Short-term flooding increases CH4 and N2O emissions from trees in a riparian forest soil-stem continuum

One of the characteristics of global climate change is the increase in extreme climate events, e.g., droughts and floods. Forest adaptation strategies to extreme climate events are the key to predict ecosystem responses to global change. Severe floods alter the hydrological regime of an ecosystem which influences biochemical processes that control greenhouse gas fluxes. We conducted a flooding experiment in a mature grey alder (Alnus incana (L.) Moench) forest to understand flux dynamics in the soil-tree-atmosphere continuum related to ecosystem N2O and CH4 turn-over. The gas exchange was determined at adjacent soil-tree-pairs: stem fluxes were measured in vertical profiles using manual static chambers and gas chromatography; soil fluxes were measured with automated chambers connected to a gas analyser. The tree stems and soil surface were net sources of N2O and CH4 during the flooding. Contrary to N2O, the increase in CH4 fluxes delayed in response to flooding. Stem N2O fluxes were lower although stem CH4 emissions were significantly higher than from soil after the flooding. Stem fluxes decreased with stem height. Our flooding experiment indicated soil water and nitrogen content as the main controlling factors of stem and soil N2O fluxes. The stems contributed up to 88% of CH4 emissions to the stem-soil continuum during the investigated period but soil N2O fluxes dominated (up to 16 times the stem fluxes) during all periods. Conclusively, stem fluxes of CH4 and N2O are essential elements in forest carbon and nitrogen cycles and must be included in relevant models.

In reaction to stressing factors, for instance, a natural or even artificial creation of anaerobic soil conditions due to flooding events, biosynthetic processes like production and consumption of CH 4 and N 2 O and their respective transportation may become modified and, thus, may lead to a change of the potential trace gas emissions [29][30][31] . Importantly, plant species react differently, depending on their anatomical and physiological predisposition/adaptation to stresses, soil characteristics 32 , seasonal effects, temperature 33 , and intensity of the stressing event in general 26 .
Nevertheless, field experiments, investigating CH 4 and N 2 O fluxes in riparian tree communities exposed to environmental stress are scarce. Thus, a large-scale forest manipulation experiment was conducted in a grey alder forest in summer 2017. The main objective of the FluxGAF ("Biogeochemical Fluxes in Grey Alder Forest") campaign was to investigate the response of the forest ecosystem to heavy overland flow. The objective was to understand the biochemical process and flux dynamics in the soil-tree-atmosphere continuum related to ecosystem CH 4 and N 2 O turnover and exchange. Therefore, we quantified CH 4 and N 2 O fluxes from stems of grey alder and adjacent soil in response to an artificial flooding event. The trace gas fluxes were measured on mature trees in an alder forest in Estonia together with forest floor CH 4 and N 2 O fluxes, and a variety of environmental parameters (including (micro)meteorological, soil and atmospheric parameters). The study site consisted of two plots: a flooded plot (FP), where 55-70 m³ of water per day was applied for two weeks in summer 2017, and a control plot (CP). The study period was divided into three periods: pre-experimental (July 24 th -August 7 th ), experimental (mimicking flooding; August 8 th -21 st ) and post-experimental (August 22 nd -September 4 th ).
We hypothesize that: (1) due to the flooding, emission of methane and nitrous oxide from both the soil and tree stems will increase while the lower parts of stems will have higher emission, (2) flooding will change the proportion of soil and stem fluxes in overall emissions while the dynamics are different for CH 4

Results
Soil physicochemical conditions and characteristics of the tree stand. The study period was characterised by mean air temperature (mean ± standard deviation) of 19.9 ± 2.1 °C, and soil temperature at a soil depth of 5 cm at the flooded and control plot, 15.1 ± 0.6 and 15.3 ± 0.6 °C, respectively (Fig. 1a). No significant differences were detected in Soil Water Content (SWC) between flooded and control plots in the pre-experimental period, although SWC was significantly higher in the flooded plot during the experimental (p < 0.001) and post-experimental periods (p < 0.05) ( Fig. 1b and Supplementary Table S1). Single rain incidents occurred and one at the beginning of post-experimental period also increased SWC in the control plot (Fig. 1a). These trends were also reflected in the water table ( Supplementary Fig. S1).
