Net ecosystem exchange of carbon dioxide on hayland with drained peat soil in central European Russia: mowing scenario analysis

ABSTRACT Haylands are the mildest option for the agricultural use of drained peatlands in terms of CO2 emissions. However, CO2 fluxes and their balance may depend on various conditions including the frequency of mowing and amount of phytomass removed. Based on field measurements of CO2 fluxes using the chamber method and monitoring of environmental factors conducted in 2018–2020 on hayland and fallow on drained peatland in central European Russia, a mathematical model of CO2 balance was built. Numerical experiments showed that mowing of hayland, irrespective of the intensity, did not lead to an increase in CO2 emissions compared to fallow. Fallow and hayland after single mowing had closely modelled net ecosystem exchange (NEE) values: 9.9 ± 2.4 and 8.5 ± 2.7 t C ha−1 season−1, respectively. Furthermore, a single mowing turned out 2.8 t ha−1 of hay (0.8 ± 0.1 t C ha−1), and 4.4 (1.4 ± 0.1) after double mowing. The modelled NEE after double mowing increased to 9.4 ± 2.9 t C ha−1 season−1. A single mowing session in early summer is recommended. Compared to other uses, e.g. arable land, mowing on hayland is a compromise between reducing CO2 emissions and gaining economic benefits from drained peatlands.


Introduction
Covering only 3% of the global land area, peatlands contain more carbon (C) than any other type of ecosystem, including forests (Parish et al. 2008). Peatlands north of 45°N store about 500 Gt of belowground carbon (Yu 2012;Loisel et al. 2014), which is nearly half of the C in the atmosphere (Friedlingstein et al. 2019). Peat C stores have accumulated over thousands of years owing to the partial decomposition of plant material under water-saturated conditions and lack of oxygen. Drainage leads to the aeration of peat soils, aerobic decomposition, and the release of soil C into the atmosphere as carbon dioxide (CO 2 ) Tiemeyer et al. 2016). Global CO 2 emissions due to the oxidative decomposition of drained peatland could emit ~1.15 Gt of CO 2 or ~3% of the total anthropogenic emissions ) and together with peat fires, ~2 Gt or ~5% of total anthropogenic emissions . The potential effect of emissions from Before drainage, a wide range of mire ecotopes were represented here, ranging from raised bogs with dwarf-shrub-sphagnum-pine vegetation to black alder swamps, some of which have been preserved in their natural state. Of the total area of the peatland (~8,200 ha), ~2,850 hectares have been drained since the late 1970s for peat mining mainly by milling, agriculture, and forestry. Drainage for peat extraction was carried out by excavating ditches with inter-ditch spacing of 40-80 m. Following the extraction of up to 1-2 m of peat, many areas were reclaimed for agriculture: some for grasslands, others for croplands (Chistotin et al. 2006;Suvorov et al. 2015). By the 2000s, an increasing proportion of such areas were no longer in use. On such abandoned areas in dry years, pockets of fire sometimes occurred. After similar fires of catastrophic degree in 2010 in central European Russia, from 2010 to 2013, the Moscow region carried out the largest programme in the northern hemisphere (by area; 73,000 ha) of rewetting fire-dangerous peatlands (Sirin et al. 2020). The programme included one part of the Dubna Peatland, where overflow and water-regulating dams, a fire-prevention pond, and other hydraulic structures were installed in order to help regulate the water regime (Sirin et al. 2020).
The measurement sites were located in the southern part of the peatland (56°40.6ʹ N, 37°48.8ʹ E), where rewetting was not carried out and drained areas were mainly used as haylands (Figure 1). At this location, the peat deposit thickness was reduced by extraction by ~0.5-1.0 m, the depth of the drainage channels ranged from 0.8 to 1.5 m, and the channel edges were overgrown with woody and shrub vegetation 2-12 m in height. Fieldwork was conducted in August 2018, May, June, August, and September 2019 and June 2020. The measurement sites were located on two adjacent inter-ditch spacings ( Figure 1). From 2015 until the end of summer 2019, the northern spacing (fallow) had not been used; then was mown on 27 September 2019 to prevent weed vegetation overgrowth. The southern spacing (hayland) was mown twice, on 28 July and 24 August 2019.

