Floodplain Biogeochemical Mosaics - A Multidimensional View of Alluvial Soils

PUBLICATIONS
Journal of Geophysical Research: Biogeosciences

RESEARCH ARTICLE Floodplain biogeochemical mosaics:
10.1002/2013JG002543
A multidimensional view of alluvial soils
Key Points: Alison P. Appling1,2, Emily S. Bernhardt3, and Jack A. Stanford4
• Vegetation influences on floodplain soil
properties decline rapidly with depth 1
University Program in Ecology, Duke University, Durham, North Carolina, USA, 2Now at Department of Natural Resources
• Subsurface C and N dynamics
reflect flood-deposited sediment and the Environment, University of New Hampshire, Durham, New Hampshire, USA, 3Department of Biology, Duke
and organic matter University, Durham, North Carolina, USA, 4Flathead Lake Biological Station, University of Montana, Polson, Montana, USA
• Whole-floodplain C and N dynamics are
sensitive to these subsurface deposits
Abstract The alluvial floodplains of large rivers are exceptionally productive and dynamic ecosystems,
Supporting Information:
characterized by a complex mosaic of vegetation at different successional stages overlying soils sorted by
• Readme historic floods. Natural floodplains are widely credited with efficiently removing nitrogen from surface waters
• Table S1 and accumulating carbon in biomass, yet very little floodplain research has examined carbon and nitrogen
• Figure S1
• Figure S2
cycling below surficial soils. We evaluated the extent to which vegetation cover could be used to predict
• Figure S3 subsurface carbon and nitrogen dynamics and to estimate whole-floodplain carbon storage and denitrification
rates. We dug soil pits under three dominant vegetation communities on a gravel-bedded floodplain in
Correspondence to: northwest Montana to the depth of the permanent water table (1–3 m). We compared depth profiles of total
A. P. Appling,
[email protected]
and dissolved carbon (C) and nitrogen (N), denitrification potentials (DEAs), organic particulates, moisture, and
pH across vegetation types. Near-surface soils (0–10 cm) of forests had larger C and N pools and DEAs than
grasslands or gravel bars, but such vegetation effects dissipated within the upper ~50 cm of soil. At depth,
Citation:
Appling, A. P., E. S. Bernhardt, and J. A.
spatial heterogeneity in carbon and nitrogen pools and fluxes depended instead on soil texture, and relatively
Stanford (2014), Floodplain biogeo- high rates of DEA and carbon storage were measured in zones of buried organic debris. Although C storage and
chemical mosaics: A multidimensional denitrification potential are generally low in subsurface soils, these deep soils might nonetheless contribute
view of alluvial soils, J. Geophys. Res.
Biogeosci., 119, 1538–1553, doi:10.1002/
substantially to whole-floodplain C storage and denitrification because of their large volume, high hydrologic
2013JG002543. connectivity, and heterogeneous biogeochemistry.
Received 28 OCT 2013

Accepted 18 JUL 2014
Accepted article online 23 JUL 2014 1. Introduction
Published online 12 AUG 2014
Floodplains are notable for their high productivity, biodiversity, and capacity to retain and transform
nutrients relative to upland ecosystems [Lowrance et al., 1984; Naiman and Decamps, 1997; Ward et al., 1999].
Floodplains are also highly heterogeneous—in structure, biota, and biogeochemistry, and in both space and
time [Amoros and Bornette, 2002; Gallardo, 2003; McClain et al., 2003; Stanford et al., 2005]. To infer whole-
floodplain biogeochemical function from field measurements, researchers must choose field and analytical
methods that either are robust to that heterogeneity or incorporate it explicitly. This is especially important in
large river floodplains where water moves freely between the river channel and a large volume of subsurface
sediment, because in these massive subsurface reservoirs, even a low level of biogeochemical activity per
volume could scale to substantial activity across the full floodplain [Findlay et al., 2011]. Current attempts to
quantify whole-floodplain ecosystem dynamics are severely constrained by a lack of information on how soil
biogeochemical properties vary with depth, and to what extent that variation can be predicted from
observations at the soil surface.
It is well understood that the biogeochemistry of near-surface floodplain soils is related to vegetation cover,
in part because both soils and vegetation are simultaneously affected by flood disturbances, sediment
deposition, channel movement, and ecological succession [Walker, 1989; Schwarz et al., 1996; Zehetner et al.,
2009]. Especially in gravel-bedded floodplains, scouring floods remove fine soils and vegetation, leaving bare
cobble or sand with low carbon and nutrient concentrations and water-holding capacity [Merigliano, 2005;
Latterell et al., 2006]. On gravel bars and banks near active channels, tree seedlings and herbaceous
vegetation enhance the deposition of fine sediments [Gurnell et al., 2001; Corenblit et al., 2007], which often
contain organic matter and rapidly accumulate nutrients [Bechtold and Naiman, 2009]. Over time these
pioneer seedlings grow to maturity and give way to later-successional floodplain forests [Fastie, 1995]. In
many floodplains an alternative successional trajectory to grass or shrub cover occurs in locations where
avulsing channels leave behind cobble fields or ephemeral channels with reduced water-holding capacity,
APPLING ET AL. ©2014. American Geophysical Union. All Rights Reserved. 1538
21698961, 2014, 8, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1002/2013JG002543 by CAPES, Wiley Online Library on [17/04/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Journal of Geophysical Research: Biogeosciences 10.1002/2013JG002543
soil organic matter, and nutrient

content [Poole et al., 2002; Van Pelt et al.,
2006]. Along both successional
trajectories, soil C and N accumulate
over time due to plant growth and
turnover, biotic N fixation, and abiotic
N deposition [Wigginton et al., 2000;
Rhoades et al., 2008].
We hypothesized that vegetation
patterns might serve as a useful proxy
for three-dimensional spatial patterns
of soil biogeochemistry, with
correlations between vegetation type
and soil C and N dynamics extending to
deeper soils as they do in upland
ecosystems [Jobbágy and Jackson,
2000]. Limited research in floodplains
and riparian soils has provided some
support for vegetation influence
throughout the rooting zone [Lowrance,
1992; Gift et al., 2010; Cierjacks et al.,
2011]. Within the deeper strata of large
river floodplains, however, there are
good reasons to expect that the
influence of surface vegetation on soil
Figure 1. Vegetation cover map of Nyack Floodplain (modified from biogeochemistry may be overshadowed
Whited et al. [2007]). Sampling locations are labeled by vegetation cover
by hydrologic and geomorphic
(T = trees, G = grass, and B = gravel bar); brackets indicate sampling blocks
(1–4). Grass areas are common along channel and forest margins but are processes. Floods heterogeneously
usually small and difficult to see in this map. redistribute fine and coarse sediments
and organic-rich debris within floodplain
sediments [Pinay et al., 1995, 2000; Tabacchi et al., 1998; Blazejewski et al., 2009], and continual subsurface water
movement may supplement, dilute, or redistribute inorganic nutrients and dissolved organic matter pools along
subsurface flow paths [Boulton et al., 1998; Devito et al., 2000; Harms and Grimm, 2010].
In this study we investigated the extent to which surface vegetation could be used to predict soil properties
through the soil profile to the depth of the permanent water table. We sampled soil pits dug within a
Montana floodplain under three dominant vegetation cover types. In each soil pit we measured soil texture,
soil C and N pools, denitrification potentials, and other biogeochemical properties; we also installed
reflectometers and lysimeters to monitor soil moisture and pore water chemistry over the growing season.
We then used these data to explore biogeochemical patterns and function at the local and whole-floodplain
scales, paying particular attention to the spatial partitioning of soil carbon storage and denitrification
potential as critical ecosystem functions for which floodplains are uniquely suited. We hypothesized that
either (H1) vegetation cover predicts many soil properties at all measured depths, with the strongest
vegetation effects in near-surface soils or (H2) vegetation cover poorly predicts subsurface soil properties
because vegetation effects are overwhelmed by geomorphic and hydrologic effects.
2. Methods
2.1. Site Description
We conducted our study on Nyack Floodplain (Figure 1), an alluvial floodplain along the Middle Fork of the
Flathead River in northwest Montana (48°27′30″ N, 113°50′ W, 1010 m elevation). Along the floodplain reach,
the river is fifth order and has a 2300 km2 catchment draining the mountains of Glacier National Park and the
Bob Marshall Wilderness Complex [Poole, 2006]. We used the floodplain boundaries defined by Whited et al.
[2007], with upstream and downstream boundaries at bedrock canyons that constrain flow to the main
channel, and lateral boundaries that include all side channels and exclude an area of modest development
(road, railway, and pasture) along the western edge of the floodplain. By this definition, the floodplain is
roughly 9 km long, 0.5–2 km wide, and 677 ha in area, with a mean elevation above the base flow water
surface of 0.65 m (standard deviation = 1.24) (Flathead Lake Biological Station, unpublished data, 2007).
As of 2004, the area of the defined floodplain included 10.4% open water and 18.1% gravel bar [Whited et al.,
2007]. The parafluvial zone of active scouring, encompassing 39% of the defined floodplain [Whited et al.,
2007], is dominated by bare gravel bars and stands of early successional cottonwoods (Populus trichocarpa)
and willows (Salix spp.), while the orthofluvial zone of gentler annual inundation (74% of the defined
floodplain area in 2008) (Flathead Lake Biological Station, unpublished data, 2007) contains mixed conifers
(primarily Picea engelmannii, P. glauca, and Pseudotsuga menziesii) and mature cottonwoods and birches
(Betula papyrifera), and the remaining area is terrace with a higher density of conifers and intermingled
cottonwoods and birches [Poole et al., 2002; Stanford et al., 2005].
Due to repeated flood disturbances across the floodplain surface, soils are relatively undeveloped fluvents
and aquepts [U.S. Department of Agriculture, 2012] with a depth of 0.5–3 m [Diehl, 2004; Helton et al., 2014].
This upper soil layer is underlain by a 6–25 m thick layer of semicompacted and compacted gravel, sand, and
clay, followed by more compacted deeper layers and bedrock at 20–200 m below the soil surface [Harrison,
2004]. Subsurface flow within the semicompacted layer is unaltered by human activity and is laterally
extensive: the alluvium conducts river water hundreds of meters from the main channel, and 5–25% of the
river water (depending on flow conditions) passes through this aquifer and rejoins the main channel within
the floodplain stretch [Helton et al., 2014].
2.2. Soil Pit Locations and Establishment

