Pinus sylvestris as a missing source of nitrous oxide and methane in boreal forest

Boreal forests comprise 73% of the world’s coniferous forests. Based on forest floor measurements, they have been considered a significant natural sink of methane (CH4) and a natural source of nitrous oxide (N2O), both of which are important greenhouse gases. However, the role of trees, especially conifers, in ecosystem N2O and CH4 exchange is only poorly understood. We show for the first time that mature Scots pine (Pinus sylvestris L.) trees consistently emit N2O and CH4 from both stems and shoots. The shoot fluxes of N2O and CH4 exceeded the stem flux rates by 16 and 41 times, respectively. Moreover, higher stem N2O and CH4 fluxes were observed from wet than from dry areas of the forest. The N2O release from boreal pine forests may thus be underestimated and the uptake of CH4 may be overestimated when ecosystem flux calculations are based solely on forest floor measurements. The contribution of pine trees to the N2O and CH4 exchange of the boreal pine forest seems to increase considerably under high soil water content, thus highlighting the urgent need to include tree-emissions in greenhouse gas emission inventories.


Results and Discussion
N 2 O fluxes. In dry field conditions typical for the studied boreal forest, we observed that P. sylvestris stems and shoots emitted N 2 O at rates (medians) of 0.023 and 0.097 μg N 2 O per m 2 of stem and projected leaf area, respectively, per hour ( Supplementary Fig. S2a), accounting for 0.11 and 1.9 mg N 2 O, respectively, after scaling up per hectare of ground area per hour (see Methods, Fig. 1a). To our knowledge, measurements of shoot fluxes of N 2 O from mature trees have never been reported, and most studies assume negligible shoot emissions compared to stem fluxes 5,9,10,12,13 . Contrary to this current understanding, the shoot fluxes of N 2 O from the studied pine trees exceeded the stem fluxes by more than 16 times. This underlines the important role of forest canopies in trace gas exchange. The N 2 O fluxes from pine trees were accompanied by forest floor flux rates reaching 2.50 μg N 2 O m −2 h −1 (24.9 mg N 2 O ha −1 h −1 ; Supplementary Fig. S2a, Fig. 1a), which agrees with previous soil N 2 O measurements in the same forest 24 . In general, boreal forest soils are characterized by low availability of mineral N 23,25 and low N deposition 23 , resulting in low soil N 2 O emissions, particularly when compared to 4 to 12 times higher emissions from temperate and tropical forests 25 .
The up-scaled N 2 O emission rates from trees, assuming the mean tree constitution and density of 1000 trees per hectare in the dry plot (see Methods), were equivalent to 8.0% of the forest floor emissions per hectare of ground area (Fig. 1a, comparison of medians). Thus, the N 2 O emissions from trees constitute a significant part of the boreal pine forest N 2 O flux. The N 2 O flux from dry areas of the studied forest, including the contribution from forest floor and pine trees, reached approximately 26.9 mg N 2 O ha −1 h −1 (8.0 g CO 2 -e ha −1 h −1 using a global warming potential [GWP] of 298 [ref. 26]), which lies within the range of the global inventory estimates of N 2 O flux rates for boreal forests 25 . Based on the shoot-to-stem N 2 O fluxes ratio of 16 at the dry area, the shoot fluxes at the wet plot could reach 3.3 mg N 2 O ha −1 h −1 versus the measured stem fluxes of 0.20 mg N 2 O ha −1 h −1 . As follows, under high soil water content typical for studied wet areas of the forest with density of 1400 trees per hectare, the contribution of pine trees could be up to 18% (based on medians comparison) of the forest floor N 2 O exchange.
Naturally, the up-scaled fluxes include uncertainties stemming from e.g. spatio-temporal variability in the fluxes, the use of mean stand density and constant shoot-to-stem flux ratio from the dry plot. The use of the constant shoot-to-stem flux ratio is justified based on the assumption that transport of N 2 O via the transpiration stream is the main driver for N 2 O emissions from the tree canopy 8 , and hence the stem emissions are directly reflected in the emissions from the canopy. Forest floor N 2 O and CH 4 exchange is often characterized by high spatial variability, which has been also found to vary with distance to the trees 27,28 , while the variation in canopy N 2 O/CH 4 exchange between individual pine trees as well as between different tree species remain unknown due to lack of canopy flux measurements. We estimated that temporal variability in the shoot, stem and forest floor  Supplementary Fig. S2. Solid lines within the boxes mark medians, broken lines denote means, boundaries indicate 25th and 75th percentiles, and the whiskers 10th and 90th percentiles. Dots mark outliers. The plotted results are the medians/means of all sampling locations from the dry plot as follows: Forest floor fluxes are determined as medians and means of measurements from three soil chambers (n = 3) with nine measurement repetitions per chamber. Stem and shoot fluxes are expressed as medians and means of measurements on three trees (n = 3) with four to six repetitions per chamber. The fluxes from the shoots, stems and from the forest floor were measured simultaneously to allow their comparison. Contribution of stems and shoots to N 2 O and CH 4 exchange are expressed as percentage of the forest floor flux. Statistically significant differences at p < 0.017 (multiple comparison -Bonferroni correction) between flux components are indicated by different capital letters above bars.
