Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.
Organic carbon burial by river meandering
partially offsets bank erosion carbon fluxes
in a discontinuous permafrost floodplain
Madison M. Douglas
1
, Gen K. Li
1
, Woodward W. Fischer
1
, Joel C. Rowland
2
, Preston C. Kemeny
1
,
A. Joshua West
3
, Jon Schwenk
2
, Anastasia P. Piliouras
2
, Austin J. Chadwick
1
, and Michael P. Lamb
1
1
Division of Geological and Planetary Science, California Institute of Technology, Pasadena, CA 91125, USA
2
Earth and Environmental Sciences Division, Los Alamos National Laboratory, Los Alamos, NM 87545, USA
3
Department of Earth Sciences, University of Southern California, Los Angeles, CA 90089, USA
Correspondence:
Madison M. Douglas (mmdougla@caltech.edu)
Received: 15 October 2021 – Discussion started: 3 November 2021
Revised: 18 March 2022 – Accepted: 12 April 2022 – Published: 10 May 2022
Abstract.
Arctic river systems erode permafrost in their banks and mobilize particulate organic carbon (OC).
Meandering rivers can entrain particulate OC from permafrost many meters below the depth of annual thaw,
potentially enabling the production of greenhouse gases. However, the amount and fate of permafrost OC that
is mobilized by river erosion is uncertain. To constrain OC fluxes due to riverbank erosion and deposition, we
collected riverbank and floodplain sediment samples along the Koyukuk River, which meanders through discon-
tinuous permafrost in the Yukon River watershed, Alaska, USA, with an average migration rate of 0.52 m yr
−
1
.
We measured sediment total OC (TOC) content, radiocarbon activity, water content, bulk density, grain size,
and floodplain stratigraphy. Radiocarbon activity and TOC content were higher in samples dominated by silt as
compared to sand, which we used to map OC content onto floodplain stratigraphy and estimate carbon fluxes
due to river meandering. Results showed that the Koyukuk River erodes and re-deposits a substantial flux of OC
each year due to its depth and high migration rate, generating a combined OC flux of a similar magnitude to the
floodplain net ecological productivity. However, sediment being eroded from cutbanks and deposited as point
bars had similar OC stocks (mean
±
1 SD of 125
.
3
±
13
.
1 kg OC m
−
2
in cutbanks versus 114
.
0
±
15
.
7 kg OC m
−
2
in point bars) whether or not the banks contained permafrost. We also observed radiocarbon-depleted biospheric
OC in both cutbanks and permafrost-free point bars. These results indicate that a substantial fraction of aged bio-
spheric OC that is liberated from floodplains by bank erosion is subsequently re-deposited in point bars rather
than being oxidized. The process of aging, erosion, and re-deposition of floodplain organic material may be
intrinsic to river–floodplain dynamics, regardless of permafrost content.
1 Introduction
The warming climate is changing Arctic landscapes, induc-
ing complex feedbacks in the global carbon cycle as per-
mafrost soils thaw (Schuur et al., 2015; Turetsky et al., 2020).
Changes in air temperature and precipitation have increased
the thickness of the active layer (ground overlying permafrost
that experiences seasonal freeze–thaw cycles), allowing res-
piration of soil organic carbon (OC) previously frozen for
thousands of years (Romanovsky et al., 2010; Isaksen et al.,
2016; Biskaborn et al., 2019). Organic carbon is also lost
from permafrost through lateral erosion by Arctic rivers – the
six largest Arctic rivers contribute
∼
3 Tg of river particulate
OC (POC) to the Arctic Ocean annually (McClelland et al.,
2016). Since a substantial portion of eroded POC is thought
to be prone to oxidation (Schreiner et al., 2014), river erosion
of POC could play an important role in the greenhouse gas
fluxes associated with permafrost thaw (Toohey et al., 2016;
Walvoord and Kurylyk, 2016).
Published by Copernicus Publications on behalf of the European Geosciences Union.
422
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
Figure 1.
Overview of sediment erosion and deposition patterns in meandering river floodplains and important variables influencing the
regional carbon cycle.
(a)
Drone photograph taken looking east across the Koyukuk River floodplain, Alaska (location marked with a white
star in Fig. 2). The river flows south toward the bottom of the image (indicated by black arrow), eroding the cutbank on the outside of the
river bend and depositing sediment on the point bar. Channel migration generates bands of higher and lower elevation sections of floodplain
called scroll bars. As the river migrates, an individual bend becomes more sinuous, eventually cutting itself off and abandoning a section of
channel, which becomes an oxbow lake.
(b)
Schematic of a meandering river floodplain, with channel geometry variables shown in black
and particulate organic carbon reservoirs and fluxes into and out of the river control volume shown in purple. The river has bankfull depth
H
and migrates laterally at rate
E
, maintaining a constant channel width. Organic carbon is stored in the river cutbanks (
C
CB
) and point bars
(
C
PB
) and is transported in the river as particulates (POC). These reservoirs are mixtures of radiocarbon-dead (Fm
=
0) petrogenic organic
carbon (OC
Petro
) and biospheric organic carbon (OC
Bio
) that has been stored in permafrost (low Fm) or been recently fixed by the biosphere
(Fm
≥
1). Fluxes of organic carbon into and out of the river control volume include cutbank erosion (
F
CB
), point bar deposition (
F
PB
),
overbank deposition (
F
OB
), and oxidation of POC and DOC (
F
OX
).
As Arctic rivers migrate laterally across permafrost flood-
plains, they can mine sediment and organics from over 10 m
below the active layer (Spencer et al., 2015; Kanevskiy et al.,
2016). Permafrost floodplains are thus an important source
of POC to rivers (Kanevskiy et al., 2016; Loiko et al., 2017;
Lininger et al., 2018; Lininger and Wohl, 2019). After mo-
bilization by a river, POC can be oxidized during transport
(Striegl et al., 2012; Denfeld et al., 2013; Serikova et al.,
2018) or re-buried in floodplains (Wang et al., 2019; Torres et
al., 2020). Alternatively, POC can be delivered downstream
to the ocean, where it may be oxidized to CO
2
, reduced to
CH
4
, or buried in deltaic sedimentary deposits (Torres et al.,
2020; Hilton et al., 2015). Riverbank erosion may be limited
by the rate of permafrost thaw (Costard et al., 2003; Ran-
driamazaoro et al., 2007; Dupeyrat et al., 2011), implying
that erosion rates could increase with warming air and river
water temperatures. Therefore, more rapid riverbank erosion
resulting from warming temperatures has the potential to in-
crease fluvial POC fluxes and oxidation, resulting in a pos-
itive feedback on the concentration of atmospheric carbon
dioxide (Striegl et al., 2012; Denfeld et al., 2013; Serikova et
al., 2018). The magnitude and timescale of this feedback are
highly uncertain but may be important to consider for pre-
dicting and mitigating impacts from anthropogenic climate
change.
