Dissolved organic carbon export in a small, disturbed peat catchment: Insights from long‐term, high‐resolution, sensor‐based monitoring

Understanding dissolved organic carbon (DOC) export dynamics from carbon‐rich environments is critical. Peatlands act as terrestrial carbon stores, and consequently supply substantial amounts of DOC to drainage. This DOC flux is temporally heterogeneous and subject to long‐ and short‐term variability. Ultrahigh temporal resolution sampling (< hourly) is still in‐frequent in peatland catchments. We used a field‐deployable —ultraviolet–visible light spectrometer (Spectro::lyser™) and monitored DOC flux from a temperate peatland over 31 months to examine seasonal and event dynamics. DOC concentration varied from 6.8 to 63.5 mg L−1, in the higher reported range for peatlands and showed clear seasonal (high‐summer, low‐winter) variability coinciding with elevated biological productivity in the peatland. Discharge was an unreliable predictor of instantaneous DOC concentration overall, with antecedent water temperatures proving the most reliable predictor overall. Discharge drove total DOC export in the catchment, where the top 10% of flow events, accounted for 41.3% of all DOC exported—increasing to 84.6% in the top 50% of flow events. Total estimated catchment DOC flux was sensitive to measurement frequency: increasing from every 30 min to daily altered export estimates by < 1%, increasing to > 10% at 1‐week intervals. The variation in estimated flux increased approximately linearly with reduced sampling frequency, reaching > 40% at monthly intervals. High‐resolution data reveal the large amount of within‐site complexity of DOC export dynamics in a temperate peatland and provide evidence on the subsequently recommended sampling frequency for the future elucidation of detailed DOC budgets in these environments.

Peatlands are substantial terrestrial C-stores and changes in the quantity of DOC export can reflect changes to the efficacy of this C-sink. Furthermore, changes in DOC export can also prime a positive feedback loop on atmospheric CO 2 concentrations (Worrall et al. 2002;Freeman et al. 2004), where DOC export can return CO 2 to the atmosphere through in-stream respiration (del Giorgio and Peters 1994;Cole et al. 2007) and photo-oxidation (Bertilsson and Tranvik 2000;Johannsson et al. 2020).
DOC stains water (de Wit et al. 2016), thus attenuating water column light and reducing photosynthesis (Williamson et al. 1999a) or, conversely, increasing health of zooplankton and some fish populations through decreased ultraviolet (UV) exposure (Williamson et al. 1999b). Regions of high peatland abundance are often sources for drinking water (Xu et al. 2018) and changes in the concentration and characteristics of DOC can affect industrial DOC removal processes (Worrall and Burt 2009). Further challenges in DOC removal from drinking water due to CO 2 and temperature enhanced concentration rises will likely exacerbate this issue (Fenner et al. 2021). From both a carbon-landscape and drinking water management perspective, detailed understanding of fluvial DOC flux is required. For a better understanding of the drivers controlling fluvial DOC dynamics, detailed temporal characterization is required, particularly as DOC export is often coupled to discharge (Q) events vs base flow (Fellman et al. 2009;Raymond and Saiers 2010;Bass et al. 2011). However, event sensitivity is often not constant and depends on antecedent catchment conditions. In many catchments preceding dry conditions can lead to elevated DOC flux in the first substantial event of the season, not repeated in subsequent events (Bass et al. 2011;Fenner and Freeman 2011). A challenge to robust characterization of Q-DOC response is their ephemeral nature (temporal and spatial) which can make traditional temporal scaling of "grab-sample" approaches ineffective.
To elucidate fine-scale patterns and controls the use of high-resolution sensors based on UV-absorbance has become more common, including campaigns in tropical regions (Waterloo et al. 2006;Bass et al. 2011Bass et al. , 2014, temperate peatlands (Grayson and Holden 2012;Jones et al. 2014), and afforested landscapes (Strohmeier et al. 2013). Given the observed global increase in surface water DOC concentrations (Monteith et al. 2007) and the anticipated continued increase in frequency and severity of precipitation events, a detailed understanding of discharge related DOC sensitivity is essential for accurate landscape-scale C-flux forecasting.
In this study, we present a 31-month sensor-generated time series of 30-min resolution DOC measurements from a peatland catchment previously influenced by disturbance during wind farm construction. We use these data to (i) evaluate the dynamics of DOC export in a peatland recovering from disturbance, (ii) elucidate short-and long-term controls of event-induced DOC export, and (iii) examine optimal measurement frequency for catchment scale DOC-export estimations. Data were collected in the 2012/2013 and 2013/2014 hydrological years, capturing two successive but hydrologically distinct summers-the very wet summer of 2012 and the dry summer of 2013.