Tree stand density was approximately 1500 trees per hectare and tree height was 19.2 ± 1.4 m in both studied plots. In addition, tree diameter at 1.3 m height showed no significant difference between FP and CP and was approximately 0.17 ± 0.03 m on both sites. www.nature.com/scientificreports www.nature.com/scientificreports/ Experimental (FP_Flooding) and post-experimental (FP_Post) periods of the flooded plot were clearly distinct from pre-experimental period of the flooded plot and all periods of the control plot (p < 0.05; Fig. 2). Pre-experimental period of the flooded plot (FP_Pre) was different from the experimental period of the control plot (CP_Flooding; p < 0.05). The flooded plot was characterised by higher NH 4 + levels during the experimental (p < 0.001) and post-experimental period (p < 0.05), and lowered NO 3 − values during the experimental period (p < 0.001, Supplementary Fig. S2). pH, phosphorous (P), nitrogen (N), organic matter and soil temperature values showed no significant differences between the two plots in different periods. Soil bulk density varied between 0.77 and 0.96 and was not significantly different between studied plots.

Fluxes of CH 4 and N 2 O and their relation to environmental factors. Different patterns emerged for
CH 4 and N 2 O emissions in the studied plots and their periods (Fig. 2). No significant differences were detected in CH 4 emissions from soil and 170 cm stem level between flooded and control plots in the pre-experimental period, although CH 4 emissions were significantly higher from soil and all stem levels in the flooded plot during the experiment (p < 0.05 for soil, p < 0.001 for all stem levels) (Supplementary Table S1). Moreover, CH 4 emissions were significantly higher from the soil and stems at 10, 80 and 170 cm level in the flooded plot compared to the control plot during the post-experimental period (p < 0.001 in all cases). N 2 O emissions from the soil and different stem heights showed more controversial results (Supplementary Table S1). Very few significant differences in N 2 O emission appeared between flooded and control plots before and during the experiment. However, N 2 O emissions from all stem heights were significantly lower in the flooded plot compared to the control plot during the post-experimental period (p < 0.001 in all cases).
Soil NO 3 − was positively and NH 4 + was negatively related to N 2 O emissions from soil and different stem heights (Fig. 2). In addition, soil moisture was negatively related to N 2 O emissions, whereas it showed positive relationships to CH 4 emissions from both the soil and stems. Soil temperature, pH, P, N, and organic matter were almost perpendicular to the CH 4 and N 2 O flux vectors, indicating little or no correlations.
Both soil and stem surfaces were net emitters of CH 4 and N 2 O (Figs. 3 and 4). At the flooded plot, we observed an increase (p < 0.05) of stem CH 4 fluxes at 10 cm from the pre-to the post-experimental period (Fig. 3a). A significant decline (p < 0.05) along the overall vertical stem profile was found during and after the experiment. On the control plot, there was significant difference at the 10 cm level before and after the experimental period only (Fig. 3b).
At the flooded plot, N 2 O fluxes from the 10 and 80 cm level of stems increased significantly with the flooding experiment and declined afterwards (p < 0.001). Fluxes from the 170 cm level decreased significantly during the post-flooding period only (p < 0.001). The N 2 O fluxes showed a diminishing trend with tree height (Fig. 3c).
At the 10 cm level in the control plot, we observed tendentially higher N 2 O fluxes without a general decrease with increasing stem height (p > 0.1). Nevertheless, the intensive rain forced occasional peaks with significant effect at the post-experimental period (Fig. 3d).
Cumulative values of mean flux for each sampled day (Fig. 5) clearly demonstrate the flooding-induced dominance of CH 4 fluxes from tree stems over soil emissions: the difference is up to 400 times. Furthermore, the vertical decrease in both CH 4     www.nature.com/scientificreports www.nature.com/scientificreports/ The relative contribution of the stems and soil to the CH 4 and N 2 O fluxes. Stem fluxes were upscaled to unit of ground area of forest and compared with related soil fluxes ( Fig. 6). At the flooded plot, CH 4 emissions from tree stems dominated 7 times more with up to 88% contribution but small uptake rates from soil surfaces balanced with stem emissions at the control area (Fig. 6a,b).