Floodplain POC stocks are vulnerable to erosion by Arc-
tic rivers (Vonk et al., 2019; Parmentier et al., 2017). For in-
stance, Lininger et al. (2018, 2019) mapped OC contents and
stocks across the Yukon Flats and found significant variabil-
ity in OC contents between riverine landforms (Lininger et
al., 2018) as well as underestimation of floodplain OC stocks
in previous data compilations (Lininger et al., 2019). Their
work built on previous studies that characterized vegetation
and permafrost succession through a time series of floodplain
surfaces that had been progressively abandoned by river mi-
gration (Shur and Jorgenson, 2007). Yet major questions re-
main about the magnitude of POC fluxes due to bank ero-
sion and bar deposition in permafrost river systems as well
as the physical processes that govern these fluxes (Lininger
and Wohl, 2019).
Alluvial rivers commonly maintain an approximately con-
stant channel width, eroding one bank while depositing sed-
iment at a commensurate rate on the opposite bank (Fig. 1a)
(Dietrich et al., 1979; Eke et al., 2014). Riverbank erosion
has been shown to contribute substantially to downstream
POC fluxes (Kanevskiy et al., 2016). However, it is unclear
to what extent the OC released by bank erosion is compen-
sated by OC burial in depositional bars as opposed to being
transported downstream or oxidized during transport within
river systems (Fig. 1b) (Wang et al., 2019; Scheingross et al.,
2021).
To quantify POC storage and mobilization, we investi-
gated the Koyukuk River in the Yukon River watershed,
Alaska, USA (Fig. 2), which is an actively meandering river
in discontinuous permafrost. We quantified OC stocks using
measurements of OC content in field samples and extrapo-
lated these across the floodplain using floodplain stratigra-
phy and correlations between grain size and OC content. We
then used a one-dimensional mass-balance model to quantify
net fluxes of OC into and out of the river due to bank erosion
Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
423
Figure 2.
Sample locations on the Koyukuk River floodplain. Loca-
tions are coded for sites where we sampled ice-cemented permafrost
versus ice-poor ground inferred to be non-permafrost. Sample sites
are located near the village of Huslia, in central Alaska, and the
river flows towards the south past the town. Sampling locations are
mapped on Landsat imagery, with the white star marking the loca-
tion of Fig. 1a (drone photo taken looking east). The inset map was
generated using the “Alaska Coast Simplified” and “Major Rivers”
shapefiles from the Alaska State Geo-Spatial Data Clearinghouse.
and bar deposition. To attribute OC to biospheric versus rock-
derived (petrogenic) sources, we used radiocarbon measure-
ments to infer the presence of a petrogenic OC end-member
and compared the range of biospheric radiocarbon compo-
sitions in permafrost and non-permafrost sediment samples
and landforms.
2 Approach
To understand cycling of POC between rivers and flood-
plains, we developed an approach to ascertain OC sources
and determined if OC eroded from river deposits was trans-
ported downstream or re-buried (Fig. 1b). Eroding banks
can source OC from modern vegetation and organic hori-
zons near the bank surface as well as deeper sediment
that may be depleted in radiocarbon. Radiocarbon pro-
vides an effective tracer of OC aging in floodplains (Galy
and Eglinton, 2011; Torres et al., 2017), but several pro-
cesses can produce depleted radiocarbon signals. First,
many Arctic permafrost deposits are relicts from colder
climatic conditions (O’Donnell et al., 2012). These de-
posits have low radiocarbon activity, expressed as fraction
modern (Fm
=
A
sample,norm
/
(0
.
95
A
Ox,norm
);
A
sample,norm
in-
dicates sample
14
C activity normalized for isotope frac-
tionation to
δ
13
C
VPDB
=−
25 ‰ (VPDB – Vienna Pee Dee
Belemnite), while
A
Ox,norm
indicates NBS Oxalic Acid
I normalized to
δ
13
C
VPDB
=−
19 ‰, with
δ
13
C
VPDB
=
(
R
sample
/R
VPDB
−
1)
×
1000 reported in per mill (‰))
(Reimer et al., 2004). If mobilized permafrost POC is re-
buried in bars without the addition of newly fixed biospheric
OC, then bar sediment should also have OC with low Fm
values inherited from permafrost carbon. Second, sediment
can contain a radiocarbon-dead, petrogenic OC component
that contributes to low Fm values (Blair et al., 2003). We ex-
pected a petrogenic OC contribution in floodplain sediments
throughout the Koyukuk River system, since the headwaters
of the Koyukuk River contain outcrops of shale bedrock rich
in kerogen (Dumoulin et al., 2004; Wilson et al., 2015; Slack
et al., 2015). Third, river–floodplain interactions generate or-
ganic carbon with low Fm values via transient OC storage,
independent of the presence of either permafrost or petro-
genic OC (Torres et al., 2020). For example, floodplain de-
posits can remain in place over millennial timescales before
being reworked by the river channel due to the stochastic na-
ture of river lateral migration (Torres et al., 2017; Repasch
et al., 2020). Therefore, radiocarbon measurements provide
insight into OC sources but require de-convolving petrogenic
OC from biospheric OC and assessing aging of OC by stor-
age in permafrost versus non-permafrost floodplain deposits.
We used sediment total OC (TOC) and Fm measurements
to calculate the Fm of the biospheric OC end-member as
well as the contribution of petrogenic OC to our samples.
This calculation allowed us to determine if low Fm val-
ues were due to a high content of radiocarbon-dead rock-
derived OC or preservation and aging of OC in permafrost or
in the river floodplain (Fig. 1b) (Scheingross et al., 2021).
Both radiocarbon-dead OC derived from bedrock erosion
(TOC
petro
) and aging of biospheric OC (TOC
bio
) in per-
mafrost and river floodplain deposits will yield sediment OC
with low Fm values (Fig. 1b). We partitioned the TOC con-
tents measured in each sample (TOC
meas
) into a two end-
member mixture of biospheric (TOC
bio
=
f
bio
×
TOC
meas
)
and petrogenic OC (TOC
petro
=
f
petro
×
TOC
meas
) fractions
(Fig. 4c) (Blair et al., 2003; Cui et al., 2016):
TOC
bio
+
TOC
petro
=
TOC
meas
,
(1)
f
bio
+
f
petro
=
TOC
bio
TOC
meas
+
TOC
petro
TOC
meas
=
1
,
(2)
where
f
bio
and
f
petro
are the fraction of organic carbon from
biospheric and petrogenic sources. Changes in the ratio of
biospheric to petrogenic OC, as well as aging of the bio-
spheric pool, will change the measured fraction modern in
sediment OC (Fm
meas
; unitless ratio) (Galy et al., 2008). By
mass balance,
TOC
meas
Fm
meas
=
TOC
bio
Fm
bio
+
TOC
petro
Fm
petro
.
(3)
The petrogenic OC end-member was assumed to be
radiocarbon-dead (Fm
petro
=
0), and Eqs. (1) and (2) substi-
tuted into Eq. (3) yield
Fm
meas
=
Fm
bio
(TOC
meas
−
TOC
petro
)
TOC
meas
.
