Boreal forests are subject to a wide range of temporally and
spatially variable environmental conditions driven by season, climate, and
disturbances such as forest harvesting and climate change. We captured
dissolved organic carbon (DOC) from surface organic (O) horizons in a boreal
forest hillslope using passive pan lysimeters in order to identify controls
and hot moments of DOC mobilization from this key C source. We specifically
addressed (1) how DOC fluxes from O horizons vary on a weekly to seasonal
basis in forest and paired harvested plots and (2) how soil temperature,
soil moisture, and water input relate to DOC flux trends in these plots over
time. The total annual DOC flux from O horizons contain contributions from
both vertical and lateral flow and was 30 % greater in the harvested plots
than in the forest plots (54 g C m
The relationship between water input to soil and DOC fluxes was seasonally dependent in both plot types. In summer, a water limitation on DOC flux existed where weekly periods of no flux alternated with periods of large fluxes at high DOC concentrations. This suggests that DOC fluxes were water-limited and that increased water fluxes over this period result in proportional increases in DOC fluxes. In contrast, a flushing of DOC from O horizons (observed as decreasing DOC concentrations) occurred during increasing water input and decreasing soil temperature in autumn, prior to snowpack development. Soils of both plot types remained snow-covered all winter, which protected soils from frost and limited percolation. The largest water input and soil water fluxes occurred during spring snowmelt but did not result in the largest fluxes of DOC, suggesting a production limitation on DOC fluxes over both the wet autumn and snowmelt periods. While future increases in annual precipitation could lead to increased DOC fluxes, the magnitude of this response will be dependent on the type and intra-annual distribution of this increased precipitation.
Boreal forests occupy 11 % of the total land surface and thus span a
variety of topographies and climate zones (Bonan and Shugart,
1989). They contain organic-matter-rich soils that store approximately
19 % of the global soil organic carbon (SOC) pool
(Pan et al., 2011). Losses of SOC from land occur
predominately through decomposition and mobilization as
The importance of upland forest SOC as a source of DOC to boreal forest
surface waters is variable among boreal regions due to differences in
connectivity driven by topography and precipitation
(McGlynn and McDonnell, 2003). In low-relief
catchments, SOC mobilized as DOC from upland forest soil may be lost as
The upper organic (O) horizons of podzols are key sources of soil DOC
(Mcdowell and Wood, 1985). The large range in values of O horizon DOC fluxes
reported from field studies in temperate and boreal forest systems (3–122 g C m
Black spruce dominates North American boreal forests
(van Cleve et al., 1983; Bona et al., 2016), and
these forests span a wide range of environmental conditions that drive
variations in SOC decomposition
(Wickland et al., 2007) and SOC
persistence across sites (Schmidt et al., 2011). Forest harvesting increases
water yield (Neary, 2016) and reduces C in the organic layers due to
reductions in litter fall and increases in soil respiration (James and
Harrison, 2016), but the extent of the impact on soil properties and
biogeochemical cycling is dependent on many interacting site-specific
variables (Kreutzweiser et al., 2008). Furthermore, while lysimeter studies
conducted in post-harvested forests found immediate increases in DOC fluxes
from O horizons (Kalbitz et al., 2004; Piirainen et al., 2002), the longer-term effects of harvesting on DOC mobilization have not been considered. We
exploited spatially (plot type) and temporally (weekly to seasonal) variable
environmental conditions in a maritime boreal black spruce hillslope site to
investigate the processes controlling DOC fluxes from O horizons. The region
receives moderately high annual precipitation (
This study was conducted in an experimental harvest site within a mature
black spruce forest at the Pynn's Brook Experimental Watershed Area (PBEWA)
located 50 km from Deer Lake, western Newfoundland and Labrador, Canada.
(48
Pynn's Brook Experimental Forest experimental design. A north-facing black spruce hillslope site divided into six
Passive pan lysimeters were installed at the interface between the O and mineral horizon. Each lysimeter footprint was 0.3 m by 0.4 m and collected water percolating through the O horizon, including both vertical and lateral flow (Fig. 1b, c), with a maximum solution collection capacity of 25 L. The lysimeters were designed using reported recommendations for achieving accurate volumetric measurements of soil leachate (Radulovich and Sollins, 1987; Titus et al., 1999). It was desirable for this study that (1) the collection pan directs leachate immediately into a deeper storage container, avoiding potential issues of sample evaporation from the collection pan; and (2) the buried storage reservoir is placed away from the collection pan so that soil and snowpack directly above and upslope from collection area are not disturbed during sample collection.