Scientific RepoRts | 6:23410 | DOI: 10.1038/srep23410 N 2 O and CH 4 fluxes was higher than the spatial variability in the dry plot, whereas in the wet plot the spatial variability dominated the fluxes. This indicates that most of the variability in the fluxes in the dominating dry areas originates from day-to-day variation, whereas the fluxes in the wet areas, which form a minority of the forest, are dominated by high small-scale variation.
The pine stem N 2 O fluxes correlated positively with forest floor fluxes (Spearman's rank correlation coefficient: ρ = 0.351, p < 0.05), indicating that the tree-emitted N 2 O could originate from soil. As N 2 O is rather water soluble 4 , and many plant species emit N 2 O irrespective of the presence of an aerenchyma system 4,5,7,8,9,13 , we hypothesize that N 2 O is absorbed by roots from the soil, transported via xylem into the above-ground tree parts, and then emitted into the atmosphere. CH 4 fluxes. Contrary to the CH 4 uptake by shoots (i.e. negative flux) found in Scots pine seedlings grown under field and laboratory conditions 20 , we observed emissions of CH 4 from both shoots and stems of mature P. sylvestris. This difference in shoot CH 4 fluxes may result from (i) different soil water content and soil temperature (not reported for the seedlings experiment 20 ), (ii) known discrepancy in emission capacities of young and mature trees 12 , and (iii) the fact that the seedlings were investigated in the absence of UV radiation 20 , which is known to stimulate CH 4 formation 17,18 . The CH 4 emission rates from pine stems and shoots were 0.005 and 0.050 μg CH 4 m −2 h −1 (medians), respectively ( Supplementary Fig. S2b). Up-scaled emission rates at stand level were 0.03 and 1.1 mg CH 4 ha −1 h −1 (Fig. 1b) assuming mean tree constitution and density of 1000 trees per hectare (see Methods). As is the case of N 2 O, pine shoots seem to be the primary tree surface emitting CH 4 into the atmosphere, given that shoot fluxes were 41 times higher than the stem fluxes. This contradicts the common assumption 5,9,10,11 that basal regions of stems are the main source of CH 4 and N 2 O from trees.
Whereas trees were a source of CH 4 , the forest floor was a sink (− 14. Fig. 1b). The estimated average pine tree CH 4 emission represented 0.8% of the forest floor uptake. The CH 4 uptake from the dry area of the studied forest (− 4.9 g CO 2 -e ha −1 h −1 using GWP of 34 [ref. 26]) is roughly 35% to 50% lower than are estimates of CH 4 uptake for boreal forests in global inventories 25,29 .
The median stem CH 4 fluxes at the wet plot (0.100 μg CH 4 m −2 h −1 ) were one order of magnitude higher than those at the dry plot (0.013 μg CH 4 m −2 h −1 ) (Fig. 2b), while the soil remained a sink for CH 4 even under high soil VWC (− 7.09 μg CH 4 Fig. 2d). Moreover, the stem-to-forest-floor CH 4 fluxes ratio increased with soil VWC, underlining the importance of pine trees at wet areas in the balance of CH 4 . Although direct measurement of shoot CH 4 flux at the wet plot was technically impossible, based on the shoot-to-stem CH 4 fluxes ratio of 41 at the dry plot, the shoot CH 4 fluxes at the wet plot were estimated to reach 24 mg CH 4 ha −1 h −1 in comparison to the stem CH 4 fluxes of 0.59 mg CH 4 ha −1 h −1 . Under high soil VWC and stand density of 1400 trees per hectare, CH 4 emissions from pine trees could, therefore, account for up to 35% of the forest floor uptake. This estimate is rather higher than in a recent study by Pangala and colleagues, who found that CH 4 emissions mediated by Alnus glutinosa and Betula pubescens contribute up to 14% to the total CH 4 fluxes from a temperate forested wetland 12 .
