Peatlands store
Aquatic carbon transfer from terrestrial ecosystems to inland waters is receiving increasing attention as it plays a major role in the watershed carbon balance (Webb et al., 2018) and in the global carbon cycle (Cole et al., 2007; Drake et al., 2017). The origin of aquatic carbon has been tracked and wetlands have been shown to be the main organic carbon suppliers to rivers at both local (Hope et al., 1997; Laudon et al., 2004; Ledesma et al., 2017) and continental scales (Hope et al., 1994; Spencer et al., 2013). Peatlands are specific wetlands which have accumulated organic matter through slow vegetation decomposition processes (Joosten and Clarke, 2002; Limpens et al., 2008). Peatlands grow under different climates (Broder et al., 2012; Dargie et al., 2017; Gorham, 1991; Page et al., 2011) and store between 20 % and 30 % of the total global soil carbon stock (Leifeld and Menichetti, 2018; Nichols and Peteet, 2019; Scharlemann et al., 2014). Stream outlets of peatlands have been monitored at different latitudes (Billett et al., 2006; Leach et al., 2016; Moore et al., 2013) in order to quantify and understand the aquatic carbon transfer between these organic-carbon-rich pools and their draining streams. Dissolved organic carbon (DOC) is a key component of these fluxes as it contributes to more than 80 % of the aquatic carbon exported from peatlands (Dinsmore et al., 2010; Hope et al., 2001; Müller et al., 2015; Roulet et al., 2007). At the outlet of peatlands, DOC is not only considered for its role in the carbon balance but also because it may be an issue for water treatment quality (Ritson, 2015) and a conveyor of potentially harmful elements along inland waters (Broder and Biester, 2017; Rothwell et al., 2007; Tipping et al., 2003).
Variability in the DOC concentration signals at the outlet of peatlands has been observed at the interannual (Fenner and Freeman, 2011; Köhler et al., 2008), the seasonal (Leach et al., 2016; Tipping et al., 2010) and even the event scales (Austnes et al., 2010; Dyson et al., 2011). DOC concentrations were found to be negatively correlated with discharge in boreal systems (Köhler et al., 2008), positively correlated with discharge in temperate areas (Clark et al., 2007) or noncorrelated with discharge in mountainous areas (Rosset et al., 2019). Temperature was also reported as an important driver of seasonal variations of DOC concentration in field (Billett et al., 2006) and mesocosm (Pastor et al., 2003) experiments since DOC production is boosted by a greater vegetation and microbial activity during warmer periods. Higher temperatures were also shown to enhance evapotranspiration from peatland resulting in a rise in DOC concentration in peat porewater and stream waters during dry summer periods (Fraser et al., 2001). Studies have highlighted that the heterogeneity of the hydraulic conductivity within peatlands (Rycroft et al., 1975) influences the water table level fluctuations (Bernard-Jannin et al., 2018; Kalbitz et al., 2002; Strack et al., 2008) and the oxygenation of the acrotelm (Freeman et al., 2001), thus driving DOC production and its transfer to streams.
DOC concentration monitoring at the outlet of peatlands has generally
consisted in a weekly or monthly stream water sampling routine (Clark et
al., 2008; Juutinen et al., 2013). Higher-frequency sampling has been
restricted to specific high-precipitation events (Austnes, 2010; Clark et
al., 2007) or snowmelt (Laudon et al., 2004). Recently, new optical in situ sensors (Rode et al., 2016) were used to track DOC concentration at a high-frequency rate (
Mountains host many small peatland areas that are often neglected in global peatland assessments but which drastically influence stream chemistry in headwater catchments (Broder and Biester, 2015; Rosset et al., 2019). The harsh mountainous climatic conditions (from the montane to the alpine belt, Holdridge et al., 1967) and the relief of those areas generate high gradients of different abiotic parameters (temperature, precipitation, hydrology) evolving along both seasonal and event (snowmelt, rainstorms) scales. In the present study, a bog and a fen in the French Pyrenees mountains were monitored for stream DOC concentration using an optical high-frequency in situ sensor placed at their outlet. The scientific objectives of this study were (1) to statistically identify the main abiotic parameters driving stream DOC concentration variability at each site, (2) to identify the temporal scale of these drivers and (3) to compare the DOC concentration patterns of two contrasted peatlands regarding their hydrological functioning.
