NET ECOSYSTEM EXCHANGE, GROSS PRIMARY PRODUCTION AND ECOSYSTEM RESPIRATION IN RIDGE-HOLLOw COMPLEX AT MUkHRINO BOG

The continuous field measurements of net ecosystem exchange (NEE) of CO2 were provided at ridge-hollow oligotrophic bog in the Middle Taiga zone of West Siberia, Russia in 2017-2018. The model of net ecosystem exchange of CO2 was suggested to describe the influence of different environmental factors on NEE and to estimate the total carbon budget of the bog over the growing season. The model uses air and soil temperature, incoming photosynthetically active radiation (PAR) and water table depth, as the key factors influencing gross primary production (GPP) and ecosystem respiration (ER). The model coefficients were calibrated using the data collected by automated soil CO2 flux system with two transparent long-term chambers placed at large hollow and small ridge sites. Experimental and modeling results showed that the Mukhrino bog acted over the study period as a carbon sink, with an average NEE of –87.7 gC m-2 at the hollow site and –50.2 gC m-2 at the ridge site. GPP was – 344.8 and –228.5 gC m-2 whereas ER was 287.6 and 140.9 gC m-2 at ridge and hollow sites, respectively. Despite of a large difference in NEE estimates between 2017 and 2018 the growing season variability of NEE were quite


INTRODUCTION
Peatland ecosystems play a significant role in the global carbon cycle, being sources and sinks of greenhouse gases (GHG) (Ciais et al. 2013;Rydin and Jeglum 2015). Despite covering a relatively small part of the Earth surface (about 3%), peatlands store a large amount of organic matter that is ranged between 500 and 700 billion tonnes of C (Page and Baird 2016;Leifeld and Menichetti 2018). In West Siberia peatlands occupy over 30% of the area (Terentieva et al. 2016;Dyukarev et al. 2011;Sheng et al. 2004). According to IPCC estimates (Ciais et al. 2013) the contribution of natural mires into total natural methane emissions ranged between 61 and 82%. The intensity of GHG fluxes is controlled by different factors including the hydrological and thermal regime of the peat deposit (Naumov 2009;Sasakawa et al. 2012;Helfter et al. 2015;Molchanov 2015;Walker et al. 2016;Glagolev et al. 2017;Veretennikova and Dyukarev 2017;Leroy et al. 2017).
The gaseous exchange between the atmosphere and the peatlands is governed by photosynthetic fixation of CO 2 from the atmosphere and by soil and vegetation respiration losses of CO 2 . The balance between them is known as the net ecosystem exchange (NEE) of CO 2 (Bubier et al. 2003;Olchev et al. 2009;Golovatskaya and Dyukarev 2012;Helfer et al. 2015). The other major gaseous emission of C into the atmosphere is accounts for methane (CH 4 ), which is produced via anoxic decay of the soil organic matter (Saunois et al. 2016). The loss of C into the fluvial system occurs via export of dissolved and particulate organic carbon, and dissolved gases (CO 2 and CH 4 ). The rise in surface air temperature (zhaojun et al. 2011;Baird et al. 2012) and the lowering of water levels causes peat drying, increase of temperature and aeration, which contributes to the intense of greenhouse gas emissions (Baird et al. 2012; The second assessment report… 2014). Peatland ecosystems in different years can also serve as both a source and a sink of carbon (Golovatskaya et al. 2008;Panzaoo et al. 2017). The variety of direct and inverse relationships existing between the components of the peatlands and the surrounding areas indicates a complex nonlinear impact of peatlands on the environment in different geographic, climatic, and geomorphological conditions (Peatlands of West Siberia 1976;Vomperskiy 1994;Ratcliffe et al. 2017;Webster et al. 2018). The quantitative estimation of the rate of carbon exchange between peatlands and the atmosphere, as well as the revealing of environmental factors affecting carbon exchange, is an important scientific issue (Sheng et al. 2004;Kabanov 2015).
High-precision measurements of carbon and GHG fluxes obtained using standardised methodologies are important for our understanding C cycle within and across ecosystems. (see Franz et al. 2018;Pavelka et al. 2018). The study of the hydrological and ecological mechanisms controlling peatland response to climate changes is critical to predict potential feedbacks on the global C cycle (Baird et al. 2012; The second assessment report… 2014). Recent field studies indicated that the peatland C balance represents a net C sink in intact peatlands in Canada (Wu et al. 2010;Munir et al. 2014;Webster et al. 2018), China (zhu et al. 2015zhou et al. 2009), Finland (Laine et al. 2019;Minkkinen et al. 2018), Ireland (Swenson et al. 2019), Scotland (Helfter et al. 2015), Germany (Günther et al. 2017), France (Leroy et al. 2017), Poland (Acosta et al. 2017), New zealand (Campbell et al. 2014), East (Runkle et al. 2013;Fleischer et al. 2016;Eckhardt et al. 2018;Davydov et al. 2018) and Western part of Russia Kurganova et al. 2011;Molchanov 2015;Ivanov et al. 2017).
Modelling approaches are useful to divide the observed NEE into gross primary production (GPP) and total ecosystem respiration (ER) components, since it provides a better diagnostic of ecosystem processes and their regulating factors (Falge et al. 2001;Widlowski et al. 2011). Carbon balance models are used to quantify the contribution of different environmental factors to GPP, ER and NEE variability, and to calculate daily and annual carbon bud-gets using the gap-filled time series. Partitioning of the NEE into GPP and ER is also needed for better understanding of interannual and spatial variability of the carbon fluxes (Sokolov et al. 2019). The fundamental ecosystem processes (including photosynthesis and respiration) are common to mires and other terrestrial ecosystems, so changes in photosynthetically active radiation (PAR), air temperature (Ta), and precipitation may affect the C cycle in peatlands, e.g. due to alterations of the growing season length, water and energy budget, vegetation composition and water table levels (WTL) (Yurova et al. 2007;Humphreys and Lafleur 2011, Grant et al 2012, Campbell et al. 2014Molchanov and Olchev 2016;Eckhardt et al. 2018).
The main purpose of this study is to asses CO 2 exchange fluxes in oligotrophic peatland complex at the Middle Taiga zone in West Siberia using field chamber measurements and developed mathematical model of NEE.