The partial soil origin of pine-emitted CH 4 is supported by strong positive correlation of stem CH 4 fluxes with forest floor CH 4 fluxes (ρ = 0.716, p < 0.001) and VWC in topsoil (ρ = 0.802, p < 0.001). In wet conditions, pine trees may therefore prevent CH 4 consumption in the upper soil layers by transporting CH 4 , produced in deeper soil, into the atmosphere. We suggest that soil-produced CH 4 is transported into the above-ground parts of P. sylvestris mainly by the transpiration stream and then released into the atmosphere predominantly via stomata 4,5,8,9 , thus explaining the higher CH 4 emissions from shoots as compared to stems. This assumption is supported by the positive correlation between the shoot CH 4 flux and transpiration (ρ = 0.626, p < 0.05), and stem CH 4 flux and sap flow (ρ = 0.390, p < 0.01). Therefore, alternative pathways, such as radial diffusion of CH 4 (and N 2 O) in stems through intercellular spaces of the ray parenchyma and a release from the stem via lenticels 11,30,31 , seem of a lesser importance.
Different mechanisms of CH 4 emissions from trees grown on dry plot as compared to those on wet plot are, however, likely. Limited soil CH 4 production in deeper mineral soil layers 32 and low mineral soil VWC (0.28 ± 0.02 m 3 m −3 ) in the studied period give an assumption of negligible soil CH 4 production in the dry plot. Moreover, approximately half of the root system of P. sylvestris is located in the top soil organic layer with the rest of the roots equally distributed to mineral soil (0-40 cm) 33 . Therefore, it is probable that part of the CH 4 emitted from trees in the dry plot originated from anaerobic production processes within the wood 14,15,16 and/or aerobic, non-microbial metabolic processes in the plant tissues 17,18 .
P. sylvestris appears to be one of the missing sources for N 2 O and CH 4 in boreal forests. N 2 O emissions from boreal pine forests may previously have been underestimated and the uptake of CH 4 overestimated. Even though our measurements indicate only potential mechanisms, and more detailed measurements of spatio-temporal variability are necessary, the pine mediated N 2 O and CH 4 emissions could account for up to 18% of forest floor N 2 O emissions and 35% of forest floor CH 4 uptake, respectively, under high soil moisture conditions. This can be crucial for the future greenhouse gas budgets of boreal pine forests, especially if precipitation and evapotranspiration patterns will change due to climate change. Our findings highlight the important, but often neglected role of upland trees in N 2 O and CH 4 exchange between the biosphere and the atmosphere and the importance of including tree emissions to the total forest ecosystem budgets of N 2 O and CH 4 . and broadleaved trees in the understorey 23,34 . The long-term annual mean temperature and precipitation are 3.5 °C and 711 mm, respectively 35 . The soil is Haplic podzol on glacial till with irregularly distributed peat soil spots 36 .

Methods
Naturally wet and dry plots (dimensions of 20 × 15 m, a distance of 100 m apart) with mean soil volumetric water content (VWC) 0.75 ± 0.016 m 3 m −3 (mean ± standard error) and 0.33 ± 0.030 m 3 m −3 , respectively, were selected. During the measurement period, soil water content was measured using an HH2 Moisture Meter and Theta Probe (type ML2x, AT Delta-T Devices, Cambridge, UK) in A-horizon corresponding to depths 0-5 cm from the soil surface, and expressed as mean of three independent measurements close to each tree and soil chamber. Soil temperature was measured continuously by a DS1921G Maxim Thermochron iButtons (Maxim Integrated, San Jose, California, USA) in A-horizon next to each soil chamber.