Field site
Drumtree water is a third order catchment in SW-Scotland (55 41 0 16 00 N, 4 23 0 37 00 W) covering an area of 5.81 km 2 and an altitude range of 195-260 m a.s.l. The Drumtree catchment drains the periphery of the Whitelee wind farm (Fig. 1) which contains 215 turbines constructed between 2006 and 2013. During construction, disturbance including forest clearance, peat excavation, access road construction, and turbine erection occurred over the Whitelee site, with the primary disturbance in Drumtree being deforestation and road construction to access eight on-site (and one off-site) turbines.
In situ sensor instrumentation DOC concentration ([DOC]) was measured at 30-min intervals using a portable, UV-visible spectrometer (Spectro:: lyser™, henceforth referred to as Spectrolyser). Water quality including pH, temperature, and conductivity (SpEC) were measured with an MP Troll 9000 hydrochemistry Sonde (In Situ). A pressure transducer (In Situ) and flow logger (ISCO) were deployed for discharge quantification. Water quality sensors were clamped to posts on a bridge, and the Spectrolyser was positioned in the channel center in a custom made, weighted steel mesh box (pore size approx. 3 cm 2 ). The Spectrolyser was positioned perpendicular to flow ensuring uninterrupted flow over the measurement window.
Sensor calibration occurred at approx. 5-week intervals. Spectrolyser maintenance involved cleaning the measurement window, carried out with a nylon brush and 10% HCl, to remove persistent staining (likely Fe or Mg). From December 2012, a cylinder of compressed N 2 was used to automatically "clean" (inhibit particle settling) the sensor window every 6-8 h, removing the necessity for manual cleaning. [DOC] data between 30 April 2013 and 05 June 2013 are of lower accuracy due to a failure of the automatic cleaning system.
The Spectrolyser can store the results from 1484 measurements ($ 1 month of data) with data unloads occurring at approximately 3-wk intervals. Alongside data unloads, a 1 liter water sample was collected in a polyethylene bottle for laboratory analysis. Within 24 h, samples were filtered (GF/F 0.7 μm) and inorganic C removed (by titration with 0.05 M H 2 SO 4 and sonification), prior to [DOC] measurement using a Thermalox™ total organic carbon (TOC) analyzer, rated to a precision of AE 3%. The [DOC] measurement was undertaken within 2 months of filtration, a period identified to be unaffected by storage (Gulliver et al. 2010). Laboratory [DOC] measurement of field samples was used to construct Spectrolyser calibration curves (Fig. 2).
Material retained on the GF/F filters was used to estimate particulate organic carbon concentration ([POC]) via loss-onignition at 500 C for 16 h, following initial drying at 105 C for 4 h. The fractional content of C in the total organic material was assumed to be 0.58 based on Schumacher (2002).
An ISCO flow logger was deployed between 05 November 2012 and 07 January 2015 to measure stream flow and extrapolate discharge. This direct measurement was used for most of the 31-month monitoring period. A 30-min resolution discharge time series was reconstructed for (i) the period from 23 May 2012 to 05 November 2012 using a linear relationship between a nearby Scottish Environmental Protection Agency monitoring station (55 36 0 22 00 N, 4 19 0 52 00 W) and discharge at Drumtee (r 2 = 0.89) and (ii) for after 07 January 2015 by using a two-order polynomial relationship established with the In Situ pressure transducer installed on 20 August 2014 (r 2 = 0.93). Consequently, uncertainty in discharge measurements for these two periods was 0.01 and 0.001 m 3 s À1 higher, respectively-equating to less than 4% and 1% of mean discharge. Periods where different methods overlapped were available, allowing for cross checking between methods. Agreement was strong and combining these methods has produced a robust, continuous, 30-min resolution discharge time series with no method-dependent step changes.