MATERIALS AND METHODS
The field measurements were provided on the international scientific field station "Mukhrino" (Yugra State University, Khanty-Mansijsk) founded in 2009 . The field station is a part of the International Network for Terrestrial Research and Monitoring in the Arctic (INTERACT, eu-interact.org) and is actively used by Russian and foreign scientists for studies of the functioning of mire ecosystems. Over the past years, the Mukhrino bog was the main object of numerous experimental study of GHG fluxes (Glagolev et al. 2011;Alekseychik et al. 2017), geochemistry and physical, chemical, and biochemical properties of peat (Stepanova and Pokrovsky 2011;Szajdak et al. 2016), mire hydrology (Bleuten and Filippov 2008), and microbiology including mycology .
The Mukhrino bog (60°54´N, 68°42´E) is located at the eastern terrace of the Irtysh River 20 km to the south from the point of its confluence with the Ob River, in the middle taiga zone of the West Siberian Lowland (Alekseychik et al. 2017 (Alekseychik et al. 2017). Automated monitoring of carbon dioxide fluxes at oligotrophic ridge-hollow complex was performed in 2017-2018 using the portable atmospheric soil measuring system (ASMS) with two transparent chambers. Automated chambers were placed at a large hollow and a small ridge. ASMS is able to measure and record simultaneously the following environmental parameters: air temperature (Ta) and humidity (RH) (at height of 2 m above the ground and at the ground surface), PAR (incoming solar radiation in the 400-700 nm spectral range), carbon dioxide content and water vapor pressure in the air samples. The system includes a two-channel gas analyzer Li-7000 (Li-COR Biogeosciences, USA) and two measuring chambers with a volume of 120 l. The chambers are closed for 5 The air for a sample is continuously pumped through the chamber and the gas analyzer during the observation period using diaphragm pump 7006zVR (Gardner Denver Thomas GmbH, Germany) with flow rate about 2 l/minute. The measurements of the concentration of CO 2 and H 2 O, Ta, RH and PAR are continuously stored in the ASMS and transferred to the web-server. The observation data were downloaded from server and processed using specially created software modules. Real-time ground water depth monitoring was conducted using a pressure transducer (Mini-diver DL501, Van Essen Instruments, Netherlands) submerged into a water at a fixed level under the surface.
The automated system operated in a measuring mode from July to August in 2017, and from May to October in 2018. The flux of CO 2 was calculated using a specialized software module developed in the Matlab R2014b (MathWorks, USA) using a linear model for changing the concentration in the chambers during the first two minutes of data sampling. Totally about 500 observations of fluxes were made at each experimental site in 2017 and more than 2500 observations -in 2018, respectively.