The peatland of Bernadouze (Fig. 1b) is situated in the eastern part of the
French Pyrenean mountains (42
The peatland of Ech (Fig. 1c) culminates at 710 m a.s.l. in the west-central part of the French Pyrenees (43
This article presents high-frequency data monitored from 1 September 2015 to 31 December 2018 at the Bernadouze site and from 22 May 2017 to 19 February 2019 at the Ech site. Precipitation (liquid and solid) and air temperature were recorded every 30 min at Bernadouze (Gascoin and Fanise, 2018) and every 60 min at Ech by automatic weather stations located respectively 300 and 15 m from the peatlands in open areas. At both sites, sensor failures prevented data
acquisition, and gap-filling models were used to complete the datasets. For
missing precipitation data in Bernadouze (27 % of the monitored timeline), a linear model (
At the outlet of each peatland, a multiparameter probe (Ysi EXO2, USA)
measured fluorescence of the dissolved organic matter (fDOM,
Water grab sampling was performed every 2 weeks at the outlet of
Bernadouze peatland and every two months at the outlet of Ech. Piezometer
wells were used to sample peat porewater on four occasions (2013, 2014,
2015, 2018) in Bernadouze and on two occasions (2017, 2019) in Ech during
stream baseflow periods. Water grab samples were collected using a manual peristaltic
pump and were directly filtered on-site using 0.22
For all samples (grab and flood samples), nonpurgeable organic carbon
(NPOC, referred to hereafter as DOC) concentration was analyzed in filtered
samples after acidification to pH 2 with a TOC-5000A analyzer (Shimadzu, Japan). The quantification limit was 1 mg L
The fluorescence of DOM (fDOM) data was explored for potential adjustments
for temperature, inner filter effect and turbidity (Downing et al., 2012; de
Oliveira et al., 2018; Watras et al., 2011). fDOM data were corrected for
temperature as described by de Oliveira et al. (2018). The inner filter
effect was adjusted at Ech for data showing absorbance values at 254 nm
higher than 0.6 (de Oliveira et al., 2018). Lastly, fDOM data recorded
during high-turbidity events (
In order to provide an overall characterization of the peatlands, a mean peat water table depth, as well as a mean water temperature, was calculated at each site by averaging peat water table depths and water temperature data at a given time from the set of piezometer probes. Calculations were performed only when all sensors were running (94 % of the time period in Bernadouze and 100 % in Ech). Hereafter, the mean water temperature in the piezometers is assimilated to peat water temperature.
Master recession curve (MRC) analyses were performed on water table and
stream level time series, using the MRCTools v3.1 software (Posavec et al., 2017). In order to characterize the hydrodynamic properties of the peat, MRC were preferred to hydraulic conductivity estimations from slug tests because they can be performed directly with the water table level datasets and repeated easily on other peatlands. The MRC represents the average recession of the water level observed when only discharge flow occurs (no recharge). An exponential master recession curve was used to adjust the observed average MRC and to define a specific recession coefficient (
Peak selections in the DOC concentration timeline were performed by running
Python 3.6 (Python Software Foundation, 2019) scripts using the function
find_peak available in the SciPy signal library (Jones et al., 2001) and the arithmetic mean of the DOC concentration signal (DOC_mean) as an input parameter. The peak selection criteria were as follows: to reach DOC_mean concentration and have a prominence higher than 0.25 times DOC_mean. Peaks occurring during an interval shorter than 12 h apart were grouped under the highest DOC concentration peak. Each DOC concentration peak was defined by the time
period delimited by the two nearest low points surrounding the peak event.