Installation of lysimeters began in July 2012 and was completed the
following spring in May 2013. Four lysimeters were installed in three plots
of each plot type for a total of 12 forest lysimeters and 12 harvested
lysimeters. The slope measured at each lysimeter was 5 %–12 % and 7 %–13 % in the forest and harvested plots, respectively. Collection began on
12 July 2013 from forest and harvested lysimeters. Synchronized sampling
from lysimeters of both plot types was carried out every 7 to 15 d from
July to January, once between January and April, and every 7 to 15 d from
April to July. Lysimeter samples were stored at 4
Lysimeter collection efficiency testing was completed on three forest lysimeters
and three harvested lysimeters following the study period. The soil on top of
and around the lysimeter catchment area was first saturated, and then the
area directly above each lysimeter was watered uniformly with 10 L of water
and the volume of solution collected by the lysimeters was retrieved. This
was repeated three times on each of the lysimeters to determine the efficiency
of the lysimeter system in collecting the leachate from the footprint of
organic soil directly above the installed pan. Lysimeter efficiency was
found to be
A tipping bucket rain gauge (RST Instruments model TR-525) was installed in
an open area at PBEWA to monitor local precipitation and air temperature.
Data from the local tipping bucket were compared with regional precipitation
reported by Environment Canada at the Deer Lake Airport (49
Snowmelt water input was estimated using changes in snow depth between each
lysimeter collection day measured near each lysimeter in both the forest and
harvested plots. The average snow depth change by plot type was multiplied
by an estimated maritime snow density of 0.343 g cm
A snow pit was analyzed for each plot type on 2 April 2014 just prior to
the onset of snowmelt. A series of 15 cm long snow cores were collected
beginning from the top of the snowpack down to the forest floor to obtain a
sample of the entire snowpack per plot type. The cores were melted,
pooled by plot type, and the DOC concentration of the pooled samples was measured to provide a mean DOC concentration in
the snow of forest and harvested plots. The snow depth of each plot,
combined with the estimated snow density (0.343 g cm
Ecosystem and soil properties of black spruce forest and adjacent
harvested plots. Values are means of 12 litterfall traps per plot type, 16
soil respiration collars per plot type, three organic (O) horizon soil
temperature and moisture probes per plot type, two mineral horizon soil
temperature and moisture probes per plot type, nine O horizon samples per plot
type used to determine thickness, % C, C stock, C : N and bulk density, one snow pit per plot type, and three seasonally distinct rain collections used
together with annual rainfall to estimate an annual C input, with standard
error in parenthesis. Results for one-way ANOVAs (litterfall, O horizon
thickness, soil % C, C stock, C : N, and soil bulk density) and
n/a: not applicable.
Throughfall was collected on an event basis using 10 buckets (0.36 m
The O horizon soil was sampled specifically for this study by taking three
Bulk density of O horizon and mineral soils was calculated using the volume and dried mass of the soil sample.
Additionally, two sets of O horizon samples were obtained for physical measurement of O horizon unsaturated and saturated hydraulic properties and water infiltration rates. Cores (5 cm diameter) were collected in triplicate at two locations in forest and harvested plots (six cores per plot type), and live moss was removed prior to analysis using a HYPROP system. The HYPROP measurements of water content and soil water tension during continuous evaporation were analyzed to obtain relationships of soil water tension and hydraulic conductivity to water content (Schindler and Muller, 2006; Schindler, 2010). A second set of cores (10 cm diameter) were collected at six locations in two forest plots for falling head infiltration (INF) analysis. These cores included the entire organic (L, F, and H) horizon and moss. Following a first round of infiltration rate measures a subset of cores were partially excluded to expose the entire H horizon, which was carefully removed before remeasuring infiltration. Forest and harvested plots had H layers with similar bulk densities, but H layers constituted much of the O horizon in harvested plots where moss cover was limited and the L and F layers were reduced in comparison to forest plots. Matrix and macropore saturation was determined for each these cores (Table 4).
