Impact of forest plantation on methane emissions from tropical peatland

Abstract Tropical peatlands are a known source of methane (CH4) to the atmosphere, but their contribution to atmospheric CH4 is poorly constrained. Since the 1980s, extensive areas of the peatlands in Southeast Asia have experienced land‐cover change to smallholder agriculture and forest plantations. This land‐cover change generally involves lowering of groundwater level (GWL), as well as modification of vegetation type, both of which potentially influence CH4 emissions. We measured CH4 exchanges at the landscape scale using eddy covariance towers over two land‐cover types in tropical peatland in Sumatra, Indonesia: (a) a natural forest and (b) an Acacia crassicarpa plantation. Annual CH4 exchanges over the natural forest (9.1 ± 0.9 g CH4 m−2 year−1) were around twice as high as those of the Acacia plantation (4.7 ± 1.5 g CH4 m−2 year−1). Results highlight that tropical peatlands are significant CH4 sources, and probably have a greater impact on global atmospheric CH4 concentrations than previously thought. Observations showed a clear diurnal variation in CH4 exchange over the natural forest where the GWL was higher than 40 cm below the ground surface. The diurnal variation in CH4 exchanges was strongly correlated with associated changes in the canopy conductance to water vapor, photosynthetic photon flux density, vapor pressure deficit, and air temperature. The absence of a comparable diurnal pattern in CH4 exchange over the Acacia plantation may be the result of the GWL being consistently below the root zone. Our results, which are among the first eddy covariance CH4 exchange data reported for any tropical peatland, should help to reduce the uncertainty in the estimation of CH4 emissions from a globally important ecosystem, provide a more complete estimate of the impact of land‐cover change on tropical peat, and develop science‐based peatland management practices that help to minimize greenhouse gas emissions.


| INTRODUC TI ON
Methane (CH 4 ) is the second most important anthropogenic greenhouse gas after carbon dioxide (CO 2 ) and its concentration is continuing to increase Nisbet et al., 2019). The global warming potential (GWP) of CH 4 is 34 times that of CO 2 on a 100 year basis when including climate-carbon feedbacks (Myhre et al., 2013). Due to its short atmospheric life span of about 10 years and relatively high GWP, there is increasing interest in reducing CH 4 emissions in order to meet global temperature targets (Collins et al., 2018). Current and future regional and global CH 4 budgets and mitigation strategies require better quantitative and process-based understanding of CH 4 sources, pathways, and removals under climate and land-use change (Saunois et al., 2016).
Since the 1980s, extensive areas of Southeast Asian peatlands have experienced land-cover changes (Miettinen, Shi, & Liew, 2016;Wijedasa et al., 2018), driven by transmigration, local population growth, and ongoing economic development. The 2015 land-cover distribution for the insular Southeast Asian peatlands reveals that half of all former peatland forest is managed as either small-holder agriculture or industrial plantation, while around 29% is characterized as intact or degraded natural peat swamp forest (Miettinen et al., 2016). The remaining 21% of the peatlands are covered by open undeveloped areas, fern, low/tall shrub, and secondary regrowth forest (Miettinen et al., 2016). Agriculture and forest plantation on peatlands require the maintenance of groundwater level (GWL) below the root zone to support the required level of productive growth. Maintaining the GWL below the surface alters the CH 4 dynamic by weakening the potential for CH 4 production and increasing the potential for CH 4 oxidation in the upper peat layers (Furukawa, Inubushi, Ali, Itang, & Tsuruta, 2005;Melling, Hatano, & Goh, 2005). Given the potential importance of tropical peatlands in global CH 4 budgets, it is important to understand any effects of land-cover changes on CH 4 emissions from tropical peatlands.
When the balance between CH 4 production and consumption is positive, CH 4 can be emitted to the atmosphere via: (a) diffusion from soil and water surfaces, (b) ebullition from water surfaces, or (c) vegetation-mediated transport through aerenchymatous and air-filled tissues in herbaceous plants and trees (Jauhiainen & Silvennoinen, 2012;Pangala et al., 2013). In addition, CH 4 can be emitted from terrestrial arthropods such as termites (Jeeva, Bignell, Eggleton, & Maryati, 1999) and plants producing CH 4 in aerobic conditions (Keppler, Hamilton, Brass, & Röckmann, 2006). The contribution of each pathway to total ecosystem CH 4 exchange varies within and among peatland ecosystems depending on surface microtopography (hummock vs. hollow), GWL, peat temperature, vegetation composition and structure, and land-use practices (Melling et al., 2005;Pangala et al., 2013). Variation in plant physiological processes driven by solar radiation might substantially influence vegetation-mediated transports as observed in northern peatlands (Kim, Verma, Billesbach, & Clement, 1998;Long, Flanagan, & Cai, 2010;Nisbet et al., 2009;van der Nat, Middelburg, van Meteren, & Wielemakers, 1998). Thus, significant spatial and temporal variability in CH 4 emissions from tropical peatlands can be anticipated, yet available data rarely allow analysis of how such variability influences annual emissions.
This leads to uncertainty in estimates of the current and future contribution of tropical peatlands to regional and global CH 4 budgets (Saunois et al., 2016).
Given these uncertainties, we need to improve our understanding of the spatiotemporal and environmental variability that drive exchange strength and direction in order to better understand the potential CH 4 exchanges that may result from any future climate or land-use change scenarios. Micrometeorological methods (such as eddy covariance) provide half-hourly measurements of turbulent CH 4 exchanges between an entire ecosystem and the atmosphere above the vegetation canopy (Aubinet et al., 2000).
Hence, eddy covariance measurements incorporate all existing CH 4 sources and removals that can vary significantly within an ecosystem in both space and time arising from variation in environmental conditions.
In the above context, we used the eddy covariance technique to measure net ecosystem CH 4 exchange over two land-covers in a single peatland hydrological unit on the Kampar Peninsula in Sumatra, Indonesia: (a) a natural forest, and (b) a forest plantation (Acacia crassicarpa). Measurements were conducted for more than four site-years (October 2016-May 2019 over the Acacia plantation and June 2017-May 2019 over the natural forest). The main objectives of this study were to: (a) determine the magnitudes of CH 4 exchanges from tropical peatlands while incorporating all existing sources and removals, and (b) understand the link between temporally varying CH 4 exchanges and associated changes in the environmental controls.
We hypothesized that a lower GWL would reduce vegetation-mediated CH 4 transport to the atmosphere in the managed peatland. We evaluated this hypothesis over timescales ranging from diurnal to annual. These results were then used to quantify the impact of Acacia plantation, considering the change in CH 4 exchanges due to the associated altered landscape, as one component of the ecosystem greenhouse gas balance. Finally, we considered the relevance of these results for tropical peatland greenhouse gas emissions reporting, climate change mitigation policies and land-use management.