In midsummer 2008, we excavated and sampled 12 soil pits from the surface to just below the water table at
Nyack Floodplain. We established sites in four blocks, which differed in their positions along the floodplain reach
(Figure 1) and the dates on which we excavated and sampled them (23–25 July for blocks 1 and 2; 13 August for
blocks 3 and 4). Each block contained one pit from each of three vegetation types: gravel bar, grass, and forest
(Figure 1). We chose these vegetation types because they represent points along multiple, often divergent
successional trajectories. We defined “gravel bar” to include sites with no vegetation or with young cottonwoods
or willows, as long as the surface substrate was gravel and sand. Grassy sites were herb-dominated clearings
where trees and shrubs were sparse or absent, and forested sites were within mixed cottonwood-spruce stands.
Sites were selected to represent the diversity of environmental conditions and soil types present on the
floodplain. River discharge during the sampling period was near base flow at 16–57 m3 s1 and fell in the 59th
quantile for the years 1940–2008 [U.S. Geological Survey, 2001]. All gravel bar sites were at low elevations above
the water table (40–100 cm) compared to grassland sites (110–250 cm) and forested sites (155–230 cm)
(Figure 2b). Our pits were roughly 2 m × 2 m wide and 1–2.7 m deep; despite this apparently large volume, the
excavations disturbed only 0.0007% of the floodplain surface.
According to nine aerial photographs spanning the period from 1945 to 2004 (data from Whited et al. [2007]), all of
our forested sites had been forested since at least 1945 (Figure 2a). Only one of our grassland sites (block 2)
had been grass covered since 1945; the others were mature forest or active river channel in 1945 and transitioned
to grass cover following disturbances. Gravel bar sites had all been gravel or active channel since at least 1986;
site 4B was classified as “regenerating forest” from 1990 to 2004 but had only 1 m tall seedlings and a gravel
surface as of 2008 (Figure 2a). We computed two estimates of stand age: the age of the oldest tree as determined
by cores from the three largest trees at each site and the years elapsed since the earliest aerial photo after which
the site was never gravel dominated in the photographic record. Because our estimates of stand age were
constrained by the availability of pioneer trees and aerial photographs, each estimate represents a lower bound
on the time since disturbance, so we take the maximum estimate for each site as the best, but still lower bound,
estimate. Forest sites were oldest (80–116 years, mean = 92), grass sites were intermediate (9–63 years,
mean = 49.5), and gravel bar sites were typically youngest (3–19 years, mean = 12.5). Note, however, that forested
sites do not necessarily progress through a grassland phase during succession.
After soil sampling, we identified two reasonably accessible soil pits (2G and 2T, Figure 1) in which we installed
ceramic cup lysimeters (Soilmoisture Equipment Corp.) at 10, 50, 100, and 200 cm depth and 0.5 inch PVC
piezometers at 290 cm depth. We also installed soil moisture probes (Campbell Scientific, Inc.) at 10, 20, 30, 50, and
Figure 2. (a) Vegetation cover history of each soil pit from aerial photos. Sites are grouped into panels by their vegetation
cover in 2008. “Mature Forest” indicates forest stands, “Regenerating Forest” indicates that trees are visible from aerial
photos (but may be mere seedlings, as for the gravel bar site in block 4), and “Grass,” “Cobble,” and “Water” indicate sites
with few or no trees. (b) Soil texture profiles extending from the soil surface (top of each bar) to 30 cm below the aquifer
water table (blue background). USDA texture categories are ordered by their silt + clay content, with more darkly colored
categories having higher silt + clay. Gravel soils are those with at least 30% gravel by volume.
100 cm and placed a Solinst Levellogger in a nearby observation well. The lysimeters, piezometers, and moisture
probes were inserted as deeply as possible into an intact wall of each soil pit (usually 30–50 cm beyond the wall
face) to reduce the influence of the soil pit excavation on measurements from these instruments. We then refilled
the soil pits and allowed the samplers to equilibrate for 6 months before monitoring hydrology and pore water
chemistry throughout the 2009 growing season.
2.3. Sample Collection

In each pit we sampled soils at 10 cm increments by driving a five-sided, 500 mL metal box into the soil pit
wall. Because water table depth rarely fluctuates by more than 10 cm d1 in the Nyack floodplain, we
did not restrict our sampling efforts to a specific time of day. Soils were returned to the lab on ice and stored
at 4°C until analysis. We conducted most soil analyses on samples from 3 to 4 key depths within each soil
profile: the top 10 cm of soil excluding surface litter (Category 1: near surface), an intermediate depth
below the A horizon but well above the water table (Category 2: intermediate), 10–20 cm above the water
table (Category 3: seasonally saturated), and 10–20 cm below the water table (Category 4: permanently
saturated). Throughout this article, we will refer to individual samples using codes of the form Block-
Vegetation-Depth. Sampling blocks and depth categories have values from 1 to 4, while vegetation is
represented by “T” for “trees,” “G” for “grass,” and “B” for “gravel bar.” For example, “3T2” refers to the soil
sample collected from the block 3 forested site at intermediate depth.
From March to August, 2009, we sampled soil pore water from our lysimeters and piezometers at 2 to 30 day
intervals, with the most frequent sampling occurring during the snowmelt period (March–April). Dry soils
prevented us from using a tension lysimeter to sample pore water after 1 June 2009 at the grassy site in
block 2, 10 cm, or after 15 July 2009 at several depths at each site. For the same reason, piezometer
collections at 290 cm were unsuccessful before 19 April 2009. We applied a vacuum pressure of 50 kPa for
sandy soils and 60 kPa for finer soils, waited 24 h, and then pumped the collected water into acid-washed glass
bottles. Samples were filtered through ashed GF/F filters on the day of collection and frozen until analysis.
2.4. Biogeochemical Analyses

