Seasonal and diurnal variation in CO fluxes from an agricultural bioenergy crop

Biogeosciences, 13, 5471–5485, 2016 www.biogeosciences.net/13/5471/2016/ doi:10.5194/bg-13-5471-2016 © Author(s) 2016. CC Attribution 3.0 License. Seasonal and diurnal variation in CO fluxes from an agricultural bioenergy crop Mari Pihlatie1,2, Üllar Rannik1, Sami Haapanala1, Olli Peltola1, Narasinha Shurpali3, Pertti J. Martikainen3, Saara Lind3, Niina Hyvönen3, Perttu Virkajärvi4, Mark Zahniser5, and Ivan Mammarella1 1Department of Physics, University of Helsinki, P.O. Box 48, 00014 University of Helsinki, Finland 2Department of Food and Environmental Sciences, P.O. Box 56, 00014 University of Helsinki, Finland 3Biogeochemistry research group, Department of Environmental and Biological Sciences, University of Eastern Finland, Yliopistoranta 1D-E, PO Box 1627, Kuopio campus, 70211, Finland 4Natural Resources Institute Finland, Green technology, Halolantie 31 A, 71750 Maaninka, Finland 5Aerodyne Research, Inc. 45 Manning Road Billerica, MA 01821-3976, USA Correspondence to: Mari Pihlatie ([email protected]) Received: 2 December 2015 – Published in Biogeosciences Discuss.: 26 January 2016 Revised: 29 August 2016 – Accepted: 31 August 2016 – Published: 4 October 2016 Abstract. Carbon monoxide (CO) is an important reactive trace gas in the atmosphere, while its sources and sinks in the biosphere are poorly understood. Soils are generally considered as a sink of CO due to microbial oxidation processes, while emissions of CO have been reported from a wide range of soil–plant systems. We measured CO fluxes using the micrometeorological eddy covariance method from a bioenergy crop (reed canary grass) in eastern Finland from April to November 2011. Continuous flux measurements allowed us to assess the seasonal and diurnal variability and to compare the CO fluxes to simultaneously measured net ecosystem exchange of CO2, N2O and heat fluxes as well as to relevant meteorological, soil and plant variables in order to investigate factors driving the CO exchange. The reed canary grass (RCG) crop was a net source of CO from mid-April to mid-June and a net sink throughout the rest of the measurement period from mid-June to November 2011, excluding a measurement break in July. CO fluxes had a distinct diurnal pattern with a net CO uptake in the night and a net CO emission during the daytime with a maximum emission at noon. This pattern was most pronounced in spring and early summer. During this period the most significant relationships were found between CO fluxes and global radiation, net radiation, sensible heat flux, soil heat flux, relative humidity, N2O flux and net ecosystem exchange. The strong positive correlation between CO fluxes and radiation suggests abiotic CO production processes, whereas the relationship between CO fluxes and net ecosystem exchange of CO2, and night-time CO fluxes and N2O emissions indicate biotic CO formation and microbial CO uptake respectively. The study shows a clear need for detailed process studies accompanied by continuous flux measurements of CO exchange to improve the understanding of the processes associated with CO exchange. 1 Introduction Carbon monoxide (CO) is an important reactive trace gas in the atmosphere, where it participates in the chemical reactions with hydroxyl radicals (OH), potentially leading to the production of the strong greenhouse gas ozone (O3). The reactions of CO and OH decrease the atmospheric capacity to oxidize atmospheric methane (CH4), hence indirectly affecting the lifetime of this important greenhouse gas. Although CO itself absorbs only a little infrared radiation from the Earth, the cumulative indirect radiative forcing of CO may be even larger than that of the third powerful greenhouse gas, nitrous oxide (N2O; Myhre et al., 2013). Anthropogenic activities related to the burning of fossil fuel and biomass (e.g. forest fires) as well as photochemical oxidation of CH4 and non-methane hydrocarbons are the main sources of CO (Duncan et al., 2007), while the reaction with OH is the major sink of CO in the atmosphere (Duncan and Logan, 2008). Published by Copernicus Publications on behalf of the European Geosciences Union. 5472 M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes Soils are globally considered as a sink for CO due to microbial oxidation processes in the soil (Conrad and Seiler, 1982; Potter et al., 1996; Whalen and Reeburgh, 2001; King and Weber, 2007). According to Conrad and Seiler (1980) the soil consumption of CO is a microbial process, it follows first-order kinetics and can take place in both aerobic and anaerobic conditions. A diverse group of soil microbes are capable of oxidizing CO. They include carboxydotrophs, methanotrophs and nitrifiers (Ferenci et al., 1975; Jones and Morita, 1983; Bender and Conrad, 1994; King and Weber, 2007), hence they potentially link CO fluxes to the exchange of CH4and N2O. In addition to CO consumption, production of CO has been found in a wide range of soils (Moxley and Smith, 1998; Gödde et al., 2000; King, 2000; Varella et al., 2004; Galbally et al., 2010; Bruhn et al., 2013; van Asperen et al., 2015), plant roots (King and Crosby, 2002; King and Hungria, 2002), living and degrading plant material (Tarr et al., 1995; Schade et al., 1999; Derendorp et al., 2011; Lee et al., 2012) and degrading organic matter (Wilks, 1959; Conrad and Seiler, 1985b). Although microbial CO formation may occur in anaerobic conditions (Funk et al., 1994; Rich and King, 1999), most often the CO production has been related to abiotic processes such as thermal, UVor visible light-induced degradation of organic matter or plant material (Conrad and Seiler, 1985b; Tarr et al., 1995; Schade et al., 1999; Derendorp et al., 2011; Lee et al., 2012; van Asperen et al., 2015; Fraser et al., 2015). Photodegradation involves direct and indirect photodegradation of e.g. litter or organic material (King et al., 2012). In the direct photodegradation, a molecule (e.g. lignin) has absorbed radiation and undergoes direct changes such as fragmentation, intramolecular rearrangement or electron transfer from or to the molecule (King et al., 2012). In the indirect photodegradation, certain photosensitizers absorb the incoming radiation and transfer the energy to other molecules such as triplet oxygen, forming reactive intermediates such as singlet oxygen, hydroxyl radical or hydrogen peroxide, which can further change the chemistry of another non-light-absorbing molecule (e.g. cellulose) or part of the same molecule where the photosensitizer resided (King et al., 2012). Indirect photodegradation may also refer to radiation-induced stimulation of microbial degradation through breaking down organic compounds making them easily available for microbial degradation (see King et al., 2012). Thermal degradation is identified as the temperature-dependent degradation of carbon in the absence of radiation and possibly oxygen (Derendorp et al., 2011; Lee et al., 2012; van Asperen et al., 2015). The separation between CO formation through thermal degradation and photodegradation is very challenging because they can both take place simultaneously and the indirect photodegradation may occur even in the absence of solar radiation if adequate thermal energy is present (Lee et al., 2012). Understanding of the biological processes leading to CO release and the importance of these sources in terrestrial ecosystems are poorly understood (Moxley and Smith, 1998; King and Crosby, 2002; Vreman et al., 2011; He and He, 2014). Formation of CO from living green plants under illumination and the presence of oxygen was already found in the late 1950s by Wilks (1959) and Siegel et al. (1962). More recently, CO has been found to be formed e.g. in plant roots (King and Crosby, 2002), in stressed plants (He and He, 2014), during heme oxidation (Engel et al., 1972; Vreman et al., 2011), in aromatic amino acid degradation processes (Hino and Tauchi, 1987) and in lipid peroxidation reactions (Wolff and Bidlack, 1976). However, the importance of these biological CO forming processes in the net CO exchange and, in general, to the global CO budget still remain largely unknown (King and Crosby, 2002). Most of the reported CO flux measurements are either short-term field experiments (e.g. Conrad and Seiler, 1985a; Funk et al., 1994; Zepp et al., 1997; Kuhlbusch et al., 1998; Moxley and Smith, 1998; Schade et al., 1999; Varella et al., 2004; Bruhn et al., 2013; van Asperen et al., 2015) or laboratory incubations with specific treatments of the soil or plant material (Tarr et al., 1995; King and Crosby, 2002; Lee et al., 2012). Both CO uptake and emissions are reported from soil– plant systems in different climatic regions, and mostly the CO fluxes range between −2 and 2 nmol m−2s−1(Conrad et al., 1988; Funk et al., 1994; Zepp et al., 1997; Moxley and Smith, 1998; Schade et al., 1999; King, 2000; King and Hungria, 2002; Varella et al., 2004; Galbally et al., 2010). Based on the available literature, there is a tendency of south-tonorth gradient with higher CO emissions from tropical and Mediterranean environments compared to boreal and temperate ecosystems (e.g. Zepp et al., 1997; Kuhlbusch et al., 1998; King, 2000; Varella et al., 2004; Galbally et al., 2010; Constant et al., 2008; Bruhn et al., 2013; van Asperen et al., 2015). However, the high variation between CO uptake and emission rates does not yet allow us to classify the ecosystem types or climatic regions. Tall tower (Andreae et al., 2015) and airborne measurements have indicated source areas of CO both in the Amazon basin (Harriss et al., 1990) and in the North American tundra (Ritter et al., 1992, 1994) suggesting a connection between high plant biomass and biological CO forming processes. To our understanding this is the first study to report long-term and continuous field measurements of CO fluxes (FCO)using the micrometeorological eddy covariance (EC) method. We measured FCO above a boreal perennial grassland ecosystem, reed canary grass (RCG), over a 7-month snow-free period in 2011 using two parallel laser absorption spectrometers. We compared the FCO with simultaneously measured fluxes of carbon dioxide (CO2), net ecosystem exchange of CO2(NEE), nitrous oxide (N2O), heat and energy as well as with relevant soil, plant and meteorological variables. Based on previous studies, we expect that the diurnal and seasonal variations in FCO are strongly dependent on radiation and temperature. On the other hand, we do not expect strong relationships between FCO and NEE or between FCO and N2O fluxes due to the limited information available Biogeosciences, 13, 5471–5485, 2016 www.biogeosciences.net/13/5471/2016/ M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes 5473 on the involvement of biological processes in FCO and challenges in separating parallel abiotic and biotic drivers of FCO. We hypothesize that a negative correlation between FCO and NEE can indicate an involvement of a biological component in CO production, and that a positive correlation between night-time FCO and N2O flux may indicate an involvement of nitrifiers in CO consumption. 