Headwater streams and ditches are important sources of atmospheric CO₂ emissions, estimated to contribute more than 70% of the total global fluvial CO₂ emissions (Raymond et al., 2013). The importance of headwaters for emitting CO₂ holds particularly true for boreal ecosystems, as they are rich in soil carbon and often closely connected with dense aquatic networks (Wallin et al., 2018). For forestry intensive areas of the Nordic and Baltic countries, a high share of these watercourses are man-made ditches, created to improve drainage and increase forest productivity during the 20^th century (Päivaänen and Hånell, 2012; Norstedt et al., 2021). Given the widespread occurrence of ditch networks combined with high concentrations and emission rates of CO₂, boreal ditches are important sources of atmospheric CO₂ (Peacock et al., 2021). As an intrinsic characteristic, boreal headwaters show close hydrochemical connectivity with adjacent soils and receive continuously high inputs of terrestrial derived carbon from which inflows are largely controlled by variations in hydrological inputs (Billett et al., 2006; Öquist et al., 2009; Crawford et al., 2013; Leith et al., 2015; Wallin et al., 2015). This carbon can enter the aquatic network directly as CO₂ that is produced in the soil via the mineralization of organic matter or from root-associated respiration (Campeau et al., 2019). In addition to terrestrial CO₂ inputs, CO₂ could also be produced in-situ in the watercourse from microbial decomposition or photochemical oxidation of organic matter transported from catchment soils (Köhler et al., 2002; Schelker et al., 2016a; Crawford et al., 2017). Aquatic CO₂ can also serve as the carbon source in primary production (photosynthesis), hence being consumed during daytime. For boreal headwaters, in-situ decomposition of organic matter is often found to be of minor importance for the observed CO₂ (Winterdahl et al., 2016) due to low residence time of the water (Catalán et al., 2016), limited light availability caused by dense tree canopies (Burrows et al., 2021), and low water temperatures (Tank et al., 2010). In addition, due to often unproductive conditions with low nutrient levels as well as restricted light availability, primary production typically also has low influence on CO₂ in boreal watercourses (Lamberti and Steinman, 1997; Roberts et al., 2007).

A critical aspect to consider when unraveling the different controls and their relative importance on CO₂ in watercourses is the time scale of interest. Different processes that control aquatic CO₂ are operating from hourly to seasonal scales (Riml et al., 2019). Thus, interpreting the primary controls on CO₂ concentration dynamics in watercourses requires continuous data collected at sufficient frequency (e.g., hourly) covering complete seasons (Wallin et al., 2020; Gómez-Gener et al., 2021). The development of sensors that monitor high-frequent CO₂ data has enabled studies that explored controls on dissolved CO₂ in watercourses draining various ecosystems (e.g., forest, agriculture, and wetlands) and across different climatic zones (e.g., boreal, alpine, subtropical, etc.) (e.g., Johnson et al., 2010; Dinsmore et al., 2013; Peter et al., 2014; Crawford et al., 2017; Rocher-Ros et al., 2020; Wallin et al., 2020). Studies performed in streams draining nutrient-poor forest ecosystems have shown that CO₂ dynamics are generally driven by variability in stream discharge (Johnson et al., 2007; Dinsmore et al., 2013; Crawford et al., 2017; Riml et al., 2019). The hydrological response in stream CO₂ is dependent on catchment-specific characteristics and variations in groundwater flow paths controlling connectivity to terrestrial CO₂ sources (Leith et al., 2015; Campeau et al., 2018). In contrast, for ditch or stream systems draining landscapes with open canopies that are fully exposed to light, in-situ metabolic processes have been found to have stronger control on aquatic CO₂. In such systems diel cycles in CO₂ can be particularly pronounced, with large concentration differences between day and night (Nimick et al., 2011; Crawford et al., 2017; Attermeyer et al., 2021; Gómez-Gener et al., 2021). These diel cycles reflect the interplay of in-situ primary production (photosynthesis) and respiration within the watercourse (or in the adjacent soils).