On each plot, six representative trees were chosen for stem flux measurements (n = 6). Shoot fluxes were measured from the upper canopy of three trees used for the stem flux measurements at the dry plot. Shoot fluxes were not measured from the wet plot as installation of a scaffold tower was technically not possible. The mean tree height, length of living crown, and stem diameter at breast height (DBH) of the selected pine trees were 18.2 ± 0.4 m, 6.47 ± 0.32 m, and 0.162 ± 0.012 m, respectively, for the wet plot. For the dry plot, these were 17.7 ± 0.5 m, 7.22 ± 0.43 m, and 0.180 ± 0.004 m, respectively. These morphological parameters did not differ significantly when comparing wet and dry plots. The stand densities were estimated to be 1000 and 1400 trees per hectare on the dry and wet plots, respectively. The stand basal area was measured directly on the plots using a rod relascope technique and was 19.5 and 26 m 2 ha −1 on dry and wet plots, respectively. Chamber systems. The stem fluxes were measured using 12 stem chambers (1 chamber per tree) enclosing the entire stem circumference 37-modified . The skeleton of the stem chamber (volume between 0.0009 and 0.0015 m 3 depending on stem diameter) was created by a flexible pipe from polyethylene-coated aluminium (Synflex, Eaton Hydraulics Group Europe, Morges, Switzerland), which was wrapped in a spiral around the stem. A tube-fitting brace was attached to this spiral and enabled fixation of inlet and outlet connectors. Teflon FEP film (0.05 mm thick, Fluorplast, Maalahti, Finland) impermeable for CH 4 and N 2 O was wrapped 1.5 to 2 times around the tube spiral to create the chamber wall, and then sealed with adhesive FEP tape. Due to the requirement of mounting the stem chambers on the basal part of the rough pine bark (around 0.2 m above the forest floor), the surface of the dead outer bark was carefully removed from the upper and basal ends of the chamber. The upper and basal ends of the Teflon foil were sealed with elastic closed cell polyethylene foam and wide flexible ties to the carefully smoothed bark surface. The results of the stem flux measurements on twelve trees were used in the comparison of the stem and forest floor fluxes between dry and wet plots (Fig. 2).
Two different shoot chamber types were used to measure fluxes of CH 4 and N 2 O: two cylindrical chambers with FEP foil walls 38 (volume 0.0054 m 3 ) and a methacrylic cylindrical shoot chamber 39 (volume 0.005 m 3 ). We did not observe any differences in flux rates obtained by these two types of chambers. The three chambers were installed in the upper canopy of the three trees on the dry plot. The air temperatures (DT 612 thermometer, CEM, Shenzhen, China) inside and outside of the chambers were regularly measured during chamber closures. To avoid overheating in the chambers, the shoot fluxes were measured only on cloudy days. The comparison of the shoot, stem and forest floor fluxes presented in Fig. 1 and Supplementary Fig. S2 is based on the measurements at the dry plot only. To compare the whole tree flux rates in dry and wet plots, we used the shoot-to-stem flux ratio from the dry plot where both shoot and stem flux measurements were performed.
In both stem and shoot chambers, the mixing of the air inside the chambers was provided by vacuum pumps (V 1500-GAS-12V standard vacuum pumps, Xavitech, Härnösand, Sweden; NMP 850.1.2. KNDC B, KNF Neuberger, Freiburg, Germany) gas-tightly connected to the chamber using Teflon tubes and stainless steel connectors (Swagelok, Ohio, USA). The chambers were non-steady-state flow-through chambers returning the air from the pump again into the chambers. Gas samples were taken with a syringe via a septum connected to the air circulation. Six gas samples (each 20 ml) were taken from the closed stem and shoot chambers at time intervals of ca 60 min over a period of 6 h. The possible under-pressure resulted from the gas sampling was compensated by the flexible foil wall. The stem chambers were flushed with ambient air for at least 30 min before sampling.
Forest floor CH 4 and N 2 O fluxes were measured using large opaque soil chambers made of aluminium 40-chamber #13 . Three chambers were placed on the dry plot (volume of ca 0.091 m 3 depending on vegetation inside the chamber, enclosed soil surface area of 0.298 m 2 ), and three on the wet plot (volume of ca 0.133 m 3 , soil area of 0.298 m 2 ). The chambers were located in the vicinity of the measured trees. The ground vegetation in the soil chambers varied among chambers depending on the soil conditions and location, and consisted of Sphagnum sp., Polytrichum sp., Dicranum polysetum, Pleurozium schreberi, Equisetum sylvaticum, Vaccinium myrtillus, Vaccinium vitis-idaea, Trientalis europaea, and several representatives of Poaceae. The placement of chamber collars took place several days before the first sampling to allow the soil to settle and avoid soil disturbances. The soil chambers were closed for ca 40 min during which gas samples (each 20 ml) were taken at 2, 5, 10, 20, 30, and 40 minutes after the chamber closure. A fan was used to mix the headspace air during the closure. Chamber headspace temperature (DT 612 thermometer, CEM, China) was regularly monitored during the measurements.