2 Materials and methods 2.1 Measurement site The measurements were conducted on a mineral agricultural field located in eastern Finland (63◦9048.6900 N, 27◦1403.2900 E), cultivated with a perennial reed canary grass (Phalaris arundinaceae, L. cv. Palaton). The measurements covered a period from snowmelt to the new snowfall, from April to November 2011. Long-term (reference period 1981– 2010) annual mean air temperature in the region is 3.2◦C and annual precipitation is 612 mm (Pirinen et al., 2012). The crop was cultivated at the beginning of June 2009. In 2011 at the beginning of the growing season (23 May, day 143), the crop was fertilized with an NPKS fertilizer containing 76 kg N ha−1(NO3-N : NH4-N =47 :53). The crop from the previous season was kept at the site over the winter (Burvall, 1997) and was harvested on 28 April (day 118; Lind et al., 2016). The spring and early summer (days 118– 160) was characterized by fast-growing crop with the crop height increasing from about 10 cm in mid-May to 1.7 m in late June (day 180), reaching the maximum height of 1.9 m in early July. The field was 6.3ha in size and from the sampling location of the EC measurement system the footprint was homogenous in all directions, extending 162, 137, 135 and 178 m to N, E, S and W respectively. There is a slight south-to-north slope in the field and the wettest area lies in the northern corner of the footprint, which often has standing water during the period of snowmelt (April). The soil at the site is classified as a Haplic Cambisol/Regosol (Hypereutric, Siltic; IUSS Working Group WRB, 2007) and the texture of the topsoil (0–28 cm) varied from clay loam to loam based on the US Department of Agriculture (USDA) textural classification system. Within the ploughing layer from the surface to about 30cm, soil pH varies from 5.4 to 6.1, and soil organic matter content varied between 3 and 11 % respectively. The average C /N ratio in the ploughing layer was 14.9 (ranging from 14.1 to 15.7). We performed footprint analysis in order to identify the source area of the flux measurements. Two limiting cases were analysed: first, a low crop representing the beginning of the campaign, and second, a canopy with 1.9m in height representing the RCG canopy after midsummer. The measurement heights 2.2 and 2.4 m were used in the analysis. In the first case, we represented the low canopy as the surface with aerodynamic roughness 0.04 m (determined from measurements), in the second case, a canopy with leaf area distribution characteristic to RCG crops was represented by a beta distribution. In both cases the sources were assumed at the soil surface. Such an assumption was made due to limited information on source–sink behaviour (see Sect. 3 below) and also in order to obtain more conservative footprint estimates. Three stability classes representing unstable (the Obukhov length L= −10m), near-neutral (L= −100 m) and stable (L= +10 m) conditions were considered. The footprint evaluation was performed by using the Lagrangian stochastic trajectory simulations (e.g. Rannik et al., 2003). The upwind distances contributing 80 % of the flux were identified for low/high canopy as follows: 53/23 m, 83/34 m and 166/60 m for unstable, near-neutral and stable stratifications respectively. The conducted footprint analysis reveals that the presence of a canopy significantly reduces the footprint extent. Note that the conservative footprint scenario with no canopy is applicable only for a short period of time due to fast canopy growth at the beginning of the campaign (see Fig. 1d). Considering that prevailing wind direction during the measurement period was from SE and SSW directions, and the wind direction interval 110–315◦contributed 90 % of the half-hour periods used in the analysis, the footprint analysis confirms that the footprint was sufficient and the measurements well represent the RCG canopy. 2.2 CO flux measurements The EC measurements were made as a part of the ICOS (Integrated Carbon Observation System) Finland programme during April to November 2011. Here we report the results of FCO calculated from the concentration measurements using two continuous-wave quantum cascade lasers: ARCW-QCL (model CW-TILDAS-CS Aerodyne Research Inc., see e.g. Zahniser et al., 2009) and LGR-CW-QCL (model N2O/CO-23d, Los Gatos Research Inc., see e.g. Provencal et al., 2005). The measurements by AR-CW-QCL extended the whole measurement period from April to November 2011 (days 110–325), whereas for LGR-CQ-QCL data are available from late summer to the end of the measurement period (days 206–330). Fluxes from the two analysers are compared; however, due to the longer data coverage, the diurnal and seasonal variation in FCO is assessed using data from AR-CW-QCL only. The AR-CW-QCL and LGR-CQ-QCL were the same as used in the study by Rannik et al. (2015) wherein four laser-based fast-response gas analysers used to measure nitrous oxide (N2O) fluxes were compared. The measurement height was 2.2m until 30 June 2011 (day 181) when the height was raised to 2.4 m due to the growth of RCG. The gas inlets of the closed-path analysers were located 10 cm below a sonic anemometer (USA-1, Metek Germany GMBH) used for measuring turbulent wind components. In addition, CO2and H2O fluxes were measured at the site with an infrared gas analyser (LI7000 – LiCor Inc., Lincoln, NE, USA) connected to a sonic anemomewww.biogeosciences.net/13/5471/2016/ Biogeosciences, 13, 5471–5485, 2016 5474 M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes ter (R3-50, Gill Solent Ltd., UK). The closed-path gas analysers were located in an air conditioned cabin at about 15 m east from the air inlet and the anemometers. This wind direction (50–110◦sector) was therefore discarded from further analysis due to possible disturbances to flux measurements. Sample lines (PTFE) were shielded and heated slightly above ambient air temperature. Sample lines were 16 m in length, their inner diameters were 4 and 8 mm, the sample airflow rates were 13.2 and 11.6 LPM (Rannik et al., 2015). Based on material testing with LGR-CW-QCL, the PTFE tubing was found inert with respect to CO in a constant-flow set-up and flow rate of 2.5 LPM (unpublished data). The EC measurements were sampled at 10 Hz frequency. Further details on the EC set-up, instrument specifications and data acquisition, can be found in Rannik et al. (2015) and Lind et al. (2016). 