Hydrological disturbances could increase dissolved organic carbon (DOC)
exports through changes in runoff and leaching, which reduces the potential
carbon sink function of peatlands. The objective of this study was to assess
the impact of hydrological restoration on hydrological processes and DOC
dynamics in a rehabilitated
In peatlands, as in many terrestrial ecosystems, DOC dynamics are controlled on the one hand by its production-to-consumption ratio in pore water and on the other hand by lateral water fluxes that drive its exports. DOC production through organic matter decomposition is known to increase with temperature (Clark et al., 2009; Freeman et al., 2001) and DOC consumption, mainly due to heterotrophic bacterial activity, also positively correlates to temperature and can lead to decreased DOC concentrations during droughts (Clark et al., 2009; Pastor et al., 2003). The export of the DOC produced in pore water is mainly controlled by peatland hydrology (Pastor et al., 2003; Strack et al., 2008), especially by the partitioning between quick near-surface flow and groundwater flow (Birkel et al., 2014). Due to the complexity of the interactions between these factors, field studies can show contradictory results regarding the effect of rewetting on DOC dynamics, with some studies reporting increasing concentrations (Hribljan et al., 2014; Strack et al., 2015) while others report decreasing concentrations (Höll et al., 2009; Wallage et al., 2006).
While changes in DOC net production resulting from WTD drawdown can be assessed through field monitoring, the relative contributions of DOC production and consumption cannot be evaluated (Strack et al., 2008). Process-based biogeochemical models can be relevant tools for understanding DOC dynamics (Evans et al., 2005) and can help identify factors controlling its production and consumption in such environments. In particular, conceptual models are appropriate because they are parsimonious in terms of their number of parameters, avoiding over-parameterization issues (Birkel et al., 2017; Seibert et al., 2009). Nevertheless, these parameters have to be adjusted to every condition through calibration and validation phases when a more physical model would require only adjusting boundary conditions. In addition, conceptual models are valid for a specific range of input data and should not be used for prediction where conditions lie out of their validation range. However, another advantage of using conceptual models is that they usually require commonly measured data (e.g. precipitation and water discharge or water level) so they can be applied to numerous study sites where such data are available, making them a suitable tool for comparing sites with different settings.
When studying DOC dynamics in peatlands, existing conceptual models are composed of a DOC module combined with a hydrological model (Birkel et al., 2014; Futter et al., 2007; Lessels et al., 2015). In these studies, the hydrological model is usually adapted to the catchment and calibrated on stream discharge. However, stream discharge in peatlands is difficult to monitor because the diffuse runoff that occurs in these flat areas can result in multiple outlets. Furthermore, while WTD is a key parameter to explain DOC dynamics (Strack et al., 2008), it is usually not considered for calibration, and water discharge is preferred instead. Therefore, while these models have proven to be well-adapted when modelling a catchment containing a peatland area (Birkel et al., 2014; Futter et al., 2007; Lessels et al., 2015), where the outlet is well-defined, they are more difficult to apply when considering a lone peatland. In this case, the model should focus on the simulation of the WTD, especially when studying DOC dynamics in peatland pore water. Furthermore, a model based on WTD can also provide interesting information about the spatial variability of the dominant hydrological processes when applied to different locations within the same peatland. Models simulating DOC dynamics are usually based on a simple mass balance and DOC production and consumption rates, usually expressed as first-order rate processes (Birkel et al., 2014; Futter et al., 2007; Lessels et al., 2015). In these cases, DOC production and consumption rates are modified using terms related to temperature and soil moisture as these two parameters control the microbial activity and peat decomposition that regulate the production and consumption of DOC in peat water.
In this study, we propose the coupling of existing WTD-dependent hydrological
model specifically developed for simulating peatland hydrology (Binet et al., 2013) with a biogeochemical
module simulating DOC production and consumption as first-order rate
processes. The hydrological model was calibrated on WTD, which is an
important driver of the DOC dynamics in peatlands. The model was applied to
two sites of a
The La Guette peatland (150 m a.s.l., 47
Location and settings of the study area. Locations of control and rewetted monitoring are indicated.