| Study area
The Kampar Peninsula is a coastal tropical peatland of around 700,000 ha ( Figure 1a). This ombrotrophic (acidic and nutrient-poor) peatland is largely formed within the past 8,000 years (Dommain, Couwenberg, & Joosten, 2011). The study area has a humid tropical climate (warm year-round) with average monthly air temperature ranging from 29 to 32°C (Badan Meteorologi, Klimatologi dan Geofisika, 1994-2017. Average annual rainfall for the last 5 years (2014-2018) is ~1,800 mm with two wet seasons (March-April and October-December) and two dry seasons (January-March and May-August). The peninsula is characterized by a large, relatively intact central forest area surrounded by a mosaic of smallholder F I G U R E 1 Land-cover map of the Kampar Peninsula, Sumatra, Indonesia and the location of research flux tower sites (a), photos of the eddy covariance instruments installed at the top of the tower at the natural forest (b), and the Acacia plantation (c), and integrated eddy covariance footprint contour lines from 10% to 80% in 10% intervals over the natural forest for June 2017-May 2019 (d), and the Acacia plantation for October 2016-May 2019 (e) agricultural land (largely oil palm, Elaeis guineensis), and industrial fiber wood plantation (largely A. crassicarpa), smaller secondary forest areas, and undeveloped open and degraded land (Figure 1a; Miettinen et al., 2016). Natural forest and Acacia plantation together occupy around 80% of the peninsula (Figure 1a).
Above-canopy eddy covariance flux towers were established at the Acacia plantation and the natural forest in 2016 and 2017, respectively, for the purpose of measuring net ecosystem CO 2 and CH 4 exchange (Figure 1b,c; note that CO 2 flux measurements will be reported separately). The terrain around the towers is flat (slope <0.05%) and land-cover and topography are homogenous for at least 3 km in all directions at both sites, ensuring a good fetch and a consistent land-cover-related signal regardless of wind direction. The relatively close proximity of the natural forest and the Acacia plantation sites (~80 km apart) within the same peatland hydrological unit avoids potentially confounding variables such as climatic differences, past natural succession, and to some extent geomorphological formation ( Figure 1a). Thus, although it is inherently difficult and expensive to replicate flux measurements using the eddy covariance technique, our sites should provide a robust and unbiased basis for evaluating the impact of land-cover change (from peat swamp forest to Acacia plantation) on CH 4 exchanges.
The natural forest is characterized as pristine peat swamp forest (Miettinen et al., 2016). The forest structure is mixed, and the canopy is uneven with the tallest canopy in a range of 28-35 m.
Tree density with diameter at breast height >5 cm was 1,343 trees per hectare. The dominant tree species of the overstory are Shorea uliginosa, Calophyllum ferrugineum, and Syzygium spp.; together they represent around 75% of the overstory vegetation (Table 1). The understory is dominated by Pandanus spp., Cyrtostachys renda, and Nepenthes spp. The forest floor is uneven with a hummock-hollow microtopography, and covered with tree debris, root mat, and leaf litter. Hollow surfaces are often 20-40 cm lower than hummock tops. The average area ratio of hollow to hummock was 3:1 around the tower. The surface peat type is fibric and the average peat thickness is ~9 ± 1 m in the area surrounding the tower. The surface peat pH is 3.6 ± 0.1 and the GWL fluctuates seasonally with the rainfall variation (see Section 3). An integrated climatologic footprint analysis (Kljun, Calanca, Rotach, & Schmid, 2015) indicated that approximately 80% of fluxes were derived within 1,200 m in the upwind direction (Figure 1d), and thus originated within the pristine peat swamp forest as characterized by Miettinen et al. (2016).
At the forest plantation, A. crassicarpa trees are grown for fiber production on a 5 year rotation from planting to harvesting. When measurements began in October 2016, the trees were already at the end of the plantation cycle. In March-April 2017, the mature trees, which had achieved an average height of 20 m, were harvested.
Replanting at a density of 1,667 trees per hectare (3 m × 2 m spacing) took place within 2 weeks after harvesting. One kg boiler wood ash per tree was applied around the seedlings during planting as per the standard operational procedure, without additional fertilizers. In May 2019, 2 years after replanting, the canopy height was ~17 m.
The ground surface within the plantation area is relatively even, without a hummock-hollow microtopography, and with very little understory vegetation. The surface peat type is hemic and the average peat thickness is ~7 ± 0.8 m in the surrounding area of the tower.
The surface peat pH is 3.4 ± 0.1. GWLs in plantation are actively managed to support the required level of productive growth via an extensive network of topographically defined water management zones, controlled by outlet sluices, and supported by large-scale and continuous rainfall and water level monitoring (Evans et al., 2019).
Water management zones comprise of ditches and canals (also used for transportation). The integrated climatologic footprint analysis (Kljun et al., 2015) indicated that (a) approximately 80% of fluxes were estimated to occur within 1,000 m in the upwind direction, and thus originated within the Acacia plantation; and (b) the water surface of ditches and canals represented 2.1% of the flux footprint ( Figure 1e).