Previous findings of low in-situ contribution to CO₂ dynamics in boreal watercourses stem mainly from studies in low-intensively or unmanaged forest systems (Marx et al., 2017; Campeau et al., 2019). In contrast, few studies (e.g., Klaus et al., 2018) have explored CO₂ dynamics in headwaters draining areas affected by clear-cut forestry, which is a common management operation for many production forests in the boreal countries. In Sweden, about 1% (or ca 230,000 ha, based on a 5-year mean 2016-2020) of the productive forest land is harvested every year (Swedish Forest Agency, 2020), and as a consequence, many forest drainage networks are affected by this practice. Following harvest, the catchment hydrology is altered due to reduced evapotranspiration leading typically to higher groundwater levels and increased runoff (Andréassian, 2004; Sørensen et al., 2009; Schelker et al., 2013). Also, dissolved organic carbon (DOC) concentrations are often increased following harvest due to higher rates of decomposition of organic matter in soils and due to greater lateral mobilization from terrestrial sources (Nieminen, 2004; Laudon et al., 2009; Schelker et al., 2012). Similarly, forest harvest often results in enhanced export of nutrients, especially nitrogen, caused by reduced uptake in vegetation and increased mineralization of organic matter (Nieminen, 2004; Schelker et al., 2016b). There is limited literature concerning the influence of forest harvesting on dissolved CO₂ concentrations and emissions in connecting aquatic systems. However, Klaus et al. (2018) found that harvest increased dissolved CO₂ concentrations in groundwater of the surrounding catchment soils but did not affect CO₂ emissions from recipient streams. To what extent this discrepancy in observed patterns between groundwater and watercourses stems from changes in the in-situ stream CO₂ controls following harvest is currently unknown.

This study aims to investigate the impact of clear-cut forestry on the dynamics of dissolved CO₂ concentrations in draining watercourses. We hypothesize that dissolved CO₂ dynamics following forest harvest are altered and become more variable on short time scales (daily) due to changes in light and nutrient regimes which in turn increase the potential for in-situ metabolism. To test this hypothesis we (1) quantified ditch CO₂ concentration levels and dynamics for a full growing season in a forest ditch within a catchment recently being clear-cut harvested, (2) identified and explored the main temporal controls and how they vary with season, and (3) compared observed CO₂ concentration patterns from the clear-cut ditch with patterns observed in a stream draining an unmanaged forest catchment in close proximity.

The study was conducted in the Trollberget Experimental Area (TEA) (64°10'N, 19°46'E), located 50 km northwest of the city of Umeå, Sweden (Figure 1A)¹. The experimental area was set up in 2018 to study the environmental impacts of different types of forest management practices on aquatic ecosystems and has been embedded within the framework of the Krycklan Catchment Study (KCS) (Laudon et al., 2021). The mean annual air temperature for the area is 2.4°C and with a mean annual precipitation of 623 mm (about 30% as snow, based on data from 1980–2020 collected at the nearby Svartberget Climate station) (Laudon et al., 2021). Across the period of the growing season, the number of sunlight hours changes drastically at these northern latitudes (from about 20 h in early June to about 8 h in October).

Within TEA, we studied a 4.4 ha large catchment (DC2) characterized by a dense ditch network (total length, 1.1 km, density, 0.025 m m⁻²) (Laudon et al., 2021, Figure 1C). The DC2 catchment is dominated by till soils (almost 100%) with well-developed podzols including a 10–20 cm humic/partly humic layer on top. Until July 2020, DC2 was completely forest covered, mainly by Norway spruce (Picea abies) and Scots pine (Pinus sylvestris). In July 2020, the catchment was completely harvested.

A second unmanaged sub-catchment (C2, located about 10 km from DC2) of the Krycklan catchment study was used in the study for comparative analysis with the clear-cut catchment of DC2 (Laudon et al., 2013; Leith et al., 2015; Figure 1B). C2 is 100% forested and slightly larger (12 ha) than DC2 but is otherwise similar in terms of soil types and forest composition. Both DC2 and C2 have been affected by historical ditching activity that occurred in the early 20^th century to improve drainage. Catchment characteristics of DC2 and C2 are shown in Supplementary Table 1.

Headwater streams and ditches are known hotspots for atmospheric CO₂ emissions, and the hydrological export of CO₂ from catchment soils is commonly found as the main source in boreal regions (Striegl and Michmerhuizen, 1998; Rasilo et al., 2012; Riml et al., 2019). However, current large-scale estimates suffer from limited information regarding how these emissions are affected by human induced disturbances. Forestry is one such major disturbance, which is known to alter a wide range of hydrological and biogeochemical processes, but the effect on CO₂ concentration dynamics and associated emissions in connected drainage networks are largely unknown.

