisol Ferralsol inthe FAO classification).
123
48 Biogeochemistry(2006)81:45–57
Soil hydraulic conductivity was evaluated in the detector, tipping out any collected water, and
situforeachsoilpitusingmini-diskinfiltrometers noting the presence or absence of overland flow.
(Zhang 1997) with 0.5 cm suction (Decagon De- The unit is then redeployed. For the present
vices, Pullman, WA, USA) with four replicate study, the OFDs were arranged in a semi-ran-
measurements at each depth: 0, 10, 25, 50 and domizedfashion,with5OFDsinstalledineachof
100 cm. Since the matrix potential equivalent to three landscape positions per watershed: plateau
0.5 cm of suction, 0.05 kPa, is indistinguishable (<2% slope), shoulder slope (2–10% slope) and
fromsaturationonsoilwatercharacteristiccurves midslope(>10%slope)foratotalof15OFDsper
(cf., Saxton 2005), we report the soil hydraulic watershed. The presence or absence of overland
conductivity determined in-situ as K. flow was checked biweekly.
s
Hydrologic instrumentation Sample collection and analysis
Each watershed was instrumented with devices Water samples were collected by hand weekly
for recording throughfall and streamflow at from groundwater springs. Samples of through-
5 minute intervals. A water-level recording fall, overland flow and leaching water were col-
device adjacent to a 90(cid:2) V-notch weir at each lected weekly during the rainy season. Spring
watershedoutletwasusedfordeterminingstream water was collected directly from tubing inserted
discharge. Initially we used pressure transducers horizontally into the spring such that emergent
(Telog Instruments, Victor, NY, USA) which groundwater could be collected prior to inter-
were subsequently replaced by water height data action with the riparian zone or the atmosphere.
loggers with thermisters for measuring water and Samples were collected monthly from ground
air temperatures (TruTrack, Christchurch New water wells in each watershed. These 8 m wells
Zealand).Atthetimeofwatersamplecollection, were constructed of 5 cm diameter PVC pipe
stream height at the weir was measured directly slotted over the lower 1.5 m, and were located
and leaves occasionally found trapped in the V- at upper and mid-slope positions in each
notch of the weir were removed, allowing cor- watershed.
rection of the logged record of stream height. Throughfall was collected in PETG bottles
Throughfall was determined from four 200 cm2 attached to stakes and topped with a 10 cm
data-logged rain gauges installed 1 m above the diameter funnel. A plug of glass wool was placed
forest floor (Pronamic, Silkeborg Denmark) con- in the base of the funnel to strain litterfall, which
nected to event data loggers (Onset Computer was removed from the funnel. Overland flow
Corp., Bourne MA, USA). samples for analysis were collected from one
The presenceofoverland flowwas determined large PVC tube in each watershed placed on the
spatially using 15 non-recording overland flow soil surface at locations of concentrated flow-
detectors (OFDs) per watershed (Elsenbeer and paths.Azero-tensionlysimeterinstalledat10 cm
Vertessy 2000; Kirkby et al. 1976). These passive depth in each watershed funneled gravity-flow
OFDs are made from 20 mm ID PVC pipe and water to PETG collection bottles. These free-
consist of a detector section and a reservoir sec- draining lysimeters served as a proxy for shallow
tion connected by a tee. The collector section is subsurface storm flow.
perforated along one side by three rows of 1 mm Water samples were filtered (Whatman GF/F
holes, which are perforated at 1 cm intervals glass fiber filters, 0.7 l m, Middlesex, UK), trea-
alongthe20 cmsectionwith5 mmbetweenrows. ted (HgCl ) and stored at 3(cid:2)C until analysis in
2
The reservoir section is inserted into an installa- pre-muffled glass vials with Teflon-lined tops.
tionholeinthesoilsuchthatsomeofthedetector DOC was determined chromatographically after
holesareincontactwiththesoilsurface;ponding combustion in a TOC analyzer (Multi N/C 3000,
of overland flow will result in water being Analytik Jena, Jena Germany).
