Trees as net sinks for methane (CH4) and nitrous oxide (N2O) in the lowland tropical rain forest on volcanic Réunion Island

Summary Trees are known to emit methane (CH4) and nitrous oxide (N2O), with tropical wetland trees being considerable CH4 sources. Little is known about CH4 and especially N2O exchange of trees growing in tropical rain forests under nonflooded conditions. We determined CH4 and N2O exchange of stems of six dominant tree species, cryptogamic stem covers, soils and volcanic surfaces at the start of the rainy season in a 400‐yr‐old tropical lowland rain forest situated on a basaltic lava flow (Réunion Island). We aimed to understand the unknown role in greenhouse gas fluxes of these atypical tropical rain forests on basaltic lava flows. The stems studied were net sinks for atmospheric CH4 and N2O, as were cryptogams, which seemed to be co‐responsible for the stem uptake. In contrast with more commonly studied rain forests, the soil and previously unexplored volcanic surfaces consumed CH4. Their N2O fluxes were negligible. Greenhouse gas uptake potential by trees and cryptogams constitutes a novel and unique finding, thus showing that plants can serve not only as emitters, but also as consumers of CH4 and N2O. The volcanic tropical lowland rain forest appears to be an important CH4 sink, as well as a possible N2O sink.


Introduction
Covering an area of c. 1730 million ha, tropical forests comprise 45% of the global forested area (D'Annunzio et al., 2017). The most extensive tropical forest type is tropical rain forest, which encompasses almost 60% of all tropical forest area (Shvidenko et al., 2005). These forests are found predominantly in South America, Africa and Asia, and are characterized as closed-canopy evergreen broadleaf forests with minimum annual temperature and precipitation of 25°C and 1500 mm, respectively (Richards, 1996). In general, tropical forests (including all kinds of forest types) constitute important carbon (C) stock (428 Gt-C in vegetation and soil) and C sink (-0.37 t-C ha À1 yr À1 ; Dalal & Allen, 2008). Furthermore, they are considered to be a natural source of nitrous oxide (N 2 O), and a natural sink and source of methane (CH 4 ), both of which are important greenhouse gases (GHG) with global warming potential of 265-298 and 28-36 over 100 yr, respectively (Myhre et al., 2013). Net N 2 O emission from tropical forest soils is estimated to be 4.76 kg ha À1 yr À1 , whereas net soil CH 4 consumption, without considering possible canopy fluxes from trees (Keppler et al., 2006), seems to be À 3.86 kg ha À1 yr À1 (Dalal & Allen, 2008). In the case of CH 4 , soils of tropical rain forests growing under submerged conditions (e.g. Amazonia) can be substantial CH 4 sources as well, as a consequence of the prevailing anaerobic conditions required for CH 4 production (Pangala et al., 2017). Tropical forest soils have the highest N 2 O emission potential among natural forest ecosystems (Dalal & Allen, 2008), and therefore tropical forests play an essential role in global N 2 O inventories.
Nitrous oxide is naturally produced in soils through a wide range of nitrogen (N) turnover processes having different soil water content optima, including aerobic nitrification, anaerobic denitrification, and also dissimilatory nitrate reduction to ammonium in suboxic conditions (Espenberg et al., 2018). The denitrification processes are the only processes known to reduce N 2 O to dinitrogen (N 2 ) (Smith et al., 2003). By contrast, CH 4 is produced by strictly anaerobic methanogenesis in water-saturated soils and is oxidized by methanotrophic bacteria (Smith et al., 2003).
Both gases can be released into the atmosphere by gas diffusion at the soil surface and by ebullition in the case of flooded areas. Plants, moreover, can contribute to ecosystem N 2 O and CH 4 exchange by: (1) taking up both gases from the soil via the root system and transporting them into the atmosphere through the transpiration stream or aerenchyma system and enlarged intercellular spaces (Rusch & Rennenberg, 1998;Machacova et al., 2013); (2) producing N 2 O and CH 4 directly in plant tissues (Smart & Bloom, 2001;Keppler et al., 2006); (3) consuming N 2 O and CH 4 from the atmosphere by a nonspecified mechanism (Sundqvist et al., 2012;Machacova et al., 2016bMachacova et al., , 2017Machacova et al., , 2019; and (4) altering the N and C turnover processes in adjacent soil (Menyailo & Hungate, 2005;Yu & Chen, 2009). Moreover, cryptogamic stem covers (i.e. photoautotrophic organisms associated with tree bark, such as lichens, liverworts, mosses or ferns) also might contribute to N 2 O and CH 4 exchange of trees and forest ecosystems (Lenhart et al., 2015;Machacova et al., 2017). Especially (but not solely) in tropical rain forests, cryptogams often grow on as much as 100% of the tree bark surface and all the way up to the crowns. These organisms remain overlooked as potential players in the trace gas exchange of forests and trees, even though they are present in the majority of tree stem chamber measurements and can, therefore, contribute to the gas exchange between the bark's surface and the chamber headspace. In summary, the net exchange of trace gases at the soil-plant-atmosphere interfaces results from a balance of simultaneously ongoing processes of gas production and consumption, gas transport within the relevant system, gas emission into the atmosphere, and gas uptake from the atmosphere, all of which together determine whether the forest compartment will be a source or sink of CH 4 and N 2 O (Barba et al., 2019).