Soil properties
Soil profiles on the fallow and hayland were excavated to a depth of 50 cm, and peat was doublesampled at 5-cm intervals for moisture, ash content, bulk density, botanical and elemental composition. Gravimetric water content was determined in sample size of 5 × 5 × 5 cm, and the ash content by igniting at 850°C. Bulk density was determined by the weighing of dry samples 5 × 5 × 5 cm size. The peat botanical composition and degree of decomposition were studied by microscopic and centrifugation methods, and C and N contents using a Vario MICRO cube elemental analyser (Elementar, Germany).
Perforated PVC pipes (ø50 mm) were installed in the centre of both inter-dich spacings (with fallow and hayland) to measure groundwater levels (GWLs). In addition, an automatic sensor Mini Diver (Eijkelkamp, the Netherlands) was installed in the fallow site, which recorded the GWL and water temperature at a frequency of 4 h from May 2019 to July 2020. A levelling profile survey was conducted through both sites, and no differences in relative position of the terrain were noted. Therefore, the sensor-recorded GWL was attributed also to the hayland site.

Vegetation cover
Vegetation cover of the investigated sites was dominated by meadow grass (Poa pratensis L.), which was actively displaced by nettle (Urtica dioica L.) in the absence of mowing on the fallow. Phytomass of plants (in terms of dry matter, g m −2 after drying at 105°C) was determined by double sampling in each field campaign. Aboveground phytomass (AP) was estimated by cutting from an area of 0.49 m 2 , and belowground phytomass (BP) by washing the plant roots from soil monoliths (10 × 10 × 20 cm) on a 2-mm mesh sieve. Note that the sum of AP and BP yielded the total phytomass of plants (P).

Carbon dioxide (CO 2 ) flux measurements
Daily dynamics of CO 2 fluxes were measured on fallow and hayland on 27-28 August 2018; 18-20 May 2017-20 June, and 27-29 August 2019; and 23-24 June 2020. Measurements were recorded during the day from 10:00 to 18:00, using the dynamic chamber method (Edwards and Sollins 1973) with the LI-820/840 gas analyser (Li-Cor, USA) in 2018 and 2019, and the static chamber method with the UGGA-28p gas analyser (Los Gatos Research, USA) on 23-24 June 2020 (see the carbon dioxide flux calculations in the electronic supplementary material).
The components of CO 2 gas exchange between the atmosphere and the ecosystem, measured on both hayland and fallow, are shown in Figure 2. Soil respiration (R soil ), including root respiration (R root ) and respiration of soil microorganisms (heterotrophic respiration, R h ) were measured at sites with removed aboveground portions of plants. Ecosystem respiration (R eco ), which is the sum of R soil and plant respiration (R plant ), was measured at sites with undisturbed vegetation during the night (without photosynthetic activity of plants). Net ecosystem exchange (NEE), also known as net CO 2 exchange between the ecosystem and the atmosphere, was also measured at the R eco sites, but during the daytime, when plants were photosynthetically active. NEE is the sum of two oppositely directed processes: the emission of CO 2 by the ecosystem into the atmosphere (R eco ) and the absorption of CO 2 by the ecosystem from the atmosphere (gross photosynthesis of plants). The gross ecosystem exchange (GEE) of CO 2 represents the total amount of CO 2 absorbed by the ecosystem because of plant photosynthesis, including the portion that plants emit by respiration. GEE was calculated as the difference between NEE and R eco .
Simultaneously with CO 2 fluxes, the soil temperature (T soil ) was measured once for 20 min at the depths of 5, 10 and 20 cm with Thermochron iButton DS1921 sensors (Dallas Semiconductor, USA), as well as the photosynthetically active radiation (PAR) and air temperature (T air ) once per minute with a Minikin QTH sensor (EMS Brno, Czech Republic).