collected in the reservoir. Determination of the Dissolvedinorganiccarbon(DIC)components
presence of overland flow consists of uncapping were determined individually for HCO––C and
3
123
Biogeochemistry(2006)81:45–57 49
dissolved CO –C. HCO––C was measured by of quickflow. Throughfall was calculated as the
2 3
titrationwith0.01NH SO topH 4.5(Neal2001). average response per 5-min interval for the four
2 4
DissolvedCO –Cwasmeasuredinthefieldusing throughfall gauges. Comparisons between water-
2
an approach for the in situ deployment of an shedsforhydrologicparametersweremadeusing
infrared gas analyzer (IRGA) in aquatic systems paired T-tests as no series correlation (e.g. auto-
modified from Tang et al. (2003) and Jassal et al. correlation) was found for the storm events
(2004).The IRGAusedwasdesignedtomeasure (Wilks 1995).
CO in harsh and humid environments using a
2
single-beam dual-wavelength, non-dispersive in-
fra-red (NDIR) silicon-based sensor (Vaisala
GMT221, Vantaa, Finland). The IRGA was fur- Results and discussion
therprotectedwithinahighporosityPTFEsleeve
thatishighlypermeabletoCO butimpermeable Activation of rapid flowpaths
2
towater,allowingdissolvedCO fromsolutionto
2
equilibrate with the headspace of the gas bench Runoff responses to rainfall generally began
within the IRGA. within 10 min in both watersheds, but the storm
Dissolved CO concentration of spring water hydrographs were more rapid for the Ultisol wa-
2
was determined by placing the PTFE-sheathed tershed. An analysis of rainy-season stormflow
IRGA within a PVC housing connected at the hydrographsforthetwowatershedsfor27storms
point of groundwater discharge prior to its found the average time to peak (T ) from the
p
emergence and subsequent outgassing to the beginningofthroughfallfortheOxisolwatershed
atmosphere. For determination of CO in stream to lag the Ultisol watershed (Fig. 2). T for the
2 p
water, thePTFE-sheathed IRGA was submerged Oxisol (52 ± 6 min) was greater than T for the
p
inthemainchannelupstreamoftheweir.Inboth Ultisol (42 ±5 min) (P = 0.001).
cases,thegasbenchwasallowedtoequilibratein Quickflow fluxes from the Ultisol were larger
situ for 10 min prior to recording the pCO con- than on the Oxisol (P = 0.017), while the
2
centration, which was adjusted to mg l–1 via quickflow component of the rainfall-runoff re-
Henry’s Law. sponses of both watersheds increased linearly for
Measurement accuracy of this instrument
is ±200 ppm CO (±0.08 mg l–1 as dissolved 0.4
2
CO –C), while precision of the method as indi-
2 Oxisol
cated by standard deviations of replicate samples Ultisol
is ±545 ppm (±0.22 mg l–1 dissolved CO –C).
2 0.3
Hydrologic and statistical analysis
y
c
n
Storm hydrographs were normalized by ue 0.2
q
e
corresponding watershed areas (Fig. 1) to allow Fr
comparison between responses for the two
watersheds. The quickflow component of storm
0.1
hydrographs was determined by separating the
base flow component from total stream flow
response to rainfall using a hydrograph line
separation technique after Hewlett and Hibbert 0.0
0.0 0.5 1.0 1.5 2.0 2.5
(1967).Alineconnectingthebeginningofstream
Tp/Tw
flow response to event precipitation with the
Fig.2 Frequencydistributionsoftimestopeakflow(T )
point at which the change in discharge on the p
normalizedbystormduration(T ).Thehydrographpeak
recession limb stabilized at a value greater than w
frequently occurs before rainfall has ended (T /T < 1)
p w
0:95ð Qt [0:95Þwas used to facilitatecalculations fortheUltisolwatershed
Qt(cid:1)1
123
50 Biogeochemistry(2006)81:45–57
increasing rainfall volumes (r2 = 0.93 for Ultisol andmasswastingofchannelbanksfortheUltisol
and r2 = 0.57 for Oxisol). The average quickflow catchment.