Recent research in various climatic zones has revealed that not only soils and herbaceous plants but also woody plants can be significant sources of N 2 O and CH 4 to the atmosphere (Machacova et al., 2013(Machacova et al., , 2016a(Machacova et al., ,b, 2019Pangala et al., 2013Pangala et al., , 2017Maier et al., 2018;Welch et al., 2018;Schindler et al., 2020). The trace gas exchange capacity of trees and their contributions to ecosystem N 2 O and CH 4 exchange seem, however, to vary considerably among tree individuals, tree species, forest ecosystem types and climatic zones, and to depend on many aspects, such as soil and site parameters, tree size, age and health conditions, environmental conditions and seasonal dynamics (Barba et al., 2019;Machacova et al., 2019). Even though the interactions between soil, vegetation and atmosphere exert a crucial role in controlling the ecosystem budget of N 2 O and CH 4 , our currentstill limited -knowledge on tree exchange of CH 4 , and especially of N 2 O, does not allow us to clearly identify common characteristics, processes, pathways, and mechanisms of N 2 O and CH 4 exchange in the soil-tree-atmosphere continuum, and to constrain the magnitudes and patterns of N 2 O and CH 4 emissions.
Wetlands and floodplains are the largest natural sources of atmospheric CH 4 in the tropics (Saunois et al., 2016). To date, calculations of N 2 O and CH 4 fluxes between wetlands and the atmosphere have been based mostly upon GHG exchange at the soil-atmosphere interface only, thus excluding the exchange activity of such other ecosystem compartments as trees and other vegetation. This approach can lead to underestimating the ecosystem fluxes (Barba et al., 2019). In general, trees growing in wetlands and floodplain forests seem to be stronger emitters of CH 4 than trees in upland forests (Pangala et al., 2013(Pangala et al., , 2017Machacova et al., 2016a,b;Jeffrey et al., 2019;Sj€ ogersten et al., 2020), whereas trees in riparian forests show emission values in between (Schindler et al., 2020). The giants for their CH 4 emission potential are angiosperms in tropical rain forests of the Amazon basin, with CH 4 emissions two to three orders of magnitude greater than those of trees growing in other tropical and temperate floodplain forests (Pangala et al., 2017;. Pangala et al. (2017) showed that these trees adapted to high soil water level are responsible for as much as half of the CH 4 emissions from the Amazon floodplain, which is the largest natural CH 4 source in the tropics.
With the exception of detailed, in-depth studies ongoing in the Amazon basin, which are today often used in estimating the overall CH 4 exchange of tropical forests, the widely distributed rain forests in other tropical continental and insular areas of Africa and Asia remain understudied. Likewise, mangroves and tropical forests on upland soils also are wholly understudied in relation to the CH 4 exchange of their woody vegetation (Pangala et al., 2013;Welch et al., 2018;Jeffrey et al., 2019;Sj€ ogersten et al., 2020). Moreover, the N 2 O exchange of mature trees growing under their natural field conditions is rarely investigated world-wide (D ıaz-Pin es et al., 2016;Machacova et al., 2017Machacova et al., , 2019Wen et al., 2017), and information for tropical regions, including tropical rain forests, is scarce (Welch et al., 2018). Without ecologically relevant studies of tree and ecosystem exchange of CH 4 and N 2 O covering the broad mosaic of tropical forest ecosystems, it is more than challenging to understand the role of woody plants in the GHG balance of tropical regions, to correctly estimate the tropical forest CH 4 and N 2 O budgets, and to predict their future development in relation to global climate change.
Accordingly, the objective of our case study was to quantify the N 2 O, CH 4 and, additionally, carbon dioxide (CO 2 , an indicator of physiological activity) exchange of dominant tree species in a tropical lowland rain forest on volcanic R eunion Island (southwestern Indian Ocean). We studied trace gas fluxes from two trees endemic for R eunion and Mauritius islands (Doratoxylon apetalum, Antirhea borbonica) and four regional endemics of Madagascar, Mauritius and R eunion (Syzygium borbonicum, Homalium paniculatum, Mimusops balata and Labourdonnaisia calophylloides). These measurements were accompanied by the investigation of GHG exchange from adjacent soil and volcanic surfaces (basaltic lava flows) and from widespread cryptogamic (bryophytic) stem covers (Pyrrhobryum spiniforme, Leucoloma capillifolium). The volcanic R eunion Island belongs to one of 36 world hotspots of biodiversity with high level of endemism (Kreft et al., 2008), and its National Park, covering 42% of the island, provides an exceptional setup for scientific experimentations in primary vegetation along a major elevational gradient (3069 m above sea level (asl), Piton des Neiges) and with proximity to a regularly active volcano (2632 m asl, Piton de La Fournaise).