Soils were sieved to isolate rock and organic particles larger than 2 mm. Roots and particulate organic matter of
any size were removed during sieving, dried at 60°C, and weighed. The volume of gravel (cobbles and
pebbles > 2 mm) was measured by water volume displacement in a graduated cylinder. The sieved soil was used
for all subsequent analyses. A 5–10 g subsample was weighed, dried at 60°C for 15 days, and reweighed to
determine soil moisture. Drying at this relatively low temperature allowed us to use the dried samples for later
CHN analysis but may have caused us to underestimate soil moisture; a later comparison of 15 soils after air drying
and after drying at 105°C for 2 days suggests that this underestimate will have rarely exceeded 3%. Soil moisture
estimates for eight samples from blocks 1 and 2 were lost due to laboratory errors and were replaced with
estimates for nearby, comparable soils collected on the same day (three samples) or replacements collected when
we returned to the field to sample blocks 3 and 4 (five samples). Soil pH was measured in 0.01 M CaCl2 solution
[Thomas, 1996] using a VWR sympHony Posi-pHlo Glass pH electrode.
KCl-extractable concentrations of NO2 + NO3 and NH4+ were assessed 7–9 days after sample collection
by shaking 5 g dry mass of soil in 50 mL 2 M KCl solution for an hour. Suspensions were filtered and analyzed
for [NO2 + NO3] and [NH4+] on a Bran + Luebbe continuous flow analyzer (cadmium reduction and
salicylate methods, respectively; SEAL Analytical Inc.).
To extract dissolved organic carbon (DOC) from soils, 50 g of field-moist soil and 150 mL of deionized water
were shaken together in 250 mL bottles for 1 h and then centrifuged and filtered. We used a Shimadzu Total
Organic Carbon Analyzer to measure both water-extractable DOC (extracted from soil in 2008) and pore water
DOC (collected from lysimeters and piezometers in 2009). Bioavailable DOC (bDOC) was determined in two
batches: blocks 1 and 2 and blocks 3 and 4. In each batch, an inoculum was prepared by pooling 10 mL from each
of the 26–27 unfiltered, centrifuged water extracts per batch. The remainder of each extract was filtered to
0.22 μm, and a 30 mL initial subsample was immediately acidified with phosphoric acid. Another 30 mL
subsample was transferred to a glass TOC vial. The samples in TOC vials received 250 μL inoculum and 50 μL
nutrient solutions (0.1% NH4NO3 and 0.1% K2HPO4) and were incubated in the dark at room temperature
for 7 days, with hand shaking on intermediate days to prevent anoxia. At the end of the incubation, extracts
were refiltered to 0.22 μm and acidified. Initial and final samples were measured on a Shimadzu Total Organic
Carbon Analyzer.
We assayed denitrification potentials (DEAs) with the acetylene block technique [Groffman et al., 1999].
We chose to assay DEA rather than in situ denitrification in part because DEAs are more consistent across
days and seasons [Groffman, 1987]. Because we were most interested in the capacity of the ecosystem to
remove a pulse of NO3-, our standard assay for all four sampling blocks included supplemental NO3- (“DEAN”;
20 mL at 100 mg N-KNO3 L1) but not organic carbon. For block 3 we also amended additional subsamples
with glucose (“DEAC”; 200 mg C-C6H12O6 L1), NO3 and glucose (“DEACN”), or neither (“DEA0”) to determine
whether organic C or NO3 was more limiting to denitrification. Further, for blocks 3 and 4 we assayed
both DEAN and DEACN to facilitate cross-study comparisons, since DEACN is the protocol most commonly
reported in the literature. All assays were conducted with at least two replicates per soil treatment
combination; block 3 soils were assayed in triplicate. We placed 10 mL of field-moist soil in a 125 mL
Erlenmeyer flask with a ground glass neck and butyl seal. We added 20 mL of deionized water containing
125 mg L1 chloramphenicol and NO3 and/or glucose in the concentrations noted above. We initiated each
incubation by repeatedly evacuating the flask headspace and flushing with N2; we then added 10 mL
acetylene (10% of headspace volume) that had been purged through sequential traps of sulfuric acid and
water. Samples were gently agitated during incubation on a rotary shaker. A syringe was used to inject 10 mL
of N2 and collect 10 mL of headspace at 0, 1, and 3 h for all soil samples. For sampling blocks 3 and 4,
headspace was also collected at 12 h for greater sensitivity to low N2O production rates. Gas samples
were stored in N2-flushed, evacuated, 9 mL glass vials with butyl septa and aluminum crimp seals
(Grace Davison Discovery Science) until analysis on a gas chromatograph with an electron capture
detector (SRI instruments, Torrance, CA).
To compute denitrification potentials, we first corrected headspace N2O concentrations for solubility in water
using a Bunsen coefficient of 0.545 at 25°C [Wilhelm et al., 1977]. For each flask we fit a linear model to
cumulative N2O production versus time. We quantified the minimum detection limit for N2O on our
gas chromatograph (GC-MDL) as the standard deviation of measurements of a low N2O standard
(0.1 ppm). We then defined a minimum detectable incubation slope (MDS) of 4 times the GC-MDL
divided by the length of the incubation in hours. We set any results below the MDS to half the MDS.
We also discarded assay results if the model fit had a p value greater than 0.05. Because the MDL for
the overall DEA method (our DEA-MDL) depended on the N2O production rate from samples of fixed
volume rather than mass, the DEA-MDL expressed per gram soil was sample specific and ranged in
practice from 0.46 to 1.0 ng N h1 (g soil)1. Reported DEA results are averages of the duplicate or
triplicate assay results for each soil sample.
Dry, archived samples from every 10 cm increment at each site (n = 191) were later analyzed for C and N
content, gravel volume, and soil texture. These samples were first sieved to 2 mm to separate gravel from soil;
organic particles were also removed. Subsamples were homogenized with mortar and pestle and then
combusted in a ThermoQuest NC 2100 CHN analyzer at 1050°C to measure total %N and 600°C to measure
%C without carbonates. As with the samples analyzed for DEA and other properties, gravel volume was
determined by volumetric displacement of water in a graduated cylinder. The texture of the sieved soil was
determined to the level of U.S. Department of Agriculture (USDA) textural class by hand. We report soil
texture as “gravel” when particles larger than 2 mm in diameter were > 30% of the volume of the sample;
otherwise, we give the USDA textural class of the sieved soil. The sieved fractions of the 102 gravel soils had
USDA textures of sandy clay loam (n = 1), sandy loam (n = 3), loamy sand (n = 5), or, most often, sand (n = 93).
2.5. Statistical Analyses