2.3 Supporting measurements A weather station located at the site monitored continuously several meteorological and soil parameters such as air temperature (Tair)and relative humidity (RH; model: HMP45C, Vaisala Inc.), precipitation (Pr; model: 52203, R.M. Young Company), global (Rglob)and net radiation (Rnet; model: CNR1, Kipp & Zonen B.V.), photosynthetically active radiation (PAR, model: SKP215, Skye instruments Ltd.), soil heat flux at 7.5 cm depth (G; model: HPF01SC, Hukseflux), soil temperatures at 2.5, 5, 10, 20 and 30 cm depths (Tsoil; model: 107, Campbell Scientific Inc.) and soil water content at 2.5, 5, 10 and 30 cm depths (SWC) (model: CS616, Campbell Scientific Inc.). All meteorological data were recorded as 30 min mean values and stored using a data logger (model: CR 3000, Campbell Scientific Inc.). Leaf area index (LAI) was measured at approximately weekly intervals during the main crop growth period using a plant canopy analyser (model: LAI-2000, LiCor). Green area index (GAI) was estimated on a weekly basis from plots adjacent to the LAI measurements according to Wilson et al. (2007) and Lind et al. (2016). The GAI measurements were conducted from three locations (1 ×1 m2)and within each from three spots (8 ×8 cm2)by counting a number of green stems (Sn)and green leaves (Ln)per unit area and measuring the green area of leaves (La)and stems (Sa). The GAI was calculated as GAI =(SnSa)+(LnLa). 2.4 Data processing and analysis The EC data processing was performed with post-processing software EddyUH (Mammarella et al., 2016). Filtering to eliminate spikes (Vickers and Mahrt, 1997) was performed according to an approach, where the high-frequency EC data were despiked by comparing two adjacent measurements. If the difference between two adjacent concentration measurements of CO was greater than 20ppb, the following point was replaced with the same value as the previous point. The spectroscopic correction due to water vapour impact on the absorption line shape was accounted for along with the dilution correction. LGR-CW-QCL automatically corrected the water vapour effect using a built-in module in the LGR data acquisition software. The same spectroscopic correction was applied to AR-CW-QCL after a software update in July 2011. Prior to this software update, the respective dilution and spectroscopic corrections to AR-CW-QCL highfrequency CO mole fraction data were performed during the post-processing phase according to Rannik et al. (2015) with the instrument-specific CO spectroscopic coefficient (b=0.28) determined in the field. Prior to calculating the turbulent fluxes, a 2-D rotation (mean lateral and vertical wind equal to zero) of sonic anemometer wind components was calculated according to Kaimal and Finnigan (1994) and all variables were linearly detrended. The EC fluxes were calculated as 30min covariances between the scalars and vertical wind velocity following commonly accepted procedures (e.g. Aubinet et al., 2000). Time lag between the concentration and vertical wind speed measurements induced by the sampling lines was determined by maximizing the covariance. Due to the larger inner diameter (8 mm) of the sampling line in LGR-CW-QCL, the resulting lag time was 4.2 s compared to that of 0.91 s for AR-CW-QCL with the sampling line inner diameter of 4 mm. The final processing was, however, done by fixing the time lag to avoid unphysical variation of lag occurring due to random flux errors. Spectral corrections were applied to account for the low and high-frequency attenuation of the covariance. The first-order response times of the EC systems were determined to be 0.07 and 0.26 s for the AR-CWQCL and LGR-CW-QCL systems respectively, following the method by Mammarella et al. (2009). This resulted in different flux correction factors mainly due to tube damping: for AR-CW-QCL the 5 and 95 percentile values of flux underestimation were 2.1 and 12.2 % and for LGR-CW-QCL they were 5.7 and 21.4 %. Data quality screening was performed according to Vickers and Mahrt (1997) to ensure exclusion of the system malfunctioning as well as unphysical and/or unusual occasions in measurements. We chose to perform tests on single time series to ensure quality of measurements used in the analysis and did not use the flux stationarity test (Foken and Wichura, 1996) because the CO fluxes are frequently small and have large relative random errors. In such cases the tests based on relative errors are not expected to perform well (e.g. Rannik et al., 2003). After quality screening, 66.0 % of the FCO data (AR-CW-QCL) were available, with data coverage of 59.2 % during the daytime and 75.9 % during the night-time. For details of the data processing and quality screening, see Rannik et al. (2015). To evaluate in detail the seasonal changes in FCO and factors affecting the fluxes, the data were divided into six periods (days 110–145 (20 April–25 May) are spring (S), days 146–160 (25 May–9 June) are early summer (ES), days 161– 181 (10–30 June) are midsummer (MS), days 205–240 (24 Biogeosciences, 13, 5471–5485, 2016 www.biogeosciences.net/13/5471/2016/ M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes 5475 Table 1. Mean, median and 25–75th percentiles of the CO fluxes (FCO, nmol m−2s−1)measured in a read canary grass (RCG) crop at Maaninka. The fluxes are separately calculated for daytime (FCO_day, sun elevation, hsun > 0) and night-time (FCO_night,hsun < 0) and as a net flux over all FCO data (net FCO)for the six measurement periods (S is spring, ES is early summer, MS is midsummer, LS is late summer, A is autumn, LA is late autumn) and over the full measurement period (All) from April to November 2011. FCO_day FCO_night net FCO Period, days mean median 25–75th mean median 25–75th mean median 25–75th percentile percentile percentile S, 110–145 0.97 0.68 −0.15 2.00 −0.64 −0.56 −0.97 −0.20 0.41 0.09 −0.57 1.28 ES, 146–160 0.24 0.08 −0.29 0.57 −0.67 −0.49 −0.72 −0.33 0.03 −0.10 −0.45 0.43 MS, 161–181 −0.07 −0.08 −0.40 0.24 −0.67 −0.52 −0.86 −0.22 −0.22 −0.18 −0.55 0.16 LS, 205–240 0.36 0.30 −0.07 0.87 −0.76 −0.49 −0.96 −0.19 −0.09 −0.04 −0.53 0.49 A, 241–295 −0.12 −0.18 −0.48 0.13 −0.66 −0.61 −0.90 −0.32 −0.44 −0.44 −0.77 −0.10 LA, 296–325 −0.62 −0.59 −0.94 −0.26 −1.05 −1.01 −1.37 −0.65 −0.92 −0.89 −1.25 −0.49 All, 110–325 0.21 0.01 −0.41 0.55 −0.77 −0.66 −1.06 −0.33 −0.25 −0.34 −0.79 0.17 July–28 August) are