Low points were located on the DOC concentration timelines by applying the
find_peak function on the negatively transformed (
Characterization of DOC concentration peaks. Peak events are identified on the DOC concentration timeline in blue. Each DOC concentration peak event is defined by an initial concentration (green points) and a maximum one (red points). DOC concentration increase is calculated by subtracting the initial from the maximum concentration. The time between 2 maximum DOC concentrations corresponds to the duration (seconds) separating two events and is used as an explanatory variable. The DOC concentration baseline (orange dotted line) corresponds to the time series defined by all the initial values of each DOC concentration peak.
In order to investigate DOC concentration variabilities (at two temporal scales: peak event and seasonal), nine explanatory variables were chosen (Table 1). Variables were calculated for each DOC concentration peak event using similar metrics to those previously described in the DOC peak characterization section (Fig. 2).
Targeted and explanatory variable description.
The variables were abiotic parameters, chosen because they have been reported in the literature to have an explanatory potential for stream DOC concentration variability (Table 1). Two categories of variables were distinguished depending on whether the process they described was related to the production of DOC within peatlands or to the transfer of DOC from peatlands to streams. After sensitivity tests and in accordance with the observations of Tunaley et al. (2018), a mean of 7 d prior to the event was defined as the best operator to characterize pre-event conditions of air and stream water temperatures.
Relationships between targeted variables (DOC_increase and DOC_initial) and the explanatory variables were investigated
using ordinary least squares (OLS) multiple regression analyses. Prior to
the analyses, variables which did not satisfy a normal distribution were log
or square root transformed to improve normality (Table 1). Multicollinearity
was assessed among all the predictors using Pearson correlation with a
threshold
Climatic variables are contrasted between the two studied areas. In 2018,
temperatures were higher in Ech than in Bernadouze with an annual mean air
temperature, water temperature and peat water temperature respectively of
11.3, 10.7 and 11.9
Precipitation and air temperature
DOC concentration was highly variable at both sites during the monitored
periods as highlighted by the numerous short DOC peak events (
Time series and DOC concentration peak metrics in Bernadouze over the 1 September 2015 to 31 December 2018 period and in Ech over the 22 May 2015 to 13 February 2019 period. Mean notations correspond to arithmetic means which are given with standard deviations.
Peak characterization (Table 2) revealed that the increases and the maxima
of DOC concentration peaks were on average 2 times higher in Ech than in
Bernadouze. However, the ratio between the mean increase and the mean
initial value of DOC concentration was higher in Bernadouze (2.3) compared
to Ech (1.9). DOC concentration peaks occurred more often at Bernadouze
compared to Ech (0.24 vs. 0.16 peak per day on average), while their duration
was slightly longer (
General mean and seasonal means of initial DOC concentrations were 2.5 and 3.1 times higher at Ech compared to Bernadouze (Table 3). However, at both sites, initial DOC concentrations showed a clear seasonal variability. The lowest values were observed in spring and the highest in autumn, while in summer and winter DOC concentration was close to the annual mean. DOC peak event frequencies also varied at the seasonal scale (Table 3). The highest frequencies were reported in autumn at both sites. The lowest peak frequencies were observed in winter at Bernadouze and in summer at Ech.