Still relatively young, the studied forest has developed on a 400-yr-old basaltic lava flow. Because young lava flows lack nutrients and organic matter, the microorganisms colonizing these newly created ecosystems fix not only N 2 , CO 2 and ammonium (NH 4 + ), but also other trace gases from the atmosphere, such as carbon monoxide (CO), hydrogen (H 2 ) and CH 4 , and drive the sequestration of N, C and other nutrients needed for further ecosystem development (King, 2003 Byloos et al., 2018). The further ecological succession is connected to the accumulation of nutrients and leads to the gradual development of various plant communities starting with algae, mosses, lichens and ferns, followed by woody shrubs, shrubs growing to trees and later mature trees, forming a dense canopy forest c. 400 yr after the destruction of all past vegetation by new lava substrate formation (Potgieter et al., 2014). This development is accompanied by the formation of a soil layer, with associated changes in soil properties and microbial community. In the case of the studied area, an unique native lowland rain forest has developed (Kirman et al., 2007).
Even though the island has a tropical climate with high annual precipitation (> 4000 mm), the studied forest is atypical compared to well-studied Amazon basin forests. It is specific for its porous volcanic bedrocks with the presence of lava holes and tubes, thin and irregular soil layer with weak water holding capacity, and steep slopes. It hosts high biodiversity, including mostly tree species endemic at the archipelago or regional levels. The aforementioned characteristics have resulted in an unique and atypical tropical rain forest ecosystem without standing water even under heavy rains. Therefore, we aimed to investigate: (1) whetherand, if so, to what extentthe tree stems growing on a lava flow exchange CH 4 and N 2 O with the atmosphere; (2) how the tree fluxes contribute to the forest GHG exchange; and (3) whether the tropical rain forest is a source or sink for CH 4 and N 2 O at the beginning of the rainy season.

Site description and study design
The experiment was conducted within the Mare Longue Nature Reserve (lat. À21°21 0 28.2024″N, long. 55°44 0 37.554″E), situated in the southeast of R eunion Island, a tropical volcanic island located in the southwestern Indian Ocean. The studied mixed forest is a tropical lowland rain forest characterized by endemic vegetation typical for the volcanic islands in the Mascarene Archipelago. The studied forest site (size 1 ha, established in 1999 by the University of R eunion Island) is located on sloping terrain at 180-200 m asl and situated on a pahoehoe basaltic lava flow c. 400 yr old (Kirman et al., 2007). The soil cover is irregular and thin, and consists mostly of organic matter, parent-rock fragments and iron oxyhydroxides (Kirman, 2003;Meunier et al., 2010). The thin A horizon (average 1 cm depth) is accompanied by thin and irregularly occurring eluvial deposits (weathered surface of the basaltic lava bedrocks; i.e. the C horizon). The chemical composition of the basalt (composed of feldspar (anorthite), olivine (forsterite) and pyroxene (augite)) and of the soil in the studied forest site can be found in Meunier et al. (2010). In the experimental plots, we distinguished soil-covered lava flow spots (further referred to as 'soil') from bare lava flow surfaces without soil coverage (further referred to as 'basaltic lava flows' or 'volcanic rocks'). As a consequence of the roughness of the soil surface, the plant roots colonize deep volumes of lava bedrocks (Meunier et al., 2010). The aboveground biomass and litterfall of the studied forest were measured in the previous study of Kirman et al. (2007). The total standing biomass (535 t ha -1 ) and annual litterfall (7.6 t ha -1 ), as well the chemical composition of the biomass (stored major elements), are similar to other tropical rain forests (Kirman et al., 2007;Meunier et al., 2010).
The long-term mean annual precipitation of the southeast of R eunion Island is 4256 mm; the mean minimum and maximum temperatures are 19.9 and 26.4°C, respectively (data from Le Baril M et eo-France station, 1981-2010. The study period (8 October-7 November 2018) was characterized by mean daily air and soil temperature (50 cm soil depth) of 20.6 AE 2.0°C and 19.3 AE 0.5°C (mean AE SD), respectively, and mean daily relative humidity of 97.4% AE 7.1%. All daily parameters were measured in the studied forest site using an RHTemp1000IS relative humidity and temperature data logger (MadgeTech, Warner, NH, USA), and Soil Matric Potential Sensors 253-L and 257-L (Campbell Scientific, Logan, UT, USA). The measurement period was a time of transition between a cool and dry season (May-October) and a warm and humid season (November-April).
The measurements of N 2 O, CH 4 and CO 2 fluxes from stems and soil were made in pairs, in the sense that the stem gas flux measurement for an individual tree directly followed the gas flux measurement of the adjacent soil to ensure measurements under similar environmental and climatic conditions. The measurements on cryptogams and basaltic lava flows were carried out in blocks. All fluxes were determined between 09:00 h and 18:00 h. One measurement set from all selected trees, cryptogams, soils and basaltic lava flows required c. 1.5-2 wk. All of the forest components were measured twice.