Parameter
Natural forest Acacia plantation Furthermore, the upper and lower mirrors of the CH 4 analyzer were manually cleaned on a biweekly basis. Dew condensation, rain, and dirty window events were excluded using an RSSI value of 20% because CH 4 data become noisy below this threshold (Chu et al., 2014;McDermitt et al., 2011). Water vapor densities were meas- Daily rainfall (mm/day) rates were measured using three and six manual bucket systems within 10 km distance from the tower location in the natural forest and the Acacia plantation, respectively. Manual bucket systems were installed 1.5 m above the ground, in an open area so that rainfall was not intercepted by the tree canopy. Soil temperature (T soil , °C) was measured at 15 cm below the peat hollow surface using temperature probe (Stevens Hydra Probe II, Stevens Water Monitoring Systems, Inc.) with three replicates at each tower site.
GWLs (m) were monitored as the water elevation relative to the ground surface (taking the base of the hollows as a datum) every 30 min using a GWL logger (Solinst Levelogger Model 3001). The GWL logger was placed in a perforated polyvinyl chloride (PVC) tube that was inserted vertically into the peat at a distance approximately 30 m away from the towers. The GWL logger also recorded temperature at 150 cm below the peat surface that is below GWL. Additionally, PVC poles were randomly distributed within a 3 km radius around the tower locations to monitor GWL fortnightly). All meteorological sensors took measurements every second and were recorded as one minute averages with a datalogger (Sutron Model 9210 XLITE, Sutron Corporation).
All measuring systems were powered using solar panels along with a rechargeable battery system (65 Watt Solar Package, SunWize Power & Battery). Owing to the large power requirement and cost of a separate CH 4 analyzer, we could not conduct CH 4 profile measurements to calculate CH 4 storage below the flux measurement height (Finnigan, 2006). In theory, accumulated CH 4 below the canopy during nighttime is likely to be released and measured by the EC system following the onset of turbulence after sunrise and the bias on annual sums should be negligible (Xu et al., 2019).

| Eddy covariance data processing
Net ecosystem CH 4 exchange (NEE-CH 4 ) was computed from the 10 Hz vertical wind velocity and CH 4 concentration data using EddyPro software (version 6.2.0, LI-COR Inc.) at a standard averaging interval of half hour period (Aubinet et al., 2000). A de-spiking procedure was applied to detect and eliminate individual out-of-range values for vertical wind velocity and CH 4 concentrations (Vickers & Mahrt, 1997). De-trending was carried out using the block averaging method. A coordinate correction was applied to force the average vertical wind velocity to zero by the planar fit method (Wilczak, Oncley, & Stage, 2001). Frequency response loss corrections were applied to compensate the flux losses at different frequencies (Massman, 2000(Massman, , 2001Moncrieff, Clement, Finnigan, & Meyers, 2004). Fluctuations in CH 4 density due to temperature (thermal expansion) and water vapor (dilution) were corrected using the Webb-Pearman-Leuning correction (Webb, Pearman, & Leuning, 1980) and spectroscopic effects taken into account by EddyPro (Burba, Anderson, & Komissarov, 2019). Differences between deploymentspecific variables, that is, sensor separation distance and instrument placement, were considered while processing the data. We adopted the standard meteorological notation whereby a positive value of NEE-CH 4 represents a net CH 4 flux to the atmosphere, and a negative value indicates net CH 4 uptake from the atmosphere (Aubinet et al., 2000). All NEE-CH 4 values in the paper are reported in mass of CH 4 per unit area per time.
After a set of quality controls, the numbers of high-quality measurements during the course of the study were 38% and 29% for the natural forest and the Acacia plantation, respectively, including measuring system malfunctions due to lightning strikes and power supply failure (Table 2). In other words, we obtained a total TA B L E 2 Summary of the percentage of half-hourly net ecosystem CH 4 exchange data that were removed using various quality control criteria and accepted high quality data
We gap-filled both low-quality and missing data due to instrument malfunction, as is commonly done in eddy covariance studies.
We applied two gap-filling approaches ( We performed MDS gap filling separately for the daytime (06:00-16:00 hr) and the nighttime (18:00-06:00 hr) data. GWL and PPFD were used during the daytime, whereas GWL and T soil above the GWL were used during the nighttime gap-filling. The emissions were similar from both methods at the natural forest (Mann-Whitney test; p = .34,  (2001). The standard deviation of three different flux values derived from friction velocity (u*) thresholds of 5th, 50th, and 95th percentiles were applied as an uncertainty due to u* threshold (σ 2 ) using the REddyProc package (Wutzler et al., 2018).
The flux uncertainty due to gap-filling (σ 3 ) was calculated as the standard deviation of the binned records used to fill the missing value (Wutzler et al., 2018). The total uncertainty in NEE-CH 4 was calculated with the law of propagation of errors (Deventer et al., 2019).
Only high quality measurements were used in the qualitative analysis  Table 4).