WTD and DOC concentrations (DOC) in pore-water were monitored in two
locations in the peatland. One is affected by the restoration work and is
called “rewetted” while the other is not affected and is called “control” (Fig. 1). The WTDs were recorded in piezometers since February 2014 at a 15 min time
step using vented-pressure probes (Orpheus mini, OTT Hydromet). Pore-water
was sampled in four wells surrounding each piezometer (each of them less than
5 m from the piezometer) during 13 campaigns that took place every 1 to 4 months between February 2014 and December 2017. The pipes were emptied
before sampling to avoid the presence of rain water and ensure that the
water sampled was representative of the peatland water. The water samples
were filtered using 0.45
Pore-water dissolved organic matter (DOM) was characterized by its
fluorescence properties through three-dimensional excitation emission
matrices (EEMs; Fellman et al., 2010) acquired with F-2500 and F-7000
spectrofluorometers (Hitachi). EEMs were recorded using a
Meteorological data were recorded at an hourly time step from a station located within the peatland between the two studied areas (Fig. 1). Rainfall was measured with a tipping bucket rain gauge and potential evapotranspiration (PET) computed with the FAO Penman–Monteith equation at an hourly time step (Allen et al., 1998) using local solar radiation, wind speed, relative humidity and temperature measurements.
The effect of hydrological conditions (dry period from 1 June to
30 November and wet period from 1 December to 31 May) and location (rewetted or control) on DOC and DOM composition were
tested using two-way ANOVA and Tukey's post-hoc tests were used to identify the
significant differences (
The modelling approach used in this study combines a conceptual hydrological model with a biogeochemical model simulating DOC dynamics. The hydrological model is based on a conceptual water-table-dependent hydrological model that has already been successfully applied in the study area (Binet et al., 2013). This model is coupled with a module based on functions describing DOC production and consumption in pore water that was developed for this study. The model is described in detail in the following subsections.
The hydrological model is based on the model described by Binet et al. (2013). It is a daily time step, reservoir model specifically developed for peatland hydrology, which integrates a WTD-dependent runoff. Compared to the original model, a few modifications were made in this study in order to improve the model. The overall structure of the new model is presented in Fig. 2.
Structure of the hydrological model, composed of the three
reservoirs of surface (Sr), macroporosity (Sm) and retention (Se). The different fluxes
are indicated in italics:
The relation between soil-water content and WTD was improved. In the original
version the user had to know the relation between WTD and soil-water content.
Now the model automatically computes the soil-water content based on the
porosity of the percolation reservoir (
With this modification, the maximum amount of water stored in the Se
reservoir (Semax in mm), which was a calibrated parameter in the original
version of the model, is now automatically computed with
A third reservoir was added, Sr (overland flow storage), in order to
differentiate the overland flow water (Sr) from the water entering the peat
macroporosity (Sm), which were not differentiated in the original model.
While it might not significantly affect the hydrological model, this was
done to prepare for the addition of the biogeochemical processes that are
different for these two reservoirs. Following the addition of the Sr
reservoir, a maximum amount of water contained in the Sm reservoir is
defined (Smmax in mm) and is computed according to
The routing was also slightly modified to take into account the addition of
the new reservoir (Sr). Water from precipitation first fills the Sm
reservoir, and the Sr reservoir starts to fill only when Sm is full
(Sm
Finally a discharge coefficient was added to compute the flow from the new
Sr reservoir, represented by
This flux is added to the total discharge which is now computed according to
Given the structure of the model,
Concerning evapotranspiration, the crop coefficient used to compute
evapotranspiration (ET) from ETP was separated into the dormant (Kc
List of the parameters used in the hydrological and the DOC models. The hydrological flux associated to each parameter is in parenthesis. Calibrated parameters and boundary condition independent (BCI) parameters are indicated.