runoff volume per event was found to be TheK valuespresentedinFig. 3arewithinthe
s
3.2 ± 0.2% of event precipitation in the Ultisol- range found for Ultisols (Acrisols) and Oxisols
dominated watershed, compared to 2.5 ± 0.3% (Ferralsols) in the humid tropics (Elsenbeer
for the Oxisol-dominated watershed (mean 2001). However, the differences in hydraulic
response ±1 SE, n = 27 storms for each conductivities between the soils in the present
watershed). Lesack (1993) determined that the study are not as dramatic as those between other
annual mean storm runoff as a percentage of Amazonian soils at La Cuenca (Acrisol) and
rainfall was 2.8% for a 23 ha watershed in the Reserva Ducke (Ferralsol) (reviewed by Elsenb-
Central Amazon, a value similar to those of this eer 2001).
study. Quickflow was less than 4% of total The Rancho Grande site, also located on the
streamflowforeachwatershedonanannualbasis. Brazilian shield and comprised of both Ultisols
Saturated conductivity (K) differed between (Godsey and Elsenbeer 2002) and Oxisols (El-
s
the Ultisol and the Oxisol. The Ultisol exhibited senbeer et al. 1999), is the Amazonian research
an initial decrease in K with depth from the soil site most directly comparable with the Juruena
s
surfaceto50 cm,whileK oftheOxisolincreased site.IncreasedK withdepthfromthesurfacewas
s s
with depth from the soil surface (Fig. 3). Declin- also observed for an Oxisol soil at Rancho
ing K with depth is likely an important factor Grande (Elsenbeer et al. 1999). In addition, our
s
contributing to more rapid responses with larger surface layer K values were quite similar to the
s
quickflow runoff volumes for the Ultisol wa- Rancho Grande forest soil of Elsenbeer et al.
tershed, though topographic differences between (1999), though lower than most studies reviewed
the watersheds would also contribute. K at the by Elsenbeer (2001).
s
soil surface was not significantly different be- Elsenbeer (2001) presents a runoff response
tween the two soils (P > 0.05). There is a feed- continuum that is useful for understanding our
back between soil hydrologic characteristics, observation that runoff responses did not differ
hydrologic responses to precipitation events and greatlybetweenthetwowatershedsinthepresent
catchment geomorphology (Gomi et al. 2002; study. Of the studies considered in that review, it
Robinson et al. 1995), which appears to have re- appearsasthoughtheJuruenaUltisolandOxisol
sulted in topographic differences between the watersheds best correspond with the Danum
catchments (Fig. 1). For example, larger volumes Acrisol (Ultisol) and the Rancho Grande Ferral-
of shallow subsurface stormflow could have re- sol (Oxisol), respectively. These sites lie together
sulted in erosion and eventual over-steepening in the intermediary group characterized by a
modest lateral subsurface component (Elsenbeer
2001).TheJuruenaUltisolofthisstudypresented
0 an anisotropy in Ks similar to that of the Danum
Ultisol (Borneo)site,whereanincreaseinK between50
20 Oxisol s
and 75 cm was also observed (Chappell et al.