(4)
A nonlinear optimization of Eq. (4) for Fm
meas
versus
TOC
meas
was used to calculate 95 % confidence intervals
https://doi.org/10.5194/esurf-10-421-2022
Earth Surf. Dynam., 10, 421–435, 2022
424
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
around Fm
bio
(effectively the mean radiocarbon activity of
biosphere-derived carbon) and the TOC
petro
content in cut-
bank and point bar sediment samples (Fig. 4c) (Hemingway
et al., 2018; Wang et al., 2019). We reported a range of fitted
Fm
bio
end-members to compare biospheric OC eroding from
cutbanks to that being deposited in point bars because cut-
banks comprise a mixture of permafrost and non-permafrost
terrain with varying Fm values that are homogenized dur-
ing transport in the river. This optimization also considers
a range of TOC
petro
content end-members for cutbanks and
point bars. We do not expect that geographic location on the
Koyukuk floodplain has a strong control on sediment OC
petro
content. While recent work found evidence for petrogenic
OC oxidation during riverine transport of sediment (Bouchez
et al., 2010; Horan et al., 2019), these studies focused on river
reaches spanning hundreds of kilometers, 1 order of mag-
nitude longer than our study reach. Even over hundreds of
kilometers, Horan et al. (2019) found that less than half of
petrogenic OC eroded from the Mackenzie River catchment
was oxidized during transport. Therefore, it is reasonable to
assume that the production and oxidation of rock-derived OC
is limited within our study reach and a single TOC
petro
end-
member is appropriate for cutbanks and another for point
bars.
3 Materials and methods
3.1 Field sampling methods
We collected samples from 33 locations along the Koyukuk
River near the village of Huslia, Alaska, during June–July
2018 (Fig. 2 inset; Fig. S1 in the Supplement). Near Hus-
lia, the mean annual air temperature is
−
3
.
6
◦
C (Nowacki et
al., 2003; Daly et al., 2015, 2018). The Koyukuk is a me-
andering river in discontinuous permafrost (portions of the
floodplain are underlain by ground below 0
◦
C while oth-
ers are not) with well-defined scroll bars (former levees)
(Mason and Mohrig, 2019) that demarcate clear spatial pat-
terns of channel lateral migration (Fig. 2) (Shur and Jorgen-
son, 2007). Bands of vegetation outline scroll bars on the
floodplain that were abandoned due to channel lateral migra-
tion and meander-bend cutoff (Fig. 1). Seasonal variations in
temperature cause an annual freeze–thaw cycle in sediment
near the ground surface across the landscape, called the ac-
tive layer, while the ground below, in areas of permafrost,
is perennially at sub-zero temperatures. To represent the di-
versity of floodplain geomorphology, permafrost occurrence,
and deposit ages, we selected eight permafrost cutbanks, six
non-permafrost cutbanks, six permafrost floodplain cores,
four non-permafrost floodplain cores and pits, and nine non-
permafrost cores and pits in transects across two point bar
complexes to characterize floodplain stratigraphy and carbon
geochemistry (Fig. 2; Tables S1 and S2 in the Supplement).
We categorized permafrost as ice-cemented sediment ob-
served during our summer field season, often containing ice
lenses and other structures indicative of permafrost (Fig. 3a
and b) (French and Shur, 2010). Permafrost cutbanks often
had an undercut marking the high water level where bank
sediment was directly thawed by the river and collapsed as
well as abundant toppled trees indicating active bank ero-
sion. We classified terrain without ice cement observed to
the depth of coring or sampling as non-permafrost (Fig. 3a
and c), although this category might also include perennial
sub-zero ground that lacked pore water to form ice cement.
Bank samples were collected by digging into cutbanks and
point bars, and cores were taken using a hand auger in non-
permafrost deposits and a Snow, Ice, and Permafrost Re-
search Establishment (SIPRE) auger in permafrost (Fig. 2).
All samples were recorded in stratigraphic columns to deter-
mine the thickness of each stratigraphic unit. Samples were
stored in sterile Whirl-pak bags and frozen within 12 h of
collection and then transported frozen back to a cold room
(
−
15
◦
C) at Caltech for laboratory analyses.
River bathymetry was characterized using a Teledyne Rio-
Pro acoustic Doppler current profiler (ADCP). We calculated
a river depth of 12.4 m as the mean of the deepest measured
value (i.e., the thalweg) for eight ADCP river cross-sectional
transects across a representative meander bend. Mean bank
erosion rates for the portion of the Koyukuk we studied were
0.52 m yr
−
1
averaged over the time period of 1978–2018
(Rowland et al., 2019). Over the same time interval, channel
width varied from 173
±
43 m in 1978 to 179
±
43 m in 2018
(median
±
1 SD), indicating a balance between cutbank ero-
sion and point bar deposition over this period since net lateral
erosion or deposition would change channel width (Fig. S2).
3.2 Laboratory analyses
Samples were transferred to pre-combusted aluminum foil,
weighed, and oven-dried at 55–60
◦
C to calculate the mass
fraction of water (
M
H
2
O
,i
). For samples taken using the
SIPRE auger with known volume, bulk density (
ρ
i
) was cal-
culated from total mass divided by volume. The samples
were gently homogenized using an agate mortar and pestle
and then split using cone and quarter or a riffle splitter, to
avoid grain size fractionation, for further analysis.
Total organic carbon content (TOC
meas
in Eq. 2), stable or-
ganic carbon isotopes, and total nitrogen (TN) content were
measured on a Costech Elemental Analyzer coupled to a
MAT 253 IRMS (isotope ratio mass spectrometer) at Los
Alamos National Laboratory (LANL). Prior to analysis, sam-
ples were ground to a powder and approximately 3 mg of
each sample was decarbonated by fumigation with HCl in sil-
ver capsules. Isotope ratios are reported relative to the VPDB
(
δ
13
C
=
(
R
sample
/R
VPDB
−
1)
×
1000; reported in per mill
(‰)), and measured blanks were below the peak detection
limit. Measurements were calibrated using laboratory stan-
dards of 2,5-Bis(5-tertbutyl-2-benzo-oxazol-2-yl) thiophene
(BBOT, Eurovector; TOC
=
72
.
53 %, measured as 69
.
59 %
±
2
.
05 %;
δ
13
C
=−
26
.
6 ‰, measured as
−
26
.
6 ‰
±
0
.
01 ‰;
Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
425
Figure 3.
Field observations of Koyukuk riverbanks and floodplain stratigraphy.
(a)
Representative stratigraphic columns from non-
permafrost (Bank 2) and permafrost (Bank 6) cutbanks.
(b)
Field photo of boundary between permafrost ice cement and the overlying
active layer in Core 4.
(c)
Thermoerosional niche formed in a permafrost cutbank, with silty permafrost overlain by a layer of peat and black
spruce trees.
(d)
Eroding riverbank without permafrost, hosting a white spruce forest with roots that reach deep into the bank sediment.