Bass et al.
Sensor-based high-resolution DOC

Estimating [DOC] from sensor absorbance
The Spectrolyser measured absorbance in the UV-visible range (from 200 nm up to 735 nm at 2.5 nm increments) in stream water flowing through a 0.5 cm pathlength. Local calibration is required to confirm applicability of internal calibration settings. Specifically, at our field site, high particulate and iron loading (Zheng et al. 2018) could affect absorbance by light scattering. We constructed an independent, site-specific calibration utilizing two wavelengths, 254 and 340 nm, to estimate [DOC] (Tipping et al. 2009). Incorporating absorbance at 735 nm was used to correct for increased absorbance due to turbidity from light-scattering particulate material or biofouling in the path length. Absorbance at 735 nm is close to zero when the measurement window is clean and water is clear (values less than 1 considered excellent and less than 10 considered good). Turbid waters and accumulated material from biofouling can increase this value (our largest recorded value was 543.7). The specific numerical approach is detailed in Eq. 1: where x is the site-specific constant derived using samples collected at the site prior to the measurement period, determined to be 0.43 for Drumtree Water. A is the absorbance value at each wavelength. For further details of this approach, see Tipping et al. (2009). Work on pairing high-resolution monitoring data alongside detailed composition characterization would allow for a more  process-based understanding. Our investigation of using the Spectrolyser spectral data to examine DOC quality ratios such as SUVA 254 and E 4 : E 6 (for aromaticity and the fulvic : humic acid ratio, respectively) yielded results substantially different from those previously measured on the site (Zheng et al. 2018). Iron presence influences spectral readings in this catchment, and this needs to be corrected for, as was done in previous work. Samples from the reported study period of previous work were not available for retroactive calibration and thus beyond the scope of this paper. However, in other cases, Spectrolyser spectral data have been used to infer molecular weight changes in the DOC pool using the spectral slope ratio (Vaughan et al. 2017) illustrating the potential of this approach to provide characterization data.
Due to the improvement in sensor in situ cleaning that occurred from December 2012, a calibration curve was produced to cover the period before and after, operating on the assumption that sensor interference should have declined after the change. Uncertainty in the sensor values decreased after December 2012 from a standard error of 5.02 to 3.60 mg L À1 (Fig. 2). Measurement uncertainty equates to an average error in export estimates of AE 6.6%.

Deriving DOC budgets and analyzing [DOC] min and max
Periods of sensor failure occurred during the monitoring period (see Fig. 3). Consequently, DOC event analysis procedures generally excluded these periods. DOC-export budgets were calculated for the 2012/2013 and 2013/2014 hydrological years from the high-resolution sensor-generated [DOC] and from the low-resolution lab-measured [DOC] time where C-export (E) (g C m À2 h À1 ) is the product of discharge (Q) (m 3 s À1 ) and [DOC] (g m À3 ), LA is the catchment land area (km 2 ), i is seconds and j any 30-min interval. Where annual export has been estimated for broad inter-annual comparison purposes, standard load estimation methods have been used (Eq. 3).
For event-based DOC analysis, we defined an event as a period of rapid discharge increase, reaching a maximum value before declining to baseflow conditions or until another period of rapid rise. To prevent within-event fluctuations in the discharge being interpreted as separate events, a condition was programmed that only one maximum value will be recorded 12-h either side of a particular data point, and if multiple maxima were identified, then the larger one was used. Thus, fluctuations in discharge over a day are treated as a single event. In total 261 individual events were identified over 2 hydrological years.  Minimum [DOC] coincided most often with the period of steepest Q increase within a 30-min period, while maximum [DOC] was more prevalent in the following 6 h after peak Q but could also occur > 40 h after.