40
1998, reviewed by Elsenbeer 2001).
m) 60
c Overlandflowwasfrequentlyobservedonboth
pth ( 80 Juruena watersheds, and was evaluated over the
e
D 100 surface of the watersheds by the responses of
overland flow detectors (OFDs). The percentage
120
ofOFDsindicatingoverlandflowvariedoverthe
140
course of the rainy season, but in a surprisingly
0 2 4 6 8 10 consistent fashion for both the Ultisol and the
Ks (cm hr - 1) Oxisol watersheds (Fig. 4, r2 = 0.73). The inter-
ceptofthelinearregressionlinerelativetothe1:1
Fig.3 Saturated hydraulic conductivity (K) in the soil
s
profile.Errorbarsare±1SE(n=4) line (Fig. 4) indicates that overland flow was in
123
Biogeochemistry(2006)81:45–57 51
s) 100 (P < 0.01). Overland flow has now been ob-
D
F served on Oxisol soils in the Amazon (present
O
e study)andinPanama(Godseyet al.2004),which
v
nsi 80 1:1 line suggests a need to reconsider the hypothesis that
o
p
es y= 0.97x +11.4 overland flow is not an important runoff produc-
% of r 60 pr 2= = 0 0.0.7037 ing mechanism for Ferrasol (e.g. Oxisol) land-
d ( scapes (Elsenbeer 2001).
e
sh A frequency analysis of the 5 min throughfall
er 40
at record over the course of 2004 indicated that
w
ol rainfall intensity is frequently sufficient to pro-
s
Ulti 20 duce Hortonian runoff across the landscape.
20 30 40 50 60 70 80 90 100
Intensitiesgreaterthan20 mm h–1wererecorded
Oxisol watershed (% of responsive OFDs)
during more than 35% of the 1024 5-min time
Fig.4 Percentages of overland flow detectors (OFDs)
intervals during 2004 for which precipitation
thatwereresponsiveintheOxisolandUltisolwatersheds
intensity was more than a drizzle (>5 mm h–1),
for each collection. The 1:1 line indicates when overland
flow was generated equally for the two watersheds. That compared to average K of surface soil of
s
the percentage of responsive OFDs is always at or above 19.8 mm h–1. More than 70% of rain events con-
the 1:1 line indicates that the spatial extent of overland
tained at least one 5-minute interval which
flow was greater for the Ultisol watershed than for the
exceeded 20.0 mm h–1 during 2004. These data
Oxisolwatershed
areillustrativethatHortonianrunoffneednotbe
considered a rare occurrence for this tropical
general more pervasive on the Ultisol watershed. forested system.
The Ultisol watershed produced 11.4% more
overland flow than the Oxisol, as determined Carbon biogeochemistry of hydrologic
from the intercept of the linear relationship of flowpaths
runoff responses for the two watersheds. Never-
theless,theresponserateforOFDsontheOxisol Quickflow and groundwater flow intersect soil
dominated watershed generally differed by less horizons with very different C characteristics
than 20% of the response rate of the Ultisol wa- (Table 1), imparting distinctive C signatures to
tershed. This indicates that overland flow is a these hydrologic flowpaths. Mean DOC concen-
feature of both Oxisol and Ultisol soils in the trationswerefoundtovarybyanorderofmagni-
Amazon, but was more prevalent for the Ultisol tude between quickflow-related flowpaths and
catchment. Among landscape positions, the groundwater-related flowpaths, with DOC
OFDs placed in the plateau and shoulder slope transported by overland flow having the highest
positions exhibited responses that were not average concentration (19.6 ± 1.7 mg l–1 DOC,
statisticallydifferentbetweenthetwocatchments, mean ±1 SE for combined Ultisol and Oxisol
while the OFDs in the midslope position consis- watershedsdata,n = 70)andemergentgroundwa-
tently indicated more overland flow for the Ulti- ter the lowest (0.50 ± 0.04 mg l–1 DOC, n = 83).
sol watershed than for the Oxisol watershed DOC concentrations in shallow subsurface flow
Table1 SoilpHinwater
Depth pH OrganicC(gkg–1)
withsoiltosolutionratio
of1:2.5andorganic Oxisol Ultisol Oxisol Ultisol
carbonatdiscretedepths.