Reduced models explaining DOC concentration during peak events
(DOC_initial and DOC_increase) at the outlet of Bernadouze and Ech peatlands. Reduced models were obtained after a backward stepwise selection procedure applied on the full model (see details in Sect. 3.6). Adjusted
Prior to multiple regression analyses, the air temperature over 7 d, the
maximum stream water level and the initial level of the water table were
excluded from the analysis because of their strong correlation with other
variables (Pearson's correlation
Relationships between
In the fen of Bernadouze the recession times in the peat ranged from 15 to
77 d, whereas in the bog of Ech they were longer, ranging from 53 to 143 d (Fig. 5). Stream recession times were shorter at both sites, reaching 4 d in Bernadouze and 9 d in Ech. Results of the OLS regressions conducted at each peat water level monitoring plot using DOC increase final models revealed that recession time influenced the model's efficiency (Fig. 5a). Piezometers characterized by shorter recession times showed greater determination coefficients
Relationship between water recession time coefficients and
To our knowledge, this is the first time that stream DOC concentration and abiotic drivers, including peat water table depth fluctuations, have been analyzed at the outlet of peatland sites on a multiyear period at this frequency (30 min). Previously, DOC concentration variability was investigated either at lower frequencies (Clark, 2005; Dawson et al., 2011) or during shorter periods (Austnes et al., 2010; Koehler et al., 2009; Tunaley et al., 2016; Worrall et al., 2002). Recently, high-frequency monitoring of nutrient dynamics in watersheds has developed and has revealed an unexpected variability of mobilization processes for these nutrients (Blaen et al., 2017; Rode et al., 2016; Tunaley et al., 2016). These acquisitions have allowed scientists to characterize the hot moments in the biogeochemical cycles of a watershed (McClain et al., 2003). A contribution of our study is to sequence extremely brief DOC concentration peaks and to statistically disentangle their event and seasonal drivers using synchronous high-frequency monitoring of climatic and hydrological parameters. The representativeness of both seasonal- and event-scale statistical models is enhanced by the large number of events (252 peaks in Bernadouze and 101 peaks in Ech) captured at all seasons (Table 2).
Clear seasonal variations in the DOC concentration baseline were observed at both sites (Fig. 3 and Table 2). The DOC concentration baseline increased in late spring, peaked in autumn, decreased during winter and reached the lowest levels in early spring. Similar seasonal DOC concentration patterns have been observed at the outlet of other peatland sites in temperate regions (Austnes, 2010; Broder and Biester, 2015; Clark et al., 2005; Tunaley et al., 2016; Worrall et al., 2006; Zheng et al., 2018) or after the snowmelt event in boreal areas (Jager et al., 2009; Köhler et al., 2008; Laudon et al., 2004; Olefeldt and Roulet, 2012; Whitfield et al., 2010).
In this study, linear regression models revealed that the seasonal variations of the DOC concentration baseline are mostly driven by peat water temperature (Table 3). At peatland sites, temperature is often identified as a DOC concentration driver at the seasonal scale (Billett et al., 2006; Clark et al., 2008; Dawson et al., 2011; Koehler et al., 2009). Warmer temperatures directly enhance DOC production by stimulating vegetation and microbial activity (Kalbitz et al., 2000; Pastor et al., 2003). Warmer temperatures are also indirectly linked to DOC production processes in temperate and northern peatlands since they often correspond to dry periods that lower water table levels. When the water table decreases, the enzymic latch (Freeman et al., 2001) is initiated on a greater volume of acrotelm (oxygenated peat) and enhances DOC production within the upper peat layers. DOC concentration relationships with peat water temperature have already been described in an acidic fen in France (Leroy et al., 2017) and in blanket peatlands from the North Pennine uplands in the UK (Clark et al., 2005); however, in these cases DOC concentrations were measured in peat porewater. A complementary study in the North Pennines (Clark et al., 2008) showed that peat porewater DOC concentrations and stream DOC concentration were strongly correlated, meaning that, by extension, the relationship between peat temperature and stream DOC concentration could be verified for these sites.