Net ecosystem exchange (NEE) modelling and model parametrisation
NEE was defined as the sum of two modelled components, R eco and GEE (here and below measured values are in plain text; depended and independent values of model are in italics). The R eco model was based on the exponential dependence of CO 2 flux on air temperature, and the dependence of CO 2 flux on GWL in the form of an optimum curve (parabolic function exponent), and proportional dependence of CO 2 flux on P. The GEE model was based on the type of saturation curve relationship between GEE (a half-saturation-constant k was determined experimentally) and PAR; and the proportional relationship of GEE to the AP. Modelling of the seasonal dynamics of R soil was also performed based on relationships with BP, T soil , and GWL. We consider R soil as the sum of respiration of BP and soil heterotrophs, thus modelling R soil allowed us to estimate the respiration rate of aboveground portions of plants R plant . R plant was calculated as the difference between the modelled R eco and R soil (see description of the model in the supplementary material).

NEE seasonal dynamics
The seasonal balance of NEE (for 3-h intervals from 15 May to 30 September) was calculated by modelling R eco and GEE based on input data (Т air , GWL, P and AP, PAR, respectively), obtained in 2019 (description of the model input data in the supplementary material). For model experiments with different frequencies of withdrawal of AP on hayland, and its influence on seasonal balance of NEE the following assumptions were adopted. The dates of mowing were 4 July for the single and 21 June 2021 August for the double mowing. For the first mowing, the speed of recovery of the AP was taken to equal that observed after mowing in July, and likewise after mowing in August for the second mowing. The choice of mowing dates was based on the AP dynamics determined in 2019, considering the maximum yield and recovery rate of AP after mowing as well as the phenological features of the plants (Oomes 1977;Gibson 2009).

Figure 2.
Components of CO 2 exchange between atmosphere and ecosystem for hayland and fallow, measured during field work. R soil is the CO 2 flux determined by the respiration of belowground phytomass (R root ) and soil heterotrophic organisms (R h ); R eco is the CO 2 flux, determined by the respiration of the soil (R soil ) and AP (R plant ); gross ecosystem exchange (GEE) is the CO 2 flux from the atmosphere into the ecosystem, determined by the gross photosynthesis of plants; NEE is the CO 2 flux between the ecosystem and the atmosphere, determined as a sum of R eco and GEE.

Descriptive statistics of input data and model stability
Descriptive statistics of field measurements (average, min., max., coefficient of variation (Cv 1 ) and n) are given in Table 1. Cv 1 is a coefficient of variation for field measurements (Table 1), Cv 2 -for model parameters (Table S1) and Cv 3 -for modelled values (Table 2). In the case of regression analysis, the significance of the correlation coefficient (α = 0.05) was determined by the F-test. When parametrising the models, their stability was checked by stochastic modelling method (see details in the supplementary material).

Meteorological conditions
The summer of 2018 (1 June-15 September) was warm and dry; average (Т air mean ), minimum (Т air min ), and maximum (Т air max ) air temperatures were 18.1, 3.5 (9 June), and 29.5°C (29 August). There were 42 rainy days, during which 131 mm of precipitation occurred ( Figure S1). The summer of 2019 was slightly cooler on average; Т air mean , Т air min , Т air max were 16.3, 6.0 (5 and 29 of August) and 30.0°C (7 June), and was relatively humid, in that, 263 mm of precipitation fell in 55 rainy days. The early summer of 2020 (1-30 of June) was characterised by abnormally high precipitation, 144 mm (compared to 76 and 112 mm in 2018 and 2019), which contributed to a GWL of 30 cm that exceeded the soil surface. The air temperature in the early summer of 2020 (18.6°C) was higher than that in 2018 (16.6) and 2019 (17.4). Various meteorological conditions during the fieldwork periods in 2018-2020 ( Figure S1) helped to parameterise the model of seasonal CO 2 balance for a wide range of conditions. Average, minimum, maximum values, coefficient of variation, number of measurements of carbon dioxide (CO 2 ) fluxes and associated environmental parameters (n 1 ), and their number used to model parameterisation (n 2 ).