| Statistical analyses
Differences between groups of data were examined using t test in

| Environmental conditions
During the course of the study, the PPFD, T air , VPD, and canopy conductance to water vapor above the canopy showed typical diurnal patterns reaching their maximum around noon (Figure 2a-h).
No significant diurnal variation in T soil below the GWL was observed at either site (Figure 2i,j). The diurnal variation in T soil above the GWL was small (<1°C) at the natural forest, due to the closed canopy and high GWL (Figure 2i). Before canopy closure, the Acacia plantation showed a clear diurnal variation (amplitude of 3°C) in the T soil above the GWL, but after canopy closure, the observed diurnal T soil above the GWL amplitude was similar to the natural forest.
Daily average T air fluctuated between 23.3 and 29.9°C as a function of rainfall and cloud cover, without showing any clear seasonality (Figure 3a, Variations in daily air temperature (a, b), soil temperature above and below groundwater level (c, d), cumulative rainfall and groundwater level (e, f), and net ecosystem CH 4 exchanges (g, h) at the natural forest (left panels) and the Acacia plantation (right panels). The vertical bar in panels (a, b, c, d) represents standard deviation. Positive value of groundwater level indicates water level above the peat surface, and negative values indicate water level below the soil surface  plantation, daily average T soil above the GWL ranged from 26.6 to 33.0°C as a function of canopy development, GWL, and cloudiness, without any clear seasonality (Figure 3c,d). The daily average T soil above and below the GWL at the natural forest was statistically different (t test; p < .05; Table 3) and around ~2°C lower than at the Acacia plantation (Figure 3c,d). The average VPD at the natural forest of 3.7 ± 1.9 hPa was significantly lower (40%) than the average of 5.6 ± 2.2 hPa at the Acacia plantation (t test; p < .05; Table 3).
At both sites, the daily cumulative rainfall was highly variable, ranging from 0 to 137 mm (Figure 3e, Asia (Cobb et al., 2017). At the Acacia plantation, GWL rose up during rain events, but remained always below the ground surface ( Figure 3f).
During the study period, the average GWL from six sampling points around the natural forest tower of −0.24 ± 0.14 m was significantly shallower than that of −0.73 ± 0.14 m from 10-20 sampling points around the Acacia plantation tower (Mann-Whitney test; p < .05).

| Net ecosystem CH 4 exchanges
At both sites, the NEE-CH 4 showed a marked peak at around 07:00-10:30 hr (Figure 4a,b), consistent with flushing of CH 4 accumulated in the vegetation canopy at night following the onset of turbulent mixing in the morning (Wong et al., 2018). NEE-CH 4 over the natural forest remained much higher than the nighttime during the remaining day hours and began to decline late in the afternoon ( Figure 4a). NEE-CH 4 over the Acacia plantation began to decline and reached levels similar to the nighttime values after around 10:30 hr ( Figure 4b). Thus, the diurnal variation in NEE-CH 4 was more pronounced over the natural forest (Figure 4a,b).
In order to avoid bias due to flushing of accumulated CH 4 , we considered nighttime NEE-CH 4 from 18:30 to 10:30 hr and daytime from 10:30 to 18:30 hr. This threshold might principally be site specific, but offered an opportunity to examine the diurnal variation in the NEE-CH 4 over our sites. Over the natural forest, daytime median NEE-CH 4 was more than three times higher (29 mg m −2 day −1 ) than at nighttime (8.4 mg m −2 day −1 ; Mann-Whitney test; p < .05; Figure 4c). Furthermore, daytime median NEE-CH 4 was almost three times higher over the natural forest than over the Acacia plantation (Mann-Whitney test; p < .05; F I G U R E 5 Response of the halfhourly net ecosystem CH 4 exchanges to canopy conductance to water vapor (a, e), photosynthetic photon flux density (b, f), vapor pressure deficit (c, g), and air temperature (d, h) at the natural forest (left panels), and the Acacia plantation (right panels). Data were binned by subgroups of 50 values of independent variable and corresponding net ecosystem CH 4 exchange rates and then averaged for the subgroup. The vertical and horizontal bars represent the standard deviation for the subgroup. Note: we excluded measurements from 7:00 to 10:30 hr to avoid the possible bias due to flushing of nighttime accumulated CH 4 . The exclusion of data may have created biases in actual response curves of both ecosystems, but this bias would not change the interpretation   Table 4).
Variation in daytime and nighttime NEE-CH 4 was positively correlated with associated changes in GWL at both sites ( Figure 6).
Notably, the relationships between NEE-CH 4 and GWL for daytime and nighttime over the natural forest were significantly different, whereas the relationships were quite similar over the Acacia plantation ( Figure 6). There was no clear relationship between NEE-CH 4 and the T soil either above or below the GWL (data not shown).
Our measurements showed that the natural forest emitted 9.1 ± 0.9 g m −2 year −1 to the atmosphere (Table 4). Annual NEE-CH 4 over the Acacia plantation were approximately 50% lower than the natural forest, at 4.7 ± 1.5 g m −2 year −1 , suggesting a net reduction of CH 4 exchanges from natural forest to Acacia plantation of −4.4 ± 1.7 g m −2 year −1 (Table 4).