Gas sampling from stems, soil and basaltic lava flows South Korea) and a neoprene sealing frame. They were gastightly affixed to the carefully smoothed bark surface c. 2 wk before measurements. Control measurements were performed to ensure that the observed fluxes did not originate from the chamber materials used. On each tree other than S. borbonicum individuals, two chambers were installed at one stem height on opposite sides of the stem and interconnected with polyurethane tubes into a single flow-through chamber system. The total enclosed stem area was 0.0108 m 2 and total internal system volume was 0.0021 m 3 . (For more details, see Machacova et al., 2015Machacova et al., , 2017Machacova et al., , 2019 Owing to the large stem diameter of S. borbonicum, three chambers (total area 0.0162 m 2 , total internal volume 0.0028 m 3 ) were installed at one stem height in the same manner on those trees. Gas-tightness of all the chambers was tested regularly using CO 2 signal and a portable gas analyzer (see the 'Flux measurements and calculations' section below). Soil N 2 O, CH 4 and CO 2 fluxes were measured using static soil chamber systems (Fig. 1b,d). The soil collars were installed in the vicinity of the investigated trees at the beginning of the research stay to reduce soil disturbances. As a result of the shallow soil layer on the lava flow, the bottom part of the collars was imbedded to a maximum 2.5 cm deep into the soil. The soil chambers were made of sewage pipes (total enclosed soil area 0.0083 m 2 , total internal system volume c. 0.0015 m 3 depending on soil depth). During measurements, the soil chambers were closed by a lid and sealed with water between the body of the chamber and the lid (Machacova et al., 2013).
The fluxes of N 2 O, CH 4 and CO 2 from basaltic lava flows were determined using the stem chambers as described above but installed directly onto the volcanic surfaces (Fig. 1c). One flowthrough chamber system consisted of one chamber (total enclosed rock area 0.0054 m 2 , total internal system volume 0.0013 m 3 ) connected to a gas analyzer (see the 'Flux measurements and calculations' section below).

Collection and incubation of cryptogamic stem covers
Pyrrhobryum spiniforme and Leucoloma capillifolium mosses were collected from the bark of S. borbonicum trees. To avoid disruption to the bark microcosm within the stem chambers, the cryptogams were collected from bark outside the measurement chambers. After collection, fresh samples were placed into plastic gas-tight incubation chamber (0.0039 m 3 volume; Lock & Lock; Fig. 1e,f) connected to a gas analyzer and incubated in the laboratory under dark conditions and low light intensity of 10-15 µmol m À2 s À1 corresponding to the low light conditions of forest understories.
The stem, soil, lava flow and incubation chamber systems were left open during the time between the individual measurements.

Flux measurements and calculations
For GHG flux measurements, the measurement chambers were closed with lids, and the concentration changes of N 2 O, CH 4 and CO 2 in chamber headspace were measured using a Gasmet DX-4015 portable Fourier transform infrared (FTIR) gas analyzer (Gasmet Technologies Oy, Vantaa, Finland; Warlo et al., 2018). A single measurement run of stem, soil and lava flow gas fluxes lasted c. 45 min. The measurement time for cryptogams was 180-240 min, depending on flux rates. The internal pump (0.002 m 3 min À1 ) of the analyzer ensured mixing of air inside the chamber systems. Every morning, zero-point calibration of the analyzer was made using N 2 (99.9992% purity).
The gas exchanges of stems, soil, lava flows and cryptogams were quantified based upon the linear changes in N 2 O, CH 4 and CO 2 concentrations in the chamber headspace over time (for examples, see Supporting Information Fig. S1; for equations used, see Machacova et al., 2016a). Decrease of gas concentration over time indicated gas uptake (i.e. negative flux), increase of gas concentration indicated gas emission (i.e. positive flux). The fluxes from tree stems and soil were further roughly scaled up to the ecosystem level based on tree and forest characteristics (stem diameter at breast height (DBH), tree height, tree density for individual tree species with trees of DBH > 0.10 m; Table 1; Machacova et al., 2016a). The scaling up of trace gas exchange of volcanic surfaces of lava flows was not possible because the exact extent of volcanic bedrock surfaces without soil cover within the studied forest was unknown. The contribution of cryptogams to the stem GHG fluxes was estimated based on upscaling of the fluxes related to g DW (dry weight) to the unit of stem surface area (Machacova et al., 2017). Syzbor, Syzygium borbonicum (n = 5); Dorape, Doratoxylon apetalum (n = 5); Antbor, Antirhea borbonica (n = 5); Hompan, Homalium paniculatum (n = 3); Mimbal, Mimusops balata (n = 3); Labcal, Labourdonnaisia calophylloides (n = 3); DBH, stem diameter at breast height. *Tree height of studied trees was estimated using allometric relationships for Mare Longue Nature Reserve kindly provided by Olivier Flores (University of R eunion Island). **Tree density is calculated for each tree species separately and for trees with DBH > 0.10 m to include only trees relevant for flux measurements.
The tree density of all six tree species studied is 372 trees ha À1 . The overall forest density, including all c. 80 tree species present, is estimated to be 2150 trees ha À1 (Kirman et al., 2007).

Statistics
The flux data were checked for normal distribution (Shapiro-Wilk test) and equality of variances in the different subpopulations. Student's t-test and one-way ANOVA for multiple comparisons were applied for normally distributed data. The nonparametric Mann-Whitney rank-sum test and Kruskal-Wallis one-way ANOVA on ranks for multiple comparisons were applied for non-normally distributed data or data with unequal variances. The exact applied tests and n values for statistical analyses are stated in the figure legends and in the 'Site description and study design' section above, respectively. The study was performed in 17 randomly selected experimental plots within the studied 1 ha forest site, and the studied trees together with the soil positions and basaltic lava flow surfaces were representatively chosen within these plots. Statistical significance for all tests was defined as P < 0.05. The statistics were run using SIGMAPLOT 11.0 (Systat Software, San Jose, CA, USA).