Complete stratigraphic sections and additional field photos are in Figs. S2 and S3.
TN
=
6
.
51 %, measured as 6
.
82 %
±
0
.
24 %), peach leaves
(1570a; TOC
=
44
.
65 %; measured as 44
.
33 %
±
0
.
96 %;
δ
13
C
=−
25
.
95 ‰, measured as
−
26
.
13 ‰
±
0
.
08 ‰; TN
=
2
.
83 %, measured as 3
.
31 %
±
1
.
27 %), and urea (Eurovec-
tor; TOC
=
20
.
00 %, measured as 17
.
98 %
±
0
.
37 %; TN
=
46
.
65 %, measured as 45
.
88 %
±
0
.
88 %) for TOC and TN
and cellulose (IAEA-C3;
δ
13
C
=−
24
.
91 ‰, measured as
−
24
.
82 ‰
±
0
.
06 ‰), sucrose (IAEA-C6;
δ
13
C
=−
10
.
8 ‰,
measured as
−
10
.
7 ‰
±
0
.
03 ‰), and oxalic acid (IAEA-C8;
δ
13
C
=−
18
.
3 ‰, measured as
−
18
.
5 ‰
±
0
.
06 ‰) for stable
OC isotopes, with uncertainties reported as 1 standard devi-
ation (
±
1 SD). Values of
δ
13
C and TN content are not dis-
cussed in the main text but are included in figures and tables
in the Supplement.
Radiocarbon content was measured on a subset of sam-
ple at the National Ocean Sciences Accelerator Mass Spec-
trometry (NOSAMS) facility in Woods Hole. Sample splits
for radiocarbon were ground to a powder and decarbonated
at Caltech in pre-combusted glassware using 1 M HCl, son-
icated for 10 min, and neutralized using 1 M NaOH. Splits
were centrifuged for 10 min, and the supernatant was re-
moved using a pipette. The samples were rinsed using 20 mL
Milli-Q water, centrifuged and decanted twice before being
lyophilized, and sent to NOSAMS to be measured for radio-
carbon activity (Fm
meas
in Eq. 3). NOSAMS also reported to-
tal organic carbon content (dry wt % with 5 % measurement
uncertainty) and organic carbon stable isotope measurements
(referenced to VPDB;
δ
13
C
=
(
R
sample
/R
VPDB
−
1)
×
1000;
reported in per mill (‰)), and these produced similar results
as LANL (Fig. S5 and Table S2). We used LANL OC con-
tents in subsequent analyses because they reported smaller
uncertainties and because we made measurements at LANL
for all samples. NOSAMS data are used only for Fm values
of the sample subset.
Sample splits for grain size analysis were placed into ster-
ile polypropylene Falcon tubes to remove carbonate and or-
ganic materials (Gee and Or, 2002). Samples were acidi-
fied overnight with 1 M HCl and then centrifuged for 15 min
at 4000 rpm and decanted; they were rinsed twice with DI
(deionized) H
2
O, centrifuged, and decanted before being
oven-dried at 55–60
◦
C; and they were then reacted with
H
2
O
2
on a hot plate at 85
◦
C to remove organics. Float-
ing pieces of organic material were removed using a micro-
spatula rinsed with DI H
2
O. Additional H
2
O
2
was added un-
til reactions ceased by visual inspection. Samples were rinsed
and centrifuged three times before oven-drying. Each sample
https://doi.org/10.5194/esurf-10-421-2022
Earth Surf. Dynam., 10, 421–435, 2022
426
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
was re-hydrated using DI H
2
O,
∼
10 mL of 10 g (NaPO
3
)
6
(sodium hexametaphosphate) per 1 L DI H
2
O was added to
prevent flocculation, and samples were sonicated for 3 min.
The samples were split while wet using a riffle splitter to
the required sediment concentration for laser diffraction, and
grain size was measured on a Malvern Mastersizer 2000,
with measurements calibrated against a laboratory silica car-
bide standard (median diameter
D
50
=
13
.
184
±
0
.
105 μm
throughout our measurements). Grain size data confirmed
our field observations of grain size that were made using a
sand card and hand lens (Table S5).
A subset of TOC and TN contents, stable OC isotopes,
and grain size data was previously published in Douglas et
al. (2021) (Table S2).
4 Results
Permafrost cutbanks and floodplains generally displayed an
organic-rich upper horizon, which extended up to 1.3 m be-
low the ground surface in peat, underlain by silt that abruptly
transitioned to sand (Fig. 3a and d; Fig. S3). The thickness
of the active layer, measured by trenching or using a 1 m per-
mafrost probe (
n
=
53), ranged from 40 cm to greater than
the length of the probe, with a median of measured values
(
n
=
38) of approximately 75 cm. Non-permafrost cutbanks
had a layer of organic topsoil overlying silt with abundant
roots and organic-rich lenses that became interbedded and
then transitioned to sand with increasing depth (Fig. 3a).
All terrain types exhibited a trend of grain size fining up-
ward, with medium sand (based on bed-material grab sam-
ples taken from a boat with a Ponar sampler) comprising the
channel-bed material. We did not observe permafrost in ac-
tive point bars, which had a thin to absent layer of organic
topsoil at the land surface underlain by sandy deposits ex-
hibiting ripple and dune cross stratification from sediment
transport and deposition. Sediment TOC content and Fm val-
ues varied with sediment size. Silt samples had higher aver-
age TOC content than sandy samples, and peat had higher
TOC content than topsoil (Fig. 4a). Although the organic
horizons overlying permafrost had a higher TOC content
than the organic horizons overlying non-permafrost deposits,
sediment samples below the organic horizon did not show
a significant difference in TOC content based on the pres-
ence or absence of permafrost for a given grain size (Fig. 4a
and b). The strong dependence of TOC content on grain size
allowed us to estimate OC stocks based on measured strati-
graphic sections.
Coarser sediment yielded lower Fm values – indicative of
older organic carbon – with silt and organic horizons hav-
ing higher Fm values (Fig. 4c). A petrogenic contribution
can explain measured differences in sediment Fm and would
be expected to be enriched in the coarser-size fraction (Galy
et al., 2007). To calculate the range of TOC
petro
and Fm
bio
end-members for cutbank and point bar sediment OC, we
fitted a nonlinear regression (nlinfit.m in Matlab 2017) be-
tween Fm
meas
and TOC
meas
using Eq. (4) and used the Jaco-
bian to calculate 95 % confidence intervals (Fig. 4c). Fitting
Fm
meas
to TOC
meas
gave a range of biospheric radiocarbon
(Fm
bio
) and petrogenic OC content (TOC
petro
) end-members.
Some cutbank samples had
δ
13
C greater than
−
20 ‰, rais-
ing concerns about incomplete decarbonation (see Table S2).
However, fitting Fm
meas
to TOC
meas
for cutbank and flood-
plain samples together but excluding samples with
δ
13
C
greater than
−
20 ‰ (
n
=
13) generated a fit with similar
end-members and confidence intervals. Therefore, due to the
small number of radiocarbon activity, we did not exclude the
high
δ
13
C samples from our analysis.