n=43forOxisoland 0–20cm 4.74± 0.05 4.72± 0.06 9.8± 0.2 10.0± 0.3
n=65forUltisolat 40–60cm 4.70± 0.04 4.76± 0.04 5.0± 0.2 5.3± 0.2
<1mdepths;n=3at 2m 5.25± 0.09 5.53± 0.14 2.0 ± 0.5 1.6± 0.2
depths>1m 4m 5.21± 0.05 5.72± 0.10 0.4± 0.1 0.5± 0.1
8m 5.04± 0.02 5.47± 0.08 0.6 ± 0.1 0.3± 0.0
123
52 Biogeochemistry(2006)81:45–57
averaged 8.8 ± 0.7 mg l–1 for the two watersheds soil (Qualls et al. 2002). Significant differences in
(n = 28). DOCconcentrationsofflowpathswerenotfound
Significant differences (P <0.05) between the forothermeasured fluxes.
two watersheds were found for DOC concentra- A general gradient is observed in DOC
tionsinoverlandflowandshallowsubsurfaceflow concentrations for both soils, decreasing with
(Fig. 5, Table 2). The DOCconcentrations in the depth from the soil surface. Soil C of the sur-
surfaceandnear-surfacefluxesthatcorrespondto face horizon (0–20 cm) was not found to be
quickflow were higher for the Ultisol catchment, significantly different between watersheds
wheremorequickflowwasalsoobserved.Thatthe (9.8 ± 0.2 g C kg–1 soil for the Oxisol (n = 42)
DOC concentrations of these fluxes can increase versus 10.0 ± 0.3 g C kg–1 soil for the Ultisol
asthevolumetricfluxoftheflowpathsincreasesis (n = 64), P = 0.38), nor between the locations of
perhaps best considered from the perspective of overlandflowandsubsurfacestormflowcollection.
the C content along the flow path. The travel The lower DOC concentrations of deeper
distance through and mean residence time within flowpaths result from numerous biogeochemical
theC-richenvironmentofthelitterlayerisgreater processes occurring within the soil matrix,
for horizontally-oriented flowpaths than where including sorption and decomposition of DOC
flowpaths are more vertically-oriented, which (Kalbitz et al. 2003; Qualls et al. 2002; Schwesig
could provide conditions allowing for increased et al. 2003). As such, groundwater-derived DOC
extraction of DOC. The more laterally-oriented has already undergone substantial processing
flowpaths typical of Ultisols (Elsenbeer 2001) compared to quickflow derived DOC that is
could result in increased DOC concentration of flushed from the soil surface and upper soil
shallowsubsurfacestormflowasthisflowpathalso horizons. Terrestrial DOC fluxes sporadically
passes through a relatively C-rich environment transported by quickflow are almost on par with
withlesssorptionopportunitiesthaninthedeeper DOCtransportedbybaseflowonanannualbasis.
Fig.5 Average(non-flow
weighted)DOC Oxisol
concentrations±1SEfor a Ultisol
Throughfall a
hydrologicflowpathson
Quickflow
theUltisolandOxisol b paths
watersheds,Juruena Overland Flow a
BrazilforSeptember
2003–April2005 b
Shallow
(Fig.5A).Different Subsurface Flow a
lettersindicatewhere a
DOCconcentrationswere Deep a
Groundwater Deeper
significantlydifferent
flow paths
betweenwatersheds a
Emergent
(P < 0.05fortwo-sample Groundwater flow a
T-test).Average A
concentrations±1SEfor
0 5 10 15 20 25 30
DICconstituentsHCO––
3 DOC (mg L- 1 )
CanddissolvedCO –C
2
arepresentedaspooled
databetweenthetwo
HCO--C
watersheds(Fig.5B) 3 Quickflow
Overland flow CO2-C paths
Emergent Deeper
Groundwater Flow flow paths
B
0 5 10 15 20 25 30
DIC (mg L- 1 )