Soil properties
Eutric Drainic Sapric Histosols were present on the fallow and hayland. The botanical composition of the top 10 cm of peat reflected the vegetation present on the site, which were predominantly cereal root remains and nettles. The deeper layers of fen wood-grass and wood peats showed predominantly alder, willow, pine, and spruce residues, while the epidermis of sedges and reeds were marked. Peat moisture (%) varied from 150 ± 84-178 ± 43 in the upper layer (0-20 cm) to 280 ± 134-340 ± 20 in the lower layer (30-50 cm) and increased consistently with depth in both areas ( Figure S2). The bulk density did not change significantly in the profile and ranged from 0.19 ± 0.06 to 0.27 ± 0.04 g cm −3 in both areas. The ash content in the upper layer was slightly lower on hayland than on fallow (24 ± 8 and 39 ± 1%) and was contrasting in the lower layer (21 ± 2 and 8 ± 1%). A slight increase in ash content at the depth of 30-50 cm on the hayland, as compared with the fallow, was poorly reflected in the C and N contents in this same horizon: 42-47 and 2.6-2.5%. The C/N ratio did not differ between both areas and slightly increased with depth: from 13.6 (0-10 cm) to 19.3 (40-50 cm) on average. Overall, fallow and hayland were characterised by nearly identical soil properties.

Dynamics of vegetation cover
AP (in terms of dry matter) in summer 2019 ( Figure 3) increased consistently from May to July, from 117 to 272 g m −2 on fallow and from 132 to 216 g m −2 on hayland. Then, AP consistently declined at both sites, reaching a minimum of 88 g m −2 in September on the fallow. On hayland after mowing on 24 August 2019, AP was 32, and 1 month later 72 g m −2 . In 2018, after the earlier mowing (28 July) the AP increased more rapidly. One month later, it reached 138 g m −2 , that is, AP nearly fully recovered to its value before mowing, 162 g m −2 (26 August 2019). The dynamics of BP in the upper 10 cm of the soil layer were similar for both fallow and hayland: growth was consistent from May to July and declined by September. However, BP values differed significantly. The maximum values for June-September were two times higher (1,257 g m −2 ) on hayland than on fallow (677 g m −2 ). Growth in the root density and BP are typical for hayland (LeCain et al. 2002;Reeder and Schuman 2002;Chang et al. 2013). It is possible that the decrease in BP value in the second half of the season was due to the excess decomposition of dead roots over the growth of living roots at the end of the vegetation period. Earlier measurements on the other fallow of the Dubna Peatland have shown an AP of 490, and BP (in the 0-10 cm layer) of 914 g m −2 (5 September 2008). This is compared to a later measurement, showing an AP of 358 g m −2 (25 April 2015) ([author(s)]), which is closer to the data obtained in the present study. In hay meadows of European Russia, AP has been recorded at 147-324 g m −2 , and BP at 647-1750 g m −2 (Vladychenskii et al. 2013), whereas hay meadows of lowland peatlands in northeastern European Russia have shown AP and BP at 185 and 1580 g m −2 (in the 0-20 cm layer), respectively (Kovshova 2006).