| High GWL supports diurnal variability in NEE-CH 4
Our results show substantial and apparent diurnal variation in the  (Brady, 1997;Sulistiyanto, 2004), and dissolved CH 4 in the root zone can be significant (100-1,500 µmol/L; Hoyt, 2017;Pangala et al., 2013). The magnitude of vegetation-mediated transport seems to be directly regulated by a well-connected root-stem pathway for the CH 4 transport, although it is strongly (if not primarily) controlled by the availability of dissolved CH 4 in the root zone (Covey & Megonigal, 2019;Pangala et al., 2013Pangala et al., , 2017Waddington, Roulet, & Swanson, 1996). At the natural forest site, S. uliginosa, C. ferrugineum, and Syzygium spp. are the dominant species; together they represent around 75% of the tall-canopy vegetation.
When the root zone is inundated, changes in biological processes in vegetation driven by solar energy input might be the most important factors controlling diurnal variation in measured NEE-CH 4 ( Figure 5a-d), as reported in northern peatlands (Chanton, Whiting, Happell, & Gerard, 1993;Garnet, Megonigal, Litchfield, & Taylor, 2005;Kim et al., 1998;Long et al., 2010;Whiting & Chanton, 1996) and recently reported over tropical peatland (Tang et al., 2018) and flooded forest (Dalmagro et al., 2019). At the natural forest, the observed positive correlation between NEE-CH 4 and canopy conductance to water vapor suggests that CH 4 could be dissolved in the water, absorbed by the roots, transported with sap flow, and emitted through the stem by effervescence (Garnet et al., 2005;Nisbet et al., 2009). In addition, the positive correlation between NEE-CH 4 and PPFD, VPD, and temperature (Figure 5b-d) may suggest vegetation-mediated transport through either diffusion or convective throughflow (Brix, Sorrell, & Orr, 1992;Chanton, Martens, Kelley, Crill, & Showers, 1992;Dacey, 1981). Our results are in line with a study in a temperate forested wetland which showed a sudden decrease in CH 4 emissions from Betula pubescens after leaf loss, suggesting physiological control on gas transport (Pangala, Hornibrook, Gowing, & Gauci, 2015). Furthermore, labile organic compounds released from root tissues during photosynthesis and respiration can then be used as substrates by methanogenic archaea, contributing to the diurnal variation in NEE-CH 4 (Chanton et al., 1995;Christensen et al., 2003).
Our study did not aim to conduct direct measurements to establish the relative importance of these different processes. Quantifying the pathway-specific emissions and improving our understanding on the impact of root distribution by depth and dissolved CH 4 concentration profile are important future study (Barba et al., 2018;Megonigal, Brewer, & Knee, 2019).
Nighttime and daytime NEE-CH 4 was positively correlated with associated changes in GWL at both sites ( Figure 6). Nighttime NEE-CH 4 can be considered as the emissions from soil and water surfaces since there would be negligible vegetation-mediated transport. A higher GWL may support larger CH 4 concentration gradients between the peat surface and the atmosphere. Thus, GWL seems to be the key indirect control on CH 4 emissions via diffusion from soil surfaces (Winton, Flanagan, & Richardson, 2017). Overall, the lack of a difference between nighttime NEE-CH 4 over the natural forest and

F I G U R E 6
The relationship between the half-hourly net ecosystem CH 4 exchange and the groundwater level. Data were binned by subgroups of 50 values of groundwater level and corresponding net ecosystem CH 4 exchange rates and then averaged for the subgroup. Note: we excluded measurements from 7:00 to 10:30 hr to avoid the possible bias due to flushing of nighttime accumulated CH 4 . The exclusion of data may have created biases in actual response curves of both ecosystems, but this bias would not change the interpretation the Acacia plantation can be attributed to the high GWL at the natural forest and to the potential presence of emissions from the water surfaces of ditches and canals in the Acacia plantation (Jauhiainen & Silvennoinen, 2012;Manning, Kho, Hill, Cornulier, & Teh, 2019). In addition, the higher soil temperature at the Acacia plantation might have increased CH 4 production , while the higher peat bulk density and the absence of a hollow-hummock microtopography at the Acacia plantation might have lowered CH 4 oxidation by increasing soil moisture content and lowering oxygen diffusion in the peat (Estop-Aragones, Knorr, & Blodau, 2012). For these reasons, the Acacia plantation seems to produce higher nighttime CH 4 emissions compared to the natural forest if the same range of GWL (−0.4 to −0.1) at both sites is considered. The difference between daytime and nighttime NEE-CH 4 can be attributed to vegetation-mediated transport. Thus, our estimated vegetation-mediated transport over the natural forest is 71% of the total daytime emissions, which is in line with the published range for tropical peatlands (Pangala et al., 2013). Overall, at the same GWL range, the natural forest emits higher CH 4 as compared to the Acacia plantation during the daytime, most likely due to the presence of CH 4 emitting trees (i.e., S. uliginosa and C. ferrugineum). However, it should be noted that only a few measurements are available for the Acacia plantation for the same range of GWL ( Figure 6).
The influence of vegetation on CH 4 emissions is strongly dependent on the GWL, and therefore, the interaction among hydrology, vegetation, and CH 4 emissions must be carefully taken into account for process-based modeling ( Figure 6). Predicted changes in rainfall amount, intensity, duration, and frequency and water management practices could affect the dynamics of hydrology in tropical peatlands (Ge et al., 2019), and thereby CH 4 emissions (Saunois et al., 2016).