Assessing sensitivity of DOC export to sampling frequency
We investigated the effect sampling resolution had on DOC export estimates. We assumed that 30-min resolution represented the best case and simulated the effect of less frequent sampling to a maximum of monthly. As sampling frequency reduced, we inserted the filtered data sets into the method-5 equation from Wallin et al. (2015) (Eq. 3) to estimate DOC loads.
where F is the load of DOC, K is the time conversion factor (dependent on sampling frequency tested), Q r is the average discharge over the defined period, C i is the instantaneous [DOC], Q i is the instantaneous discharge, and n is the number of sampling points. For our frequency sensitivity analysis, an uninterrupted period of data (23 May 2012 to 30 December 2013) was used. Larger sites with more predictable solute controls can often be modeled using loading models such as LOADEST (USGS). However, these regression-based models often specifically struggle with small, flashy catchments over short timescales (Aulenbach 2013).

Statistical analysis
Variable relationships were investigated with an ordinary least squared linear regression technique, with significance

Bass et al.
Sensor-based high-resolution DOC defined as p < 0.05. Models were evaluated based on r 2 , confidence intervals, and standard error of the regression coefficients. To zero the baseflow data and remove small inconsistencies for individual event extraction, the R package "Baseline" was used. Data processing and visualizations were carried out in the statistical environment R using RStudio version 4.0.2 and IBM SPSS v25.

Results
Catchment hydrological, water quality, and carbon time series Discharge ranged from 0.001 to 5.3 m 3 s À1 (Fig. 3), with similar mean and median discharge values (0.23 and 0.11 m 3 s À1 , respectively) indicating discharge is generally low with substantial relative event size increase (Table 1). Rising limb discharge increased faster than the falling limb, with the median time between event start and maximum discharge being 10 h and from maximum to baseflow being 19.75 h.
Repeated event flow was a key feature of the hydrograph. We defined 261 events, often, without recovery to baseflow condition between peaks. This pattern was particularly evident in 2012, characterized by an unusually wet summer. The early months of 2013 (particularly February and March) were characterized by few events and prolonged baseflow periods. Discharge values greater than 1 m 3 s À1 occurred 10% of the time (Fig. 4) and conditions considered "base-flow" 50% of the time.
Stream hydrochemistry varied with event flow. The pH declined within a few hours of event start to reach a minimum value of 5-6 increasing after, albeit at a slower rate. Minimum SpEC was correlated with higher discharge values and, similarly to pH, exhibited a more rapid response on the ascending vs. descending limb of the hydrograph. SpEC and pH co-varied reaching max and min values at the same times.
The concentration of DOC varied between 6.8 and 63.5 mg L À1 (Fig. 3), recorded on 13 February 2013 and 12 July 2012, respectively, with an average value of 29.4 AE 11.5 mg L À1 . Seasonality in [DOC] occurred with higher values measured in the third quarter (July-September) and lower values in the first quarter (December-February), corresponding to summer and winter periods, respectively. The [POC] ranged from 0.9 to 23.5 mg L À1 (Fig. 3), recorded on 18 October 2012 and 16 June 2012, respectively. The peak value recorded (23.5 mg L À1 ) appeared anomalously high, with the next largest value being 5.8 mg L À1 . However, given the low temporal resolution of POC sampling, it is possible similar magnitude peaks were missed.
The [POC] did not correlate strongly with event flow (or seasonality). However, [DOC] generally increased rapidly with event flow by up to 22.1 mg L À1 (16 June 2012), though the average event-induced increase was less (5.5 AE 4.3 mg L À1 ). The average time between event initiation and peak DOC was variable, with an average of 15.5 AE 24.9 h. Although peak [DOC] broadly corresponded to high Q, in 61.3% of the recorded events, an initial significant decline in [DOC] occurred between event onset and peak Q, often reaching a minimum value in this period. These [DOC] minima were consistently recorded when Q was rising most rapidly (Fig. 5a), with no equivalent consistent pattern observed for maximum [DOC] values (Fig. 5b). Specifically, 40% of DOC minima values occurred during the same 30-min period as peak-Q, and 65% within 2 h of the period of most rapid Q increase. Counter clockwise hysteresis loops dominated events at Drumtree, associated with an initial decline in [DOC] prior to peak discharge and maximum [DOC] after peak discharge (Fig. 6).