This study, coupling high-frequency stream DOC concentration and water table
depth monitoring at both peatland sites, revealed that peat water table
increase is a strong predictor of stream DOC concentration increase at the
event scale (Table 3 and Fig. 4b). Until now, stream DOC concentration
variability at the event scale has been investigated in terms of discharge
but rarely in terms of peat water table variation. Several studies have
reported stream DOC concentration increases at the outlet of peatlands
during flood events (Austnes, 2010; Ryder et al., 2014; Tranvik and Jansson,
2002; Yang et al., 2015), whereas others showed dilution during high flow
events (Clark et al., 2007; Grayson and Holden, 2012; Laudon et al., 2004;
Worrall et al., 2002). At the outlet of peatlands, nonlinear discharge–DOC-concentration relationships have been reported (Roulet et al., 2007; Tunaley
et al., 2016) and modeled (Birkel et al., 2017); this seems to be the case at our sites where stream water level explains the variability of DOC increases during flood events only poorly (Bernadouze) or not at all (Ech) (
The link between DOC dynamics and peat water table has been largely investigated at the seasonal scale (Kalbitz et al., 2002; Strack et al., 2008; Hribljan et al., 2014) or in mesocosm experiments (Pastor et al., 2003; Blodau et al., 2004). The peat water table is usually considered a DOC production driver as it controls the oxygenated acrotelm volume (Billett et al., 2006; Freeman et al., 2001; Ritson et al., 2017). Therefore, different studies attempted to quantify the effect of water table position on DOC production rate in peatlands. In fen and bog mesocosms, Pastor et al. (2003) observed no DOC concentration variation in the stream water after long-term peat water table decreases. Contrastingly, increasing DOC concentrations were observed during the rewetting phase of the acrotelm at fen sites in Germany (Kalbitz et al., 2002), in Canada (Strack et al., 2008) and in the USA (Hribljan et al., 2014). Clark et al. (2009) reported similar observations after rewetting peat cores in controlled laboratory conditions. Our results are in line with these studies. Moreover, thanks to the high-frequency survey, they highlight, in addition to DOC production processes, specific hydrodynamic processes driving DOC export from peatlands at the event scale.
The correlation between DOC peak and water table increase can have different origins. First, the water table increase could create a piston flow that expels prefloodwater (Małoszewski et al., 1983), enriched in DOC. At our sites, the delay (a few hours) between stream discharge peaks, peat water table increase and DOC peaks suggests that DOC concentration peaks are not directly related to water pressure. As previously observed in peat-dominated headwater catchments by Rodgers et al. (2005), this observation rejects the piston flow hypothesis. Secondly, as DOC is mostly produced in the oxygenated and unsaturated peat volume above the water table (Billett et al., 2006; Freeman et al., 2001; Ritson et al., 2017), it can be flushed by floodwater during a flood event (Boyer et al., 1997). Due to the exponential decrease in hydraulic conductivity properties (Rycroft et al., 1975), prefloodwater under the water table is less mobile than floodwater located above (Quinton et al., 2008). Our data support this second hypothesis, with a very fast increase in DOC concentrations and a rapid DOC concentration recession in the same order of magnitude as subsurface flow recession. This does not exclude the possibility that a fraction of prefloodwaters may reach the stream during the recession time, but this mixing process is minor compared to flushing processes. These mechanisms are in line with the two-layer hydrology–biogeochemistry model developed by Birkel et al. (2017) at the outlet of a peatland. Moreover, this model emphasizes the positive relationship between the stream DOC concentration and the water table connection to the upper soil horizon. Following this second hypothesis, if DOC production was the limiting factor, the linear regression (Fig. 4b) should show a plateau for the high value of water table increase. This is not the case. Thus the limiting factor appears to be the amount of water brought by floods and ultimately the full saturation of the peat. These observations support the practices for degraded peatland restoration, where a general rise in the water table is recommended to limit water table increases and the DOC concentration peaks induced at their outlets (Höll et al., 2009; Strack and Zuback, 2013).
The higher DOC concentration observed in summer could be explained by
evapotranspiration processes that concentrate solutes in stream water.
However, the evapotranspiration rates in these mountainous environments are
low (
In Bernadouze, DOC concentration remained extremely low when the fen was snow-covered, and it did not drop drastically during the spring snowmelt as has been observed in boreal areas (Laudon et al., 2004; Leach et al., 2016). This pattern can be explained by (1) the low initial DOC concentration, which prevents a clear dilution being observed during the snowmelt event, and (2) the snowmelt regime in this Pyrenean catchment, which may be less sudden than in boreal regions and occurs from the early snow deposit to the beginning of the growing season, continuously diluting the low winter DOC production within the peatland.