123
Biogeochemistry(2006)81:45–57 53
Table2 Hydrologic fluxes and dissolved carbon concentrations of quickflow and deeper flowpaths for forested Juruena
headwatercatchments,September2003–April2005
Parameter* Oxisol Ultisol
Streamdischarge(ls–1)
Avg. 0.42 1.08
Min.daily 0.12 0.2
Max.instantaneous 39 48
Quickflow(%) 2.5± 0.34 3.2± 0.3
Baseflow(%) 97.5± 0.3 96.8± 0.3
DOC(mgl–1)
TF 14.1± 4.0 13.5± 3.9
OLF 10.7± 1.6 25.2± 2.2
SSF 7.5± 1.1 10.1± 0.8
DGW 1.0± 0.2 1.1± 0.5
EGW 0.51± 0.05 0.47± 0.05
CO -C(mgl–1)
2
EGW 20.5± 1.4 19.0± 1.3
*Quickflow and baseflow fluxes expressed as percentage of event precipitation, TF=throughfall, OLF=overland flow,
SSF=subsurfacestormflow,DGW=deepgroundwater,EGW=emergentgroundwateratsprings
Quickflow DOC concentrations on the order of of soil CO derived from root respiration versus
2
10 mg l–1aretransportedbyapproximately4%of that derived from microbial respiration remains
streamflow, while the remaining 96% of stream- unresolved (Davidson and Trumbore 1995; Tur-
flow originates as low-DOC (~0.5 mg l–1) base pin 1920), the soil atmosphere in the Amazon
flow. Mayorga et al. (2005) showed that much of reaches values for pCO of over 60,000 ppmv at
2
the CO lost to outgassing from large rivers and depth (Davidson and Trumbore 1995), resulting
2
wetlands in the Amazon is mineralized from a in large concentrations of dissolved CO in
2
rapidly cycling pool of young terrestrial organic groundwater (Richey et al. 2002).
C, with an older, more recalcitrant pool of DOC The measured CO concentrations were com-
2
comprising an additional component of riverine pared with theoretical CO concentrations calcu-
2
DOC. The results of the present study suggest lated from pH and alkalinity determinations, as:
that the punctuated input of DOC via quickflow
flowpathsisalikelymechanismforthetransferof
½CO (aq)(cid:2)¼½HCO(cid:1)(cid:2)½Hþ(cid:2)=K
young,allochthonousCfromthelandscapetothe 2 3 1
Amazon River system. where K1 ¼10(cid:1)6:3
Deeper hydrologic flowpaths were found to be
important C pathways, but for inorganic Crather Dissolution of CO in water results in hydra-
2
than DOC. Dissolved CO in the groundwater tionofCO asCO (aq),aswellastrueH CO via
2 2 2 2 3
that supplies base flow was found to be super- protolysis (Stumm and Morgan 1981). We may
saturated with respect to the atmosphere ignore H CO for the purposes of comparing in
2 3
(Table 2), and did not vary significantly between situdeterminationsofCO withcalculatedvalues
2
watersheds, averaging 19.9 ± 1.8 mg l–1 CO –C since the ratio of CO (aq) to H CO is 650 at
2 2 2 3
(mean ±1 SE for combined Ultisol and Oxisol 25(cid:2)C (Butler 1982).
watersheds data, n = 27, equivalent to Measured and calculated values for dissolved
48,700 ppmv pCO ). CO derived from root and CO were found to show good agreement, with
2 2 2
microbial respiration builds up in deeper soil calculated concentrations generally slightly
horizons as a result of increasing diffusional dis- higher than the measured values, reflecting the
tancewithdepth(DavidsonandTrumbore1995), tendency for slight overestimation of alkalinity
which can then be dissolved by percolating water inherent to endpoint titrations (Mackereth et al.