The computation of the following processes remained unchanged, with infiltration
from Sm to Se (ISe), percolation (
The modified hydrological model is now controlled by nine parameters (Table 1).
Three input parameters describe the peat structure (
To simulate DOC dynamics, a module was developed based on first-order
production and loss and mass balance, similar to what can be found in the
literature (Birkel et al., 2014; Lessels et
al., 2015). Production and loss are computed in the Se and Sm reservoirs,
only since the main biogeochemical processes linked to DOC dynamics occur in
soil storage and no reaction takes place in the Sr reservoir. DOC production
was based on a production coefficient and two additional modifiers based on
soil-water content and air temperature, as usually considered in DOC
production models (Birkel et al., 2014; Futter et al., 2007; Lessels et al., 2015). The effect of the
temperature was based on a Q
DOC loss, corresponding to mineralization and sorption, was based on a loss
coefficient linked to air temperature in the same way as DOC production. DOC
loss is computed according to
Finally, the mass balance of DOC is computed in the Sm and Se reservoirs at
the daily time step
The DOC model is controlled by four parameters (Table 1). Two input parameters
(SOC and DOC
The hydrological and biogeochemical model parameters were calibrated for
each piezometer in the peatland for the wettest period (1 April 2014 to
1 April 2015) and the driest period (1 October 2016 to 15 December 2017)
considering the available data. The model was validated over a period with a more
intermediate condition (1 April 2015 to 1 April 2016). The period from
1 May 2016 to 30 September 2016 was not simulated because exceptionally heavy
rainfall (return period of about 50 years) occurred on 31 May 2016, causing
extensive flooding in the whole region. The definition of the model is not
suitable for these exceptional events because the water coming from the
river during the flood is not taken into account in the model. However, the
flood was not expected to impact DOC in the peat profile since it was
already saturated with rain water when the flood of the river reached the
peatland. In addition, it has to be noted that the model is able to
represent less exceptional events as long as the flood does not reach the
peatland (estimated at a 10 to 20-year return period in our case).
Water and DOC balance computed for the simulated period
(1 April 2014 to 1 April 2016 and 1 October 2016 to 10 December 2017) in rewetted and control areas.
Calibrated parameters and efficiency criteria for the different periods of calibration and validation. Ranges of parameters used for autocalibration are also indicated.
Simulated and observed pore-water DOC in control and rewetted sites. Observations are the average of four samples for each sampling date. Error bars indicate standard deviation.
The parameters were calibrated with a Nelder–Mead algorithm (Varadhan et al., 2016), implemented in the R software
(R Core Team, 2012) using the Nash–Sutcliffe coefficient on the
water-table depth (NS, Nash and Sutcliffe, 1970) as the
objective function for the hydrological module and the root-mean-square
error (RMSE) for the DOC concentrations in Sm. NS was chosen for the
hydrological model because it can take the large variation of the water
table into account while RMSE was chosen for the DOC model because DOC
variations are not very large and the RMSE provides a quantitative estimate of the
error. In addition, the coefficient of determination multiplied by the slope
of the regression (Br2; Krause et al., 2005) was
computed for both the hydrological and DOC model to better assess the
quality of the simulations. The hydrological model was calibrated following
a multi-site strategy. The parameters independent of the location within the
peatland were kept similar for both sites (Kc
The mean annual precipitation (
The average of the DOC measurements was
The PARAFAC analysis revealed three main components characterizing the DOM
(Fig. 4b). According to the review by Fellman et al. (2010), the first component (ex 360, em 466) can be described
as having a high molecular weight and being humic, and it is referred to here under its original
name as C. The second component (ex 330, em 407) can be described as having a
low molecular weight and is referred to here as
The best simulated and the observed WTD dynamics are shown in Fig. 3. NS and
Br2 were greater than 0.10 and 0.24 for calibration periods and reached
values greater than 0.10 and 0.39 for validation periods, respectively. The
RMSE ranged between 1 and 9 cm and no drop in the model performance was
observed for the validation period, compared to the calibration ones (Table 3). The model performed better during the wettest year in the control area and
better during the intermediate and the driest years in the rewetted area. The
important point is that the model was able to reproduce two different WTD
dynamics using the same input data (i.e. rainfall and PET). These
differences are explained by the modification in calibrated parameter
values. As the evapotranspiration coefficient and maximum infiltration rates
were the same for each site, the differences are driven by the discharge
coefficients. The values of the three discharge coefficients (
Sensitivity rank of the parameters of the hydrological model.