CO 2 fluxes and environmental parameters
The statistical characteristics of R eco , R soil , and NEE for all measurement periods are listed in Table 1. Average R eco (gCO 2 m −2 h −1 ± σ; n see in the Table 1) for fallow (2.5 ± 1.2) and hayland (1.5 ± 1.3) differed slightly. The Cv 1 of R eco for fallow was twice that of hayland which testifies to higher variability of fluxes during both daily and seasonal cycles. Likely, both the lower R eco values with higher Cv 1 on hayland could be due to the variability of AP, in that immediately after mowing, an important component of ecosystem respiration, R plant , generally becomes close to zero. Smaller average AP leads to greater temperature fluctuations in the surface air layer of hayland, causing a periodic decreasing of R plant ; for example, at night and in other periods with a large drop in air temperature. Relatively large AP, unlike smaller AP, forms a special microclimate near the soil surface. R soil (gCO 2 m −2 h −1 ± σ) for fallow (1.4 ± 0.6) and hayland (1.6 ± 0.5) were identical and were characterised by small variability during measurements. This is likely owing to relatively consistent temperature fluctuation in the soil, which is an important factor of flow variability. R soil for hayland was comparable with R eco , which confirms the assumption of a small contribution of respiration of aboveground parts of plants, especially after mowing.
NEE (gCO 2 m −2 h −1 ± σ) was characterised by significant fluctuations for both fallow (1.1 ± 1.6) and hayland (0.4 ± 1.5). Likely, the higher Cv 1 for hayland was caused by the influence of mowing on not only R plant , but also on photosynthesis. On one hand, immediately after mowing, CO 2 assimilation drops sharply and NEE increases. On the other hand, as AP recovery progresses, NEE decreases due to photosynthesis growth. The NEE values agreed well with previous results on haylands and fallows of peat soils (gCO 2 m −2 h −1 ): 1-2 (Elsgaard et al. 2012), 0.1-0.5 (Eickenscheidt et al. 2015), from −3.6 to 0 (Karki et al. 2019) and others.

Model input data
The input data for the model were T soil and T air , GWL, and PAR measured in 2019 ( Figure S3). The maximum daily T air and GWL occurred in May and June, while in July GWL decreased, and the temperature had dropped. In August and before the beginning of September, the T air changed slightly, but dropped in the second half of September. GWL increased again in the second half of August due to heavy precipitation, and then declined smoothly until late summer.
The values of AP during the season were used in numerical experiments to assess the impact of the mowing regime on NEE (Figure 4). Maximum productivity of haymaking at single mowing reached 2.8 t ha −1 of hay (at 17% moisture) or 0.8 ± 0.1 t С ha −1 (at 45% C content in dry phytomass), which is comparable with previous data (Oomes 1977;Perkins 1978;Gibson 2009). Double-mowing productivity reached 4.4 t С ha −1 (2.4 t ha −1 in the first mowing and 2.0 in the second) of hay or 1.4 ± 0.1 t С ha −1 . If we take the withdrawable biomass at single mowing for 100%, at double mowing it will be 169%. For double mowing, the value of withdrawn phytomass almost twice exceeds its natural growth in the absence of mowing. In the absence of fertiliser application, this intensity may lead to gradual impoverishment of the soil (due to alienation with biomass) and lower hay productivity (Oomes 1977;Gibson 2009).
The estimated recovery rate of AP after mowing varied depending on the mowing time; in the case of single mowing (04.07) in 44 days (by 17.08), and in the case of double mowing (21.06 and 21.08) in 48 and 35 days (by 08.08 and 25.09). The recovery rate of AP after mowing can be underestimated for single mowing (in July) and the first mowing (in the case of double mowing) in June, as it is based on field measurements made after mowing in August. In natural conditions, AP is likely to recover more rapidly; this is because for June and July, T soil , T air , daylight length, and PAR were longer than in August and September when the rate of recovery was evaluated in the field. Lowering the rate of recovery of AP after mowing also suggests an overestimation of the impact on hayland, which would be lower under natural conditions.