Litterfall was collected using four 0.34 m
Three soil temperature and moisture probes per plot type (Decagon
Temporal variation of soil respiration
Measurements of soil respiration were made at biweekly intervals for the
snow-free growing seasons (May–November) in 2013–2015. Four collars
consisting of a 7 cm section of 10 cm inside diameter PVC pipe were inserted
into the ground 8 months prior to the start of measurement in four forest
plots and four harvested plots. Soil respiration rate and soil temperature
were measured every 2 weeks using a LI-6400-09 soil chamber and a
penetration soil temperature probe, both attached to an LI-6400 portable
All statistical analyses were performed using RStudio version 1.0.136.
Pearson correlations between lysimeter-captured dissolved organic
carbon concentrations (mg DOC L
Two-way ANOVA results examining the effect of water input, season, and their interaction on DOC fluxes. Data plotted in Fig. 4.
Correlation testing was used to assess the relationships among data from lysimeter collections (DOC flux, water flux, and DOC concentration) and mean soil temperature, mean soil moisture, and daily water input (Table 2) across 30 collection days. Multiple regressions were not used due to the multi-collinearity of many of the predictor variables, which affected the estimated regression parameters. Individual correlations, however, were assessed to evaluate the strength of relationships among variables within the dataset.
A linear mixed effects model was used to examine the effects of plot type, sample year (2013–2015), and their interaction on soil respiration. The interaction term was further analyzed with a post hoc least square means test. Linear interpolation was used to calculate cumulative soil respiration for the snow-free growing season during the period of 2013–2015. A multiple linear regression was used to explain the dependence of soil respiration on soil temperature, moisture, and the soil temperature by soil moisture interaction.
Soil bulk density was not different between the forest and harvested plots
for either O or mineral soil horizons (Table 1). However, O horizon depth
was almost twice as great in the forest plots compared with the harvested
plots (means of 8.17 and 4.26 cm, respectively; Table 1). This resulted in
an estimated 78 % greater O horizon SOC stock in forest plots relative to
harvested plots (2390 and 1340 g C m
Average soil hydraulic parameters of organic horizons. Data were
obtained from the HYPROP (HP) evaporation apparatus for unsaturated conditions
and falling head infiltration (INF) tests for matrix-saturated and
totally saturated (macropore infiltration) conditions. Both tests were made
on intact cores and standard deviations are provided in parentheses (
The temporal range in instantaneous
There was no overall significant difference in soil respiration between plot types for the 2013–2015 growing season estimates; however, there was a significant plot type by sample year interaction effect on soil respiration (Table S3). The multiple comparisons found that soil respiration in the harvested plot was lower relative to that in the forest plot for 2014 and 2015 growing seasons but not 2013 (Tables S4 and S5). Soil respiration was positively related with soil temperature but negatively related with soil moisture content, and the presence of a soil temperature by soil a moisture interactive effect on soil respiration in the regression analysis indicated the effects of soil temperature on soil respiration had been modified by soil moisture (Table S6).
The local mean annual air temperature over the July 2013–July 2014 study
period was
The O horizons in the harvested plots were generally warmer and thinner than
those in the forest plots (Table 1, Fig. 2b; forest plot range:
1.1 to 16
The O and mineral horizons were consistently wetter in harvested plots
relative to the forest plots over the duration of the study (Fig. 2c), but
given the high variability and few measurement replicates (
The mean annual volume-weighted DOC concentration collected by lysimeters
was 29.4 and 26.1 mg C L
The mean DOC concentration in the snowpack, measured immediately prior to
snowmelt on 2 April 2014, was 7.5 and 3.3 mg C L
The mean annual O horizon water flux was 2040 L m
Longer periods of soil drying and low rainfall, occurring predominately during summer, corresponded with periods of little to no water flux and, consequently, little to no DOC flux in both harvested and forest plots (Fig. 2b, d, e; shaded areas). In contrast, periods of relatively high moisture and consistent rainfall, occurring predominately in autumn, corresponded with high and consistent water and DOC fluxes. During spring snowmelt, however, when the DOC concentration was relatively low, the largest water fluxes did not result in the largest fluxes of DOC (Fig. 2; 8 April to 1 May 2014). The highest DOC flux over the study period was observed in early autumn when a large rain event followed a warm period of soil drying. Soil water fluxes were negatively correlated with soil temperature (Table 2a), and there was a strong positive correlation between water input and both soil water and DOC fluxes measured in both plot types (Table 2c). There was an interaction between season and water input on DOC fluxes (Table 3), where a linear relationship between water input and DOC fluxes was observed in the summer (Fig. 4a), but DOC fluxes exhibited a tapering off in autumn and snowmelt when water input to soil was high (Fig. 4b, c).