Stems and soils were N 2 O emitters while no consumption was observed. Soil dominated N 2 O flux ratio on both plots with up to 99%, accordingly up to 145 times more. However, stem fluxes contributed 12% at the control plot's post-experimental period (Fig. 6c,d).

Discussion
The relevance of the experimental set-up. The choice of the experimental area depended on many environmental factors. One of them was a slight slope which allowed slow movement of the added water and, most importantly, did not influence groundwater level and quality on the adjacent control plot (Supplementary Fig. S4). The impact of the slope and a 1 m high dyke, which separated the flooded and control plots, can be seen in the similar characteristics of FP and CP during the pre-experimental period (Supplementary Table S1, Figs. 4 and 5).

Main controllers of CH 4 and N 2 O fluxes.
Methane is produced in anoxic soils and sediments, while well-drained soils act as a sink for atmospheric CH 4 due to methane oxidation, both processes are controlled by different microorganisms. The main environmental factors controlling the CH 4 emission in soils are the availability and quality of carbon, soil temperature and water content 34 . Since in our study area, the carbon content of the soil is not a limiting factor (Fig. 2), soil water content (water table) and temperature determine the most variation of CH 4 fluxes in soil and stems (Fig. 2, Supplementary Table S1). N 2 O is produced in the soil mainly via nitrification under aerobic conditions, where ammonia is oxidised, and by denitrification, which occurs under anaerobic conditions, where nitrate is sequentially reduced to nitrite, N 2 O and pure molecular nitrogen (N 2 ) 35 . The gaseous nitrogen losses directly depend on soil nitrate content and soil moisture which affects oxygen availability in the soil 36  Under laboratory conditions, Unger et al. 31 showed that alternating oxic/suboxic and anaerobic conditions, which are similar to our field study soil moisture, can coherently change NH 4 + and NO 3 concentrations. Also, www.nature.com/scientificreports www.nature.com/scientificreports/ under laboratory conditions, Klemedtsson et al. 39 showed the relationships between soil moisture and N 2 O production during nitrification and denitrification. Nevertheless, variation in N 2 O concentrations does not stringently correlate to variation of denitrifying activity. N 2 O can be produced by a range of organisms, further, denitrifiers such as bacteria or archaea may both produce and consume N 2 O. Nitrate may limit denitrification in forest ecosystems 40 . In our study, a decline of nitrate was coherent with the increase of ammonia in the soil (Supplementary Fig. S2).
In conclusion, the flash flooding significantly enhanced both nitrification and denitrification processes in the soil. N 2 O could be produced from both processes. Further, flooding can accelerate production of both CH 4 and N 2 O that may lead to increased emissions from forest ecosystems.
Comparison of results with outcomes from similar studies. Although our full-size flooding experiment during a reliable timeframe is the first attempt to measure both CH 4 and N 2 O emissions from soil and tree stems simultaneously, there are analogous experiments conducted in laboratory conditions and/or in mesocosms. Rusch and Rennenberg 11 considered an increase of N 2 O emissions from 3 years old black alder seedlings immediately after flooding had started but no effect on CH 4 emissions but vice versa results after 40 days, in particular, N 2 O fluxes were below the detection limit, but CH 4 increased enormously. Further, they found the efflux decrease with stem height from 0 to 2 m. On the other hand, Keppler et al. 41 concluded after greenhouse experiments that even plants may produce CH 4 in situ, covering 10-30% of the world's total CH 4 emissions. However, Terazawa et al., 42 concluded soil temperature and water table depth as possible environmental factors controlling stem CH 4 emissions". Mander et al. 43 , for instance, investigated the impact of flooding, using a pulsing groundwater level on GHG fluxes from the soil and found an increase of CH 4 emissions but a decrease of N 2 O emissions after flooding. However, our field experiment on flooding the Alnus incana forest indicated a "chimney effect" similar to Rice et al. 44 , i.e. soil microbes produce CH 4 which is transported via roots, stems, and leaves to be released to the atmosphere. Comparing to a mesocosm experiment with Alnus glutinosa seedlings under laboratory conditions www.nature.com/scientificreports www.nature.com/scientificreports/ conducted by Machacova et al. 12 we found an increase of flooding-induced N 2 O stem emissions of up to 10 times (instead of a factor of 740 by Machacova et al.) but CH 4 emissions in our experiment were up to 100 times higher. To compare with other studies, our flooding-induced CH 4 emissions from alder stems were significantly higher, especially when comparing them with the soil fluxes. Pitz et al. 45 measured methane fluxes from tree stems and soils along a habitat gradient and found mean stem CH 4 emissions of 68.8 ± 13.0 (mean ± standard error), 180.7 ± 55.2 and 567.9 ± 174.5 g CH 4 -C m −2 h −1 for the upland, transitional and wetland habitats, respectively. In the same time, mean soil methane fluxes in the upland, transitional and wetland were −64.8 ± 6.2, 7.4 ± 25.0 and 190.0 ± 123.0 g CH 4 -C m −2 h −1 , respectively.