| GWL controls seasonal variability in NEE-CH 4
The seasonal variation is controlled by the GWL driven by rainfall. Our results show higher NEE-CH 4 during the wet season as compared to the dry season. Other eddy covariance studies in tropical peatlands have reported a similar seasonal pattern in CH 4 emissions (Sakabe et al., 2018;Wong et al., 2018). A study in Amazonian peatland reported lower soil-CH 4 emissions in the wet season as compared to the dry season, where the GWL was 54 cm above the peat surface during the wet season (Teh, Murphy, Berrio, Boom, & Page, 2017). If GWL rises above a limit, soil CH 4 emissions can decrease with flooding depth as gas diffusion may be restricted more as hydrostatic pressure increases along with increasing flooding depth (Ishikura et al., 2019).
Furthermore, the standing water can enhance CH 4 oxidation because it would increase dissolved oxygen and prolong traveling time of CH 4 to the atmosphere (Strack, Waddington, & Tuittila, 2004). Notably, Teh et al. (2017) only reported emissions from soil and water surfaces and did not measure vegetation-mediated transport which can be significant in Amazonian wetlands (Pangala et al., 2017). This highlights that seasonality differs from one pathway to another; thus, caution should be taken when modeling seasonality in CH 4 emissions from tropical peatlands.
In northern peatlands, temperature exerts a strong effect on seasonal variation in CH 4 emissions with an exponential dependence via its influence over enzyme kinetics of CH 4 production and plant growth and development Rinne et al., 2007;Tagesson et al., 2012). Observed fluctuations in both T air and T soil in this study are much smaller than those of northern peatlands.
During the study period, the T soil below the GWL varied within a very narrow range (~2°C) at both sites. This suggests that variation in T soil would have only a minor effect (if any) on variation in NEE-CH 4 . Furthermore, T soil tended to be higher when GWLs were lower; thus, it is difficult to determine the independent effect (if any) that a change in temperature had on CH 4 production and oxidation . For example, if CH 4 oxidation above the GWL increased more rapidly (due to the combination of a deeper aerobic zone and higher rates of microbial activity at a higher temperature) than rates of CH 4 production below the GWL, the net effect of warmer and drier conditions would be a lower NEE-CH 4 .
Our results suggest that the effects of changing rainfall and land management on peat hydrology will be more important than rising temperature as a driver of changes in tropical peatland CH 4 balance in the future.

| Low GWL reduces NEE-CH 4 over the Acacia plantation
At the Acacia plantation, the lower GWL leads to an aerobic root zone (indeed, this is the specific aim of water management in the plantation, to support Acacia growth) which is likely to reduce (but not eliminate) CH 4 production and transport. Firstly, aerobic conditions are unfavorable to methanogens and promote methanotrophy Moore & Roulet, 1993;Strack et al., 2004).
Secondly, as most of Acacia roots are mainly restricted above GWL in the aerated peat layer, this may result in inadequate CH 4 in the root zone to be taken and transported to the atmosphere. But given the GWL fluctuation, it is possible that when GWL rises after a heavy rain event, some portion of the root system will be below GWL, at least for a few days. However, our measurements over the Acacia plantation do not show a diurnal variation in NEE-CH 4 , and this may confirm that the root system remained above the GWL. Finally, it is likely that a substantial fraction of CH 4 emission from the Acacia plantation area could be occurring from the open water surface of the ditch and canal network (Evans, Renou-Wilson, & Strack, 2016;Jauhiainen & Silvennoinen, 2012;Manning et al., 2019), and therefore subject to different environmental controls (Deshmukh et al., 2014). The CH 4 uptake rates in the Acacia plantation are similar to those previously reported over tropical peatlands during the dry season (Sakabe et al., 2018). The CH 4 uptakes might be due to methanotrophy in the aerobic upper peat layer (Arai et al., 2014).