All statistical analyses were performed in R version 2.15.0 [R Development Core Team, 2010]. To assess the
effects of depth category and vegetation cover on soil properties, we ran factorial analyses of variance
(ANOVAs) for linear models of the form
y abi ¼ V a þ Db þ ½V Dab þ Bi þ εabi
εabi ~ Nð0; σ2 Þ
where yab is the soil property of interest (log transformed as necessary to achieve a normal distribution;
Table S1 in the supporting information), Va = vegetation effects, Db = depth effects, Bi = sampling block
effects, εab = independent error terms, a = vegetation type, b = sample depth, and i = block. Significant
main effects and interactions render Tukey’s honest significant difference (HSD) tests uninterpretable for
pairwise treatment comparisons. We therefore ran these post hoc comparisons on models of subsets of
the data that were restricted to single-vegetation cover types (for depth comparisons) or depth
categories (for cover type comparisons).
For soil %C and %N, which were measured not just at the four representative depths but also at 10 cm intervals
throughout the profile, we explicitly tested soil texture effects by modeling soil %C, %N, and C:N ratios as
functions of ln(depth), vegetation cover, and silt + clay (expressed as a mass fraction of the sieved soil). Since
block effects were nonsignificant in the discrete-depth models, we omitted them here. P values were
heteroscedasticity corrected by the Huber-White method using the “car” package in R [Fox and Weisberg, 2011].
We also identified disruptions in general depth trends (“local maxima” or “hot spots”) by the arbitrary but
consistent criterion that a %C or %N value should be more than 50% higher than the mean of the three 10 cm
increments overlying that depth increment and should exceed 0.5%C or 0.05%N.
Treatment responses in the DEA limitation assay were assessed with one-way ANOVA and Tukey’s HSD test. To
investigate in situ predictors of DEAN, we used bidirectional stepwise regression with the Akaike information
criterion to simplify and fit a model of the form, ln(DEAN) ~ ln(%C) + ln(NO3) + moisture. We fit our model to
data for near-surface soils only (n = 12) because most DEA values above the MDL were at this depth. DEAs and
candidate predictor variables were log transformed as necessary to achieve more normal distributions.
P values for model coefficients were again computed from heteroscedasticity-corrected covariance
matrices with the car package.
To visualize samples in multidimensional biogeochemical space, we ran a principal component analysis (PCA)
using those soil properties that were most directly related to elemental pools and fluxes: the immobile pools
represented by soil %C and %N; dissolved pools of DOC, NO3, and NH4+; soil water content; and
denitrification potential. Before running the PCA, we standardized each variable by subtracting the mean and
dividing by the standard deviation.
We modeled temporal patterns in soil pore water [DOC] and [NO3] using fixed, categorical effects for
site, sampling depth, and sampling date. Of the 346 total samples, four DOC measurements and five
NO3 measurements were identified and excluded as outliers based on Grubb’s tests performed with
the “outliers” package in R [Komsta, 2011]. We assessed differences across sites and depths using
ANOVAs and Tukey’s HSD test, and we report pairwise differences only when the p value for a pairwise
comparison is less than 0.05.
2.6. Floodplain-Scale C Storage and Denitrification
We explored the floodplain-scale implications of vegetation effects and subsurface soil heterogeneity
for two biogeochemical functions, soil C storage and potential denitrification of river-borne NO3. We
first used ArcGIS, version 10 (Environmental Systems Research Institute, Inc., Redlands, California) to
produce maps of these biogeochemical functions in near-surface soil layers. We began with a map in
which vegetation cover was classified into patches of Cobble, Grass, Regenerating Forest, or Mature
Forest based on aerial photographs (Figure 1) (data from Whited et al. [2007]). We then assigned each
patch the vegetation-specific means we had observed for soil C storage and denitrification potential.
For this analysis only, we split our gravel bar sites into Cobble (sites 1B, 2B, and 3B) and Regenerating
Forest (site 4B) to be consistent with the classifications used by Whited et al. [2007]. Because 4B is less
developed than some other areas in the Regenerating Forest class, this approach may underestimate
the mean C and denitrification potential of that class.
We evaluated soil C storage in the upper 100 cm because many studies report soil properties to a
maximum of 100 cm depth. We evaluated potential denitrification in the top 10 cm, focusing on the
denitrification of riverine NO3 (as distinct from NO3 that originates in floodplain soils). Denitrification
of riverine NO3 is a function of three factors: the spatial distribution of denitrification potential, the
duration of contact between river-borne NO3 and floodplain soils, and the volume of soil in contact
with river water. We therefore assumed that near-surface denitrification of river-borne NO3 occurs only
in the top 10 cm of soil (where DEAs are greatest), only during the ~1 month of inundation per year
(further assuming that NO3 is nonlimiting during this period), and only in that area of the floodplain
(74%) that was inundated during the moderately large flood of 2008 (Flathead Lake Biological Station,
unpublished data, 2007). Restricting denitrification to the flooded area changes not only the relevant
soil volume but also the proportion of each vegetation cover within that soil: The percent cover by
vegetation type is 44% mature forest, 27% regenerating forest, 1% grass, and 28% cobble in the flooded
area, compared to 58%, 21%, 1%, and 20%, respectively, in the full floodplain.
We also scaled our estimates to the whole floodplain by multiplying the total area of each vegetation cover type
by the average C storage or DEA value for that cover type. We calculated the uncertainty in these whole-
floodplain estimates by propagating the standard errors for each vegetation-specific soil property distribution
through the multiplication by cover type areas and the sum over all cover types as follows:
hX i1=2
SEðy tot Þ ¼ Av 2 SEðy v Þ2
where ytot is the total soil property value for the whole floodplain, SE is the standard error function, Av is
the area covered by cover type v, and yv is the mean soil property value for that cover type. We could not
directly calculate the standard errors for Regenerating Forest because we had only one site (4B) in that
category; we therefore used the largest standard error from the other three categories as a conservative
estimate of the standard error for Regenerating Forest.
a
Table 1. Overall Treatment Effects and Pairwise Treatment Comparisons by Soil Property
Soil Property Vegetation Depth Vegetation × Depth Block Vegetation|Depth Depth|Vegetation
Soil C (%) 0.001 (8.4) 0.000 (14.9) 0.010 (3.4) 0.130 (2.0) T > G > B|1 1 > 3 = 4|T; 1 > 3 = 4|G
1
DOC (μg C (g soil) ) 0.000 (17.5) 0.000 (16.5) 0.051 (2.4) 0.014 (4.2) T > G > B|1 1 > 3|T; 1 > 3 = 4|G
1 1
bDOC (mg C d (g soil) ) 0.004 (7.0) 0.003 (6.1) 0.025 (2.9) 0.143 (2.0) T > G = B|1; T > G|4 1 > 3|T
Soil N (%) 0.001 (9.1) 0.000 (15.5) 0.008 (3.6) 0.865 (0.2) T > G > B|1 1 > 2 = 3 = 4|T; 1 > 3 = 4|G
1
NO2 + NO3 (μg N (g soil) ) 0.435 (0.9) 0.089 (2.4) 0.669 (0.7) 0.100 (2.3) B > G|2 (none)
+ 1
NH4 (μg N (g soil) ) 0.095 (2.6) 0.009 (4.7) 0.595 (0.8) 0.143 (2.0) (none) (none)
1 1
DEA (ng N h (g soil) ) 0.000 (13.6) 0.000 (25.6) 0.000 (9.9) 0.151 (1.9) T > G = B|1 1 > 2 = 3 = 4|T
Moisture (mass %) 0.033 (3.8) 0.000 (11.8) 0.086 (2.1) 0.032 (3.3) T > B|1; T > B|3 1 < 4|G; 1 = 3 < 2 = 4|B
pH 0.791 (0.2) 0.113 (2.2) 0.475 (0.9) 0.097 (2.3) (none) (none)
1
Fine roots (g (g soil) ) 0.139 (2.1) 0.000 (13.8) 0.065 (2.2) 0.346 (1.1) T = G > B|1 1 > 3 = 4,2 > 4|T; 1 = 2 > 3 = 4|G
1
POM (g (g soil) ) 0.008 (5.6) 0.000 (18.5) 0.863 (0.4) 0.027 (3.5) T > B|1 1 > 3 = 4|T; 1 > 3 = 4|G; 1 > 4|B
Rock volume (%) 0.000 (15.8) 0.000 (19.2) 0.062 (2.3) 0.233 (1.5) T = G < B|1 1 = 2 < 4|T; 1 = 2 < 3 = 4|G; 2 < 4|B
a
Overall treatment effects are reported as p value (F statistic). Treatment effects are reported whenever the p value for that pairwise comparison is less than 0.05.
“|” can be read as “given” or “within.” Vegetation categories are B (gravel bar), G (grass), and T (trees). Depth categories are 1 (near surface; top 10 cm),
2 (intermediate depth), 3 (seasonally saturated), and 4 (near-permanently saturated).
3. Results
3.1. Vegetation, Depth, and Texture Effects on Soil Biogeochemical Properties
Pairwise comparisons between vegetation covers and between depth categories reveal that most pools and
fluxes decline rapidly with depth, such that vegetation effects are negligible at all but the shallowest depths
(Tables 1 and 2). The near surface of forested sites, grassland sites, and gravel bars held 5.1%, 1.5%, and 0.3% C,
respectively (Table 2). The C:N of soil organic matter was highest in the forested sites and lowest in the
gravel bars, such that %N of soil organic matter was 0.26%, 0.11%, and 0.03% N in forests, grasslands, and
gravel bars (Table 2). The near-surface effects were strong enough that across all sampled depths, vegetation
cover significantly predicted 8 of 12 soil properties, including soil %C, water-extractable organic C (DOC),
bioavailable DOC (bDOC), soil %N, denitrification potential (DEA), moisture, particulate organic matter (POM),
and rock volume (Table 1). However, depth effects were even stronger; effects of depth category across
all vegetation covers were significant for 10 of 12 measured soil properties in factorial ANOVAs,
including the above properties plus [NH4+] and fine root mass. Almost all of the significant differences
between vegetation categories occurred in near-surface soils (Table 1), a pattern that can be explained
by a general decline with depth in C pools, N pools, and DEAs, and an increase with depth in soil
moisture and rock volume (Table 2). The interaction of vegetation and depth was significant for soil %C,
soil %N, bDOC, and DEA (Table 1).
Soil %C and %N at 10 cm intervals throughout each soil profile (Figure 3a) were consistent with patterns
inferred from the analysis by vegetation and depth categories. Integrating our soil C data at 10 cm
resolution for each profile, we estimate that soil C in the top 1 m differs significantly by vegetation cover
with 0.79 kg C m3 (SE = 0.16) in gravel bars, 5.1 kg C m3 (SE = 1.6) under grass, and 9.5 kg C m3 (SE = 1.8)
under forest. Total soil %N in each 10 cm increment was strongly correlated with soil %C (Pearson’s r = 0.91
for ln(%C) ~ ln(%N)) and had a nearly identical spatial pattern, so we only show %C and the C:N ratio in
Figure 3. Soil %C, %N, and C:N ratio all declined significantly with depth and increased significantly with the
silt + clay fraction. The regression equations are as follows:
lnð%CÞ ¼ 1:06*** – 0:33 lnðdepthÞ*** þ 0:01grass þ 0:37 forest** þ 2:99 ðsilt þ clayÞ***
lnð%NÞ ¼ 3:49*** – 0:21 lnðdepthÞ*** þ 0:12 grass þ 0:28 forest** þ 2:33 ðsilt þ clayÞ***
lnðC:NÞ ¼ 2:43*** – 0:12 lnðdepthÞ*** 0:11 grass þ 0:09 forest þ 0:66 ðsilt þ clayÞ***
where grass and forest are dummy variables (1 or 0); significant terms are asterisked (* = p < 0.05,
** = p < 0.01, and *** = p < 0.001); and R2 = 0.75 for %C, R2 = 0.77 for %N, and R2 = 0.37 for C:N ratio.
The significance of silt + clay in the C and N models is consistent with the observation that local maxima
in %C and %N never occurred in gravel or sand, occurred once in sandy loam (site 4B at 35 cm), and
most often occurred in loam (3G at 95 cm, 4G at 45 cm, 2T at 245 cm), silt loam (4G at 55 cm, 2T at
115 cm, 2T at 225–235 cm), or clay loam (2T at 75 cm).
a
Table 2. Soil Properties by Vegetation Cover and Depth
+
DOC bDOC NO2 + NO3 NH4
1 1 1 1 1
Vegetation Cover Depth Category Soil C (%) (μg (g soil) ) (μg C d (g soil) ) Soil N (%) (μg N (g soil) ) (μg N (g soil) )
Forest Near surface 5.11 (0.76) 120.7 (46.3) 3.70 (2.19) 0.262 (0.037) 0.81 (0.42) 5.25 (3.11)
Intermediate 1.09 (0.41) 59.0 (48.4) 0.52 (0.07) 0.077 (0.023) 0.36 (0.07) 1.58 (0.50)
Seasonally saturated 0.36 (0.14) 4.9 (1.3) 0.26 (0.08) 0.035 (0.008) 0.10 (0.01) 0.47 (0.11)
Permanently saturated 0.52 (0.35) 12.3 (4.7) 1.02 (0.24) 0.038 (0.014) 0.44 (0.25) 3.66 (3.08)
Grass Near surface 1.47 (0.39) 19.6 (3.5) 0.59 (0.15) 0.111 (0.026) 0.99 (0.66) 1.25 (0.40)
Intermediate 0.67 (0.28) 10.4 (4.1) 0.61 (0.30) 0.054 (0.016) 0.22 (0.02) 1.00 (0.39)
Gravel bar Near surface 0.25 (0.03) 5.3 (1.2) 0.27 (0.06) 0.031 (0.005) 0.48 (0.14) 1.46 (0.31)
Intermediate 0.71 (0.53) 7.2 (1.4) 0.10 (NA) 0.050 (0.028) 0.62 (0.10) 1.73 (1.51)
DEA Moisture Fine roots POM Rock volume