late summer (LS), days 241–295 (29 August–23 October) are autumn (A) and days 296–325 (24 October–21 November) are late autumn (LA)). The division into these periods was based on seasonal changes in crop growth and development or changes in FCO and temperature, while the lengths of the periods were kept as similar in length as possible. Also, FCO were not measured during an instrumental break between days 181 and 204. To compare diurnal changes in the FCO, the data were further divided into daytime (FCO_day)and night-time (FCO_night)data. We used sun elevation angle h<0 for night-time and h> 0 for daytime. Pearson correlations between daytime and night-time halfhour average fluxes and other measured parameters were determined. Data processing was performed with Matlab version R2014a (The MathWorks, Inc., United States) and the statistical testing with IBM SPSS statistics 23 (IBM Corporation, United States). To evaluate the gross CO emission during the daytime (gross daytime CO emission), we calculated the gross daytime CO emission in two ways (1) by assuming an equivalent CO uptake for daytime and night-time (constant uptake) and (2) by taking into account temperature dependency (Q10 of 1.8) in CO uptake according to Whalen and Reeburgh (2001). Based on a constant CO uptake, the gross daytime CO emission was calculated by subtracting the night-time FCO (FCO_night)from the daytime FCO (FCO_day), presented in Table 1. The uptake CO fluxes refers to the estimated CO uptake taking place during the day, based on measured CO uptake values at night. The temperature-corrected daytime CO uptake (daytime CO uptake, Q10 1.8) is calculated by extrapolating the measured night-time CO fluxes (FCO_night; Table 1) using the difference between day and night soil temperatures (2.5 cm depth; 1tsoil)and the Q10 value of 1.8 (Whalen and Reeburgh, 2001). The temperature-dependent daytime CO uptake (R2) was solved from the equation Q10 =R2 R110 (T 2−T1), where Q10 is 1.8 (Whalen and Reeburgh, 2001), R1 is the night-time FCO (net FCO_night; nmol m−2s−1), and T2−T1 is the temperature difference between daytime (T2) and night-time (T1) soil temperature at 2.5 cm depth (◦C). The temperature-corrected gross daytime CO emissions (gross daytime CO emission; Q10 1.8) was estimated by subtracting the temperature-corrected daytime CO uptake (daytime CO uptake, Q10 1.8) from the daytime FCO (FCO_day). These gross CO emission and uptake rates were estimated for each of the six measurement periods and are presented in Table 2. 3 Results 3.1 Seasonal variation The RCG field was a net source of CO from mid-April in the spring to mid-June (days 110–160), after which the site turned to a net sink until the end of the measurement period in November 2011 (days 161–325; Fig. 1f). Cumulative CO flux (cumulative FCO)curves, calculated by cumulating the half-hourly fluxes, show that the site was a net sink of CO over the 7-month measurement period (Fig. 1f). During daytime, the net CO fluxes (FCO_day)were positive in spring and early summer (days 110–160) and again during late summer (days 205–240). These daytime emissions were highest in spring (Table 1). Night-time CO fluxes (FCO_night)were negative (CO uptake) throughout the whole measurement period with a trend of increasing CO consumption towards late autumn (Table 1). The spring emission period (days 110–145) covered a time (days 110–118) with a standing dry crop from the previous year. The old crop was harvested on 28 of April (day www.biogeosciences.net/13/5471/2016/ Biogeosciences, 13, 5471–5485, 2016 5476 M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes Table 2. Mean, median and 25–75th percentiles of the estimated gross daytime CO emission (gross daytime CO emission, nmol m−2s−1), temperature-corrected daytime CO uptake (daytime CO uptake; Q10 1.8) and temperature-corrected gross daytime CO emission (gross daytime CO emission; Q10 1.8) calculated for the read canary grass (RCG) crop at Maaninka. The CO emission and uptake rates are calculated for six measurement periods (S is spring, ES is early summer, MS is midsummer, LS is late summer, A is autumn, LA is late autumn) and over the full measurement period (all) from April to November 2011. The estimated gross daytime CO emission is calculated in two ways: (1) assuming a constant CO uptake and (2) assuming temperature-dependent CO uptake. Gross daytime CO emission based on a constant CO uptake (way 1, Sect. 2.4) refers to the difference between daytime fluxes (FCO_day)and night-time fluxes (FCO_night) presented in Table 1. The temperature-corrected gross daytime CO emission (gross daytime CO emission; Q10, 1.8) refers to the difference between daytime fluxes (FCO_day; Table 1) and daytime CO uptake (Q10, 1.8). The daytime CO uptake (daytime CO uptake; Q10, 1.8) is calculated by extrapolating the night-time CO fluxes (FCO_night)to daytime using the difference between day and night soil temperatures (2.5 cm depth; 1tsoil)and the Q10 value of 1.8 (Whalen and Reeburgh, 2001), as described in Sect. 2.4. Gross daytime CO emission 1tsoil Daytime CO uptake (Q10, 1.8) Gross daytime CO emission (Q10, 1.8) Period, DOY mean median 25th–75th Tday-Tnight mean median 25th–75th mean median 25th–75th percentile percentile percentile S, 110–145 1.61 1.24 0.83 2.20 2.1 −1.24 −1.09 −1.89 −0.39 2.22 1.76 1.74 2.39 ES, 145–160 0.91 0.57 0.43 0.91 1.2 −1.27 −0.92 −1.36 −0.63 1.51 1.00 1.06 1.20 MS, 160–181 0.59 0.45 0.46 0.46 0.7 −1.23 −0.96 −1.58 −0.41 1.15 0.89 1.18 0.65 LS, 205–240 1.12 0.79 0.89 1.07 0.9 −1.42 −0.91 −1.78 −0.36 1.77 1.21 1.71 1.24 A, 240–295 0.54 0.42 0.41 0.45 1.0 −1.24 −1.13 −1.68 −0.59 1.11 0.95 1.19 0.72 LA, 295–325 0.42 0.42 0.43 0.39 0.3 −1.90 −1.84 −2.49 −1.18 1.28 1.25 1.56 0.92 ALL, 110–325 0.98 0.68 0.65 0.88 3.5 −1.58 −1.37 −2.19 −0.68 1.79 1.38 1.78 1.23 118), after which the ground consisted mainly of short dead plant material and litter and a slowly sprouting new RCG. The second emission period in early summer (days 