Gas analyses. Gas samples from stem, shoot, and soil chambers were taken in 20 ml syringes (BD syringe, Franklin Lakes, New Jersey, USA) and immediately transferred to the evacuated 12 ml glass vials (Labco, Ceredigion, UK), then stored at 4 °C. The gas samples were analysed by an Agilent 7890A gas chromatograph (GC) (Agilent Technologies, Santa Clara, California, USA) equipped with a flame ionization detector (FID) and an electron capture detector (ECD) for CH 4 and N 2 O analyses, respectively 40 . Briefly, CH 4 was detected by FID (300 °C) supplied with synthetic air (450 ml min −1 ) and hydrogen (H 2 , 45 ml min −1 ) and with nitrogen (N 2 , 5 ml min −1 ) as a make-up gas. N 2 O was detected using the ECD (380 °C) supplied with argon/methane (15 ml min −1 ) as a make-up gas. Helium (He, 45 ml min −1 ) was used in both cases as a carrier gas. Columns Porapak Q 80-100 Mesh and Hayesep Q 80-100 Mesh (Agilent Technologies, USA) were used for water vapour removal and gas separation. Oven temperature was kept at 60 °C. Retention times for CH 4 where The stem surface area was estimated as a smooth cylinder around the bark because the micro-topography of the bark (very rough surface) makes any other methods ambiguous. The projected leaf area of shoots enclosed in the shoot chambers was determined by applying a destructive method at the end of the measurement campaign using an LI-3000 portable area meter (Li-Cor, Lincoln, Nebraska, USA). The flux rates of N 2 O and CH 4 were further estimated for the entire stem and projected needle area of each tree using the following parameters: The stem surface area (3.6-6.2 m 2 per tree) was calculated as the lateral surface area of a right circular cone using the stem diameter at breast height (DBH) and the tree height. The needle biomass was determined using an allometric biomass equation (based on DBH, tree height, and length of living crown) for Scots pine 41-equation n. 27 and used to calculate the entire projected needle area of each tree (10-31 m 2 per tree) by multiplying the biomass weight with specific leaf area determined for P. sylvestris at the SMEAR II station 42 . The CH 4 and N 2 O fluxes from stems, shoots, and forest floor were scaled up to 1 hectare of 50-year-old boreal pine stand using the estimated forest density and stand basal area (see chapter "Site description").
The flux estimates and upscaling to a stand level are saddled with uncertainties arising from sampling and gas analyses 43 , variables in Equation 1, and application of allometric relationships for estimation of total leaf area 41 and from calculation of stem area per tree. At a stand level, uncertainties of flux rates are thus particularly given by spatio-temporal variability in the fluxes, heterogeneity of tree morphological parameters (height, length of living crown, stem diameter), and stand heterogeneity (tree density per hectare and species composition etc.). In addition, fast changes in transpiration and sap flow rates induced by dynamic light environment under variable sky conditions have also a potential to substantially influence the gas exchange over longer periods 44 . However, here such factors are of minor importance, as the measurements were predominantly conducted during overcast days.
Ancillary measurements. The following continuously measured variables at the SMEAR II experimental station were used for correlation analyses: a) soil water content (TDR-100, Campbell Scientific, North Logan, Utah, USA) 45,46 and b) soil temperature (Philips KTY81, NXP, Eindhoven, Netherlands) 46 , both in four soil horizons (O-, A-, B-and C-horizon; corresponding to depths of − 4-0, 0-5, 5-23 and 23-60 cm from the mineral soil surface); c) air temperature at 4.2 m height within the forest stand (Pt100 sensors), d) photosynthetic photon flux density at 23 m height (Li-190SZ, Li-Cor, USA); on P. sylvestris: e) stem sap flow using the Granier-type heat dissipation method 47,48 at a height of about 2 m; and f) shoot transpiration at the top canopy with dynamic enclosures 49 .
Statistics. Datasets were tested for normal distribution (Shapiro-Wilk test) and homogeneity of variances in different subpopulations. The flux data were assumed independent. Because of non-normally distributed data and/or data with unequal variances, the non-parametric Mann-Whitney rank sum test was run at p < 0.05 to test the statistical significance a) among flux rates from stems, shoots, and forest floor; and b) between flux rates from dry and wet plots.
Correlation analyses (a) between stem, shoot, and forest floor flux rates of N 2 O or CH 4 , and (b) between the trace gases flux rates and micro-climatic and other tree parameters were performed using non-parametric correlation analyses (Spearman's rank correlation). The statistical significance was defined at p < 0.05. SigmaPlot 11.0 (Systat Software, San Jose, California, USA) was used for statistical analyses.