1 1 1 1
Vegetation Cover Depth Category (ng N h (g soil) ) (mass %) pH (mg (g soil) ) (mg (g soil) ) (%)
Forest Near surface 58.26 (5.98) 14.7 (1.9) 6.92 (0.15) 4.71 (0.76) 15.89 (5.61) 0.1 (0.1)
Intermediate 0.41 (0.02) 14.3 (4.3) 7.31 (0.13) 0.52 (0.33) 1.12 (0.43) 0.8 (0.8)
Grass Near surface 9.87 (5.84) 8.1 (3.3) 7.16 (0.05) 2.11 (0.93) 2.10 (0.76) 0.3 (0.3)
Intermediate 0.39 (0.06) 14.2 (6.3) 7.12 (0.14) 0.11 (0.04) 0.50 (0.48) 15.2 (15.1)
Gravel bar Near surface 0.30 (0.01) 4.2 (1.1) 7.22 (0.08) 0.14 (0.10) 0.41 (0.23) 48.5 (1.6)
Intermediate 0.93 (0.65) 24.6 (3.3) 7.22 (0.03) 0.24 (0.24) 0.36 (0.35) 26.4 (26.4)
a
Values are reported as mean (standard error). A denominator of “g soil” refers to grams of oven-dried soil.
3.2. Denitrification Potentials

For the most part denitrification potential estimates followed the same trend as %C and %N, with
the highest DEAs observed in the surficial soils of our forested sites (58 ng N h1 (g soil)1, SE = 6),
significantly lower DEAs in grassland sites (9.9 ng N h1 (g soil)1, SE = 5.8), and near-negligible rates in
our gravel bar sites (0.3 ng N h1 (g soil)1, SE = 0.01). DEACN was correlated with and ~25% higher than
DEAN (DEACN = 0.0015 + 1.25 × DEAN, R2 = 0.99, n = 6, p < 0.001) for the samples in which we assayed
both DEAN and DEACN (25 soils total; same six above MDL for both DEAN and DEACN). The stepwise
regression on near-surface soils retained all three terms, yielding ln(DEAN) = 0.60 + 1.10 × ln(%C)
0.21 × ln(NO3) + 0.17 × moisture (n = 12 with eight above MDL). Only the coefficient for ln(%C) was
significant (p = 0.04), but the coefficient for moisture had a relatively low p value (p = 0.13). The
coefficient for NO3 was both nonsignificant and negative. In contrast to this nonsignificance, three
samples in our DEA limitation assay had DEA0 = DEAC < MDL < DEAN = DEACN, indicating that DEA was
limited by NO3 supply (forest and grass at 0–10 cm and forest at 10–20 cm below the water table;
Figure S3 in the supporting information). Denitrification was below our method detection limit (MDL) for
all treatments in the other five of eight samples in the limitation assay.
3.3. Ordination of Biogeochemical Properties

A principal component analysis (PCA) of total and dissolved C and N pools, DEAs, and soil moisture shows
distinct separations among vegetation covers in the top 10 cm and a separate clustering of samples from
depth categories 3 and 4 (10–20 cm above and below the August water table, respectively); each of these
four groups is outlined in Figure 4. Soils at intermediate depths (codes ending in 2) were intermediately
located on the PCA, falling between the clusters of near-surface samples (0–10 cm; “B1,” “G1,” and “T1” groups
in the figure) and deep ones (group “Depths 3 & 4”). The forested and grassland sites at 0–10 cm are
Figure 3. Depth profiles of soil %C and C:N ratios. Data for %N are not shown because %C and %N are strongly correlated
(r = 0.91) and their depth profiles are nearly identical. Symbol shapes indicate sampling block, and fill colors indicate soil
texture category.
separated from the deep and gravel bar sites along PC axis 1, which is most strongly associated with the
factors %C, %N, DOC, NH4+, and DEA. Samples from depth categories 3 and 4 are separated from shallow
gravel bar samples primarily along PC axis 2, which is most strongly associated with soil moisture. The
remaining environmental factor, NO3, falls evenly between PC axes 1 and 2.
Three individual soil samples have been circled in the PCA because they lie far from the other samples with
the same vegetation cover and depth, i.e., they are local maxima with respect to several soil properties. The
block 4 gravel bar sample at 30–40 cm depth (sample 4B2) had the highest soil N (0.078%) and C (1.2%), the
greatest fine root mass (0.48 μg (g soil)1), the most NH4+ (3.2 μg N (g soil)1), and the highest DEA
(1.6 ng N h1 (g soil)1) of all gravel bar samples. The block 2 forest sample at 240–250 cm (sample 2T4) had
higher total N and organic C content (0.076% N, 1.55% C) than surrounding soils. Finally, the block 3 forest
sample at 170–180 cm (sample 3T3) had a high DEA of 3.6 ng N h1 (g soil)1. Each of these samples also
occurred at a local maximum within the %C and/or %N depth profiles (Figure 3) and tended to have a higher
silt + clay fraction than other soils in that depth-vegetation combination (Figures 2b and 4).
3.4. Spatiotemporal Patterns in Pore Water Chemistry

Like most C and N pools and fluxes in our 2008 soil samples, pore water data collected from two sites in 2009
showed overall declines in [DOC] and [NO3] with depth (Figure 5). After fixed effects for the sampling dates
were applied, [DOC] in pore water of the forested profile (2T) at 10 cm was higher than at 290 cm (p < 0.05),
and [DOC] was significantly higher at

50 cm than at any other depth
(p < 0.05). [DOC] in the grassy site was
likewise higher at 10 cm than at any
other depth (p < 0.05), and [DOC] at
100 cm exceeded [DOC] at 290 cm
(p < 0.05). Nitrate in the forested site
also declined consistently with depth,
with [NO3] at 10 cm greater than
[NO3] at 50 cm greater than [NO3]
at all other depths (p < 0.05). In
contrast to [NO3] at the forest site
and [DOC] at the grassy site, [NO3]
at the grassy site was highest at
depth: [NO3] at 100, 200, and
290 cm significantly exceeded
[NO3] at 10 cm (p < 0.05).
Between-site differences in mean
Figure 4. PCA of key soil properties for all sites and depths. Samples are speci-
concentrations were generally
fied with code BVD, where B is sampling block (1–4), V is vegetation cover type
consistent with the vegetation effects
(trees, grass, or gravel bar), and D is depth category (1 is top 10 cm, and 4 is
we observed in the soil samples, in
10–20 cm below water table). Arrow labels indicate the mapping of variables to

PCA axes: NO23 is KCl-extractable NO2 + NO3 , DOC is water-extractable
that the forested site had higher
[NO3] than the grassy site at 10 cm
dissolved organic C, DEA is N-amended denitrification potential, PctN and PctC
+
are soil %N and %C, NH4 is KCl-extractable NH4 , and Mois is gravimetric soil
and higher [DOC] at every depth
moisture. Dashed lines define sample groupings by vegetation cover and depth
(p < 0.05 for each test). However,
as indicated by the italic, gray labels. Fill colors indicate soil texture category.
[NO3] in the grassy site was higher

than [NO3 ] in the forested site at 100, 200, and 290 cm, demonstrating that the concentrations of mobile
solutes can be substantially decoupled from total nutrient concentrations in these soils.
Water infiltration and evapotranspiration were associated with several temporal patterns at our monitoring
sites. At the seasonal scale, [DOC] and [NO3] at many depths became more dilute as snow melted into the soils
(first 2 weeks of April) and more concentrated as soils dried through the summer (Figure 5; see also Figures S1
and S2 in the supporting information). At finer temporal scales, soil moisture pulses occurred daily during
the snowmelt period and with rain events throughout the spring and summer. These pulses propagated
through the gravel-dominated grass profile within hours, while the propagation was more evident over days
to weeks in the more finely textured forest profile (Figures 5a, 5d, and S2). Moisture pulses likely corresponded
to the downward transport of DOC and NO3 to subsurface soils, but concentration changes were typically
difficult to detect or quantify during even the largest pulses (9 April and 17 May, Figure 5). However, after the
May rainstorm, [DOC] did increase by up to 40% at some depths in the grassy site (Figures 5c and 5f).
3.5. Floodplain-Scale C Storage and Denitrification