146–160) was characterized by fast-growing RCG crop, high fertilizerinduced N2O emissions (Shurpali et al., 2016), increasing air and soil temperatures, growing leaf area and increasing NEE (Fig. 1). After the crop had reached its maximum height of 1.9 m in mid-June (around day 160), the site started to act as a net sink of CO, followed by a period of net daytime emissions during late summer in July-August (days 205–240). The autumn (A, LA) was characterized by decreasing daytime FCO (FCO_day)and slowly dropping air and soil temperatures, decreasing radiation intensity and decreasing photosynthetic activity of the crop (less negative NEE; Fig. 1). Comparison of the two gas analysers, AR-CW-QCL and LGR-CW-QCL, during the period when both were operational (days 205–325), shows that the measured FCO agree reasonably well (Fig. 1f). A correlation scatter plot of the FCO from LGR-CW-QCL against FCO of AR-CW-QCL results in a correlation coefficient of 0.95 and a slope of 0.96 (data not shown). According to this comparison, LGR-CWQCL shows slightly (4 %) smaller fluxes compared to ARCW-QCL; however, the difference between the two analysers is very small, giving us confidence in the use of either analyser in further analysis. 3.2 Diurnal variation The FCO had a distinct diurnal pattern with an uptake in the night-time and an emission during the daytime with maximum emissions at noon (Fig. 2). This pattern was most pronounced during the spring, on days 110–145, when the maximum daytime CO emissions reached 2.7 nmol m−2s−1 (Fig. 2). The net FCO was positive (emission) in spring and early summer, after which the night-time uptake dominated, making the site a net sink of CO (Fig. 2, Table 1.). Night-time FCO show a near constant uptake of CO over the whole measurement period with a mean of −0.77 nmol m−2s−1over the whole measurement period (Fig. 2, Table 1.). The diurnal FCO over the six measurement periods closely followed the daily pattern of Rglob with a maximum FCO (emission) at around noon and minimum FCO (highest uptake) at midnight (Figs. 2 and 3). The highest radiation intensity was reached during the early summer (days 146–160), while the maximum FCO were observed in spring (days 110– 145; Figs. 2 and 3). Diurnal variation in soil temperature was highest in spring and early summer and always peaked during the afternoon (Fig. 3). Compared to the FCO, the diurnal variation in CO2exchange, expressed here as NEE, was very small during spring (days 110–145; Fig. 4). A rapid increase in LAI and GAI at around day 150 (Fig. 1d) led to an increase in CO2uptake during daytime, which is seen in a distinct diurnal pattern with high CO2uptake (negative NEE) during daytime and a small positive NEE during night-time (Fig. 4). Maximum NEE values were reached during mid-June (days 161–181) after which the NEE slowly decreased and the CO2uptake disappeared by mid-October (day 290; Figs. 1 and 4). During early summer, the fluxes of N2O followed a similar daily pattern as that of FCO with higher daytime N2O emissions compared to night-time fluxes (Shurpali et al., 2016). This period of high N2O emissions (days 143–158) was a direct response to the NPKS fertilizer application on 23 May, and it lasted for about 15 days. After this, an opposite diurnal Biogeosciences, 13, 5471–5485, 2016 www.biogeosciences.net/13/5471/2016/ M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes 5477 Table 3. Pearson correlation matrix for half-hour daytime CO fluxes (FCO_day)during six periods (S is spring, ES is early summer, MS is midsummer, LS is late summer, A is autumn, LA is late autumn) at the reed canary grass crop in Maaninka. MCO is CO mixing ratio, NEE is net ecosystem exchange, RESP is ecosystem respiration, FN2Ois N2O flux, His sensible heat flux, LE is latent heat flux, Tair is air temperature, Rglob is global radiation, Rnet is net radiation, Gis soil heat flux, Tsoil is soil temperature at 2.5 cm, SWC is soil water content at 2.5 cm. FCO_day FCO_day FCO_day FCO_day FCO_day FCO_day S, 110–145 nES, 146–160 nMS, 161–180 nLS, 205–240 nA, 241–295 nLA, 296–325 n MCO 0.080∗711 0.128∗∗ 510 −0.116∗436 −0.074 488 0.038 851 −0.284∗∗ 288 NEE −0.188∗∗ 711 −0.469∗∗ 510 −0.308∗∗ 436 −0.488∗∗ 488 −0.237∗∗ 850 −0.25∗∗ 288 RESP 0.015 711 0.274∗∗ 510 0.272∗∗ 436 0.257∗∗ 488 0.198∗∗ 850 0.077 288 FN2O−0.219∗∗ 669 0.000 453 −0.293∗∗ 426 −0.026 478 −0.085∗850 −0.172∗∗ 287 H0.729∗∗ 711 0.329∗∗ 510 0.234∗∗ 436 0.427∗∗ 488 0.132∗∗ 851 −0.076 288 LE 0.402∗∗ 418 0.398∗∗ 401 0.514∗∗ 224 0.625∗∗ 307 0.317∗∗ 573 0.289∗∗ 185 RH −0.537∗∗ 711 −0.176∗∗ 510 −0.303∗∗ 436 −0.434∗∗ 488 −0.081∗851 −0.179∗∗ 288 Tair 0.425∗∗ 711 0.344∗∗ 510 0.36∗∗ 436 0.433∗∗ 488 0.241∗∗ 851 0.073 288 Rglob 0.760∗∗ 711 0.498∗∗ 510 0.373∗∗ 436 0.549∗∗ 488 0.265∗∗ 851 0.256∗∗ 288 Rnet 0.760∗∗ 711 0.515∗∗ 510 0.376∗∗ 436 0.558∗∗ 488 0.277∗∗ 851 0.218∗∗ 288 G0.575∗∗ 711 0.473∗∗ 510 0.406∗∗ 436 0.485∗∗ 488 0.247∗∗ 851 0.033 288 Tsoil 0.191∗∗ 711 0.282∗∗ 510 0.318∗∗ 436 0.358∗∗ 488 0.206∗∗ 851 0.071 288 SWC −0.099∗∗ 711 0.033 510 0.095∗436 0.086 488 −0.105∗∗ 851 0.095 288 ∗∗ Correlation is significant at the 0.01 level (2-tailed). ∗Correlation is significant at the 0.05 level (2-tailed). pattern was observed during which the N2O emissions were on average 50% higher during the night than during the day (Shurpali et al., 2016). The gross daytime CO emissions were estimated in two ways: (1) assuming an equal CO uptake during day and night (constant uptake) and (2) accounting for temperature-dependent CO uptake according to Whalen and Reeburgh (2001). The gross CO emissions calculated in either way, show that in the daytime the site emitted CO throughout the whole measurement period with the highest emissions in spring and late summer (Table 2). During midsummer and autumn the daytime emissions were markedly smaller and less than half of the emissions in spring. The smallest gross CO emissions were measured in late autumn (Table 2). When the temperature dependency in the CO uptake was taken into account, using a Q10 value of 1.8 (Whalen and Reeburgh, 2001), both the daytime CO uptake (daytime CO uptake, Q10, 1.8), and the daytime emission (daytime CO emission, Q10, 1.8) were almost twice as high as the rates without the temperature correction (Table 2). 