Model of ecosystem respiration (R eco ), soil respiration (R soil ) and gross ecosystem CO 2 exchange (GEE)
The models used the dependencies of R eco , R soil , and GEE from temperature, GWL, and PAR ( Figure 5). R soil and R eco were characterised by different temperature variabilities (Table S1, Q 10 ), which could be explained by the peculiarities of the reaction of respiration components indicative to each parameter. In the first case, the reaction involves the R h and R root , and, in the second case involves the R h , R root and R plant . Therefore, R plant makes a significant contribution to the R eco (especially on fallow) and  is known to rapidly grow with air temperature, which leads to a larger Q 10 . In general, the obtained Q 10 corresponds to the data for both R soil and R eco (Tjoelker et al. 2001;Elsgaard et al. 2012;Liu et al. 2017, etc.).
The dependence of relative GEE (per unit of AP) on PAR is represented by a classic saturation curve ( Figure 5), with a saturation constant of 78.9 µmol m −2 s −1 , and did not differ between fallow and hayland.
The dependence of flux from GWL (attributed to the phytomass r soil = R soil /BP and r eco = R eco /P) was similar for both fallow and hayland, in that growth at GWL increased up to ~0.5 m ( Figure 5) and reached the maximum at GWL ~0.6-0.4 m. Maximum CO 2 emissions at the achievement of certain 'optimal' GWL on drained peatlands has been previously noted (Zaidelman and Shvarov 2000). The dependence of fluxes for the range of GWL from −0.4 m below the soil surface up to 0.4 m above the soil surface was obtained under the conditions of high GWL observed in 2020. The curve of flux dependence on GWL in this range was smoothly reduced to zero at 0.2-0.3 m above the soil surface.
The value of fluxes differed significantly between r eco and r soil as well as between hayland and fallow. For fallow, r eco was 2 times higher than for hayland, and r soil -for 4 times. Moreover, the difference between r soil and r eco varied for fallow and hayland. In the first case, r soil exceeded r eco by 4 times, in the second, by 2 times. Lower r soil for hayland may be associated with the oppression of plant roots during mowing. Thus, even though there was an increase in the density of roots under continuous mowing, soil respiration did not increase.
The best results in the simulation showed that the values of R eco (R 2 at α = 0.05) between the measured and calculated values were 0.44 for fallow (p < 0.0003) and 0.59 for hayland (p < 0.04); and for GEE were R 2 = 0.57 (p < 0.0002) and 0.77 (p < 0.00001), respectively. No significant correlation was found for R soil ( Figure 6); therefore, the dependence of R soil and T soil on GWL requires additional study.