In Bernadouze, contrary to the initial hypothesis (Table 1), the time between peaks was a negative significant predictor in both seasonal and event DOC concentration models (Table 3). This is considered to be an indirect consequence of the seasonal temperature control on DOC concentration. Indeed, snow cover and the low temperatures associated with high water table positions prevent the occurrence of DOC peaks in winter, creating large time gaps between two events (Table 2) of low initial values. In contrast, DOC production is amplified during warmer periods, resulting in more frequent stream DOC concentration peaks starting at higher initial values. In Ech, where average annual temperatures are higher and snow cover is reduced, the initial hypothesis was verified since DOC concentrations were stronger in autumn after the long summer times between peaks (Table 2). However, the variable was not significant enough to be integrated in any final model (Table 3).
Spatial analysis of water table variation within the peatland revealed that the studied sites are composed of several peat units, characterized by contrasted recession times. In these mountainous peatlands, recession times are related to DOC dynamics, driving model efficiency between DOC concentration increase and peat water table rise and explaining DOC concentration in peat porewater.
In the present study, both stream and peat porewater DOC concentrations were higher at Ech compared to Bernadouze (Tables 2 and S3 and Fig. 5b). This is consistent with mesocosm (Pastor et al., 2003) or field (Chanton et al., 2008; Chasar et al., 2000; Moore, 1988) peat porewater observations which highlighted higher concentrations in bogs compared to fens.
Sphagnum species, which are dominant on bogs, usually produce relatively less labile and reactive DOC than vascular plants, which are more abundant on fens (Chanton et al., 2008). Lower pH values are expected to reduce DOC solubility (Clark et al., 2005). However, these relationships are not observed at our sites. As can be seen on Fig. 5b, peat porewater DOC concentrations are related to MRC, with higher concentrations being associated with longer recession times. Beyond this hydrologic control, other parameters, such as residence time and vegetation cover, linked to bog and fen conditions, influence DOC concentration levels in peat porewater.
On average the bog of Ech presented a longer recession time (111 d) than the fen of Bernadouze (20 d). However, a large variability is observed within each site. For instance, a specific unit in the fen of Bernadouze was characterized by a long recession time of 77 d. This unit shows surface bog vegetation and topographic patterns but is surrounded by typical fen units characterized by shorter recession times (Fig. 5). Thus a peatland complex must be considered a patchwork of different units and not as a uniform peat entity.
At the event scale, the univariate model between DOC concentration and peat water table increase showed a non-negligible intercept at Ech contrasting with the model of Bernadouze (Fig. 4b). This means that, in Ech, DOC concentration increases can occur without water table increases. In this case, DOC is transferred from the upper peat layers via fast runoff flows without any water table level fluctuation. Such a phenomenon is consistent with the lower hydraulic conductivities (longer recession times) measured in bogs (Fig. 5). In contrast, DOC stored in the upper peat layers of fen units is transferred to the stream by fast percolating water raising water table levels and supplying subsurface flows (Fig. 6). This explains why the DOC increase model based on peat water table increase is particularly efficient for fen units characterized by short recession times (Fig. 5a). Recession times, used as proxies of the hydraulic conductivity, also explain the differences in peat porewater DOC concentration observed between bog and fen sites. In the fen, recession times are short, meaning that the upper peat layers are rapidly washed by precipitations, inducing sudden DOC pool depletions of the peat porewater (Fig. 3c). At the bog site, DOC stored in the upper peat layers is slowly released to the stream after precipitation events and contributes to maintaining a high stream baseline (Fig. 3c) and peat porewater DOC concentrations (Fig. 5a).
Schematic overview of a peatland complex. Size of the arrows corresponds to DOC quantity mobilized from distinct peatland units. The DOC concentration observed in the stream depends on the contribution of the different peat units within the peatland complex.