3.3 Driving factors for CO fluxes The most pronounced relationships between FCO and other measured scalars were found for the daytime data (sun elevation h> 0) during the two emission periods in the spring and early summer (Table 3, Fig. 5). Furthermore, the strongest correlations were found in spring between FCO_day and Rglob (r=0.760, p< 0.01), Rnet (r=0.760, p< 0.01), H (r=0.729, p< 0.01) and G(r=0.575, p< 0.01). These positive correlations remained significant but became weaker towards the end of the measurement period (Table 3, Fig. 5). Strong negative correlations were found in spring between FCO_day and RH (r= −0.537, p< 0.01), and during the early summer with NEE (r= −0.469, p< 0.01), while the correlation between daytime FCO and MCO,FN2Oor ecosystem respiration (RESP) were very weak throughout the 7month measurement period (Table 3). Night-time (h< 0) FCO (FCO_night)correlated weakly with FN2O(r= −0.336, p< 0.01), H(r=0.315, p< 0.01), and LE (r=-0.241, p< 0.05) in the spring and with SWC (r=0.308, p< 0.01) during early summer (Table 4). A strong negative correlation was found between FCO_night and FN2Oduring midsummer (r= −0.607, p< 0.01) and late autumn (r= −0.514, p< 0.01) and a positive correlation was found between FCO_night and LE (r=0.459, p< 0.05) during midsummer (Table 4). 4 Discussion Based on the 7-month EC flux measurements at the RCG crop, we demonstrate that the EC method is suitable for measuring CO fluxes (FCO)from a perennial agricultural crop. We show that the soil–plant system acted as a net source of CO in spring and early summer and a net sink of CO over the late summer and autumn, and that the FCO had a clear diurnal pattern, with net CO emissions in the daytime and net CO uptake at night. This source–sink pattern existed over the whole measurement period with decreasing net emissions towards the end of the autumn. To our knowledge, similar longterm and continuous FCO data series measured by the EC method over any ecosystem type does not exist, hence this study is unique in bringing new insight to the understanding of short-term diurnal and long-term seasonal FCO dynamics at ecosystem level. Combining the continuous FCO data www.biogeosciences.net/13/5471/2016/ Biogeosciences, 13, 5471–5485, 2016 5478 M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes Table 4. Pearson correlation matrix for half-hour night-time CO fluxes (FCO_night)during six periods (S is spring, ES is early summer, MS is midsummer, LS is late summer, A is autumn, LA is late autumn) at the reed canary grass crop in Maaninka. MCO is CO mixing ratio, NEE is net ecosystem exchange, RESP is ecosystem respiration, FN2Ois N2O flux, His sensible heat flux, LE is latent heat flux, Tair is air temperature, Rglob is global radiation, Rnet is net radiation, Gis soil heat flux, Tsoil is soil temperature at 2.5 cm, SWC is soil water content at 2.5 cm. FCO_night FCO_night FCO_night FCO_night FCO_night FCO_night S, 110–145 nES, 146–160 nMS, 161–180 nLS, 205–240 nA, 241–295 nLA, 296–325 n MCO −0.045 380 −0.043 142 −0.279∗∗ 134 −0.165∗∗ 324 −0.110∗∗ 1149 −0.041 700 NEE 0.069 380 −0.167∗142 −0.118 134 −0.049 324 0.024∗∗ 1149 0.025 700 RESP 0.056 380 0.015 142 −0.006∗∗ 134 0.125∗∗ 324 0.062∗1149 0.072 700 FN2O−0.336∗∗ 350 0.034 120 −0.607∗∗ 126 −0.197∗∗ 307 0.009 1140 −0.514∗∗ 696 H0.315∗∗ 380 0.170∗142 0.002 134 0.051 324 −0.021∗∗ 1149 0.080∗700 LE −0.241∗74 0.099 72 0.459∗20 −0.078 62 0.135∗∗ 453 0.161∗∗ 279 RH 0.027 380 −0.016 142 −0.057 134 −0.12∗∗ 324 −0.033 1149 −0.041∗∗ 700 Tair 0.107∗380 −0.013 142 0.092 134 0.249∗∗ 324 0.138∗∗ 1149 0.098∗∗ 700 Rglob 0.077 380 0.118 142 −0.096 134 −0.02 324 −0.001 1149 −0.041∗∗ 700 Rnet 0.011 380 0.111 142 0.026 134 0.087 324 0.043 1149 −0.053∗∗ 700 G0.050 380 0.029 142 0.121 134 0.207∗∗ 324 0.175∗∗ 1149 0.162∗∗ 700 Tsoil 0.075 380 −0.146 142 −0.035 134 0.167∗∗ 324 0.038 1149 0.117∗∗ 700 SWC 0.043 380 0.308∗∗ 142 0.212 134 0.138∗324 0.093∗∗ 1149 0.008 700 ∗∗ Correlation is significant at the 0.01 level (2-tailed). ∗Correlation is significant at the 0.05 level (2-tailed). Table 5. Reported CO fluxes measured in different ecosystems and climatic regions, using chambers (transparent or dark), micrometeorological flux gradient or eddy covariance methods and the reported data period, measurement frequency and the moment of the measurements. Reference Ecosystem, climate, country Measurement method Data period, FCO measurement frequency, (nmol m−2s−1) moment of measurement Zepp et al. (1997) Black spruce forest, boreal, Manitoba, Canada Chambers, transparent 3 months, weekly, daytime −1.06 Zepp et al. (1997) Jack pine forest, boreal, Manitoba, Canada Chambers, transparent 3 months, weekly, daytime −0.58 King (2000) Pine forest, North-east, Walpole, Maine, USA Chambers, dark 1.3 years, biweekly, daytime 1.12 King (2000) Mixed hardwood-coniferous forest, Walpole, Maine, USA Chambers, dark 1.3 years, biweekly, daytime 0.62 King (2000) Pine forest, Griffin, Georgia, USA Chambers, dark 1 year, bimonthly, daytime −0.21 King (2000) Pine forest, Tifton, Georgia, USA Chambers, dark 1 year, bimonthly, daytime −0.95 Kuhlbusch et al. (1998) Black spruce, boreal, Manitoba, Canada Chambers, dark 1 year, bimonthly, daytime −1.11 Galbally et al. (2010) Mallee, Eucalyptus sp. Ecosystem, tropical, Australia Chambers, transparent 1 year, bimonthly, daytime 0.61 Kisselle et al. (2002) Cerrado, campo sujo, tropical, Brazil Chambers, transparent 1 year, monthly, daytime 3.16 Kisselle et al. (2002) Cerrado, stricto sensu, tropical, Brazil Chambers, transparent 1 year, monthly, daytime 2.66 Varella et al. (2004) Natural cerrado, tropical, Brazil Chambers, transparent 1.5 years, monthly, daytime 1.91 Varella et al. (2004) Pasture (Brachiaria brizantha), tropical, Brazil