Carbon exports and sensitivity to temporal sampling frequency
Using the high-resolution discharge and concentration data, TOC export was 24.7 g C m À2 yr À1 . DOC comprised 91.3% of total export with POC contributing 8.7%. Laboratory sample-derived TOC export was 22.6 g C m À2 yr À1 , only 8.5% smaller than the Spectrolyser-derived budget. Export was discharge dependent, with the highest and lowest 10% of flows accounting for 41.3% and 0.02% of TOC export, respectively (Fig. 4); 84.6% of TOC export occurs in the top 50% of discharge conditions. TOC export varied between hydrological years with export in the 2013/2014 period exceeding 2012/2013 by 19.8% and 28.5% for Spectrolyser and laboratory sampling, respectively. This elevated TOC export is a consequence of an average [DOC] in 2013/2014 being 21.1% higher, coupled to 9.4% more water discharged over the hydrologic year. Estimated C-export depended on the sampling resolution used vs. 30-min resolution measurements; sampling frequencies less than once per day yielded < AE 1% difference in export estimates. At weekly frequency, this increased to AE11%, AE 21% at bi-weekly intervals, and finally > AE 40% at monthly resolutions and beyond (Fig. 7).

Nondischarge drivers of DOC concentration
Although at broad scales [DOC] and discharge dictated total C export, controls on instantaneous [DOC] are highly variable (Fig. 8). An inherent temporal variability is evident in [DOC] independent of discharge and most other potential nondischarge drivers. Antecedent water temperature had the strongest fit to [DOC]. We examined the [DOC]-antecedent water temperature relationship at daily intervals from 1 to 42 d assessing the regression fit (r 2 ; Supporting Information Fig. S1). Average water temperature over the preceding 35-d period had the strongest relationship with [DOC] (r 2 = 0.69). These results indicate a significant temperature related control on [DOC] concentrations. Additional environmental or hydrological factors had little predictive power when considered in isolation (Fig. 8a-e). The patterns of the scatterplots suggest more complex, nonlinear relationships exist.

Discussion
In this study, we show that variability in DOC export dynamics, even in a highly homogenous peatland catchments is high. DOC concentration was poorly predicted by discharge, instead being linked to antecedent climate conditions in the catchment. Uncertainty in catchment scale C-export estimates was relatively stable at temporal resolutions under once per 48 h, increasingly substantially at lower resolutions.