Mean annual lysimeter-collected variables. Volume-weighted
dissolved organic carbon (DOC) concentration
Seasonal relationship between dissolved organic carbon (DOC)
fluxes and water input to the soil in mature forest (F) and harvested (H)
plots. Seasons are designated as summer
This study revealed a 30 % greater annual mobilization of DOC from O
horizons in 10-year-old harvested plots compared with forest plots. This was
despite lower O horizon SOC stocks and C inputs from aboveground litter in
harvested plots (Table 1). Annually, the larger flux of DOC in the harvested
plots correlated to a larger annual input of water to the soil surface,
larger fluxes of water through thinner O horizons, and warmer mean annual
soil temperature. On weekly to monthly timescales, both forest and
harvested O horizon DOC flux patterns mirrored those of water fluxes, while
the contribution of DOC concentration variations to observed temporal
differences was less evident in DOC flux patterns (Fig. 2d, e, f). This is
additionally described in both plot types by a strong positive relationship
between water input to the forest floor (as rainfall, throughfall, and/or
snowmelt) and DOC flux, but with no relationship between DOC flux and soil
temperature (Table 2). Therefore, across both forest and harvested
landscapes characterized by different surface soil and ecosystem properties,
water input to soil is a dominant control over O horizon DOC mobilization
dynamics on varying time and spatial scales. Increases in DOC fluxes from O
horizons immediately following and up to 5 years after boreal forest
harvesting were previously documented by lysimeter studies (Piirainen et
al., 2002; Kalbitz et al., 2004). However, to our knowledge this is the
first study to demonstrate a longer-lasting (10-year) harvesting effect on
DOC fluxes. Harvesting results in sites becoming
To establish water input as a main driver of regional O horizon DOC flux variability, regional C budget models should be parameterized to reflect the spatial heterogeneity in mean annual precipitation (MAP) that exists across the boreal zone. This is supported by our results, as well as prior correlations between MAP and annual DOC fluxes across ecosystems (Michalzik et al., 2001), and is especially relevant given the large range in MAP that exists across boreal ecoregions (for example, Canada's boreal ecoregions 173–1492 mm; A National Ecological Framework for Canada, 1999). Furthermore, studies examining controls on DOC content in soils at depth focus on delivery of DOC from O to mineral horizons and the subsequent mineral–OM interactions that control soil C sequestration (Clarke et al., 2007; Fröberg et al., 2011; Kalbitz et al., 2004; Rosenqvist et al., 2010). Associated conceptual models assume vertical fluxes of water and DOC (e.g. Kaiser and Kalbitz, 2012). Vertically dominated O to mineral horizon DOC fluxes may occur in some boreal systems, and they may be relevant at larger spatial scales in low-relief landscapes. In our moss-mantled hillslopes, however, event-specific lateral flow was likely important in over half of the measurements made as water collected by lysimeters located at the base of the O horizon exceeded total precipitation or snowmelt over the lysimeter footprint on 17 of 30 collection dates in the forest plots and on 18 of 30 in the harvested plots. Although passive lysimeters do potentially disrupt natural soil hydrological conditions, the soil hydraulic properties of the O horizons (Table 4), combined with continuous field measurements of O horizon soil moisture, indicate that these lysimeters captured a combination of vertical and lateral flow during many precipitation events. Water fluxes measured exceeded the total precipitation or snowmelt over the lysimeter footprint only when matric saturation of the O horizon had been reached and macropore flow was initiated (Fig. 5). At soil moisture contents above matric saturation, capillary forces are ineffective and water flows uninhibited through the macropores of O horizons, flowing downslope at the base of the O horizon due to the lower hydraulic conductivity of the underlying mineral horizons. This phenomena likely drove the pipe throughflow observed at the O–mineral-horizon interface in a boreal forest hillslope during snowmelt, resulting in the delivery of highly acidic surface soil water to lakes (Roberge and Plamondon, 1987). Lateral transport of water and solutes as facilitated by macropore flow is recognized as a potentially important feature controlling landscape transport of solutes in forest hillslope and stream catchment studies (Kaiser and Guggenberger, 2005; van Verseveld et al., 2008; Terajima and Moriizumi, 2013; Laine-Kaulio et al., 2014). While modelling of water and solute transport continues to evolve and incorporate macropore flow (Beven and German, 1982, 2013), models are limited to modelling of mineral soil and do not explicitly define porous O horizons that are typically an important source of DOC in boreal forest landscapes. We advise that direct measurement and incorporation of the specific hydrologic role of O horizons is essential because they represent both a hydrologically unique layer and a hotspot for DOC mobilization. This will improve estimates of DOC mobilization and redistribution dynamics at the landscape scale.