Simulated DOC exports for control and rewetted sites.
Simulated and observed pore-water DOCs are shown in Fig. 5. The simulations
presented a RMSE
Calibrated parameters and efficiency of the DOC model.
In this study, observed water-table dynamics were used to better understand
the dominant hydrological processes taking place in the two locations of a
restored peatland (rewetted and control) by calibrating a conceptual model.
Though simple (six calibrated parameters), the model was able to reproduce the
specific water-table dynamics in each location of the studied area using
the same input data (precipitation and potential evapotranspiration). This
difference in observed water-table dynamics (17 cm of difference for the
maximum water-table drawdown) is reflected in the calibrated parameter
values for each location (Table 3). In addition, and in order to better
assess the dominant processes, a sensitivity analysis of the model was
performed for each location (Table 4). The results indicate that the most
sensitive parameters are Kc
A module simulating DOC production and loss was added to the hydrological
model in order to better understand DOC dynamics in the two peatland
locations, with RMSE between 1.6 and 10.8 mg L
Long-term studies have reported decreasing pore-water DOC more than 10 years after a restoration operation took place (Höll
et al., 2009; Wallage et al., 2006), while others observed increasing DOC
after restoration (Hribljan et
al., 2014; Strack et al., 2015). Glatzel et al. (2003)
observed an increase in pore-water DOC following a drain blocking
operation but predicted a decrease in DOC over time due to a depletion of
easily decomposable organic matter in the peat. In this study, the results
indicate that during the three years following a restoration operation,
DOCs were higher in the rewetted site than in the control location during the
dry period (from 1 June to 30 November), while they were
similar during the wet period. In addition, the difference in DOC dynamics
are also reflected in DOM quality inferred from its fluorescence properties,
with a greater increase in low molecular weight compounds (component
The main difference in DOC is observed during the dry period, when the water-table dynamics are different between the two locations. This would confirm that hydrology, especially the magnitude of the water-table drawdown, might be a major factor controlling DOC dynamics in the peatland. Indeed, the higher WTD in the dry period in the rewetted site is related to a higher DOC than in the control site where the WTD is lower. A larger proportion of low aromatic DOC is also observed during the same period in the rewetted site than in the control site. Therefore, we propose to explain the differences in DOC by the difference in water-table drawdown in the dry period. When the water-table drawdown is small (high water table), more DOC is produced from the top peat layer containing more recent and easily biodegradable organic matter than when the water-table drawdown is more severe (low water table). In addition, anaerobic conditions in the rewetted site would lead to the less efficient decomposition of organic matter, increasing the production of water-soluble intermediate metabolites (Kalbitz et al., 2000; Strack et al., 2008). An increase in DOC in the rewetted location can also be explained by an increase in the photic zone, potentially supporting the algae photosynthate production that enhances the DOC release into the water column, as suggested by Hribljan et al. (2014). However, the latter hypothesis is the least probable in our case, since no ponding water is observed in summer in the study area. The ability of the model to reproduce pore-water DOC dynamics can be attributed to its consideration of the water-table drawdown, which is expressed in the model through the use of soil moisture (based on water level in the Sm and Se reservoirs) as a production rate modifier. Finally, while this study focuses on the hydrological control on DOC dynamics, it is important to note that other factors not directly integrated in the model are also known to affect DOC exports such as the pH and redox state (e.g. Grybos et al., 2009; Knorr, 2013).