[DOC] and C-export
This research highlights significant heterogeneity in the [DOC] and associated export rates, and a variable relationship between hydrology and [DOC]. Peatlands and high organic content soils have high lateral C-flow (van den Berg et al. 2012;Jennings et al. 2020;Fenner et al. 2021). Interest in their coupled hydrological-chemical dynamics has increased as their role in downstream C-transport, long-term storage of organic-C, and associated challenges treating high-[DOC] water for human consumption has been elucidated.
DOC concentrations in the Drumtree catchment were high, but within the reported range for peatlands. The Glanamong River (Ireland), whose catchment is 77% upland peat had [DOC] ranging from 3 to 14 mg L À1 (Jennings et al. 2020), substantially lower than Drumtree. However, Eaglesham Moorland (part of the Whitelee wind farm) soil water [DOC] ranges from 25.5 to 74.9 mg L À1 (van den Berg et al. 2012). More recent work at multiple (20) sites on Eaglesham Moorland recorded [DOC] ranging from 6.1 to 137.0 mg L À1 (Heal et al. 2020). DOC concentrations at Drumtree were higher than drained, nondrained, and restored peatlands in far-north Scotland, which ranged from 3.8 mg L À1 (nondrained) to 42.2 mg L À1 (drained) (Pickard et al. 2022). The nondrained sites, likely to reflect the peat status most closely at Drumtree, had lower TOC export (7.9-9.2 g C m À2 yr À1 ). However, Drumtree DOC flux was within the range reported in the nearby Auchencorth Moss (19.3 AE 4.6 g C m À2 yr À1 ), a smaller (3.2 km 2 ) catchment used primarily for sheep grazing (Dinsmore et al. 2013).
Use of high-temporal resolution sensors has improved our understanding of fluvial solute dynamics (Waldron et al. 2009;Vaughan et al. 2019;Jennings et al. 2020). Here, high-resolution data have a minimal impact on overall estimates of C-export when the frequency remains higher than once every 2 d (< 2%). However, this potential inaccuracy increases beyond this resolution, reaching over 10% and 21% for 1-and 2-week samplings, respectively. At a sampling interval of once per month difference in estimates exceeded 40% and any robust flux estimates are likely impossible. Measurement resolution is a critical consideration given even relatively high sampling frequencies of 2 d are still rare.
A Spectrolyser deployment in the Lehstenbach catchment (SE Germany) showed different DOC-export estimates than manual spot sampling, between 32% smaller and 67% greater when sampling frequency fell below 2-d resolution (Strohmeier et al. 2013). This sensitivity is greater than we observed, potentially due to variable event hysteresis: in the Lehstenbach catchment [DOC] increases synchronously with discharge, whereas at Drumtree counter-clockwise hysteresis loops dominate (Fig. 6), meaning an initial slow concentration increase or dilution, followed by a concentration rise, often after peak discharge. This hysteresis pattern can dampen discharge-related changes and reduce the magnitude of event-induced [DOC] change-and subsequently suppress the magnitude and rapidity of DOC flux peaks. In addition, significant seasonal variation in [DOC] occurred at Drumtree and not at Lehstenbach, raising the possibility that in catchments where event-induced variability exceeds seasonal variability, short-term high-resolution data are more important. Indeed, in a tropical forest catchment with lower annual temperature variability, reducing sampling frequency from 30 min to weekly missed up to 78% of C-export (Bass et al. 2011). Site-specific characteristics and hydrological connectivity need to be considered when assessing optimal monitoring protocols, but evidence from this and other studies suggests sampling frequencies below 2-5 d risk under or over estimating DOC fluxes by 10-80%.

Controls on the [DOC] response
Using temporally high-resolution sensors equips researchers to better understand and model fluvial [DOC] dynamics during discharge events (Jones et al. 2014;Vaughan et al. 2019;Liu et al. 2021). At Drumtree data show that 41.3% of DOC export occurs in only the top 10% of catchment discharge, with 84.6% of all DOC export in the top 50% of discharge scenarios (Fig. 4). Most export occurring during a few, infrequent floods have regularly been measured. For example, the largest 10% of flows accounted for up to 50% of annual DOC export (Hinton et al. 1997;Clark et al. 2007), 86% of exports accounted for by the top 47% of flows (Inamder et al. 2004;Raymond and Saiers 2010), and in extreme cases (draining coastal tropical forest regions) 85% of the DOC can be exported in only the top 9% of flows (Bass et al. 2011).
Although instantaneous discharge cannot predict spot-DOC concentration, maximum [DOC] was broadly discharge driven at Drumtree, with most (66%) of the peak [DOC] values occurring after, but associated with, peak flow (Fig. 6). Minimum [DOC], however, primarily occurred (79%) in the period between event start and maximum discharge. The initial decrease in [DOC] likely reflects an inflow of surficial, low-[DOC] water related to the precipitation event itself prior to infiltration and associated soil DOC mobilization in the deeper peat layers. Indeed, evidence from previous work suggests peatland [DOC]-Q dynamics can be explained through a two end-member mixing model including rainwater and catotelm stored water as the "new" and "old" sources, respectively (Worrall et al. 2002). However, it should be noted that  (Futter et al. 2008;Raymond and Saiers 2010), but not exclusively (Vaughan et al. 2017).
[DOC] at Drumtree varied seasonally and was highest in summer. Antecedent temperature conditions potentially drive this seasonal trend and have been observed in other catchments (Wallin et al. 2015), reflected in the relationship between [DOC] and mean water temperature over the preceding time period, reaching a maximum fit at 840-h (Fig. S1). Temperatures may restrict DOC export due to processes such as dissolution, desorption, and microbial degradation of organic matter. Higher temperatures may also reduce DOC-export through reduced catchment wetness and discharge (Raymond and Oh 2007). Our measured summer increase in [DOC] and associated export suggests a seasonal overlay on event response. The relationship between [DOC] and the 35-d average antecedent water temperature is stronger at low [DOC] values (Fig. 8), breaking down somewhat above values of 8-10 mg L À1 . This suggests a background temperature control on baseline DOC, but that this control is overridden during significant discharge events, that generally correspond to higher [DOC].
Although we can detect both long-term (seasonal) and short-term (event) drivers on DOC dynamics, we can also see substantial interannual variability between hydrologic years. Specifically, both average discharge and water temperature were higher in the 2013/2014 hydrological year (0.24 AE 0.35 m 3 s À1 ; 11.5 AE 4.4 C) vs. the 2012/2013 hydrological year (0.22 AE 0.32 m 3 s À1 ; 7.7 AE 5.6 C)-with correspondingly an increase in average [DOC] and subsequent cumulative export (Fig. 3).
Long-term data show that DOC is increasing in most catchment types (Freeman et al. 2001;Monteith et al. 2007;Larsen et al. 2011) due to climate changes, pH-induced solubility changes, and hydrological changes. On average DOM exports from northern ecosystems increased by 27% during the period 2002-2016(de Wit et al. 2021. Some of the most substantial of these changes were recorded in peatlands (Freeman et al. 2001), making refining our understanding of long-and short-term DOM-drivers critical.