| Potential effects of GWL on CH 4 production and oxidation
Variation in soil redox conditions driven by GWL fluctuation plays an essential role in influencing not only the quantity but also the quality of organic substrate used by the methanogenic archaea for CH 4 production (Girkin et al., 2018;Hoyos-Santillan et al., 2016;Reiche, Gleixner, & Küsel, 2010;Winton et al., 2017). Higher GWLs promote CH 4 production in a relatively large portion of the peat column and restrict the zone in which aerobic CH 4 oxidation can occur (Moore & Roulet, 1993;Moore et al., 2011;Strack et al., 2004). In contrast, lower GWL would narrow the zone of CH 4 production in the peat column and further supporting aerobic CH 4 oxidation above the GWL. In tropical peatlands, the availability of labile organic matter is largely limited to near-surface peat, for example, via root exudation and leaching from fresh litter (Brady, 1997;Könönen et al., 2016). When GWLs are low, most of this labile organic matter will be aerobically decomposed to CO 2 (Itoh, Okimoto, Hirano, & Kusin, 2017) and unavailable for CH 4 production. Therefore, when GWLs are low only organic matter with a greater aromatic content derived from the deeper peat would be available for anaerobic decomposition, restricting CH 4 production (Sakabe et al., 2018). In the Acacia plantation, most of the labile organic matter supplied from harvested vegetation residues (leaf litter, small branches, and roots) and boiler wood ash might be restricted above GWL in the surface peat layer and expected to be aerobically decomposed to CO 2 (Jauhiainen, Hooijer, & Page, 2012). Therefore, the effect (if any) of harvested vegetation residues and boiler wood ash on CH 4 production would be minor.

| Comparison of NEE-CH 4 with other studies
Our annual NEE-CH 4 over the natural forest are in the same range as those measured using the eddy covariance technique above a tropical peatland in the presence of CH 4 -transporting trees (10.0-14.4 g m −2 year −1 ; Tang et al., 2018;Wong et al., 2018). In the absence of CH 4 -transporting trees, a study in a tropical peatland reported no significant diurnal pattern in NEE-CH 4 (Sakabe et al., 2018) and far lower annual CH 4 emissions (0.12-0.23 g m −2 year −1 ), despite similar GWLs to our forest site. The chamber-based total ecosystem flux including tree CH 4 emissions to an average height of 15 m based on the power function relationship from a tropical peatland is lower than our results over the natural forest (Pangala et al., 2013). Despite the higher GWL as compared to our study, the lower emissions in Pangala et al. (2013) can be attributed to (a) a lower hollow to hummock area ratio (1:1), as CH 4 emissions from hollows can be up to 50 times higher as compared to hummocks (Pangala et al., 2013); and (b) possible underestimation of vegetation-mediated transport. Emissions from young trees exceed those of mature trees by orders of magnitudes (Pangala, Gowing, Hornibrook, & Gauci, 2014), but Pangala et al. (2013) reported on emissions from mature trees. Also shoots can emit up to 10 times more than stems in a boreal forest (Machacova et al., 2016), but were not included in Pangala et al. (2013). Furthermore, entire trees may release CH 4 , albeit at the lower rates from their higher portions.
Our annual NEE-CH 4 over the natural forest are around two times higher than the IPCC CH 4 emissions factor for rewetted tropical peatland, derived from undrained sites (Blain et al., 2014). This difference could be attributable to vegetation-mediated transport, which was not captured by most of the studies used to derive the IPCC CH 4 emission factor (Blain et al., 2014). Our annual NEE-CH 4 over the natural forest are nevertheless lower than those reported from Amazonian peatlands (Teh et al., 2017) and floodplain wetlands (Dalmagro et al., 2019;Pangala et al., 2017). In Amazonian peatlands, CH 4 production is greater owing to high nutrient status and soil pH, and low recalcitrant carbon (Wassmann et al., 1992). In addition, methanotrophy is generally less effective because of increased anoxic and stratified, water-submerged sediments (Bartlett et al., 1988;Devol & Rickey, 1990).
Our annual NEE-CH 4 over the natural forest is similar to emissions from northern bogs (average = 9.5 g m −2 year −1 ) and around two times lower than CH 4 emissions from northern fens (average = 20.5 g m −2 year −1 ; Abdalla et al., 2016). Higher temperatures in tropical peatlands favor greater humification, selective removal of reactive labile carbohydrates, and accumulation of aromatic content leading to a highly recalcitrant residual peat (Brady, 1997;Hodgkins et al., 2018). This results in low substrate availability for CH 4 production in the woody peat where there is a high aromatic lignin content (Miyajima, Wada, Hanba, & Vijarnsorn, 1997;Sakabe et al., 2018).
In northern peatlands, peat is mainly derived from mosses, sedges, and herbs which contain a high carbohydrate and lower aromatic content (Hodgkins et al., 2018). This supports higher CH 4 production in northern peatlands, despite lower temperatures (Sundh, Nilsson, Granberg, & Svensson, 1994;Updegraff, Pastor, Bridgham, & Johnston, 1995). Tropical peatlands also typically have higher vertical and lateral recharge rates, driven by higher hydraulic conductivity than northern peatlands , making them susceptible to rapid flushing of the dissolved CH 4 after rainfall. This could limit CH 4 accumulation in near-surface porewaters, reducing the potential for diffusion, ebullition, and vegetation-mediated transport, but could increase emissions via drainage waters. In contrast, lower vertical and lateral recharge rates in northern peatlands support the buildup of dissolved CH 4 concentrations, and result in substantial ebullition and a high CH 4 concentration near the surface soil causing high diffusive and vegetation-mediated transport (Hoyt, 2017).
Our annual NEE-CH 4 at the Acacia plantation is around 18 times higher than the IPCC CH 4 soil-derived emission factor for this category, which is mostly based on soil CH 4 flux measurements (Drösler et al., 2014 (Myhre et al., 2013), this implies a CH 4 emission of 3.1 t CO 2 eq ha −1 year −1 from natural forest. Applying a long-term peat accumulated CO 2 rate of around 2.6 t CO 2 ha −1 year −1 since their formation (Dommain et al., 2011), the 100 year net warming impact for tropical peatland would be 0.5 t CO 2 eq ha −1 year −1 .
Over longer time-horizons, the shorter atmospheric lifetime of CH 4 compared to CO 2 means that an ecosystem that is in approximate greenhouse gas balance based on 100 year net warming impact will have a net cooling impact if it acts as a sustained CO 2 sink and a steady CH 4 source (Allen et al., 2018;Frolking, Roulet, & Fuglestvedt, 2006). However, according to the current IPCC assessment, tropical peatlands are in approximate CO 2 balance (Drösler et al., 2014); therefore, the net warming impact value would be 3.1 t CO 2 ha −1 year −1 . Nevertheless, our data confirm that CH 4 emissions from tropical peatlands should be included in landscape level greenhouse gas budgets (Miettinen, Hooijer, Vernimmen, Liew, & Page, 2017;Wijedasa et al., 2018).