The model enables DOC exports to be estimated for each location (Table 2).
The results (values) are in the range reported in the literature (from 4.2
to 18.9 g-
The model developed in this study follows a parsimonious coupled hydrology-biogeochemistry model philosophy (Birkel et al., 2014, 2017; Lessels et al., 2015). By keeping parameterization to a minimum, it was able to identify factors controlling WTD and DOC dynamics in the two contrasting sites of the studied peatland with a relatively low requirement of input data (precipitation, potential evapotranspiration and temperature). Contrary to similar models, here hydrology is calibrated on WTD instead of on stream discharge. This way, the model proves to be a relevant tool to be applied in flat areas where catchment delineation is highly uncertain and outlets difficult to monitor. It is also useful to explore the hydrology of areas located within the same peatland by performing a multi-site calibration. However, it is necessary to perform an uncertainty analysis to better assess confidence in the computed fluxes when no data are available. The careful application of the model highlights the impact of hydrological restoration on hydrology and DOC dynamics that would have been difficult to study with models calibrated on stream discharge and are applicable at the catchment scale only. In addition, the DOC model developed in this study has shown good results in modelling pore-water DOC dynamics, meaning that the two-calibrated parameter model is adapted to simulate DOC dynamics in peatland ecosystems. Therefore, if applied to several WTD time series, it could provide spatial information by identifying the main areas of DOC production within a peatland. This model could also be applied to longer time series and different study sites to assess the effect of hydrological restoration over longer periods, and the dominant controlling factors in peatlands with different settings.
A conceptual hydrological model, developed especially for peatland and
calibrated on WTD, has been combined with a simple DOC production–loss model
and applied to two locations of a peatland, one which was affected by
hydrological restoration. The application of this model has shown the
following:
The hydrological restoration was found to impact water balance by
increasing fast superficial drainage, compared to slow deep drainage. The intensity of the maximum water-table drawdown was found to be the main
factor in controlling pore-water DOC dynamics in the peatland. A higher DOC in the rewetted location was linked to differences in DOM
composition. Simulated DOC exports were within the same order of magnitude for rewetted and
control locations in a short-term period (3 years). Water partitioning between fast superficial drainage and slow deep drainage
controls DOC sources as well as the temporal dynamics of DOC exports.
These results suggest that hydrological restoration does not affect short term DOC fluxes in peatland. In addition, this study has shown that the proposed conceptual hydrological and biogeochemical model can provide relevant information about water balance and the factors controlling element cycling processes in peatlands. The application of a WTD-based model is a relevant alternative to a discharge calibrated catchment model when the outlet is not easily identifiable or when seeking spatial information within peatlands.
Data and model code are available on request from the authors.
FLD, SG and SB designed the study site restoration and monitoring. LBJ, SG, FLD, FL and LP helped with instrumentation and data collection. CD, NJ and RZ helped with fluorescence analysis and data interpretation. LBJ and SB developed the model. LBJ performed simulations and data analysis. LBJ prepared the draft of the manuscript. FL, SG, CD, NJ, FLD and SB helped improve the final manuscript.
The authors declare that they have no conflict of interest.
This paper is a contribution of the Labex VOLTAIRE (ANR-10-LABX-100- 01) and of the PIVOTS project (ARD 2020 of the Centre Val de Loire region, CPER and FEDER). This study was undertaken in the framework of the Service National d'Observation Tourbières (French Peatland Observatory), accredited by the INSU/CNRS. The authors would like to thank Audrey Guirimand-Dufour and Franck Le Moing for their help in fluorescence analysis, Nathalie Lottier for the DOC analysis, Elizabeth Rowley-Jolivet for the revision of the English version and the reviewers for their helpful comments and suggestions. Edited by: Insa Neuweiler Reviewed by: Michel Bechtold and one anonymous referee