Chambers, transparent 1.5 years, monthly, daytime 1.20 King (2000) Cropland, corn, Walpole, Maine, USA Chambers, dark 1.3 years, biweekly, daytime 2.19 King (2000) Cropland, sorghum/wheat, Griffin, Georgia, USA Chambers, dark 1 year, bimonthly, daytime 1.16 King (2000) Cropland, cotton/peanuts/winter wheat, Tifton, Georgia, USA Chambers, dark 1 year, bimonthly, daytime 1.03 Galbally et al. (2010) Cropland, wheat, tropical, Australia Chambers, transparent 1 year, bimonthly, daytime 0.98 Constant et al. (2008) Grassland, boreal, Quebec, Canada Flux gradient 1 year, diurnal cycle −2.11 Bruhn et al. (2013) Grassland, temperate, Denmark Chambers, dark 2 months, monthly, daytime −0.78 Bruhn et al. (2013) Grassland, temperate, Denmark Chambers, transparent 2 months, monthly, daytime 0.36 van Asperen et al. (2015) Grassland, Mediterranean, Italy Chambers, transparent 5 weeks, summer, diurnal cycle 0.35 van Asperen et al. (2015) Grassland, Mediterranean, Italy Flux gradient 1 month, 30 min, diurnal cycle 1.74 this study Grassland, reed canary grass, boreal, Finland Eddy covariance 7 months, 30 min, diurnal cycle −0.25 with simultaneously measured CO2, N2O and energy fluxes as well as meteorological and soil variables allowed us to distinguish driving variables of the FCO and demonstrate the suitability of the EC method to analyse ecosystem-level CO exchange dynamics. Due to the fact that the EC method measures net fluxes, we cannot directly separate different processes such as CO production and consumption. However, based on process understanding and our data, we made an assumption that most of the CO production takes place during daytime and that the night-time CO uptake is due to microbial activity. After these assumptions, we divided the data into daytime and night-time periods in order to analyse seasonal changes in dependencies between CO emissions and uptake and their driving variables. Cumulative FCO over the whole 7-month measurement period showed that the RCG crop was a net sink of CO. This cumulative FCO estimation may be biased due to the instrumental break during July (days 181–205), during which we do not have an estimate of the CO fluxes. Also, due to the fact that the data processing removed more daytime values (40.8 % reBiogeosciences, 13, 5471–5485, 2016 www.biogeosciences.net/13/5471/2016/ M. Pihlatie et al.: Seasonal and diurnal variation in CO fluxes 5479 Figure 1. (a) Daily mean air and soil temperatures, (b) global radiation sum (Rglob),(c) daily precipitation sum (Pr)and soil water content (SWC), (d) weekly leaf area index (LAI; black) and green area index (GAI; grey), (e) net ecosystem exchange of CO2(NEE), and (f) cumulative CO fluxes calculated from half-hour mean CO fluxes (cumulative FCO; black lines) and daytime mean CO fluxes (FCO_day; grey) over the 7-month measurement period in a reed canary grass crop. Measurement periods (S is spring, ES is early summer, MS is midsummer, LS is late summer, A is autumn, LA is late autumn) are separated by solid lines. moved) compared to night-time data (24.1 % removed), the night-time CO uptake is weighing more in the cumulative flux estimation, potentially leading to smaller and more negative net fluxes than estimated based on an equal number of flux data from daytime and night-time. We tested a simple statistical gap-filling method to obtain a balanced number of daytime and night-time data. However, as this gap-filling did not change the interpretation of the results, and as we do not have an appropriate process model to account for uptake and emission processes, we decided not to present these results. Based on seasonal variation, we can divide the FCO into a distinct emission period and an uptake period. During the emission period (days 110–160), the soil–plant system was a strong source of CO during the daytime and a small sink during night-time. Furthermore, the emission period was diTime (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4S, 110–145 Time (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4ES, 146–160 Time (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4MS, 161–181 Time (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4LS, 205–240 Time (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4A, 241–295 Time (h) 0 6 12 18 24 FCO (nmol m-2 s-1) -2 0 2 4LA, 296–325 Figure 2. Diurnal cycle of half-hour mean CO fluxes (FCO, nmol m−2s−1)from the reed canary grass crop from six distinct periods during the April to November 2011. Grey areas indicate the moment of sunrise and sunset, and the vertical bars indicate ±1 SD of the fluxes. vided into a spring emission period (days 110–145) and an early summer emission period (days 146–160), which differed from each other based on the daytime CO emission rates and relationships with other measured variables such as radiation and NEE. The highest CO emissions were observed soon after the snowmelt during spring from April to early May when the air and soil temperatures were rather low, the crop was not yet actively photosynthesizing (low LAI, low NEE) and radiation intensity was already rather high. As suggested by King (2000), the elevated springtime CO emissions probably resulted from the degradation of the last year’s readily available crop and litter, which have been shown to be a significant source of CO (King, 2000; King et al., 2012; Lee et al., 2012). Decreasing amounts of this readily degradable litter also partly explains the decreasing trend in CO emissions during spring and early summer (King, 2000). In general, the FCO rates from the RCG crop in this study fall into the same range as those reported from different natural and managed ecosystems across the different climatic regions (Table 5). There is a tendency of higher CO emissions from tropical and Mediterranean ecosystems compared to northern and boreal ecosystems. The data comparison also indicates net CO uptake from forest ecosystems (Zepp et www.biogeosciences.net/13/5471/2016/ Biogeosciences, 13, 5471–5485, 2016