Conclusions
As we improve our high-resolution monitoring of fluvial DOC, it becomes clear that broad-scale drivers such as hydrology and season are often only partially effective in enabling prediction. Targeted study is necessary of the underlying processes that are primarily controlled by hydrology or season, such as biological activity, climatic change, or anthropogenic disturbance. This research demonstrated the overarching seasonal controls and event-induced complexity on DOC export dynamics from a peatland catchment. Although such highresolution data are often desirable, financial and logistical factors can make it difficult to collect, so understanding its importance is of use. Here, we observe that while highresolution monitoring is needed for event-based dynamics, for long-term C-flux estimates the sampling frequency can be reduced significantly (every 30 min to every 48 h) with a < 8% impact on accuracy. Beyond this sampling resolution, confidence in catchment scale fluxes decreases as uncertainties become potentially consequential, and so how much this impacts on the specific research objective must be considered. This degree of temporal frequency sensitivity supports other studies, but the degree of temporal sensitivity seems to vary between catchments, and this should be considered.
A broad connection between DOC export and discharge was measured, though [DOC]-Q relationships were heterogeneous. This general connection, alongside a partial decoupling from hydrological behavior occurred in other peat catchments (Worrall et al. 2002) with current technologies allowing a further refinement of these observations. Consequently, while broadscale temporal patterns may be deducible using regression-based discharge relationships, in small, flashy catchments they may be insufficient to account for short-term (but often substantial) variability. To account for these potential complications, future deployments of high-resolution sensors should not be limited to a short-term, event-based focus. Rather, they should imprint their short-term potential over prolonged periods to temporally characterize fluvial systems over time.
Critical in future considerations of peatland C-export is that DOC is not a homogenous component, and the lability matters for export. Seasonal drivers for example may potentially drive DOC to a greater autochthonous (labile) composition vs. event-induced variability which may act more to simply mobilize the existing, predominantly allochthonous source. Instruments such as the Spectrolyser, appropriately calibrated, may be able to provide such detail.
This work demonstrates the potentially critical importance of high-resolution monitoring for accurate land-scape scale dissolved-C export estimates in small, flashy catchments, even ones with a relatively homogenous land use. The specific balance required between long-and short-term monitoring techniques will be site dependent, but we present a case that a thorough consideration of the sampling requirements are necessary and potentially consequential.

Data availability statement
The time series data supporting this study are available from the corresponding author, A.M.B., upon reasonable request.