(Caronet al.1998).Whiletherelativemagnitude 1978). The mean of the absolute value of the
123
54 Biogeochemistry(2006)81:45–57
residual (expressed as [CO –C /CO – Decreases in CO concentration with longitu-
2 calculated 2 2
C ]) was found to be 0.17 ± 0.05 (mean ±1 dinal stream distance in headwater streams have
measured
SE). A sensitivity analysis of the measured also been shown by Palmer et al. (2001), who
parameters used to determine CO –C noted a large decline in free CO between upper
2 calculated 2
showed that varying the pH by ±0.01 resulted in and lower sampling sites, and Finlay (2003) who
±2.3% variations in CO –C , while vary- observed dissolved CO concentrations to
2 calculated 2
ing the volume of H SO used to determine decrease rapidly downstream from a spring in a
2 4
alkalinity by ±0.05 ml resulted in ±6.3% vari- forested catchment. The landscape organization
ability in CO –C . As such, the 17% mean of headwater catchments results in focused
2 calculated
residual between measured and calculated groundwater discharge at springs, with diffuse
CO –C is on par within the analytical sensitivity groundwater discharge across streambeds along
2
of the methods employed, given that the toler- the stream network (National Research Council
ances of the variables used for CO –C 2004). As a consequence, the relative contribu-
2 calculated
are multiplicative. tion of groundwater discharge to total stream
CO concentrations in stream water at discharge decreases as the distance from stream
2
watershed outlets were found to be substantially source increases. The CO concentration at each
2
less than that of springs. CO –C in the stream pointinastreamrepresentsthebalanceofinputs,
2
draining the Oxisol catchment averaged 33% of losses to the atmosphere, and CO generated
2
the CO –C concentration in the Oxisol spring. in-stream via mineralization of DOC (Jones and
2
CO –C in the stream draining the Ultisol catch- Mulholland 1998).
2
ment averaged 12% of the CO –C concentration Dissolved CO concentrations in throughfall
2 2
in the Ultisol spring. While this indicates sub- and overland flow are at or near atmospheric
stantial outgassing of CO from emergent concentration (~370 ppm; 0.15 mg l–1 CO –C)
2 2
groundwater occurring in the upper reaches of andthereforetwoordersofmagnitudelowerthan
both headwater streams (Johnson et al. in prep.), in groundwater. An additional difference
the differences in CO concentrations between between the quickflow and deeper flowpaths was
2
the two streams may be due to physical differ- thatofHCO––C,whichwasfoundtobegreaterin
3
ences in the streams themselves rather than dif- emergent groundwater than in overland flow
ferencesinwaterquality.Thestreamdrainingthe (1.6 ± 0.1vs.1.0 ± 0.1 mg l–1HCO––C,means±1
3
Ultisol catchment was sampled 50 m below its SE with n = 70 and n = 40 respectively). DIC
source, while the stream draining the Oxisol transported bygroundwaterflowwas foundtobe
catchment was sampled 20 m below its source, predominantly in the dissolved CO –C form
2
which resulted from geomorphological charac- (92%) for these acidic, highly-weathered catch-
teristics (e.g. stream constrictions) that favored ments, while quickflow DIC is largely comprised
construction of weirs at different distances below of the HCO––C form (87%). Geogenic DIC
3
springs. It should be noted that the outgassing resulting from carbonate weathering is negligible
occurring in the headwater reaches of streams is in these highly weathered and acidic soils,
driven by concentration gradients between whereas the bicarbonate that is present results
groundwater and the atmosphere, which is en- from buffering of dissolved CO derived from
2
hanced by the turbulent mixing within shallow root and microbial respiration.
headwater streams. This process occurs upstream In synthesizing the contrasting depth versus
of, and is in addition to, the mineralization of concentration relationships for DOC and CO in
2
terrestrial C that Richey et al. (2002) found to the soil profile, we see conceptually that quick-
driveoutgassingfromlargeAmazonianriversand flow flowpaths intersect the zone of relatively
wetlands, and should be considered as an addi- high DOC concentrations, while slower and dee-
tional C flux to the atmosphere beyond that per flowpaths intersect with zones of high CO
2
computed for the central Amazon River system concentrations (Fig. 6). Since streamflow in the
(J. Richey, Pers. Comm.). study catchments is predominantly derived from
123