Lysimeter-captured water fluxes versus water input over the
lysimeter footprint in harvested
Despite the control of water input rate on DOC fluxes, the relationship between DOC flux and water flux varied at the seasonal scale (Fig. 4; Table 3). Soils of both plot types appeared to be flushed of DOC during periods of high, continual leaching and low temperatures (Fig. 2), suggesting that the seasonally variable production of DOC and/or water-soluble organic carbon (WSOC) is an important secondary control. Some field studies have shown that soil DOC concentrations remain constant and do not become more dilute with increasing soil water fluxes, suggesting that the pool of WSOC is not easily exhausted in those systems (Kalbitz et al., 2007; Klotzbücher et al., 2014). This leads to proportional increases in DOC flux with increasing water flux and, therefore, a water limitation on DOC mobilization. While summer (Fig. 4a), and likely winter, DOC fluxes in this study were similarly water-limited, autumn and spring snowmelt fluxes exhibited a tapering off of DOC fluxes during periods of highest water input (Fig. 4b, c), suggesting a production limitation during autumn and snowmelt.
DOC flux was calculated as the product of DOC concentration and solution volume for each measurement period; therefore, the highest periods of DOC flux occur when conditions support relatively high values of both terms. This occurred most frequently during late summer/early autumn and ecologically requires the combination of (1) the production of water-soluble organic carbon or DOC via temperature-sensitive mechanisms such as soil organic matter (SOM) and/or litter decomposition rhizodeposition, and microbial biomass turnover (Christ and David, 1996; Kalbitz et al., 2007; Weintraub et al., 2007); and (2) sufficient water inputs to result in a soil water flux that mobilizes or extracts DOC from O horizons. Soil water fluxes were negatively correlated with soil temperature in this study (Table 3a), likely driven by the seasonal temperature dependence of net water input and evapotranspiration, while DOC concentration was positively correlated with soil temperature. Therefore, the seasonality of DOC flux involves an interactive temperature effect, where temperature-dependent biogeochemical processes and temperature-dependent soil water fluxes interact to form seasonally unique combinations or scenarios important to a predictive understanding of these fluxes.
Fluxes of water and DOC were dynamic on the weekly to monthly scale during
all seasons except winter (Fig. 2e, f), revealing that flux conditions can
occur at all times of the year in these sites, except during periods of
deep, consistent snowpack, which limits water input to the soil and,
consequently, DOC mobilization. Summer also exhibited a water limitation on
DOC mobilization but on a shorter timescale, alternating between weekly
periods of no water and DOC flux and periods of large water and DOC fluxes.