The 95 % confidence intervals for Fm
bio
of the cutbanks
and point bars overlapped with Fm values from centimeter-
scale wood fragments collected from bank samples and cores
(Fm
=
0
.
2319
±
0
.
0015 to 0
.
9843
±
0
.
0027, equivalent to ra-
diocarbon ages of 11 750
±
55 to 125
±
20 yr BP). Since wood
and plant debris is devoid of petrogenic OC, its Fm directly
reflects storage and aging in these deposits. Therefore, we
inferred that non-permafrost point bars also likely contained
some aged biospheric OC.
5 Analysis: organic carbon cycling by river
meandering
5.1 Carbon mass balance for a meandering river
To evaluate particulate OC fluxes into and out of the
Koyukuk River, we used a mass-balance model applicable
to single-threaded, meandering rivers (Fig. 1b), neglecting
fluxes due to dissolved OC and wood and plant debris. Our
model includes vertical variations in floodplain structure and
their corresponding OC stocks, following similar floodplain–
river exchange models (Lauer and Parker, 2008). While other
models exist that incorporate more complex boundary condi-
tions and sediment tracking (Lauer and Parker, 2008; Mal-
mon et al., 2003; Lauer and Willenbring, 2010), we sought
the simplest possible framework that could utilize our field
data to constrain carbon fluxes. We considered POC fluxes
into the river due to cutbank erosion (
F
CB
; kg yr
−
1
) and
out of the river due to POC being deposited in point bars
(
F
PB
; kg yr
−
1
) or overbank deposits (
F
OB
; kg yr
−
1
) or oxi-
dized during transport and released to the atmosphere as CO
2
(
F
OX
; kg yr
−
1
; Fig. 1b) (Striegl et al., 2012; Denfeld et al.,
2013; Serikova et al., 2018). This net budget is represented
by
d(POC)
d
t
=
F
CB
−
F
PB
−
F
OB
−
F
OX
.
(5)
In the subsequent sections, we estimate the organic carbon
stocks to find
F
CB
and
F
PB
in Eq. (5) and then discuss the
relative magnitudes of
F
OB
and
F
OX
.
Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
427
Figure 4.
Floodplain sediment geochemistry results.
(a)
Total organic carbon versus median sediment grain size, with organic horizons split
into ice-rich permafrost peat and non-permafrost topsoil, with 1 SD error bars. The horizontal lines indicate the mean and shaded region the
standard error of the mean for the peat (
n
=
5, blue shading), topsoil (
n
=
2, red shading), silt (
D
50
<
0
.
63 mm,
n
=
14, gray shading), and
sand (
D
50
>
0
.
63 mm,
n
=
7, gray shading) grain size classes.
(b)
Radiocarbon activity (reported as fraction modern, Fm) versus median
grain size, with 1 SD error bars and shaded regions indicating the mean and standard error of the mean for peat (
n
=
3), topsoil (
n
=
1), silt
(
n
=
13), and sand (
n
=
7).
(c)
Sediment sample fraction modern (Fm
meas
) plotted against TOC content (TOC
meas
) and fit using Eq. (4)
to calculate end-members for biospheric radiocarbon fraction modern (Fm
bio
) and petrogenic organic carbon content (TOC
petro
). The 95 %
confidence intervals (CI) for cutbanks and point bars are shaded in blue and yellow, with the horizontal upper bound on the point bar CI
representing TOC
petro
=
0
.
0 wt %. Black lines denote mixing between representative values of TOC
petro
and Fm
bio
. The range of wood and
plant debris Fm values is plotted on the right
y
axis, indicating the likely range of biospheric end-members.
5.2 Floodplain organic carbon stocks
To quantify the fluxes of carbon in and out of the river due
to bank erosion and bar deposition, we first needed to esti-
mate the carbon stocks in the floodplain. Our approach was
to first take advantage of particle-size correlations with TOC
content (Fig. 4a and b), as discussed in detail below, to es-
timate carbon contents for stratigraphic units where we only
had grain size information. This process increased our sam-
ple size from 9 to 30 complete stratigraphic sections. Next,
we used our mapping of floodplain stratigraphy and grain
size to estimate carbon stocks integrated over a characteris-
tic depth of the floodplain. We produced this analysis using
two different characteristic depths for comparison. A depth
of 1 m was used for comparison to previous studies that of-
ten only sampled in the top meter of the floodplain (Hugelius
et al., 2014). The second depth we used was the depth of the
Koyukuk River (12.4 m) because ultimately this is the thick-
ness of floodplain material that is being eroded and deposited
by the river. In Sect. 5.3, these depth-integrated carbon con-
centrations are used to estimate carbon fluxes due to bank
erosion and bar deposition.
Measured stratigraphic sections were divided into four
units (Fig. S4): sand (
D
50
>
63 μm), mud (
D
50
<
63 μm),
topsoil (organic horizons overlying non-permafrost sedi-
ment), and peat (organic horizons overlying permafrost).
These stratigraphic units correlated with distinct magni-
tudes of mean TOC content (
c
i
) and mass fraction of wa-
ter (
M
H
2
O
,i
). We found the average TOC value from each
unit and assigned these average values to the corresponding
https://doi.org/10.5194/esurf-10-421-2022
Earth Surf. Dynam., 10, 421–435, 2022
428
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
units for beds where we measured grain size but did not mea-
sure TOC. We quantified the uncertainty in
c
i
and
M
H
2
O
,i
us-
ing Gaussian error propagation of 1 standard deviation (Ta-
bles S2–S4).
To estimate carbon stocks, total OC measurements and es-
timated values for each unit (Fig. 4a; Figs. S6 and S7) were
integrated both over 1 m depth below the surface (Fig. 5a)
and over a depth equivalent to the bankfull river depth
(12.4 m; Fig. 5b). We calculated the depth-integrated OC
stock using
C
CB
=
n
∑
i
=
1
ρ
i
×
H
i
×
c
i
(1
−
M
H
2
O
,i
)
.
(6)
We accounted for
n
beds of the four stratigraphic units in
each measured stratigraphic section, where
ρ
i
is the mean
unit bulk density (kg wet sediment per m
3
),
H
i
is the unit
thickness (m),
c
i
is the mass fraction of OC in the unit (kg OC
per kg dry sediment of each unit), and
M
H
2
O
,i
is the mass
fraction of water in the unit (kg H
2
O per kg wet sediment of
each unit).
M
H
2
O
,i
+
M
dry
,i
=
1, with
M
dry
,i
being the mass
fraction of dry sediment in the unit (kg dry sediment per kg
wet sediment of each unit). Bulk densities measured from
cores for mineral (mean
±
SD of 989
±
323 kg m
−
3
,
n
=
7)
and organic (905
±
49 kg m
−
3
,
n
=
2) horizons were the same
within uncertainty (Table S2). Therefore, we used a constant
mean bulk density (
ρ
i
=
971
±
283) across all stratigraphic
units (Table S3).