Here we observed a mean clear-cut ditch CO₂ concentration (2.47 mg C L⁻¹) that was relatively high compared to both what was observed in the comparing forested catchment (C2), but also compared to other high-resolution monitoring studies of forested headwaters found in the literature. For example, mean CO₂ concentration levels found in a study of streams draining different boreal and temperate forest ecosystems were generally lower (range of means: 0.73–2.13 mg C L⁻¹) than the mean of the current study (Dinsmore et al., 2013). Furthermore, expressed as partial pressure (pCO₂), the range found in our study (461–7,183 μatm) encompassed the full pCO₂ range found by Crawford et al. (2017) covering multiple ecosystem types from alpine tundra (434–536 μatm) to temperate forests (2,815–6,225 μatm). The overall seasonal CO₂ concentration pattern found, characterized by a summer peak in CO₂, is typically observed across different types of ecosystems (i.e., arctic tundra, boreal forest, temperate forest, temperate peatlands, and alpine regions) (Crawford et al., 2017). The observed seasonal CO₂ concentration pattern suggests a respiratory source further supported by the close relationship between mean daily CO₂ concentration and water temperature (Figure 3A). Respiration is strongly controlled by temperature (Del Giorgio and Williams, 2005; Yvon-Durocher et al., 2012), and microbial mineralization of soil organic material is known to increase after clear-cut harvest due to increased soil temperatures caused by the absence of shading trees (Liski et al., 1998; Schelker et al., 2013).

In addition to the observed overall seasonal patterns, the high-frequency measurements allowed us to capture ditch CO₂ concentration dynamics on short timescales (hourly or daily). For a majority of the study period, a clear diel signal was recorded with large day-to-night differences in CO₂ concentration (mean and medium Δ: 0.35 and 0.30 mg C L⁻¹, respectively corresponding to mean and median ΔpCO₂ of 1,105 and 885 μatm). These diel CO₂ cycles were particularly pronounced in amplitude during June and July (reaching up to 1.1 mg C L⁻¹ or 4,078 μatm) but became more constrained toward the autumn. The mean of observed daily CO₂ amplitudes was comparatively high in relation to other continuous CO₂ measurements in low-productive arctic and alpine streams exhibiting diel fluctuations. For example, Rocher-Ros et al. (2020) registered a mean summer amplitude of about 900 μatm in a Swedish arctic tundra stream, and Peter et al. (2014) found a mean diel CO₂ amplitude of about 370 μatm in an alpine stream, with amplitude values that peaked at 845 μatm in the summer during extended base flow. In contrast, much higher diel CO₂ amplitudes were found in a nutrient rich agricultural stream in Sweden (medium amplitude: 2.03 mg C L⁻¹, 2,974 μatm) (Wallin et al., 2020). We suggest that the pronounced diel cycles found in the ditch of the current study were driven by aquatic primary production consuming CO₂ during daytime, as the minimum concentrations were recorded during mid-day and with a gradual shift in timing toward the afternoon during autumn (Rocher-Ros et al., 2020, 2021; Gómez-Gener et al., 2021). The pronounced diel patterns we observed further suggests that the temporal control on ditch CO₂ has changed after clear-cut as a consequence of the elevated DOC and nutrient concentrations (compared to the forested catchment, C2) as well as due to increased light availability. Diel dynamics of the observed amplitude are typically observed in open canopy systems and is attributed to primary production largely driven by high light exposure (Crawford et al., 2017; Gómez-Gener et al., 2021).

Hydrology (i.e., variations in discharge) commonly plays an important role in regulating stream CO₂ dynamics across different environments although with site-specific CO₂-Q relationships (Dinsmore et al., 2013; Riml et al., 2019; Wallin et al., 2020). In our case study, the influence of variations in discharge on growing season ditch CO₂ concentrations was complex and not easy to disentangle from the metabolic diel dynamics. Median daily ditch CO₂ concentration was found negatively related to median daily discharge during spring and summer (May to August) but not during autumn (September to October) (Figure 5). However, the slopes of the monthly CO₂-Q relationships indicate a general “chemostatic” response in relation to variable discharge, implying a relatively low hydrological influence on ditch CO₂ concentrations. This suggests that (1) the terrestrial (or in-ditch) source for CO₂ is relatively stable in its hydrological connectivity, or (2) that non-hydrological processes counterbalance any variations in CO₂ caused by a variable discharge (Rehn et al., 2023). The response in CO₂ following individual hydrological events was in contrast highly variable. The CO₂-Q hysteresis plots were in many cases influenced by the diel CO₂ fluctuations leading to tangled hysteresis shapes, making it hard to extract information. Only one extreme hydrological event (i.e., event no. 8, Figure 6) showed an unequivocal and straightforward response in CO₂ concentrations from the analysis of the hysteresis loops. It is worth noting that this event registered both the highest incoming SR as well as the second highest total precipitation and had a runoff peak during day hours (Supplementary Table 2), when CO₂ is consumed due to high primary production rates. Other runoff events (some with comparable intensity) had a significant impact on the CO₂ level, but either had runoff peaks during night hours (e.g., events no. 4 and 13) or occurred in the autumn (e.g., events no. 16 and 18), when the SR is low. As a result, the simultaneous metabolic signals made the hysteresis plots complex. We suggest that the different hydrological responses on CO₂ are related to the timing of an event, during what season the event occurs, whether the event follows an extended dry period, or when during the day (day or night) the runoff peaks. The absence of a clear response in CO₂ concentration for most of the events, suggests that variations in runoff did not have a major control on ditch CO₂ dynamics and were instead overridden by the stronger light and temperature induced metabolic control operating at the diel timescale (Bernal et al., 2022).