Fluxes from other tree compartments can have an effect on CH 4 and N 2 O balances. Machacova et al. 46 show that mature Scots pine trees consistently emit CH 4 and N 2 O from both stems and shoots. The shoot fluxes of CH 4 and N 2 O exceeded the stem flux rates by 41 and 16 times, respectively. Therefore, further investigations at canopy level are very important.
Likewise, full-year investigations of gas emissions might change the results. For instance, Machacova et al. 47 found a clear N 2 O emission peak from stems of boreal trees during the vegetation season/summer while there was no effect of soil water content on tree N 2 O fluxes in the vegetation season.
In future climate, the frequency of both flash floods and drought events is expected to be increased [48][49][50][51] . Although trees and other plants indicate the "forest ecosystem", their role within gas cycles is somehow underestimated 12 . Therefore, understanding the capacity of ecosystems to adapt to such environmental modifications is the key to predict ecosystem responses to global change 52 .

Conclusions
Our experimental flooding induced changes of water regime and consequent dynamics of CH 4 and N 2 O pathways: emission of these gases from both soil and tree stem increased significantly. Thus, our first hypothesis has been supported. As well, the lower parts of stems showed higher emission compared with the higher positions. Likewise, the second hypothesis could be supported: we saw that flooding changed the proportion of soil and stem fluxes in overall emissions showing significantly different patterns for CH 4 and N 2 O. Methane fluxes from tree stems were observed up to 100 times higher than related fluxes from soil chambers. In contradiction, nitrous oxide fluxes from the soil surface were up to 16 times higher than stem N 2 O fluxes close to the ground at the post-experimental period. Furthermore, the stems contributed up to 88% of CH 4 emissions to the stem-soil continuum during the investigated period but soil fluxes dominated the N 2 O contribution during all periods when upscaling for forest area unit. A substantial link was shown between N 2 O fluxes from the stems and soil with soil water content and nitrogen availability as the main controlling factors.
Our results convince us that the N 2 O and CH 4 exchange of riparian trees should be included in forest ecosystem GHGs budgets and forest-ecosystem process models. In addition, extreme climate events (flooding, drought) are significantly altering not only the soil fluxes, but also the stem fluxes. However, emission from tree stems of different species may vary significantly and further studies are needed for comprehensive flux estimates for different forest types.

Material and Methods
Site description and experimental design. The measurements were performed at the experimental site of Agali (58°17′N; 27°17′E) situated in eastern Estonia, 10 km west of Lake Peipus. The studied forest is a 40-year old hemiboreal Filipendula type grey alder (Alnus incana (L.) Moench) forest stand in a former agricultural Gleysol.