We estimate that total C storage in the top 1 m is 37 × 103 Mg C (SE = 6.8) or 54 Mg C ha1 (SE = 10.0) over the
floodplain surface. Because soil C storage is greater in vegetated soils than in gravel bars, and vegetated soils are
often farther from the river channel than gravel bars, soil C tends to be concentrated away from the river
channel (Figure 6a). In contrast, denitrification activity is concentrated at an intermediate distance from the river
channel: far enough from the channel that the relatively C-rich soils of forested patches dominate, but near
enough to receive surface floodwaters in most years (Figure 6b). Annual denitrification potential for the flooded
near-surface soils (top 10 cm) is 5970 kg N yr1 (SE = 1240) or 8.8 kg N ha1 yr1 (SE = 1.8).
4. Discussion
4.1. Vegetation Predicts Biogeochemistry in the Near Surface But Not at Depth
Contrary to our first hypothesis, the predictive capacity of vegetation cover was negligible throughout the
subsurface, despite significant effects at the surface. We were surprised that vegetation effects dissipated so
Figure 5. Soil moisture and soil pore water chemistry in the spring and early summer of 2009 at the (a–c) forested site and (d–f)
grassland site of sampling block 2. Each line represents a sampling depth: moisture was measured at 10, 20, 30, 50, 100, and
200 cm, while pore water was sampled at 10, 50, 100, 200, and 290 cm. For complete data through August 2009, plus additional
hydrologic data, see Figures S1 and S2 in the supporting information.
rapidly with depth given that soils and vegetation develop together in these successional systems [Fonda,
1974; Walker, 1989; Corenblit et al., 2007]. Root mass, organic matter pools, and denitrification potentials were
generally low below ~40 cm (Table 1 and Figure 4). Similarly, total C and N in the vegetated profiles declined
to low, often indistinguishable values over ~50 cm (Figure 3). If vegetation effects are indeed insignificant
beyond 40–50 cm depth, then vegetation predicts biogeochemistry in a very small fraction of the floodplain’s
total soil profile, which averages 7 m deep.
4.2. Soil Texture Predicts Biogeochemical Variation and Hot Spots at All Depths
The poor predictive capacity of vegetation cover in the subsurface was due not only to the lower mean %C, %N,
organic matter, and denitrification potential at depth but also to the presence of local maxima (hot spots)
whose locations were less related to the overlying vegetation cover than to buried deposits of organic
debris and more finely textured soils. Soil textural differences also helped to explain differences in water
and solute movement through the subsurface. A strong relationship between soil texture and soil
properties in floodplains has been reported elsewhere [Pinay et al., 1992, 2000; Bechtold and Naiman, 2006;
McIntyre et al., 2009; Noe et al., 2013] and implies that subsurface soil biogeochemistry ultimately reflects
the drivers of soil texture patterns: erosion and deposition events from the historic to geologic past.
This texture-biogeochemistry relationship is consistent with our second hypothesis, that subsurface
biogeochemistry is largely driven by geomorphic and hydrologic events, and allows us to refine that
hypothesis: Erosion and deposition in floodplains not only decouple surface vegetation from subsurface
soil properties but also generate hot spots in an otherwise C-poor, low-activity, homogeneous matrix.
At short time scales, soil texture can drive large differences in the temporal patterns of pore water solute
concentrations through its effects on hydraulic conductivity and soil water-holding capacity. The two sites
where we monitored soil moisture and

pore water chemistry (block 2 forest
and grassland) were just 25 m apart but
differed markedly in their soil textural
profiles (Figure 2b) and in their
patterns of [DOC] and [NO3] over
time (Figure 5). Soil texture also
constrained rates of water
movement: Moisture pulses to the
subsurface of the coarse-textured
profile were immediate and short
lived, while in the finer-textured
profile, those pulses were lagged,
longer lasting, and more quickly
dampened with depth (Figures 5a
and 5d). Other work indicates that in
soils below the water table, texture
differences cause the preferential
flow of aquifer water along specific
flow paths [Poole et al., 2002],
illustrating yet another way that
texture drives biogeochemistry.
Relatively fine textures characterized
each of the three subsurface
samples where we found hot spots
in multiple biogeochemical
properties (circled in Figure 4).
Further supporting the importance of
soil texture variations to biogeochemical
Figure 6. Mosaics of (a) soil carbon storage to 1 m depth (top legend) and patterns, our 10 cm resolution soil

(b) potential denitrification of riverine NO3 in the top 10 cm (bottom samples showed significantly higher
legend). Biogeochemical properties are assigned by the vegetation means %C, %N, and C:N in soils with higher silt
for each soil property.
and clay fractions, even after accounting
for depth and vegetation effects.
Although much of the subsurface matrix is nutrient poor and inactive, hot spots within that matrix are abundant:
we found an additional four locations with locally elevated C and N in the 10 cm resolution soil samples, for a total
of seven hot spots in just 12 soil pits. The frequency of hot spots in our soil analyses is consistent with groundwater
sampling surveys of the Nyack, where Reid [2007] reported that sediments pumped from the upper 1 m of the
aquifer via PVC sampling wells were C rich (5–50 kg C m3) in 3 of those 45 wells.
4.3. Deep Soils May Matter