While we detected no relationship between DOC flux and soil moisture using
the whole dataset (Table 3b), antecedent soil moisture can affect the
proportion of the water input that results in a water and DOC flux in the
summer when soil drying–rewetting cycles were common (Fig. 2; grey shaded
bars), although this does not appear to be a driving factor throughout the
year in these plots. In summer, when
With continuous leaching and decreasing soil temperatures, late autumn water inputs resulted in a decrease in DOC concentrations and DOC fluxes, such that soils appear to be flushed of the WSOC or DOC pool just prior to snowpack development. Thus, the availability of the extractable DOC pool in these soils during the snowpack and subsequent snowmelt period was likely much reduced by high autumn water input at low soil temperatures. Spring snowmelt captured during this study year followed a winter of constant snow cover and contributed approximately 31 % of the annual water input to the soil, and 20 % of the annual DOC flux, but occurred over a period that represented only 13 % of the year. Despite representing the largest hydrological event during this study year, the large water flux over a short time period, combined with relatively low soil temperatures and previously flushed soils, resulted in dilute leachate (low DOC concentration) and a smaller contribution to the annual DOC flux in relation to early autumn fluxes.
This study shows that DOC flux variation is well described by water flux variation but that gradual flushing of O horizons occurs during consistent leaching events throughout autumn as soil temperatures decrease. These seasonal trends suggest that the projected increases in precipitation at mid- to high latitudes in the Northern Hemisphere (Kirtman et al., 2013) can result in proportional increases in DOC fluxes in the summer and early autumn when soil temperatures are warm but that DOC or water-soluble organic carbon pools are depleted during seasonal decreases in soil temperature. In order for increasing water fluxes to result in increased losses of DOC, they must therefore be met with increased production of DOC/WSOC – a process dependent on how increases in precipitation are seasonally distributed. Two potential mechanisms of increased WSOC production that are linked to reductions in snowpack are the increased occurrence of winter rainfall and soil frost. No soil freezing occurred under the consistently deep snowpack conditions observed during winter in this study. With warm winter conditions expected to become more frequent in northern regions, melting and reforming of the snowpack over winter will have consequences for soil exposure and frost, as well as the frequency and magnitude of wintertime water flux events. Similar to soil drying–rewetting events (Fierer and Schimel, 2002), soil freeze–thaw cycles have been shown to increase soil DOC concentrations by disturbing soil, root, and microbial structures (Haei et al., 2013; Schimel and Clein, 1996). Increased winter rainfall and midwinter snowmelt events that drive larger winter soil water fluxes, in combination with soil freeze–thaw events that increase production of WSOC, can therefore contribute to future increases in wintertime mobilization of DOC. Changing snowpack dynamics is therefore one possible mechanism of increasing river DOC export trends in northern temperate watersheds that are specifically attributed to increases in wintertime DOC exports (Huntington et al., 2016). These results suggest that the effect of climate change on boreal forest DOC fluxes will depend on the redistribution of seasonal precipitation and changes to precipitation form. In addition, this study highlights that defining macropore-driven lateral water flow dynamics, particularly at the O-to-mineral-horizon interface, can help define the role of DOC at the landscape scale.
All data are included in the paper tables and the Supplement.
The supplement related to this article is available online at:
KAE and SEZ designed the study with input from KLB. KLB and KAE designed the lysimeters and planned their installation as well as the installation of all environmental monitoring equipment. KLB collected and analyzed the lysimeter, environmental monitoring, and soil properties data. XZ contributed the soil respiration data and analysis. KP contributed soil hydrology data and interpretations. KB prepared the paper, with editing from SEZ and KAE and further contributions on final drafts from XZ and KP.
The authors declare that they have no conflict of interest.
Special thanks for field assistance provided by individuals at the Atlantic Forestry Centre (Corner Brook) of Natural Resources Canada: Andrea Skinner, Darrell Harris, and Gordon Butt; and Memorial University, Grenfell campus: Sarah Thompson and Danny Pink, as well for laboratory assistance provided by Jamie Warren at Memorial University, St. John's campus. The throughfall carbon inputs were estimated based on collections made and analyzed by Alex Newman in 2015.
This research has been supported by the Centre for Forest Science and Innovation (Agrifoods and Forestry, Government of Newfoundland and Labrador), the Natural Sciences and Engineering Research Council (NSERC) Strategic Partnerships Grants (grant no. 479224-15), and the Canada Research Chairs Program.
This paper was edited by Frank Hagedorn and reviewed by L. Thieme and two anonymous referees.