Seasonal dynamics of R eco , GEE and NEE
The seasonal dynamics of simulated R eco , GEE and NEE (defined as the sum of R eco and GEE) on fallow and hayland under different mowing modes are shown in Figure S4. The total daily R eco (gС m −2 day −1 ) on fallow was largest; from 15 to 35 in warm and dry periods (July) to 35-70 in hot and humid periods (May, June, and August). For hayland in July, R eco (g С m −2 day −1 ) reached 7-15 in the absence of mowing or after single mowing, and 5-14 after double mowing, and in June 28-36 g С m −2 day −1 , irrespective of the mowing regime. In August, in the absence of mowing and after single mowing R eco was 6-28, and after double mowing was 6-26 g С m −2 day −1 . In all cases, immediately after mowing, R eco was reduced by 12%. Figure 6. Relationship between modelled and observed R eco , R soil and GEE (g CO 2 m −2 h −1 ) for fallow (dark) and hayland (light).
Seasonal dynamics of modelled R eco on both fallow and hayland coincided with the change of Т air (R 2 = 0.7, p < 0.0001) and was substantially weaker in terms of GWL (R 2 = 0.3 on fallow and 0.2 on hayland, p < 0.0001) and daily Т air (R 2 = 0.9, p < 0.0001). The middle of summer was colder and drier than the beginning, and drier than the end. Maximum R eco showed typical values for the increases of GWL in May and late August, as well as the increase of Т air . In the middle of summer, GWL dropped to the range of −90 to −95 cm, and in May and late August, increased to −56 cm, which is the optimum value for R eco . With further increase of GWL, which was noted in 2020 (GWL was 15-20 cm or more above the surface), R eco fell sharply to the range of 2-3 g C m −2 day −1 both on fallow and hayland.
GEE changed little on fallow during the season (−26 to −29 g C m −2 day −1 ) with a slight increase at the beginning and end of the season. These are known to be periods of relatively small AP of plants, which can effectively photosynthesise (i.e. without self-shading). After reaching an AP of 200 g m −2 , GEE on fallow began to decline. Oppositely, on hayland in the absence of mowing, maximum GEE (−15 g C m −2 day −1 ) occurred in July, during the peak of AP, which was, however, 26% lower than in the fallow (and comparable to that in the fallow in June). Based on this, we postulate that mowing probably has a significant impact on GEE. GEE (g C m −2 day −1 ) decreased by a factor of 3.5 (from −14 to −5) after one-time mowing, although AP decreased by a factor of 6.5, and by a factor of 3.5 (from −14 to −4.5) and a factor of 3 (from −11 to −4) after mowing two times. At the same time, the AP decreased by 6.5 and 5.5 times, respectively, which emphasises the nonlinear relationship between the AP and GEE.
NEE values indicated carbon loss during warm and humid summer months, and occasionally observed small assimilation during drier and cooler periods. This could be due to a corresponding phenomenon in which, if the temperature drops, the R eco drops and the GEE stays close, leading to a drop in NEE. Similarly, low GWL reduced the R eco , which also could have led to a drop in NEE. At the same time, if GWL is increased above the soil surface (as in 2020) R eco would reduce to almost zero.
According to our results, both fallow and hayland are sources of CO 2 in the atmosphere. For hayland after single mowing, NEE changed briefly from a periodically observed, small assimilation (up to −5 g C m −2 day −1 ) to a larger CO 2 emission; however, this was only for the period of recovery of AP. During mowing, R plant also decreased, which together with the larger emission, led to a slight shift of the season's total NEE towards carbon loss by the ecosystem.

CO 2 balance at different cutting regime
According to the modelling results, the total R eco for the season (t С ha −1 season −1 ± σ; σ calculated at n = 1000 replicates of bootstrap) was 48.0 ± 3.5 for fallow and 21.9 ± 4.9 for hayland without mowing (Table 2). After single and double mowing, the R eco values for hayland were 21.7 ± 4.9 and 21.1 ± 4.7 t С ha −1 season −1 , respectively. Integral R plant (t С ha −1 season −1 ) for fallow without mowing was 28.0 ± 3.1 and much less for hayland, 3.3 ± 0.8; and was 3.1 ± 0.8 for single and 2.5 ± 0.6 for double mowing. The significant difference between R plant for fallow and hayland was likely caused by the long period (>10 years) of mowing on the hayland. This was also confirmed by the reduction of R plant with the increase in mowing frequency on the hayland.
R soil (t С ha −1 season −1 ± σ) differed slightly for fallow and hayland (20.0 ± 0.7 and 18.6 ± 0.4, respectively), irrespective of the mode of mowing the latter. This is partly due to the accepted assumption of AP invariability before and after mowing. Close R soil values for fallow and hayland were obtained, despite the twofold superiority of AP on the latter (677 and 1,257 g m −2 for fallow and hayland, respectively). The growth of root density in the soil during mowing did not lead to an increase in soil respiration. This may have been due either to the suppression of plant respiration during regular mowing, or the growth of CO 2 binding by the better-developed root systems in the hayland (LeCain et al. 2002;Reeder and Schuman 2002;Soussana et al. 2007). GEE (t С ha −1 season −1 ± σ) differed significantly for fallow (−38.1 ± 6.2) compared to hayland (−15.6 ± 1.5). After single mowing, GEE was reduced to −13.2 ± 1.2, and after double mowing to −11.7 ± 1.0 t С ha −1 season −1 . Despite vegetation re-growth after mowing, plants apparently did not have sufficient time to assimilate more CO 2 with mowing rather than without mowing, due to the reduction of AP.
After single mowing, the phytomass withdrawal of hayland was 0.8 ± 0.1 t С ha −1 season −1 , and NEE changes showed an increase in carbon loss by 2.2 ± 1.3 in comparison with the absence of mowing (Figure 7). After double mowing, the phytomass withdrawal increased to 1.4 ± 0.1, and NEE was 3.1 ± 1.8 t С ha −1 season −1 . In comparison with fallow, which had prolonged absence of use, the withdrawal of 0.8 ± 0.1 and 1.4 ± 0.1 tС ha −1 season −1 when mowing on hayland did not lead to an increase in carbon loss due to CO 2 emissions (i.e. growth of NEE).
For each ton of withdrawn phytomass per hectare of hayland, carbon losses increased by 2.8 ± 1.8 t С ha −1 season −1 with single and 2.2 ± 1.3 t С ha −1 season −1 with double mowing. Nevertheless, considering the amount of the withdrawn phytomass, single mowing is more justified in terms of the growth of CO 2 emissions than double mowing. The obtained results are a conservative estimate; other data (Maljanen et al. 2001) exist on the fall of NEE after mowing for the period of phytomass recovery and the overall reduction of CO 2 emissions when using peat soils for haymaking. To make better use of haylands on drained peat soils and reduce CO 2 emissions, one-time mowing is recommended (Abril and Bucher 1999) without applying fertilisers. The use of fertilisers may increase not only CO 2 emissions but also N 2 O (Grønlund et al. 2006), which may offset the positive effect of mowing. The carbon dioxide balance of hayfields on drained peatlands cannot be compared with virgin peatbogs. However, haylands are one of the smallest sources of CO 2 among drained peatlands (Shurpali et al. 2009;Maljanen et al. 2010).