Our repeated observations of hot spots in soil C, N, and DEAs raise the question of whether subsurface soils
might contribute substantially, despite their decoupling from biological activity at the soil surface, to whole-
floodplain properties such as total C storage or denitrification. Subsurface hot spots of denitrification and carbon
storage have been observed in many riparian ecosystems [Jones et al., 1995; Sobczak et al., 1998; Kukulak et al.,
2002; Hill and Cardaci, 2004; Hill et al., 2004; Gurwick et al., 2008], and there have been calls to investigate the
importance of such hot spots at the ecosystem scale [McClain et al., 2003; Blazejewski et al., 2009; Groffman et al.,
2009]. However, studies of deep soils are challenging. Researchers seeking to estimate whole-ecosystem
biogeochemical properties or functions in floodplains must balance practical limitations, such as the
environmental and financial costs of deep soil sampling or geophysical sensing of the subsurface, against
the likely importance of such deposits as indicated by preliminary measurements. To that end, here we
explore the sensitivity of whole-ecosystem C storage and denitrification to the presence and abundance of
biogeochemical hot spots in the subsurface of Nyack Floodplain.
We found evidence for numerous organic deposits in the subsurface (this study and Reid [2007]) that might
be important to C storage at the floodplain scale. To see their potential contribution, suppose that all such
deposits contain 16.7 kg C m3 (equal to storage at the block 2 forest site at 220–250 cm), that most of the
subsurface contains 0.50 kg C m3 (the average %C of all sand-and-gravel samples below 1 m in our soil
profiles), and that the volume of the subsurface is 47 × 106 m3 (677 ha in area and approximately 7 m deep;
see section 2.1). Then if organic deposits occupied just 1.7% of the entire subsurface volume, then floodplain
aquifer sediments (>1 m deep) would contain as much carbon as the upper meter of soil (37 × 103 Mg C).
The actual volume of organic deposits is unknown, but 1.7% would be compatible with our observations of
local maxima in seven locations in 12 soil pits or Reid’s observations of C-rich sediment in 3 of 45 well samples
[Reid, 2007]. If our estimate of floodplain volume is too high by a factor of n, or if hot spots tend to occur
only in the upper 1/nth of that volume, then we may be underestimating the necessary density of hot spots
by that same factor n. Conversely, we may be overestimating the necessary hot spot volume because we
have assumed that hot spots have the C content of organic soil rather than woody debris, which is also buried
in the floodplain. Our current best estimate suggests that fully half of the whole-floodplain soil C pool would
be below 1 m if hot spots filled just 1.7% of the subsurface.
Empirical studies demonstrate that sand-and-gravel aquifers can host substantial denitrification due to their
low redox potentials, stable temperatures, and high residence times [Sjodin et al., 1997; Puckett and Cowdery,
2002]. At Nyack Floodplain, DEAs were typically below our detection limit in samples near the water table;
however, our observations of organic C-rich deposits and the strong relationship between soil C and
denitrification potential suggest that hot spots in denitrification could exist throughout the aquifer. To
evaluate their importance, suppose that subsurface denitrification occurs year round but only in sparsely
distributed hot spots with the same denitrification potential as the block 3 forest site at 170–180 cm
(1680 ng N h1 L1). Then if hot spots were present at a density of just 0.85% by volume, subsurface
denitrification potential (below 10 cm) would match the 8.8 kg N ha1 yr1 potential of the near surface
(upper 10 cm). Hot spot denitrification rates could be even higher at those times and places where a rising
water table meets a C-rich inclusion [Baker and Vervier, 2004; Harms and Grimm, 2008]. Actual denitrification
rates, as opposed to potentials, will also depend on the NO3 content of pore water [Groffman et al., 2009],
but given that subsurface water in this system is highly connected to river channel water, and therefore
continually replenished with NO3, we suggest that a substantial contribution of the subsurface to whole-
floodplain denitrification is plausible.
5. Conclusions
The dominant controls and predictors of soil properties change abruptly with depth in floodplain soils. Vegetation
cover predicts soil properties in the upper 10–50 cm of the soil profile. In the much thicker layer of deeper soil,
however, soil properties are better explained by soil texture, which acts both as a correlate of organic matter
deposits and as a direct control on moisture retention and hydraulic conductivity. Subsurface deposits of fine-
textured and organic-rich soils are embedded in a largely inactive soil matrix; even so, whole-floodplain rates of
soil C storage, denitrification, and other processes could be considerably augmented by these hot spots.
Although vegetation and soil texture predict soil properties at different depths and by different mechanisms,
these two predictors are themselves ultimately driven by the same force: floodwater movement. Large,
infrequent floods cause the scouring and successional resetting of floodplain plant communities. Floodwaters
also erode, sort, and deposit floodplain soils and thereby drive subsurface patterns in soil texture.
The timing and magnitude of flooding events are undergoing major shifts. Climate change is altering
catchment precipitation and snowmelt regimes, and humans are directly altering river hydrographs with
dams, water withdrawals, and land use change. Such changes offer ongoing opportunities to study the
relationship between water regimes and the biogeochemical mosaic of floodplain soils. They also offer
motivation for such studies, because floods and other water inputs are strongly tied to floodplain structure
and function as we have explored here. Understanding the many ways in which water movement
structures soil texture and biogeochemistry—by altering patterns of solute transport, water availability,
plant species composition, soil fertility, and organic matter burial—is essential for understanding both
fine-scale patterns and whole-ecosystem function, both now and under future water regimes, in these
highly heterogeneous ecosystems.
Acknowledgments References
Data from this paper are available from
Amoros, C., and G. Bornette (2002), Connectivity and biocomplexity in waterbodies of riverine floodplains, Freshwater Biol., 47(4), 761–776.
the first author by request. We thank
Baker, M. A., and P. Vervier (2004), Hydrological variability, organic matter supply and denitrification in the Garonne River ecosystem,
Meredith Wright, Ashley Helton, and
Freshwater Biol., 49(2), 181–190, doi:10.1046/j.1365-2426.2003.01175.x.
Hannah Jordt for their assistance in the
Bechtold, J., and R. Naiman (2009), A quantitative model of soil organic matter accumulation during floodplain primary succession,
field and lab. Kristen Olsen, Jim Craft,
Ecosystems, 12(8), 1352–1368, doi:10.1007/s10021-009-9294-9.
Diane Whited, Eric Anderson, Mark
Bechtold, S., and R. J. Naiman (2006), Soil texture and nitrogen mineralization potential across a riparian toposequence in a semi-arid
Potter, Ana-Maria Tenekedjieva,
savanna, Soil Biol. Biochem., 38(6), 1325–1333, doi:10.1016/j.soilbio.2005.09.028.
Fernando Bonassi, and Joseph Lucas
Blazejewski, G. A., M. H. Stolt, A. J. Gold, N. Gurwick, and P. M. Groffman (2009), Spatial distribution of carbon in the subsurface of riparian
provided valuable advice on field, lab,
zones, Soil Sci. Soc. Am. J., 73(5), 1733–1740, doi:10.2136/sssaj2007.0386.
and data analysis methods. The
Boulton, A. J., S. Findlay, P. Marmonier, E. H. Stanley, and H. M. Valett (1998), The functional significance of the hyporheic zone in streams and
Dalimata family generously allowed us
rivers, Annu. Rev. Ecol. Syst., 29, 59–81, doi:10.2307/221702.
to collect samples on their property,
Cierjacks, A., B. Kleinschmit, I. Kowarik, M. Graf, and F. Lang (2011), Organic matter distribution in floodplains can be predicted using spatial
and we especially thank Steve, Ben, and
and vegetation structure data, River Res. Appl., 27, 1048–1057, doi:10.1002/rra.1409.
Vince Dalimata for skillfully excavating
Corenblit, D., E. Tabacchi, J. Steiger, and A. M. Gurnell (2007), Reciprocal interactions and adjustments between fluvial landforms and vegetation
the soil pits. Ben Colman, Brian Lutz,
dynamics in river corridors: A review of complementary approaches, Earth Sci. Rev., 84(1–2), 56–86, doi:10.1016/j.earscirev.2007.05.004.
Eileen Thorsos, Rob Jackson, and others
Devito, K. J., D. Fitzgerald, A. R. Hill, and R. Aravena (2000), Nitrate dynamics in relation to lithology and hydrologic flow path in a river riparian
at Duke provided helpful comments on
zone, J. Environ. Qual., 29(4), 1075–1084.
the manuscript as did Klement Tockner,
Diehl, C. (2004), Hydrogeological characteristics and groundwater-river exchange in a gravel-dominated floodplain, Middle Fork of the
Stephen Carpenter, and several anon-
Flathead River, northwestern Montana, Masters thesis, Univ. of Montana, Missoula, Mont.
ymous reviewers. This project was
Fastie, C. L. (1995), Causes and ecosystem consequences of multiple pathways of primary succession at Glacier Bay, Alaska, Ecology, 76(6),
funded by the Gordon and Betty Moore
1899–1916, doi:10.2307/1940722.
Foundation under the auspices of the
Findlay, S., et al. (2011), Cross-stream comparison of substrate-specific denitrification potential, Biogeochemistry, 104(1), 381–392,
Salmonid Rivers Observatory Network
doi:10.1007/s10533-010-9512-8.
(SaRON). Alison Appling was supported
Fonda, R. W. (1974), Forest succession in relation to river terrace development in Olympic National Park, Washington, Ecology, 55(5), 927–942.
by the United States Environmental
Fox, J., and S. Weisberg (2011), An R Companion to Applied Regression, 2nd ed., Sage, Thousand Oaks, Calif.
Protection Agency (EPA) under the
Gallardo, A. (2003), Spatial variability of soil properties in a floodplain forest in northwest Spain, Ecosystems, 6(6), 564–576,
Science to Achieve Results (STAR)
doi:10.1007/s10021-003-0198-9.
Graduate Fellowship Program,