Mathematical modelling
To obtain continuous data records, to extrapolate them to other periods when experimental data are missing and to calculate the annual carbon budget of the ecosystem, a model of total ecosystem carbon exchange was proposed (Dyukarev 2017). The measured total NEE ( Fig.1) was partitioned into the incoming (GPP) and expenditure (ER) components (Mäkelä et al. 2004;Laine et al. 2009;Kandel et al. 2013;Campbell et al. 2015).

NEE = ER -GPP;
(1) GPP = f W × f PAR ; ( GPP is defined as the total amount of the carbon fixed in the process of photosynthesis by plants in an ecosystem, while NEE refers to GPP minus ER. ER is the result of plants and soil respiration, where soil respiration is the sum of autotrophic respiration (roots) and heterotrophic respiration (soil biota).
It is well known that the photosynthetic response under low light intensities is characterized by a linear response and photosynthetic saturation is observed at high light intensities (Pessarakli 2005). A rectangular hyperbolic function f PAR (4) is used for the light response of NEE in daytime (Mäkelä et al. 2004;Laine et al. 2009).
where α is the initial slope of the light response curve at low light (photosynthetic efficiency) (mg μmol -1 ), G m is the theoretical maximum rate of photosynthesis at infinite PAR (photosynthetic capacity) (mg m -2 h -1 ). Carbon dioxide fluxes are given in mg of CO 2 per m 2 per hour. PAR is measured in μmol m -2 s -1 . Possible GPP limitation at high air temperatures was not accounted.

02|2019
The total ecosystem respiration was modelled using an exponential equation f T (5) widely used for explanation of ER variation (Kandel et al. 2013;Campbell et al. 2015). The shapes of functional dependence of C fluxes from environmental variables are shown in Fig. 1.
where Ta is air temperature ( o C), E 0 is the reference ecosystem respiration (mg m -2 h -1 ) at T a = 0 o C and WTL = 0, k T is coefficient describing the respiration temperature response ( o C -1 ).
The rate of photosynthesis and respiration of Sphagnum mosses are well correlated with peat moisture (Molchanov and Olchev 2016;Taylor et al. 2016). Changes in water table depth strongly influence GPP (Grant et al. 2012;Pugh et al. 2018), ER (Helfter et al. 2015), and heterotrophic respiration (Eckhardt et al. 2018). Additional factor f W (6) characterises the influence of WTL on GPP and ER fluxes and can be expressed in a simple linear form: where k W is a parameter of sensitivity of GPP and ER to variation of WTL (cm -1 ).
NEE is negative when the value of GPP exceeds the ER value and there is a net removal of carbon dioxide from the atmosphere. NEE is positive when the ER value exceeds the GPP value and the carbon dioxide is released from the ecosystem into the atmosphere.
The model has been calibrated using all available data set on carbon dioxide fluxes in 2017 and 2018. Two-step procedure of model calibration was developed to estimate model parameters. At the first step, model parameters were calculated for each studied site (ridge and hollow) using all the available data for 2017-2018. At the second step, k T , G m , k W were fixed and the E 0 and α were calculated for each month of the study period separately. Multi-objective optimization procedure was performed in the Matlab software using fminsearch function. The minimum of the unconstrained multivariable function was found using derivative-free optimization method (Lagarias et al. 1998). Root-meansquare error was used as a minimizing function (Dyukarev 2017).

Environmental conditions
During the study period (May -October 2017 and 2018) environmental conditions characterized by large seasonal and diurnal variability (Fig. 2). The depth of the water table level is gradually decreased from the early spring after snowmelt to the end of summer (Fig. 2). Rapid rise of WTL occurs after heavy rains. The total amount of precipitation over the growing season in 2018 (315 mm) was higher than in 2017 (288 mm) and it is resulted in higher WTL in 2018.  Fig.3. A diurnal course of fluxes is quite similar for entire period of measurements. The daily pattern of carbon dioxide fluxes is characterized by a clear maximum at night hours (from 11 p.m to 1 a.m.) when CO 2 is released into the atmosphere, and minimum from 10 a.m. to 1 p.m., when CO 2 uptake by plants exceeds the ecosystem respiration . Night hours are characterized by positive fluxes whereas negative fluxes are observed from early morning (4 -6 a.m.) until late evening (6 -8 p.m.).