The artificial flooding was conducted in summer 2017 and was divided into three periods: a pre-experimental (July 24 th -August 07 th ), experimental (August 8 th -21 st ), and a post-experimental (August 22 nd -September 4 th ) period. Within the forest, two experimental plots were established, a flooded plot (FP, 40 × 40 m), where water was pumped into using an irrigation pipe system mimicking the intensive-rain induced overland flow, and a control plot (CP, 20 × 20 m). The plots were separated by a 1 m high natural dyke preventing the spread of water into the control plot. Slight slope (1%) parallel to the dyke guaranteed isolated groundwater dynamics in the plots ( Supplementary Fig. S4). Every experiment day, 55 to 70 m 3 of pond water (pH 7.41, total C 76 mg/l, total N 1.8 mg/l, NH 4 + -N 0.79 mg/l, NO 3 − -N 32.6 µg/l) was applied to the FP by a fire truck. The total water amount of 875 m 3 was distributed over 1,600 m² forest area, which is an equivalent to 547 mm precipitation. The long-term average annual precipitation of the region is 650 mm, and the average temperature is 17.0 °C in July and −6.7 °C in January. The duration of the growing season is typically 175-180 days from mid-April to October 53 .
Nine representative mature grey alder trees were selected for stem flux measurements at the FP, four at the CP ( Supplementary Fig. S4). Soil fluxes were investigated close to each selected tree. The dominating ground vegetation at both plots was Alnus incana ((L.) Moench), Filipendula ulmaria ((L.) Maxim.), Prunus padus (L.), and Rubus idaeus (L.).
Stem flux measurements. The tree fluxes were measured manually: five measurement sets in the pre-, six in the experimental and four in the post-experimental period. The representative trees were equipped with static closed tree stem chamber systems for stem flux measurements 28 . The chambers were installed at the bottom part of the tree (approximately 10 cm above the soil). In addition, a vertical profile of the stem fluxes (measurements at three stem heights of approximately 10, 80 and 170 cm above the ground) was studied in 6 and 2 trees at the FP and CP, respectively. The chambers were installed in June 2017 one month prior to the campaign. The rectangular shape stem chambers were made of transparent plastic containers, including removable airtight lids (Lock & Lock, South Korea). The bottom was cut and hot-glued with a neoprene band. The chambers were sealed with non-acid silicone to the smoothed stem surface and tested for airtightness. Two chambers per profile were set randomly across 180° and interconnected with tubes into one system (total volume of 0.00119 m³) covering 0.0108 m² of stem surface. A pump (Thomas, Germany, model 1410VD, 12 V) was used to homogenize the gas concentration prior to sampling. Chamber systems remained open between each sampling campaign. During 15 measurement campaigns, four gas samples (25 ml) were collected from each chamber system via septum in a 60 min interval: www.nature.com/scientificreports www.nature.com/scientificreports/ 0/60/120/180 min sequence (sampling time between 12:00 and 16:00) and stored in pre-evacuated (0.3 bar) 12 ml coated gas-tight vials (LabCo International, United Kingdom). The gas samples were analysed in the laboratory at University of Tartu within 2 weeks using gas chromatography (GC-2014; Shimadzu, Japan) equipped with an electron capture detector for detection of N 2 O and a flame ionization detector for CH 4 . The gas samples were injected automatically using Loftfield autosampler (Loftfield / Germany). For gas-chromatographical settings see Soosaar et al. 54 .
Fluxes were quantified on a linear approach according to change of CH 4 and N 2 O concentrations in the chamber headspace over time, using the equation according to Livingston and Hutchison 55 . Soil flux measurements. Soil fluxes were measured using automatic dynamic chambers located close to each measurement tree and installed in June 2017. Nine chambers were situated at the FP, four at the CP. Every PVC made soil chamber covered a 0.16 m² soil surface, containing a volume of 0.032 m³. To avoid stratification of gas inside of the chamber, air with a constant flow rate 1.8 L/min was circulated within a closed loop between the chamber and gas analyzer unit during the measurements by a diaphragm pump. The air sample was taken from the top of the chamber headspace and pumped back by distributing it to each side of the chamber. For the measurements, the soil chambers were closed automatically for a duration of 9 minutes each. Flushing time of the whole system with ambient air between measurement periods was 1 minute. Thus, there were approximately 12 measurements per chamber per day. A Picarro G2508 (Picarro Inc., United States) gas analyzer using cavity ring-down spectroscopy (CRDS) technology was used to monitor CO 2 , CH 4 and N 2 O gas concentrations in the frequency of approximately 1.17 measurements per second. The chambers were connected to the gas analyzer using a multiplexer.