The importance of in-situ metabolic processes controlling CO₂ dynamics was finally supported by the event based and control integrated PCA analysis (Figure 7). Both the magnitude in CO₂ concentration and range in diel CO₂ concentration amplitude were closely related to both daily total SR and mean water temperature. In contrast, the PCA displayed low influence on CO₂ by any of the hydrological metrics. The elevated in-situ control on ditch CO₂ following forest harvest was further evident when comparing continuous data collected from DC2 with the completely forested catchment C2 included in the KCS and located within 10 km from DC2. C2 is representative for the conditions at DC2 as they were prior to the forest harvest and is used as one of two forest control catchments within the experimental design of the TEA. From the comparative analysis between DC2 and C2 clear differences were observed, both in concentration magnitude and amplitude of the diel CO₂ cycles, but also in diel water temperature patterns. This suggests that the collective conditions after clear-cut, with elevated solar radiation exposure and increased DOC and nutrient concentrations alter the in-situ ditch ecosystem function, and by that enhancing the importance of temperature- and light-induced metabolic control on the CO₂ dynamics. Furthermore, the slope of the logCO₂-logQ relationship observed in C2 was more negative than in DC2 (−0.12 and −0.08, respectively) suggesting a higher runoff control on CO₂ concentration (Supplementary Figure 9). This comparative part of the study further supports our hypothesis that in-situ metabolism is a key driver of aquatic CO₂ dynamics in clear-cut catchments.

We acknowledge that the current study only represents a single ditch and observed patterns are likely site-specific. However, we believe our finding of an increased metabolic control on CO₂ dynamics in forest ditches and streams following clear-cut harvest should be valid across regions with similar climatic conditions and forest management. The increased short-term CO₂ concentration dynamics following forest harvest will also lead to altered emissions patterns. To what extent these altered patterns will influence total annual emissions is uncertain and will require detailed investigations. Klaus et al. (2018) found that despite significant increases of CO₂ in groundwater of clear-cut affected catchment soils, no change in GHG (including CO₂) fluxes in adjacent streams were detected within 3 years after the treatment. The authors explained the mismatch in patterns between ground- and stream water with that the trees left in the riparian zones most likely acted as an effective buffer zone mitigating stream GHG emissions. The findings of the current study suggest that increased aquatic productivity might play a role in consuming the elevated soil CO₂ export following forest harvest. Thus, inorganic C will be converted into organic forms shortly after being transported across the soil-water interface.

We conclude that CO₂ concentration dynamics in forest ditches affected by clear-cut harvest are driven by a complex interplay of light and hydrologically induced processes. Despite the common perception of forest ditches as nutrient-poor systems typically showing low metabolic rates, our findings suggest that metabolism, and primary production specifically, exerts significant control on short-term ditch CO₂ concentration dynamics. In contrast, variations in discharge displayed a comparatively less dominant influence on the variation in CO₂ concentrations. The high CO₂ concentration dynamics and the associated metabolic controls should be considered when scaling CO₂ emissions across boreal landscapes impacted by clear-cut forestry. To improve our understanding of these processes, we recommend that future studies combine measures of C export/emission with in-situ metabolism and that these are conducted over longer time scales (i.e., >single growing season). Overall, our results emphasize the need for more comprehensive and detailed investigations of the factors regulating CO₂ dynamics in forest ditches and their implications for the landscape-scale C budgets.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AZ and MW designed the study. AZ carried out most of the fieldwork, analyzed the data, and wrote the first draft of the manuscript. MW, KE, EL, EM, and HL further provided scientific insight to the analysis and interpretation of the data. All authors commented on earlier versions of this paper.