GES
The NEE rate changed the sign from positive (release) to negative (uptake) in May even at low air temperatures and remained in the time at relatively low level. CO 2 fluxes increased during the first half of summer whereas GPP and ER reached maximum values at mid-July. ER in August-October is lower than in mid-summer due to decreased air and soil temperatures, but it is higher than at the beginning of the growing season because of a large amount of plant litter and mortmass.
The diurnal pattern of measured CO 2 fluxes on the hollow site in different summer months of 2017, does not vary significantly (Fig. 3). The ridge site is characterized in turn by a slightly decreased CO 2 absorption before noon and increased nocturnal emission from June to August. The early spring in 2017 resulted in a long growing season and, consequently, early onset of the development of vascular plants.  Dwarf shrubs and herbs available at the ridge site are characterized by higher green biomass than sedge at the hollow site. Therefore, both CO 2 uptake and emission fluxes at the ridge site have higher absolute NEE values during the entire growing season, excepting May.
Obtained results are well agreed with measured fluxes at peatlands in other geographical regions. In particular, NEE at peatbog in the south taiga in the European part of Russia (Ivanov et al., 2017) in summer period 2014 was positive (+200 mg m -2 h -1 ) at hummocks and negative (-79 mg m -2 h -1 ) at hollow sites. Under very dry and hot weather conditions in year 2015, CO 2 balance in hummock and hollow was positive with NEE reached +220 and +31 mg m -2 h -1 , respectively. Ecosystem respiration was significantly higher (300-700 mg m -2 h -1 ) than the ER rates obtained in our study. According to our estimates, CO 2 emission at the ridge site was 2-4 times higher in comparison with estimations in a forested peatbog of the southern taiga in West Siberia (Golovatskaya and Dyukarev 2012). NEE measured in a patterned peatland in Ireland (Laine et al., 2006) have showed, that the absolute flux rates were some higher at hummocks and lower at hollows. The daytime average NEE at the sites were -1700 and -330 mg m -2 h -1 at hummock and hollow; and the average night time fluxes were +300 and +50 mg m -2 h -1 , respectively.

Model calibration
The adequate projection of carbon cycle by an ecosystem-level model requires accurate calibration of model input parameters (Wu et al. 2010). The NEE rate measured by automated system was partitioned into ecosystem respiration ER and GPP using suggested NEE model. At the first step of the model calibration all available observation data for the years 2017 and 2018 were used. The 3299 and 3190 observations were used in total for each experimental site (ridge and hollow).
The results of calibration showed a great difference between key model parameters for both experimental sites. The temperature sensitivity coefficient (k T ) for ER rate for the ridge site was about two times smaller than for the hollow site, but at the same time the reference respiration (E 0 ) at the ridge site was 5.4 times higher (Table 1). The photosynthetic efficiency (α) and photosynthetic capacity (G m ) obtained for the ridge site were about 1.7 times higher than corresponding parameter for the hollow site, likely due to difference in green biomass amount. The effect of WTL on CO 2 fluxes at the hollow site was higher comparing with the ridge site. The model calibrated for the whole data set allows reproducing adequately the fast diurnal variations of CO 2 fluxes, but the projected diurnal variations are significantly lower than the variation obtained from observation data. Mean error (difference between modeled and observed data) was small resulting in high correlation between simulated and observed fluxes (R>0.94, R 2 >0.84, significant at p<0.05), although the mean absolute error (MAE) was quite high (Table 1).
In the second step of the model calibration, k T , G m and k W were taken to be constant and equal to the values obtained after the first calibration step ( Table 2). The parameters E 0 and α were calculated for each month of the study period separately. The number of observations used for model calibration varies from 110 in June 2017 to 661 in September 2018.
The parameters α and E 0 increase during the growing season simultaneously with plant biomass development until the mid-summer (Table 2). The maximum values of photosynthetic efficiency (5.14 mg μmol -1 ) were obtained for July 2018 at the ridge site. The maximum values of α for the hollow site were by 20-60% lower than for the ridge. Seasonal course of photosynthetic efficiency is well pronounced and α value for May is significantly lower than α value for middle of the growing season, and about two times lower than the value for the end of the season (October). The ER rate growth is mainly influenced by increased autotrophic respiration rates due to raised biomass amount and increased soil temperatures. E 0 for the hollow site was significantly smaller than the value at the ridge site due to lower biomass amount and reduced contribution of leaf and root respiration.
The model parameters α and E 0 for ridge and hollow sites were related with monthly air temperature T m using exponential model: These equations were used for projection of α and E 0 for May, September and October 2017 taking into account monthly air temperatures (Table 2).