Methane exchange of forest compartments
The soil of the tropical lowland rain forest consistently consumed CH 4 from the atmosphere (-83.3 AE 9.8 µg-CH 4 m À2 h À1 (soil area), mean AE SE) at the beginning of the rainy season. Uniform CH 4 consumption was observed at all subsites with low, nonsignificant spatial flux heterogeneity (Fig. 2a). The volcanic surfaces of basaltic lava flows also showed consistent uptake of CH 4 from the atmosphere (À107 AE 21.6 µg-CH 4 m À2 h À1 (rock area), Fig. 2a). As the similar consumption rates of irregularly spread soil layers and of volcanic rocks are expressed per 'ground' area, and both surface types seem to be equally distributed in the studied forest, we can confidently say that the detected CH 4 consumption of both forest compartments equally and significantly contributes to CH 4 exchange of the forest ground area (Fig. 2a). The forest floor CH 4 fluxes can therefore be estimated using 50:50 proportion of soil and volcanic rock coverage. All mature studied tree stems were net sinks of CH 4 from the atmosphere (À15.6 AE 2.0 µg-CH 4 m À2 h À1 (stem area)). The CH 4 consumption was uniform across the various tree species and did not show significant species-specific variability (Fig. 3a). Moreover, no significant changes in stem CH 4 consumption were observed with respect to stem height aboveground (Fig. 4a). The uptake of CH 4 by tree stems was not closely connected to stem CO 2 efflux (Fig. 5a). To better understand the mechanisms behind the observed CH 4 uptake by trees, the GHG exchange of widespread cryptogamic stem cover was measured. We estimated how much CH 4 is taken up by the active cryptogams per stem area unit when fully covered with these organisms. As a result, cryptogams were identified as net sinks of CH 4 (-8.3 AE 3.0 µg-CH 4 m À2 h À1 (stem area); À0.047 AE 0.016 µg g À1 h À1 (DW); Fig. 3a). All main forest compartments studied therefore consumed CH 4 from the atmosphere.
In order to estimate the contribution of the tree stems to the tropical rain forest CH 4 exchange, the detected stem and soil consumption rates were scaled up to ecosystem level (Fig. 6a,b). The six tree species studied (joint tree density of c. 370 trees ha À1 ) consumed in total -59.0 AE 17.9 mg-CH 4 ha À1 h À1 (ground area) (i.e. the sum of mean tree species consumption rates and SEs) and contributed 7.1% to the soil CH 4 uptake (-833.1 AE 97.5 mg-CH 4 ha À1 h À1 ; Fig. 6b). As there are > 80 tree species present in the studied forest (total tree density of c. 2150 trees ha À1 ; Kirman et al., 2007), we estimate that the overall contribution of all tree stems present to the soil CH 4 uptake might be 41% when assuming similar CH 4 uptake potential by all tree species present. This tropical rain forest with its > 80 tree species seems to be a strong sink for CH 4 at the beginning of the rainy season (À1176 mg-CH 4 ha À1 h À1 ).

Nitrous oxide exchange of forest compartments
The soil was on average a weak source of N 2 O (1.8 AE 2.5 µg-N 2 O m À2 h À1 ; Fig. 2b). The detected exchange potential of volcanic surfaces of basaltic lava flows was negligible (0.024 AE 0.871 µg-N 2 O m À2 h À1 ). The exchange of soil and Dots mark outliers. As indicated by the letter 'a' above each bar, there were no statistically significant differences among fluxes in adjacent soil of individual tree species at P < 0.05. One-way ANOVA was applied for CH 4 fluxes and Kruskal-Wallis one-way ANOVA on ranks was used for N 2 O and CO 2 fluxes.

Research
New Phytologist volcanic surfaces showed high spatial heterogeneity without any clear trend, including both N 2 O consumption and emission. Moreover, the majority of the exchange rates were very low.
In contrast, clear uptake of N 2 O by tree stems was observed. Thus, these are net sinks of N 2 O from the atmosphere (-3.0 AE 0.8 µg-N 2 O m À2 h À1 ; Fig. 3b). The lowest N 2 O consumption was detected for Syzygium borbonicum and the highest for Homalium paniculatum. The uptake of N 2 O by tree stems was negatively correlated with stem CO 2 efflux (Fig. 5b). Similar to the situation for CH 4 , no significant changes in stem N 2 O exchange were observed within the vertical stem profile (Fig. 4b).
The cryptogams were a clear sink not only of CH 4 , but also of N 2 O (À1.9 AE 0.7 µg-N 2 O m À2 h À1 (stem area); À0.011 AE 0.004 µg g À1 h À1 (DW); Fig. 3b). Our test measurements confirmed that the observed cryptogamic N 2 O and CH 4 uptake could not be explained by gas dilution in air humidity within the incubation chambers (Machacova et al., 2017).