Flux calculations. The air temperature in the chamber was measured and used to convert the concentrations from ppm[v] to mg m −3 according to the ideal gas law before flux calculation 56 . Soil fluxes were calculated for a selected time window (150 seconds), after discarding initial 90 seconds to exclude the initial stabilisation period, using the linear model. The linear models is based on the assumption of a linear relationship between concentrations inside the chamber headspace and time. Fluxes were calculated using the equation, according to Livingston and Hutchison 55 .
To compare the contribution of soil and stems, the stem fluxes were up-scaled to hectare of ground area based on average tree diameter, stem surface area, tree density, and stand basal area estimated for each period and plot. A cylindric shape of tree stem was assumed. To estimate average stem emissions, fitted regression curves for different periods were made between the stem emissions and height of the measurements as previously done by Sjögersten et al. 57 . The regression model parameters are reported in Supplementary Table S2.

Data quality check.
Fluxes were quantified on a linear approach according to change of CO 2 , CH 4 and N 2 O concentrations in the chamber headspace over time. A data quality control was applied based on R 2 values of linear fit for CO 2 measurements. When the R 2 value for CO 2 efflux was above 0.9, the conditions inside the chamber were applicable, and the calculations for both CH 4 and N 2 O gases were also accepted in spite of their R 2 values. Calculations with lower R 2 value (R 2 < 0.9) were removed from the database. Ancillary measurements. Automatic groundwater level data loggers (Hobo U20L-04, Onset Computer Corporation, USA) were installed in groundwater wells. Soil temperature (107, CAMPBELL SCIENTIFIC. INC, USA) and soil moisture sensors (ML3 ThetaProbe, Delta-T Devices, United Kingdom) were installed at 0-10 cm soil depth close to adjacent tree spots. During five campaigns (two in the pre-experimental period, two in experimental period, and one in post-experimental period) composite topsoil samples with soil corer at depth 0-10 cm were taken for physical and chemical analysis using standard methods 58 .
Statistical analysis. Principal Component Analysis (PCA) was performed on real and imputed data sets of soil physicochemical and gas flux measurements using the R package ade4 v. 1.7-13 59 . For PCA, missing data patterns were identified and explored using the packages "naniar" v. 0.4.2 60 and "VIM" v. 4.8.0 61 (Supplementary  Table S3). In addition, R package "missMDA" v. 1.14 62 was applied to impute missing data values in case of PCA. Differences in PCA between the plots and periods were evaluated using PERMANOVA with 9999 permutations and pairwise comparisons were corrected with Bonferroni method, using the R package vegan v. 2.5-6 63 . To evaluate the significance of the differences between the plots and their periods with respect to physicochemical variables values, multivariate linear models were constructed, and models were tested after 9999 permutations using the anova function in R package mvabund v. 4.0.1 64 . Linear mixed-effects models (LMM) were applied to investigate differences in soil moisture and gas flux measurements between the control and flooded plots using the R method "lmer" incorporating both temporal (sampling days) and spatial (different trees) effects as random effects (v. 1.1-21) 65 , while p-values were calculated in order to confirm the significance of the relationships using the R package lmerTest v. 3.0-1 66 . The same mixed effects method was used to test statistical significance between different periodic groups of soil and stem fluxes. LMM was also used to detect differences between stem fluxes of different stem heights in different periods. For exploring LMMs fit, inspection of residuals patterns for the model was used as diagnostic tool 65,67 . To meet the analysis assumptions, rank transformation of the data was performed and LMM on the rank-transformed response variables are analysed and reported 68 . The regression models for stem heights and stem fluxes of different periods were estimated by nonlinear least squares, using the R function "nls. " R version 3.6.1 (R Development Core Team, 2019) was used to conduct all the statistical analyses. The significance level (alpha) considered for all the tests was 0.05.