Conclusion
Abandoned haylands on drained peatlands continue to emit CO 2 to the atmosphere due to the mineralisation of the peat soils. Based on field measurements of CO 2 fluxes (i.e. soil respiration, ecosystem respiration and NEE) and monitoring of environmental factors (e.g. air and soil temperature, PAR, and groundwater levels) conducted in 2018-2020 on hayland and fallow in drained peatland in the centre of European Russia, a mathematical model of CO 2 balance was built. The numerical experiments showed that mowing, irrespective of the intensity, did not lead to increased CO 2 emissions in hayland compared to fallow. Fallow and hayland after single mowing had closely modelled NEE values; 9.9 ± 2.4 and 8.5 ± 2.7 t С ha −1 season −1 , respectively. Moreover, single mowing turned out 2.8 t ha −1 of hay (17% moisture), which was a withdrawal of 0.8 ± 0.1 g С ha −1 (with a carbon content of 45% in the dry phytomass). After double mowing, the amount of hay obtained increased by 4.4 t ha −1 (first mowing: 2.4; second mowing: 2.0), while the phytomass withdrawal increased to 1.4 ± 0.1, and the modelled NEE to 3.1 ± 1.8 t С ha −1 season −1 . At the same time, double mowing could lead to gradual soil depletion in terms of nutrients (due to the exclusion of biomass), a decrease in productivity, and could necessitate fertilisation, which may increase greenhouse gas emissions. Under these conditions, a single mowing session in the early summer is recommended. Compared to other uses, such as arable land, mowing on hayland is a compromise between reducing CO 2 emissions and gaining economic benefits from drained peatlands. In addition, returning drained peatlands to user control would be expected to reduce peat fire risks, which can have significant environmental and economic impacts.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the Russian Science Foundation project 19-74-20185 and the project 'Restoration of peatlands in Russia for fire prevention and climate change mitigation', funded by the International Climate Initiative of the Federal Environment Ministry and managed through the German development bank KfW (project no. 11 III 040 RUS K Restoration of peatlands). . NEE (t C ha −1 season −1 ) and BP (t C ha −1 ) on fallow and hayland; aboveground phytomass at different mowing modes on hayland (t C ha −1 ). a Belowground phytomass was assumed unchanged for different mowing regimes mode including that of its absence.