Gift, D. M., P. M. Groffman, S. S. Kaushal, and P. M. Mayer (2010), Denitrification potential, root biomass, and organic matter in degraded and
Agreement FP-91693501. EPA has not
restored urban riparian zones, Restor. Ecol., 18(1), 113–120, doi:10.1111/j.1526-100X.2008.00438.x.
officially endorsed this publication, and
Groffman, P. M. (1987), Nitrification and denitrification in soil—A comparison of enzyme assay, incubation and enumeration methods, Plant
the views expressed herein may not
Soil, 97(3), 445–450.
reflect the views of the EPA.
Groffman, P. M., K. Butterbach-Bahl, R. W. Fulweiler, A. J. Gold, J. L. Morse, E. K. Stander, C. Tague, C. Tonitto, and P. Vidon (2009), Challenges to
incorporating spatially and temporally explicit phenomena (hotspots and hot moments) in denitrification models, Biogeochemistry,
93(1–2), 49–77, doi:10.1007/s10533-008-9277-5.
Groffman, P. N., E. A. Holland, D. D. Myrold, G. P. Robertson, and X. Zou (1999), Denitrification, in Standard Soil Methods for Long-Term
Ecological Research, edited by G. P. Robertson et al., pp. 272–288, Oxford Univ. Press, Oxford.
Gurnell, A. M., G. E. Petts, D. M. Hannah, B. P. G. Smith, P. J. Edwards, J. Kollmann, J. V. Ward, and K. Tockner (2001), Riparian
vegetation and island formation along the gravel-bed Fiume Tagliamento, Italy, Earth Surf. Processes Landforms, 26(1), 31–62,
doi:10.1002/1096-9837(200101)26:1<31:aid-esp155>3.0.co;2-y.
Gurwick, N. P., P. M. Groffman, J. B. Yavitt, A. J. Gold, G. Blazejewski, and M. Stolt (2008), Microbially available carbon in buried riparian soils in
a glaciated landscape, Soil Biol. Biochem., 40(1), 85–96, doi:10.1016/j.soilbio.2007.07.007.
Harms, T. K., and N. B. Grimm (2008), Hot spots and hot moments of carbon and nitrogen dynamics in a semiarid riparian zone, J. Geophys.
Res., 113, G01020, doi:10.1029/2007JG000588.
Harms, T. K., and N. B. Grimm (2010), Influence of the hydrologic regime on resource availability in a semi-arid stream-riparian corridor,
Ecohydrology, 3(3), 349–359, doi:10.1002/eco.119.
Harrison, N. E. (2004), Gravity, radar and seismic investigations to help determine geologic, hydrologic, and biologic relations in the Nyack
Valley, northwestern Montana, M.S. thesis, The Univ. of Montana, Missoula, Mont.
Helton, A. M., G. C. Poole, R. A. Payn, C. Izurieta, and J. A. Stanford (2014), Relative influences of the river channel, floodplain surface, and alluvial
aquifer on simulated hydrologic residence time in a montane river floodplain, Geomorphology, 205, 17–26, doi:10.1016/j.geomorph.2012.01.004.
Hill, A. R., and M. Cardaci (2004), Denitrification and organic carbon availability in riparian wetland soils and subsurface sediments, Soil Sci.
Soc. Am. J., 68(1), 320–325.
Hill, A. R., P. G. F. Vidon, and J. Langat (2004), Denitrification potential in relation to lithology in five headwater riparian zones, J. Environ. Qual.,
33(3), 911–919.
Jobbágy, E. G., and R. B. Jackson (2000), The vertical distribution of soil organic carbon and its relation to climate and vegetation, Ecol. Appl.,
10(2), 423–436.
Jones, J. B., S. G. Fisher, and N. B. Grimm (1995), Vertical hydrologic exchange and ecosystem metabolism in a Sonoran Desert stream,
Ecology, 76(3), 942–952.
Komsta, L. (2011), Outliers: Tests for outliers, R package version 0.14. [Available at http://CRAN.R-project.org/package=outliers.]
Kukulak, J., A. Pazdur, and T. Kuc (2002), Radiocarbon dated wood debris in floodplain deposits of the San River in the Bieszczady Mountains,
Geochronometria, 21, 129–136.
Latterell, J. J., J. Scott Bechtold, T. C. O’Keefe, R. Van Pelt, and R. J. Naiman (2006), Dynamic patch mosaics and channel movement in an
unconfined river valley of the Olympic Mountains, Freshwater Biol., 51(3), 523–544, doi:10.1111/j.1365-2427.2006.01513.x.
Lowrance, R. (1992), Groundwater nitrate and denitrification in a coastal-plain riparian forest, J. Environ. Qual., 21(3), 401–405.
Lowrance, R., R. Todd, J. Fail, O. Hendrickson, R. Leonard, and L. Asmussen (1984), Riparian forests as nutrient filters in agricultural watersheds,
BioScience, 34(6), 374–377.
McClain, M. E., et al. (2003), Biogeochemical hot spots and hot moments at the interface of terrestrial and aquatic ecosystems, Ecosystems, 6,
301–312.
McIntyre, R. E. S., M. A. Adams, and P. F. Grierson (2009), Nitrogen mineralization potential in rewetted soils from a semi-arid stream landscape,
north-west Australia, J. Arid Environ., 73(1), 48–54, doi:10.1016/j.jaridenv.2008.09.011.
Merigliano, M. F. (2005), Cottonwood understory zonation and its relation to floodplain stratigraphy, Wetlands, 25(2), 356–374.
Naiman, R. J., and H. Decamps (1997), The ecology of interfaces: Riparian zones, Annu. Rev. Ecol. Syst., 28, 621–658.
Noe, G. B., C. R. Hupp, and N. B. Rybicki (2013), Hydrogeomorphology influences soil nitrogen and phosphorus mineralization in floodplain
wetlands, Ecosystems, 16(1), 75–94, doi:10.1007/s10021-012-9597-0.
Pinay, G., A. Fabre, P. Vervier, and F. Gazelle (1992), Control of C, N, P distribution in soils of riparian forests, Restor. Ecol., 6(3), 121–132,
doi:10.1007/BF00130025.
Pinay, G., C. Ruffinoni, and A. Fabre (1995), Nitrogen cycling in two riparian forest soils under different geomorphic conditions,
Biogeochemistry, 30(1), 9–29.
Pinay, G., V. J. Black, A. M. Planty-Tabacchi, B. Gumiero, and H. Decamps (2000), Geomorphic control of denitrification in large river floodplain
soils, Biogeochemistry, 50(2), 163–182.
Poole, G. C. (2006), Multiscale geomorphic drivers of groundwater flow paths: Subsurface hydrologic dynamics and hyporheic habitat
diversity, J. North Am. Benthol. Soc., 25(2), 288–303.
Poole, G. C., J. A. Stanford, C. A. Frissell, and S. W. Running (2002), Three-dimensional mapping of geomorphic controls on flood-plain
hydrology and connectivity from aerial photos, Geomorphology, 48(4), 329–347.
Puckett, L. J., and T. K. Cowdery (2002), Transport and fate of nitrate in a glacial outwash aquifer in relation to ground water age, land use
practices, and redox processes, J. Environ. Qual., 31(3), 782–796.
R Development Core Team (2010), R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria.
Reid, B. (2007), Energy flow in a floodplain aquifer ecosystem, PhD thesis, Univ. of Montana, Missoula, Mont.
Rhoades, C., D. Binkley, H. Oskarsson, and R. Stottlemyer (2008), Soil nitrogen accretion along a floodplain terrace chronosequence in
northwest Alaska: Influence of the nitrogen-fixing shrub Shepherdia canadensis, Ecoscience, 15(2), 223–230, doi:10.2980/15-2-3027.
Schwarz, W. L., G. P. Malanson, and F. H. Weirich (1996), Effect of landscape position on the sediment chemistry of abandoned-channel
wetlands, Landscape Ecol., 11(1), 27–38, doi:10.1007/BF02087111.
Sjodin, A. L., W. M. Lewis Jr., and J. F. Saunders III (1997), Denitrification as a component of the nitrogen budget for a large plains river,
Biogeochemistry, 39(3), 327–342, doi:10.2307/1469082.
Sobczak, W., L. Hedin, and M. Klug (1998), Relationships between bacterial productivity and organic carbon at a soil-stream interface,
Hydrobiologia, 386(1), 45–53, doi:10.1023/A:1003583813445.
Stanford, J. A., M. S. Lorang, and F. R. Hauer (2005), The shifting habitat mosaic of river ecosystems, Verh. Int. Ver. Theor. Angew. Limnol., 29,
123–136.
Tabacchi, E., D. L. Correll, R. Hauer, G. Pinay, A.-M. Planty-Tabacchi, and R. C. Wissmar (1998), Development, maintenance and role of riparian
vegetation in the river landscape, Freshwater Biol., 40(3), 497–516, doi:10.1046/j.1365-2427.1998.00381.x.
Thomas, G. W. (1996), Soil pH and soil acidity, in Methods of Soil Analysis, Part 3: Chemical Methods, vol. 5, edited by D. L. Sparks, pp. 475–490,
Soil Science Society of America, Madison, Wis.
U.S. Department of Agriculture, Soil Survey Staff of Natural Resources Conservation Service (2012), Web soil survey. [Available at
http://websoilsurvey.nrcs.usda.gov.] (Accessed 26 October 2013)
U.S. Geological Survey, National Water Information System (2001), Water data for the nation. [Available at http://waterdata.usgs.gov/mt/
nwis/uv?site_no=12358500.] (Accessed 15 August 2011)
Van Pelt, R., T. C. O’Keefe, J. J. Latterell, and R. J. Naiman (2006), Riparian forest stand development along the Queets River in Olympic National
Park, Washington, Ecol. Monogr., 76(2), 277–298, doi:10.1890/05-0753.
Walker, L. R. (1989), Soil nitrogen changes during primary succession on a floodplain in Alaska, USA, Arct. Alp. Res., 21(4), 341–349.
Ward, J. V., K. Tockner, and F. Schiemer (1999), Biodiversity of floodplain river ecosystems: Ecotones and connectivity1, Regul. Rivers Res.
Mgmt., 15(1–3), 125–139, doi:10.1002/(SICI)1099-1646(199901/06)15:1/3<125::AID-RRR523>3.0.CO;2-E.
Whited, D. C., M. S. Lorang, M. J. Harner, F. R. Hauer, J. S. Kimball, and J. A. Stanford (2007), Climate, hydrologic disturbance, and succession:
Drivers of floodplain pattern, Ecology, 88(4), 940–953.
Wigginton, J. D., B. G. Lockaby, and C. C. Trettin (2000), Soil organic matter formation and sequestration across a forested floodplain
chronosequence, Ecol. Eng., 15(Supplement 1), S141–S155.
Wilhelm, E., R. Battino, and R. J. Wilcock (1977), Low-pressure solubility of gases in liquid water, Chem. Rev., 77(2), 219–262, doi:10.1021/
cr60306a003.
Zehetner, F., G. J. Lair, and M. H. Gerzabek (2009), Rapid carbon accretion and organic matter pool stabilization in riverine floodplain soils,
Global Biogeochem. Cycles, 23, GB4004, doi:10.1029/2009GB003481.

Floodplain Biogeochemical Mosaics - A Multidimensional View of Alluvial Soils

Uploaded by

Copyright:

Available Formats

Floodplain Biogeochemical Mosaics - A Multidimensional View of Alluvial Soils

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Floodplain Biogeochemical Mosaics - A Multidimensional View of Alluvial Soils

Uploaded by

Copyright:

Available Formats

PUBLICATIONS

Journal of Geophysical Research: Biogeosciences

Received 28 OCT 2013

soil organic matter, and nutrient

2.2. Soil Pit Locations and Establishment

2.3. Sample Collection

2.4. Biogeochemical Analyses

2.5. Statistical Analyses

DEA Moisture Fine roots POM Rock volume

3.2. Denitriﬁcation Potentials

3.3. Ordination of Biogeochemical Properties

3.4. Spatiotemporal Patterns in Pore Water Chemistry

and [DOC] was signiﬁcantly higher at

3.5. Floodplain-Scale C Storage and Denitriﬁcation

where we monitored soil moisture and

4.3. Deep Soils May Matter

You might also like