Measurement and sampling were only possible on the ex-
posed section of the riverbank, above the water level. Ex-
posed sections represented 7 %–47 % of total bank height (as
measured from channel thalweg to bank top). We assumed
all sediment below the base of our stratigraphic sections con-
sisted of sand, which was supported by our measurements of
grab samples of the active channel and cores of the flood-
plain beyond 2 m depth (Fig. S3) and was consistent with
downward-coarsening trends widely observed in meandering
rivers and their deposits (Tables S3–S4) (Miall, 2013).
Estimated permafrost cutbank and floodplain OC
stocks integrated to 1 m depth were 31
.
1
±
9
.
8 kg OC m
−
2
(mean
±
1 SD
of
OC
stocks;
n
=
14),
while
non-
permafrost cutbanks, floodplains, and point bars contained
23
.
3
±
4
.
8 kg OC m
−
2
(
n
=
10) (Fig. 5a). The Mann–
Whitney
U
test found that OC stocks in permafrost and non-
permafrost deposits had similar organic content distributions
(
p
=
0
.
1669). Grouping results by terrain type, permafrost
and non-permafrost cutbanks had 30
.
2
±
9
.
2 kg OC m
−
2
(
n
=
11), permafrost and non-permafrost floodplains had
28
.
8
±
8
.
3 kg OC m
−
2
(
n
=
9), and non-permafrost point
bars had 19
.
4
±
5
.
2 kg OC m
−
2
(
n
=
4). The Mann–Whitney
U
test could not reject the null hypothesis of cutbank and
floodplain OC stocks being drawn from the same distribution
at 5 % confidence (
p
=
0
.
7891), but the test found weak
evidence for point bars having distinctly lower OC stocks
(
p
=
0
.
0503 for floodplains versus point bars,
p
=
0
.
0601
for point bars versus cutbanks). Therefore, floodplains and
cutbanks generally had higher OC stocks in their upper
1 m of sediment than point bars, but we did not observe a
significant difference in 1 m OC stocks between permafrost
and non-permafrost deposits (Fig. 5a).
Estimated
permafrost
cutbank
and
floodplain
OC stocks integrated over the channel depth were
125
.
1
±
14
.
9 kg OC m
−
2
(mean
±
1 SD of OC stocks;
n
=
14), while non-permafrost cutbanks, floodplains,
and point bars contained 116
.
1
±
11
.
4 kg OC m
−
2
(
n
=
10)
(Fig. 5b). The Mann–Whitney
U
test could not reject the null
hypothesis that OC stocks in permafrost and non-permafrost
deposits had the same organic content distributions
(
p
=
0
.
3641). Grouping results by terrain type, permafrost
and non-permafrost cutbanks had 125
.
3
±
13
.
1 kg OC m
−
2
(
n
=
11), permafrost and non-permafrost floodplains had
121
.
0
±
13
.
5 kg OC m
−
2
(
n
=
9), and non-permafrost point
bars had 114
.
0
±
15
.
7 kg OC m
−
2
(
n
=
4). Again, the
Mann–Whitney
U
test could not reject the null hypothesis
of all landform OC stocks being drawn from the same
distribution at 5 % confidence (
p
=
0
.
3619 for floodplains
versus cutbanks,
p
=
0
.
8252 for floodplains versus point
bars,
p
=
0
.
2799 for point bars versus cutbanks). Therefore,
the distribution of OC stocks integrated to channel depth
for cutbanks was indistinguishable from the distribution of
measured stocks of newly deposited point bars (Fig. 5b).
5.3 Carbon fluxes from river meandering
We used the OC stocks calculated to channel depth to quan-
tify POC fluxes due to lateral channel migration (
F
CB
and
F
PB
in Eq. 5). We averaged the lateral migration rate over
83 km river length comprising eight meander bends (Fig. 2)
to capture the characteristic sediment transport distances be-
tween depositional events (Pizzuto et al., 2014), variation in
local erosion rate due to channel curvature (Sylvester et al.,
2019; Howard and Knutson, 1984), and the formation of cut-
offs and oxbow lakes. We calculated the mean bank erosion
rate by averaging the area of floodplain eroded (1.60 km
2
)
and accreted (1.85 km
2
) from previously published erosion
masks generated using Landsat imagery (Rowland et al.,
2019). Dividing this area by the length of the channel reach
centerline (82.823 km) and the measurement interval for the
erosion masks (2018–1978) resulted in a mean lateral migra-
tion rate of 0.52 m yr
−
1
.
We approximated the flux into the river due to cutbank
erosion as
F
CB
=
L
×
E
×
C
CB
, where
L
is a unit river reach
length (1 m),
E
is the bank erosion rate (0.52 m yr
−
1
), and
C
CB
is the cutbank carbon stock (kg OC m
−
2
). The point
bar carbon flux was similarly calculated using
F
PB
=
L
×
E
×
C
PB
, where
C
PB
is the carbon stock of the point bar
(kg OC m
−
2
). Using OC stocks integrated to channel depth,
we estimated fluxes of POC due to bank erosion as
F
CB
=
65
.
2
±
7
.
3 kg OC yr
−
1
and due to point bar deposition as
F
PB
=
59
.
3
±
8
.
2 kg OC yr
−
1
(Fig. 5c). This result means that
Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
429
Figure 5.
Carbon cycling due to river meandering.
(a)
Total organic carbon (OC) in each stratigraphic column integrated to 1 m below surface,
with unmeasured portions of the section assumed to be sand; horizontal lines indicate the mean and shaded regions 1 SD for the complete
dataset.
(b)
Total organic carbon in each stratigraphic column integrated to mean channel depth (12.4 m) using the same assumptions and
uncertainty.
(c)
The net OC flux due to channel migration is comparable to floodplain net ecological productivity (NEP), and both are zero
within uncertainty. The net flux of OC into the river due to erosion of cutbanks and out of the river due to sediment deposition in point bars in
the Koyukuk River is calculated as the mean OC stock for each landform (with
±
1 SD OC stock uncertainty for that landform) multiplied by
an average channel migration rate for a 1 m downstream section of riverbank. The cutbank and point bar fluxes are differenced to calculate
the net bank erosion flux. Floodplain NEP is calculated for a 10 km wide, 1 m downstream distance section of floodplain using previously
reported regional NEP and uncertainties (Potter et al., 2013).
OC fluxes due to bank erosion and bar deposition were equal
within uncertainty.
We used radiocarbon measurements to evaluate if (1) the
OC being eroded from cutbanks was oxidized during trans-
port (
F
OX
), (2) the eroded OC was re-deposited in bars via
lateral accretion (
F
PB
) or overbank deposits (
F
OB
), or (3)
new biospheric OC was being added to point bars and flood-
plains by vegetation growth after sediment deposition. Simi-
lar to TOC and TN contents, Fm displayed a trend of higher
values for finer grain sizes – a pattern consistent with prior
findings that reflects the greater proportional petrogenic OC
contribution in coarser material (Hilton et al., 2015; Galy
et al., 2007). Coarser sediment tended to have lower TOC
content, potentially indicating that low Fm values are in part
due to a greater fraction of petrogenic OC (
f
petro
). When we
fit a range of mixing models to assess sediment biospheric
radiocarbon activity, we found that sediment from cutbanks
and point bars had similar ranges of potential biospheric OC
end-members (Fig. 4c). This observation matched the range
of aged wood and plant debris found at sediment sampling
locations.