The scaled-up rates of N 2 O consumption by tree stems of the six dominant tree species studied (À11.9 AE 6.3 mg- Fig. 6c,d), thus decreasing the weak source strength of the soil. (Please note that the usage of the negative percentage in case of N 2 O is due to comparison of bidirectional fluxesnegative fluxes (i.e. N 2 O uptake by stems) with positive fluxes (i.e. N 2 O emission by soil; equal to 100%). In the case of CH 4 , the percentage contribution of the stem fluxes to the soil fluxes is positive, as bothtree stems and soilare sinks for CH 4 .) Under the assumption of all 80 tree species in the forest being N 2 O consumers at similar uptake rates, however, we expect that the overall N 2 O uptake potential of all trees (-68.9 mg-N 2 O ha À1 h À1 ) might represent even -374% of the soil N 2 O exchange, thus turning the tropical rain forest into a sink of N 2 O (-50.45 mg-N 2 O ha À1 h À1 ).  3)), from all studied trees and tree species ('All trees', n = 24), and from cryptogams (Pyrrhobryum spiniforme, Leucoloma capillifolium; n = 4). The trace gas exchange of cryptogams is presented as mean and median of all measurements under low light and dark conditions, because CH 4 and N 2 O fluxes did not differ in relation to light conditions. The mean and median fluxes of CO 2 in cryptogams include both CO 2 emission and uptake measured under dark and light conditions, resulting in low gas exchange (more details in manuscript text). All fluxes, including fluxes from cryptogams, are expressed per m 2 of stem area. Positive fluxes indicate trace gas emission and negative fluxes trace gas uptake. The box boundaries mark the 25 th and 75 th percentiles and whiskers the 10 th and 90 th percentiles. Dots mark outliers. Statistically significant differences among fluxes in individual tree species at P < 0.05 are indicated by different letters above bars. One-way ANOVA was applied for CH 4 and N 2 O fluxes and Kruskal-Wallis one-way ANOVA on ranks was used for CO 2 fluxes.

Carbon dioxide exchange of forest compartments
The measured CO 2 exchange of forest compartments is an ancillary parameter (i.e. an indicator of physiological activity) helping to understand the CH 4 and N 2 O exchange in the soil-tree-ecosystem-atmosphere continuum. The CO 2 emissions from soil substantially exceeded those from volcanic surfaces of lava flows (650 AE 95 and 35.5 AE 4.7 mg-CO 2 m À2 h À1 , respectively; Fig. 2c). Tree stems emitted 295 AE 73 mg-CO 2 m À2 h À1 , with Homalium paniculatum being the strongest emitter (Fig. 3c). The stem CO 2 efflux was uniform across the vertical stem profile (Fig. 4c). As expected, photoautotrophic cryptogamic stem covers consumed CO 2 under light conditions (À15.6 AE 2.5 mg-CO 2 m À2 h À1 (stem area); -0.094 AE 0.027 mg g À1 h À1 (DW)) and emitted CO 2 under dark conditions (14.8 AE 2.9 mg-CO 2 m À2 h À1 (stem area); 0.087 AE 0.010 mg g À1 h À1 (DW); NB Fig. 3c shows total mean and median values from both incubation conditions).

Discussion
All studied forest compartments are net methane sinks Soil and basaltic lava flows Soil (i.e. soil on soil-covered lava flow spots) and basaltic lava flows (i.e. bare lava flow surfaces without soil coverage) consistently consumed methane (CH 4 ) from the atmosphere, even though the measurements were done at the start of the rainy season. The forest floor is covered by irregular and thin soil (Kirman et al., 2007), likely preventing the formation and preservation of anaerobic soil conditions for methanogenic CH 4 production and thus supporting methanotrophic CH 4 oxidation (Smith et al., 2003). The steep slope further promotes rapid rainwater runoff from the forest into the ocean. Moreover, the high porosity typical for the volcanics forming the underlying rocks of the forests and their typical irregular distribution with the presence of hollows and lava tubes also contributes to rapid rainwater runoff and oxygenation of the whole system. To the best of our knowledge, the ability of basaltic lava flows to exchange CH 4 with the atmosphere has never been reported. We speculate that the CH 4 oxidation occurs in volcanic pores covered with a thin layer of biofilm consisting of a broad spectrum of microbial communities, including (inter alia) methanotrophs (e.g. Gomez-Alvarez et al., 2007;Byloos et al., 2018). Methanotrophs seem to colonize the newly formed lava already within 40 yr, but the colonization is closely related to the development of plant and microbial communities and local climatic conditions (King & Nanba, 2008). We expect that volcanic surfaces' low water absorption and retention capacity (Gomez-Alvarez et al., 2007;Byloos et al., 2018) promote aerobic conditions and support, inter alia, CH 4 -oxidizing microorganisms leading to the observed CH 4 uptake by volcanic surfaces.
Soil CH 4 flux measurements on lava flows are rare. The only study conducted on relatively fresh volcanic deposits detected soil CH 4 uptake in a tropical rain forest (c. 300 yr old) on the islands of Hawaii (King, 2003). The CH 4 consumption rate in that study (reaching -75 µg-CH 4 m À2 h À1 ) is in accordance with the findings Positive fluxes indicate trace gas emission and negative fluxes trace gas uptake. The box boundaries mark the 25 th and 75 th percentiles and whiskers the 10 th and 90 th percentiles. As indicated by the letter 'a' above each bar, there were no statistically significant differences in fluxes among stem heights at P < 0.05. One-way ANOVA was applied for CH 4 and N 2 O fluxes and Kruskal-Wallis one-way ANOVA on ranks was used for CO 2 fluxes.