| Impact of Acacia plantation on CH 4 emissions
We present here an assessment of the impact of forest plantation on CH 4 emissions associated with the altered landscape (i.e., Acacia plantation). By definition, the impact represents the actual CH 4 exchange with the atmosphere in addition to the exchange that existed in the pre-existing natural landscape, and thus represents the exchange that can be directly attributed to the creation and existence of the Acacia plantation (Prairie et al., 2018;Teodoru et al., 2012).
Our measurements indicate that both studied ecosystems in the tropical peatland functioned as net CH 4 sources to the atmosphere (Table 4, Figure 7). Therefore, our results indicated that the impact of the Acacia plantation was to reduce CH 4 emissions by 4.4 ± 1.7 g m −2 year −1 (  Figure 7). If we apply a 100 year GWP of 34 (Myhre et al., 2013), this implies an emission reduction of 1.5 t CO 2 eq ha −1 year −1 .
For comparison, the IPCC's Tier 1 default emission factor for CO 2 from Acacia plantation on tropical peat is 73 t CO 2 ha −1 year −1 (Drösler et al., 2014), which is larger than the natural forest. Measurements of net ecosystem CO 2 exchanges over the natural forest and the Acacia plantation are being conducted (C. Deshmukh, unpublished data); results of this ongoing study will be published in due course, following the completion of one 5 year Acacia plantation cycle, and will also take into consideration the biomass harvested from the plantation.
These measurements will lead to a better understanding of the climate footprint of Acacia plantation Petrescu et al., 2015).
The estimated impact of Acacia plantation on CH 4 exchange related to land-cover change that we present here is by no means invariant in time and space. In addition to variations related to natural hydrology, the impact is also likely to vary with actual water F I G U R E 7 Impact of the Acacia plantation on net ecosystem CH 4 exchange from tropical peatland management practices in plantation landscapes. Furthermore, results presented here are specific for Acacia plantation; thus, caution should be taken when extrapolating to other agriculture in the region (e.g., sago, oil palm, rubber plantations, etc.) with different water management practices and fertilizer applications (Hergoualc'h & Verchot, 2012). To evaluate the impact of land-cover change on global peatland CH 4 emissions, more ecosystem-scale flux measurement studies are needed.
In conclusion, our half-hourly multi-year NEE-CH 4 measurements directly captured and integrated "hot spot and hot moment" dynamics of all known and unknown sources and removals in the studied ecosystems. The observed high variability in NEE-CH 4 suggests complex nonlinear process-level controls on CH 4 exchange between tropical peatlands and the atmosphere. Our results provide some of the first reliable information on the magnitudes of CH 4 exchange at a tropical peatland ecosystem scale, demonstrating that traditional manual soil chamber techniques provide an incomplete picture of the total CH 4 flux, and improving mechanistic understanding based on high temporal resolution measurements of NEE-CH 4 and key environmental variables such as the sensitivity of emissions to GWL.
Our data indicate that the Acacia plantation on tropical peatland results in significant reductions in CH 4 emissions compared to the natural system, although the associated cooling impact is likely to be smaller than the accompanying warming impact of higher CO 2 and nitrous oxide emissions. More ecosystem-scale measurements are needed to fully evaluate the effect of land-cover change on the greenhouse gas balance, at a larger number of sites and over long time periods, in order to develop science-based, climate-smart management practices for tropical peatlands.

ACK N OWLED G EM ENTS
The authors thank Asia Pacific Resources International Ltd (APRIL) and Riau Ecosystem Restoration (RER) for providing financial and logistic support. The contributions of CE, SP, VG, SS, FA, and AL form part of their role to the Independent Peat Expert Working Group (IPEWG), which was set up by APRIL to provide objective sciencebased advice on peatland management. ARD acknowledges financial supports from APRIL provided for consultation on eddy covariance data protocols and analysis.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.