Our mass-balance calculation and the presence of aged
Fm
bio
in newly deposited point bars both support the hypoth-
esis that a significant fraction of OC eroded from cutbanks is
re-deposited in the floodplain and not oxidized during trans-
port. In addition to point bar deposition, OC could be lost
from the river via overbank deposition (
F
OB
). In this case,
one would expect the carbon stocks to increase on floodplain
surfaces of increasing age due to the deposition of silt units
near the surface. Our measurements did indicate a slight in-
crease in 1 m OC stocks between recently deposited point
bars and floodplain inferred to be older based on their dis-
tance to the river (Fig. 5a), but they did not show a signifi-
cant increase in OC stock when integrated to channel depth
(Fig. 5b). One possible explanation could be that
F
OB
is sub-
stantial but that this carbon has been remineralized and lost
to the atmosphere. To constrain the frequency of overbank
flooding along the Koyukuk River near Huslia, we exam-
ined the Landsat image record and did not find instances
of overbank flooding. Ice jams, where floating ice piles up
and causes high water during spring break up along Arc-
tic rivers, occurred only four times near Huslia from 1967–
2019, and in these cases, overbank flooding did not occur
(White and Eames, 1999). Therefore, historical records sug-
gest that sediment fluxes due to overbank sediment deposi-
tion are relatively minor compared to fluxes due to channel
migration. Our stratigraphic observations showing a similar
thickness of capping silt units in floodplain stratigraphy (with
a mean of 1.29 m for cutbank, 0.92 m for floodplain, and
1.55 m for point bar samples; Table S4), and the low mass
fraction of siliciclastic sediment in organic horizons (based
on high mass fraction TOC; Fig. 4a) also indicated that over-
bank deposition of sediment on the distal floodplain is rela-
tively small.
Rather than additional OC from overbank flows, flood-
plains do appear to accumulate additional OC from biomass
production. We observed an increase in organic horizon
thickness, from a mean of 0.06 m in point bars to 0.45 m in
cutbanks and 0.44 m in floodplain deposits, primarily driven
by increasing thickness of peat horizons (Table S4). The
increase in organic horizon thickness can explain the cut-
bank and floodplain OC stocks summed to 1 m depth being
slightly higher than the point bar 1 m OC stocks. Since OC
stocks summed to channel depth were statistically similar be-
https://doi.org/10.5194/esurf-10-421-2022
Earth Surf. Dynam., 10, 421–435, 2022
430
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
tween landforms, we expected that there was some oxidation
of modern, labile OC during fluvial transport that was re-
placed after sediment was deposited in a point bar by biomass
production. In spite of biospheric OC input to floodplain
sediment through the growth of peat (on permafrost) and
an organic-rich topsoil (on non-permafrost), observations of
sediment containing old radiocarbon in both cutbanks and
point bars indicate that point bar OC has been eroded from
upstream and subsequently re-deposited, generating a reser-
voir of OC that has been aged by sediment storage along the
Koyukuk River.
6 Discussion
Our mass-balance model indicated that channel migra-
tion generated substantial fluxes of OC into the river (
>
50 kg OC yr
−
1
from cutbank erosion). If we assumed that
all OC in point bars was deposited with river sediment, the
calculated OC fluxes due to bank erosion and bar deposi-
tion balanced each other within uncertainty (Fig. 5c). How-
ever, our radiocarbon analyses indicated that a portion of
the biospheric OC in point bars was fixed after deposition
by local vegetation. This was reflected in slightly higher
1 m OC stocks in cutbanks and floodplain deposits versus
point bars. If we instead assumed that around half of OC
in eroding cutbanks was oxidized during river transport,
we calculated the river must transport downstream or oxi-
dize
>
30 kg OC yr
−
1
per meter of river reach. For compar-
ison, measurements of floodplain net ecological productiv-
ity (NEP) – the rate of OC fixation minus respiration – in-
dicated that an equivalent 10 km wide, 1 m long river reach
would emit 12
.
1
±
39
.
9 kg OC yr
−
1
(mean
±
1 SD) (Potter et
al., 2013). Therefore, the large depth (
>
10 m) and migration
rates (0.52 m yr
−
1
) of the Koyukuk River allow fluxes due
to bank erosion and deposition to exceed floodplain NEP, de-
spite the far smaller land area of erosion and deposition along
the riverbanks compared to the expansive floodplain. In addi-
tion, our results indicate that
∼
75 % of OC liberated by bank
erosion comes from below the top meter. Therefore, large
downstream OC fluxes from river migration can be attributed
to rapid exposure and mobilization of a deep OC reservoir
not readily accessible by top–down thaw.
The channel migration rates we measured reflect the river
area eroded versus deposited from 1978–2018, and these mi-
gration rates are influenced by the cutoff of a narrow river
reach that decreases channel length but slightly increases
average width (Fig. S2). Autogenic processes such as river
response to cutoffs and re-visiting areas of the floodplain
more or less frequently may cause transient changes in down-
stream OC fluxes along the Koyukuk. However, sparse obser-
vations indicate very high excess dissolved CO
2
and methane
in Koyukuk River water, supporting that there is significant
OC oxidation during transport (
F
OX
) (Striegl et al., 2012).
Overall, significant work remains to understand the partition-
ing of OC loss between the dissolved and particulate loads as
well as between petrogenic versus biospheric POC, particu-
larly since DOC concentration and lability varies seasonally
in the headwaters of the Koyukuk River (O’Donnell et al.,
2010).
Our results indicated less variability in OC stocks across
the Koyukuk River floodplain than previous work by
Lininger et al. (2019), who found significant variations in
OC stocks between geomorphic units in the Yukon Flats.
Lininger et al. (2019) report OC stocks to a depth of 1 m
along the Yukon River and its tributaries and extrapolated
the deepest measured mineral OC contents to 1 m based on
similar OC content in a few samples taken at depth along cut-
banks. Similar to their results, we found that newly deposited
point bars without a thick organic horizon had slightly lower
OC stocks for the upper 1 m of sediment. Our results also
agree with Lininger et al. (2019) that the coarser sediment
fraction contributes significant OC and that floodplain sedi-
ments can store OC for thousands of years between riverine
transport events. However, we found little variation with ge-
omorphic unit for OC stocks calculated to the channel depth
(12.4 m). Though we included organic horizons extending
below 1 m, the majority of our OC budget used to calcu-
late fluxes due to channel migration was comprised of the
more massive sandy deposits with low OC content. These
differences point to the importance of river depth relative to
the depth of significant floodplain biospheric OC production
and the grain size of the floodplain material at depth. We
hypothesize that cutbank and point bar OC stocks will be
similar for rivers with coarser sediment and channels much
deeper than the active layer and rooting depth of vegetation.