Monthly CO 2 fluxes
Time series of gap-filled modeled ER, GPP and NEE fluxes were integrated for each month of the study period. Annual variability of monthly carbon fluxes for ridge and hollow sites is shown in Fig 4. The largest ER efflux was measured in July 2017 at the ridge site -97.3 gC m -2 . Respiration rate at the hollow site reached maximum values in July too, and they were somewhat lower than at the ridge site -42.1 gC m -2 . In May, the total respiration at both sites were similar and does not exceed 17.9 gC m -2 in 2018 and 5.6 gC m-2 in 2017 because of lower temperatures. June and August were characterized by moderate respiration fluxes ranging between 13.8 and 75.0 gC m -2 . The more intense emission was obtained for the ridge site where various vascular species strongly contribute to autotrophic part of respiration and thicker acrotelm layer promotes aerobic  Fig.2). The hollow site is characterized by faster autumn decrease of GPP comparing with ridge site.
The ratio of GPP to ER is used to estimate the fraction of assimilated carbon that was consumed by the plants (Falge et al., 2002). Analysis of the GPP/ER ratio for the entire measuring period showed that the ratio was 1.2 for the ridge site and 1.6 for the hollow site respectively. Both sites acted as a sink of carbon dioxide from the atmosphere. Over the two years of flux measurements, the average annual uptake of CO 2 was 87.7 gC m -2 at the hollow and 50.2 gC m -2 at the ridge site at Mukhrino bog and the NEE rates were quite similar to findings at other peatland sites. In particular, two years of measurements of CO 2 fluxes in the Stordalen palsa mire (a nutrient poor permafrost peatland) in Sweden showed that the mire was a net sink of carbon, with average annual uptake of −46 gC m −2 per year (Olefeldt et al. 2012). The results of two years flux measurements in a boreal minerogenic oligotrophic mire in northern Sweden (Nilsson et al., 2008) showed the peatbog was also a net carbon sink with annual net uptake of about −55 gC m −2 . McVeigh et al. (2014) reported about the average 10-years annual CO 2 uptake of −55.7 ± 18.9 gC m −2 in Atlantic blanket bog in Glencar, southwest Ireland. The results of 11 year flux measurements in a temperate lowland peatland in central Scotland (Helfter et al. 2015) showed a very high variation of annual NEE rate that is ranged between −5.2 and −36.9 gC m −2 yr -1 .
The differences in microtopographic features between hummocks and hollows and its statistically significant influence on the total ER, but not on GPP, were found by Wu et al. (2010) at ombrotrophic MerBlue bog. NEE rates estimated at the hummock and hollow sites were -66 ÷ +19 gC m −2 yr -1 and -146 ÷ -260 gC m −2 yr -1 , respectively. The chamber estimates of NEE at patterned blanked bog (Laine et al. 2006) found that the annual NEE of the driest peatbog sites was about 130% larger than the NEE rate at the wet sites, indicating a large spatial variation that can be found in NEE rates within a quite uniform peatbog ecosystem.

CONCLUSION
The results of field measurements of CO 2 fluxes at ridge-hollow complex bog in combination with suggested mathematical model allowed us to estimate adequately the NEE, ER and GPP rates for ridge and hollow sites at oligotrophic bog in Middle Taiga zone of West Siberia. The cumulative CO 2 uptake rates exceed cumulative respiration rates at both experimental sites. The two year average growing season NEE at the hollow site was 1.7 times higher (87.7 gC m -2 ) than at the ridge site (50.2 gC m -2 ). GPP and ER rates at the ridge site were higher than at the hollow site. The influence of key environmental factors (air temperature, incoming photosynthetically active radiation and water table depth) on CO 2 fluxes at each ecosystem was very different. It is claimed by differences in model parameters describing ER and GPP response to changed ambient characteristics. The suggested NEE model is a promising tool to describe the NEE partitioning into GPP and ER, and to better understand the biogeochemical processes in mire ecosystems in order to find new possibilities to extrapolate the data of local observations to peatland ecosystems of Western Siberia.

ACkNOwLEDGMENTS
This study was supported by the project АААА-А17-117013050031-8 and grant 13-01-20/39 of the Yugra State University. The field works at Mukhrino field station was funded by Russian Fund for Basic Researches and Government of the Khanty-Mansiysk Autonomous region according to the research project 18-44-860017.