Thus, stream DOC concentration modeling at the outlet of peatlands must account for different proportions of fen-like or bog-like units in peatland complexes to reflect the real seasonal and event DOC concentration variability. Every unit supplies DOC to the stream at a different rate depending on its volume, distance from the stream and recession time (Fig. 6). This end-member mix concurs with the model of Binet et al. (2013) describing event and seasonal water table variability in peatlands using a double porosity parametrization. In that sense, recession time appears as a physical parameter able to characterize peatland units beyond the binary typology of bog or fen. This would surely improve the efficiency of hydrological and biogeochemical models. In the case of peatland complexes characterized by long recession times, further investigations of peatland runoffs and subsurface flows are needed, analyzing denser and stream directed piezometer transects in order to build stronger DOC concentration models.
This study reports a statistical analysis of the stream DOC concentration variability at the outlet of two mountainous peatlands. Multiyear in situ high-frequency (30 min) monitoring revealed that at both sites DOC concentration time series can be decomposed into a seasonal baseline interrupted by many short, intense peaks of higher concentrations. At the seasonal scale, DOC concentration baseline variations are mainly explained by peat water temperature, which controls integrative DOC production processes within the peatland. During the hot moments of peak events, DOC concentrations are well explained at both sites by water table increases within the peatlands. Recession time is a relevant parameter to explain peat porewater DOC concentration and the different model performances observed between bog and fen sites. Recession time assessments in different locations on the two studied sites showed that peatlands are composed of different units presenting contrasted hydraulic conductivities. Thus, peatlands should not be considered to be uniform landscapes. Distinct peatland units within the same peatland complex contribute differently to the DOC transfer processes to inland waters. Recession time assessment in piezometers appears to be a simple and promising tool to investigate hydrological processes occurring in peatlands over time and space. Indeed, water table time series are often underused and only account for a seasonal mean or minimum depth. Assessing recession times on peatlands is a first step to taking peatland water table dynamics into consideration and to explaining potentially related biogeochemical processes.
The data used in this paper are described and available on the Pangaea data repository (
The supplement related to this article is available online at:
TR, SB and LG designed the study. SB and LG obtained the funds. TR, JMA, SB and LG contributed to the site instrumentation and data acquisition. TR and ELR supervised and compiled the data. TR, FR, SB and LG analyzed the data and discussed the results. TR, SB and LG designed the study. SB and LG obtained the funds. TR, JMA, SB and LG contributed to the site instrumentation and data acquisition. TR and ELR supervised and compiled the data. TR, FR, SB and LG analyzed the data and discussed the results. TR wrote the first draft of the paper, which was complemented by significant contributions from all authors.
The authors declare that they have no conflict of interest.
The authors wish to thank Lionel Plagnet for permitting the access to the peatland of Ech; Virginie Payré-Suc, Frédéric Julien, Didier Lambrigot and Wendy Amblas for assisting in stream organic carbon concentration analysis; François De Vleeschouwer, Deonie Allen, Pilar Durantez Jimenez and Thierry Camboulive for assisting in water sampling; Guilhem Susong and the Regional Natural Reserve of Pibeste-Aoulhet for piezometer maintenance; Simon Gascoin, Pascal Fanise, the CESBIO laboratory and the OSR Toulouse for providing the meteorological data; Didier Galop for assisting in site preparation and communication with local policy makers; and Elisabeth Rowley-Jolivet for English language assistance on an earlier version of the manuscript.
This project was cofunded by the LabEx DRIIHM French program “Investissements d'Avenir” (grant no. ANR-11-LABX-0010), which is managed by the ANR and funds the PhD of Thomas Rosset; and LabEx DRIIHM OHM Haut Vicdessos/Haute Vallée des Gaves, INTERREG V POCTEFA REPLIM (project no. EFA056/15) PIRAGUA (project no. EFA210/16) OPCC (project no. EFA082/15), and ANR JCJC TRAM (grant no. ANR JCJC 15-CE01-008 TRAM), which funded the investigations at both sites.
This paper was edited by Ji-Hyung Park and reviewed by two anonymous referees.