In contrast, OC stocks in floodplains of fine-grained, shallow
rivers might have a higher fraction of their OC oxidized after
erosion from cutbanks and replaced after deposition in point
bars.
The presence of aged biospheric OC in newly deposited,
non-permafrost point bars along the Koyukuk River illus-
trated that floodplains are important reservoirs of aged OC
in sediments both with and without permafrost. Rivers tend
to rework younger floodplain deposits faster than older flood-
plain deposits, and this can yield a heavy-tailed distribution
of deposit ages and carbon storage over thousands of years
(Torres et al., 2017). Our results supported the idea that a
fraction of particulate OC has experienced transient mobi-
lization and deposition and thus becomes naturally aged dur-
ing transport through the river–floodplain system. Therefore,
particulate OC with old radiocarbon signatures might be at-
tributed to OC storage in floodplains and may not be a diag-
nostic indicator of permafrost thaw. One might expect better
preservation of carbon stocks in permafrost deposits. How-
ever, our field observations of bank sediment rapidly chang-
ing color from gray to orange when exposed to air imply that
thawed floodplain sediments may be anoxic, which would
reduce rates of organic matter respiration in non-permafrost
deposits. When coupled with cold mean annual temperatures,
Earth Surf. Dynam., 10, 421–435, 2022
https://doi.org/10.5194/esurf-10-421-2022
M. M. Douglas et al.: Organic carbon burial by river meandering partially offsets bank erosion carbon fluxes
431
anoxic non-permafrost terrain might be similarly effective
as permafrost in preserving and aging biospheric OC stocks
(Davidson et al., 2006). Thus, transient storage of particles
in floodplains, potentially for thousands of years (Repasch et
al., 2020; Torres et al., 2020), may delay or diffuse down-
stream signals of perturbations to the watershed’s carbon cy-
cle before reaching long-term monitoring stations at river
mouths or sediment depocenters (McClelland et al., 2016;
Holmes et al., 2012).
Climate change is expected to cause a decrease or dis-
appearance of permafrost, which might alter rates of POC
oxidation (
F
OX
), overbank deposition (
F
OB
), and ultimately
downstream riverine POC fluxes. Permafrost thaw is also hy-
pothesized to increase river lateral migration rates (Costard
et al., 2003), although such changes have yet to be system-
atically documented. For the Koyukuk River, higher chan-
nel migration rates should, with all else equal, increase the
magnitude of OC fluxes due to erosion and deposition and
thereby decrease the residence time and age of OC within the
floodplain, but possibly with no net change in OC fluxes from
the floodplain to the river. However, if, for example, climate
change increases the relative importance of overbank deposi-
tion of OC-rich mud (higher
F
OB
) relative to sand bar accre-
tion, then this change would cause a permanent increase in
floodplain OC stocks, with associated decreases in OC river
fluxes during the transient period of floodplain grain size fin-
ing. In contrast, an increase in channel lateral migration rela-
tive to overbank flooding would cause floodplains to become
sandier and floodplain OC stocks to decline. Furthermore,
climate change is altering flood discharge and frequency
(Koch et al., 2013; Vonk et al., 2019; Walvoord and Kurylyk,
2016) as well as sediment supply, often associated with thaw
slumps (Kokelj et al., 2013; Lantz and Kokelj, 2008; Malone
et al., 2013; Shakil et al., 2020). Increases in flood magnitude
could cause channel widening (Ashmore and Church, 2001;
Walvoord and Kurylyk, 2016), which would increase cut-
bank OC fluxes relative to point bar fluxes (
F
CB
> F
PB
), cre-
ating a transient increase in riverine OC flux. We expect that
changes in floodplain hydrology and sedimentation due to
climate change will alter downstream particulate OC fluxes
and floodplain OC storage along deep, meandering Arctic
rivers similar to the Koyukuk. In the process, sediment depo-
sition in river bars should preserve radiocarbon-depleted OC
and dampen positive feedbacks due to POC being released
from permafrost by riverbank erosion as the climate warms.
7 Conclusions
To evaluate the role of riverbank erosion and bar deposition
in liberating organic carbon (OC) from permafrost flood-
plains, we conducted a field campaign along the Koyukuk
River in central Alaska, taking samples of riverbank and
floodplain sedimentary deposits. Finer bank sediment had
a systematically higher TOC content and Fm values than
coarser sands. We combined measurements on individual
samples with measured floodplain stratigraphic columns to
calculate OC stocks for cutbanks, point bars, and floodplains
summed to both 1 m below the surface and extrapolated to
the 12.4 m river channel depth. We found that cutbanks had
slightly higher OC stocks than point bars at shallow depths.
However, OC stocks integrated to river channel depth did
not significantly vary between river cutbanks, floodplain, and
point bars or with the presence or absence of permafrost.
As the Koyukuk River migrates, it is able to rapidly erode
this deep OC reservoir, generating substantial OC fluxes
from bank erosion and bar deposition. Net OC fluxes due to
river migration are of the same order of magnitude as flood-
plain net ecological productivity, despite the river occupy-
ing a small fraction of the land surface. Our results indicate
that floodplain processes generated an aged biospheric radio-
carbon signature in newly deposited point bars, and varia-
tions in sediment Fm with grain size may be due to mix-
ing with a petrogenic end-member. We conclude that a por-
tion of biospheric OC that was eroded from cutbanks was
preserved through transport and deposition. The presence
of radiocarbon-depleted sediment in non-permafrost deposits
indicates that aged POC in Arctic rivers is not a unique
indicator for the presence of permafrost. Our results high-
light that Arctic floodplains are significant reservoirs of OC,
and their stratigraphic architecture and morphology influ-
ence POC fluxes and radiocarbon ages transmitted down-
stream. Therefore, sediment deposition in river bars should
dampen positive feedbacks due to POC being released from
permafrost by riverbank erosion as the climate warms.
Data availability.
All datasets are included in the paper and Sup-
plement.
Supplement.
The supplement related to this article is available
online at: https://doi.org/10.5194/esurf-10-421-2022-supplement.
Author contributions.
MPL, JCR, WWF, AJW, GKL, and
MMD conceptualized the study. MPL, AJW, JCR and GKL deter-
mined the methodology. MMD, GKL, JCR, PCK, AJW, JS, APP,
AJC, and MPL collected field data. MMD, GKL, PCK, and AJW as-
sisted with geochemistry. MPL supervised the work. MMD con-
ducted data analysis and wrote the original draft, and all authors
contributed to the review and editing of the paper.
Competing interests.
One author is a member of the editorial
board of
Earth Surface Dynamics
. The peer-review process was
guided by an independent editor, and the authors also have no other
competing interests to declare.
https://doi.org/10.5194/esurf-10-421-2022
Earth Surf. Dynam., 10, 421–435, 2022