Research
New Phytologist of our study (-83.3 µg-CH 4 m À2 h À1 ). The similar CH 4 consumption potentials of soil and volcanic surfaces were accompanied by soil carbon dioxide (CO 2 ) emissions substantially exceeding those from volcanic surfaces (Fig. 2c). By comparison, the volcanic deposits in Hawaii produced twice as much CO 2 (58.8 mg-CO 2 m À2 h À1 ; King, 2003). The very low CO 2 emissions of volcanic surfaces in R eunion might be explained by the low availability of organic carbon (C) and resulting low respiration rates (King, 2003).
Tree stems and cryptogamic stem covers Our study clearly determined all mature tree stems of the dominant tree species to be net sinks of CH 4 (Fig. 3a). To the best of our knowledge, such consistent consumption of CH 4 by tree stems constitutes an unique finding, because trees are generally known as CH 4 sources with high individual variability in CH 4 emission rates, depending on tree species, climatic zone, forest ecosystem type, environmental and meteorological conditions, and seasonal dynamics (Barba et al., 2019;. In particular, stems of Amazonia trees that are well adapted to inundation are regarded as extraordinarily strong CH 4 sources (Pangala et al., 2017). Trees from other tropical areas seem also to play important roles in forest CH 4 budgets, even with much lower detected emissions than in Amazonia (Pangala et al., 2013;Welch et al., 2018;Jeffrey et al., 2019;Sj€ ogersten et al., 2020). It is expected that CH 4 emitted from these trees mostly originates from the deeper soil layers, as reported for the only study done on tropical upland soils (Welch et al., 2018). In that case, trees were observed to serve as 'chimneys' for soil-produced CH 4 to reach the atmosphere without being oxidized in wellaerated soil surface layers.
Only boreal tree species are known to be capable of taking up CH 4 from the atmosphere by their leaves and stems (Sundqvist et al., 2012;Machacova et al., 2016b), although the stem uptake potential was detected only in the dormant season and was rather negligible (Machacova et al., 2016b). Welch et al. (2018) detected irregular CH 4 uptake by stems of Simarouba amara in a tropical upland forest in Panama during the transition from dry to wet season. Those authors assumed a diffusion gradient from the atmosphere via tree stems into the soil with predominant CH 4 oxidation, which might result in tree stem uptake of trace gases.
Our incubation experiments with the most abundant cryptogams on studied trees consistently revealed that all freshly collected cryptogams (bryophytes) were net CH 4 sinks independently of light conditions (Fig. 3a). The estimated CH 4 uptake rates per area unit were of the same order of magnitude as stem uptake rates measured under field conditions (Fig. 3a). Therefore, it seems that the epiphytes are co-responsible for the observed CH 4 uptake by tropical trees on R eunion. The only known study on cryptogamic CH 4 exchange presents various cryptogamic species as small CH 4 sources (0.28 ng-CH 4 g À1 h À1 , Lenhart et al., 2015; vs -47 AE 16 ng-CH 4 g À1 h À1 , this study). We hypothesize that CH 4oxidising microorganisms (i.e. methanotrophs) can be involved in the CH 4 consumption observed by trees and cryptogams. Future analyses of wood and cryptogams samples for the pmoA gene controlling CH 4 oxidation could bring more explanation to this topic.
In conclusion, the studied tropical rain forest located on a basaltic lava flow seems to be a strong net sink of CH 4 , with soil and volcanic surfaces predominantly being responsible for the CH 4 uptake. We dare to say that, owing to the reasons stated above, no elevated water table levels should be expected during the rainy season, and the studied tropical rain forest on lava flow might be a substantial sink for CH 4 all through the year. The stems of the six tree species studied contributed 7.1% to the forest floor CH 4 uptake (Fig. 6b). As there are > 80 tree species present with high tree density, we estimate that the overall contribution of all tree stems to the forest floor CH 4 uptake might be as much as 41% and the overall forest uptake potential can reach to À1176 mg-CH 4 ha À1 h À1 . The trees growing in the tropical rain forest on lava flows, therefore, are significant contributors to ecosystem CH 4 uptake. Theoretically, the contribution can be even greater, as foliage, known from the first available studies as important contributors to ecosystem CH 4 exchange (Machacova et al., 2016a), is not included in this case study.
Tree stems and cryptogamic stem covers are net nitrous oxide sinks By contrast with their relationship to CH 4 , soil and basaltic lava flows do not significantly exchange nitrous oxide (N 2 O) with the atmosphere (Fig. 2b). The reason might be nitrogen (N) limitation in the thin soil layer, thus resulting in very limited biomass increment in the studied forest over the past 10 yr (O. Flores, pers. comm.; Bennici, 2015). Moreover, the young lava flow is probably still poor in N-rich nutrients and organic compounds (King, 2003;Gomez-Alvarez et al., 2007).
Nevertheless, the studied trees were identified as net sinks of N 2 O (Fig. 3b). Their uptake potential was weakly and negatively correlated with stem CO 2 efflux (Fig. 5b), which means that low N 2 O consumption by tree stems was associated with low stem CO 2 efflux. The net stem CO 2 efflux, an indicator of tree physiological activity, is a result of stem respiration, radial CO 2 diffusion from the transpiration stream and CO 2 re-fixation on the stem surface (Aubrey & Teskey, 2009;H€ oltt€ a & Kolari, 2009;Bloemen et al., 2013). A physiological dependence of N 2 O exchange has been observed in boreal trees, whose seasonality in stem N 2 O release followed the tree physiological activity, particularly processes of CO 2 uptake and release (Machacova et al., 2019). A closer connection between plant N 2 O and CO 2 fluxes was detected early under a wide range of controlled environmental conditions in species belonging to cryptogamic covers (Lenhart et al., 2015;Machacova et al., 2017) and Spermatophyta (Machacova et al., 2017;Lenhart et al., 2019).
Similar to tree stems, cryptogams covering the studied tree stems up to the crowns also were detected as net N 2 O sinks, and their N 2 O uptake rates were of the same order of magnitude (Fig. 3b). These findings are unique, as the majority of studies on mature trees under natural field conditions present trees only as weak N 2 O emitters (Machacova et al., 2019). The only known study presenting trees (European beech, temperate mountain forest) as substantial net sinks of N 2 O showed cryptogamic stem covers to be co-responsible for the observed N 2 O uptake by tree stems (Machacova et al., 2017). We hypothesize that the N 2 O, which is taken up by the cryptogams, might be reduced to N 2 in the final step of the denitrification pathway. Some irregular N 2 O uptake by trees also has been observed by boreal trees (Machacova et al., 2019) and tropical upland trees (Welch et al., 2018). In general, trees seem to exchange N 2 O (and CH 4 ) with the atmosphere in both directions under certain conditions. However, the processes and mechanisms behind and fate of the N 2 O (and CH 4 ) taken up by the tree stems and cryptogams are still unknown and warrant further investigation.
The studied tree species contributed À64% to soil N 2 O exchange and, therefore, reduced the weak soil N 2 O emission potential (Fig. 6d). We expect an overall N 2 O uptake potential of all tree species in the forest might represent as much as À374% of the soil exchange, thus turning the tropical rain forest into a strong sink of N 2 O (À50.45 mg-N 2 O ha À1 h À1 ).

Conclusion and future perspectives
This study shows for the first time that native trees in tropical lowland rain forest growing on a basaltic lava flow can be important sinks of atmospheric CH 4 and N 2 O and can contribute as much as 41% and À374% to the soil CH 4 and N 2 O exchange, respectively. Also, cryptogamic stem covers were identified as substantial CH 4 and N 2 O sinks. This 'plant' uptake was accompanied by strong CH 4 uptake by soil and volcanic surfaces of the basaltic lava flow.
In relation to ecosystem CH 4 exchange, the lowland rain forests growing on lava flows seem to behave differently from more commonly studied rain forests, not acting as strong CH 4 sources but rather as significant CH 4 sinks. More studies in different tropical Fluxes of methane (CH 4 ) (a, b), nitrous oxide (N 2 O) (c, d) and carbon dioxide (CO 2 ) (e, f) from tree stems and adjacent soil expressed per stem or soil surface area unit (a, c, e) and scaled up to unit ground area of the tropical rain forest (b, d, f). The fluxes are expressed as medians (solid lines) and means (dashed lines) of measurements from all studied trees of six tree species (n = 24) and all studied soil positions (n = 24). Positive fluxes indicate trace gas emission; negative fluxes trace gas uptake. The box boundaries mark the 25 th and 75 th percentiles and whiskers the 10 th and 90 th percentiles. Statistically significant differences in fluxes at tree stem and soil level at P < 0.05 are indicated by different letters above the bars. Mann-Whitney rank-sum test was applied for the gas flux pairs. The contributions of stem fluxes of six tree species to the soil fluxes (equal to 100%) are expressed as percentages of the soil flux.

Research
New Phytologist volcanic regions are needed to understand the trace gas exchange potential and mechanisms of these specific ecosystems.
Because our case study was limited by time, the results need to be extended by examining the seasonal variability in CH 4 and N 2 O exchange for a larger set of tree species, further including fluxes from leaves, and simultaneous measurements of soil and plant chemical and hydrological properties following the previous work of Meunier et al. (2010), and seeking to understand the role of cryptogams and basaltic lava flows in the CH 4 and N 2 O exchange under different climatic conditions.
Our results highlight the importance of previously unexplored lowland tropical rain forests on basaltic lava flows in CH 4 and N 2 O uptake and, therefore, can be used to improve national greenhouse gas (GHG) inventories. As the studied lowland tropical rain forest in Mare Longue Nature Reserve falls within the range of tropical rain forests (biogeochemistry, biomass increment, litterfall; Kirman et al., 2007;Meunier et al., 2010), this study can be justified as a model ecosystem of lowland tropical rain forests. These forests are the most species-rich ecosystems in the world and are of high importance in regulating C and water cycling in the global climate system. However, they are under the greatest pressures from anthropogenic disturbances (acceleration of deforestation of original lowland tropical rain forests, impacts of global warming and drought events) and start to occur only as isolated forest fragments with decreasing biodiversity (Turner & Corlett, 1996;Corlett, 2011;Corlett & Primack, 2011). Confirming the unequivocal CH 4 and N 2 O uptake potential by future studies would bolster the argument that the protection of such dynamic ecosystems in volcanic regions around the world should be strengthened not only because of their great biodiversity but also for their contribution to reducing GHGs in the atmosphere.