Model Development Model Development

Abstract: It is estimated that tropical forest soils contribute 6.2 Tg yr(-1) (28 %) to global methane (CH4) uptake, which is large enough to alter CH4 accumulation in the atmosphere if significant ...


Introduction
Methane (CH 4 ) is an important atmospheric trace gas because it influences both the energy and the oxidant balance of the earth's atmosphere.Presently, the atmospheric concentration of CH 4 is about 1800 ppbv, which accounts for about 0.48 W m −2 of the total anthropogenic radiative forcing (Denman et al., 2007).About 75 % of the global CH 4 source strength, which is about 600 Tg yr −1 , originates from biogenic sources wherein CH 4 is exclusively produced by methanogenic microorganisms (Conrad, 1989).Although CH 4 is primarily produced in wetland soils, CH 4 production can also occur in upland soils during high rainfall or wet season, for example in anaerobic microsites inside soil Published by Copernicus Publications on behalf of the European Geosciences Union.5368 E. Veldkamp et al.: Nitrogen-limited methane uptake in tropical forest soils aggregates (Keller and Reiners, 1994).In well-aerated soils, CH 4 is oxidized by methanotrophic microorganisms and CH 4 oxidation normally exceeds production, which results in a net CH 4 uptake.The largest biogenic sink of atmospheric CH 4 is through uptake by upland soils, which contributes about 5 % to the total removal of CH 4 from the atmosphere (Reeburgh, 2003).
Tropical ecosystems play an important role in the production and uptake of atmospheric CH 4 (Keller and Matson, 1994).In tropical forest areas, known wetland sources of CH 4 production do not suffice to explain the observed high CH 4 concentrations over Neotropical forests (Frankenberg et al., 2008), and some "canopy" wetlands may contribute significantly to the CH 4 production (Martinson et al., 2010).Most tropical forests grow on well-drained upland soils that are too dry to emit CH 4 but act instead as an important sink for atmospheric CH 4 (Kiese et al., 2003).In a review where measurements were stratified according to climatic zone, ecosystem and soil texture, the total global CH 4 uptake was estimated at 22.4 Tg yr −1 , of which 9.2 Tg yr −1 (41 %) occurred in tropical ecosystems (Dutaur and Verchot, 2007).The contribution of tropical forest soils to global CH 4 uptake was estimated at 6.2 Tg yr −1 (28 %), which is large enough to alter the CH 4 accumulation in the atmosphere if significant changes would occur to this sink.
CH 4 fluxes at the soil surface are the result of methanogenesis and CH 4 oxidation, which can occur simultaneously in aerated soils (Yavitt et al., 1995).The microorganisms involved in CH 4 oxidation are methanotrophic bacteria and ammonium-oxidizing bacteria.Most methanotrophic bacteria use CH 4 as their only source of carbon and energy and all use methane monooxygenase in the first step of CH 4 oxidation (Hanson and Hanson, 1996).Methanotrophic bacteria are separated into Type I and II according to their biochemical pathways of oxidizing CH 4 .Type I methanotrophs are generally non-N-fixing organisms, while Type II methanotrophs can fix atmospheric N 2 but can also assimilate mineral N (Hanson and Hanson, 1996).Depending on the CH 4 concentration that they live on, two groups of methanotrophs can be distinguished: one group contains "low affinity" methanotrophs which are adapted for growth at high CH 4 concentrations (e.g. in rice fields), and the other group contains "high affinity" methanotrophs which are able to make use of the atmospheric CH 4 concentrations (around 1.8 ppm).Ammonium-oxidizing bacteria can also oxidize CH 4 through the enzyme ammonia monooxygenase, which can react with CH 4 instead of NH + 4 (Bédard and Knowles, 1989).The increased use of nitrogen (N) fertilizers, fossil fuel, and cultivation of N-fixing crops have more than doubled the amount of "reactive" nitrogen (N r ) cycling worldwide (Vitousek et al., 1997).In the past decades, this has led to enhanced N r input in forest ecosystems, especially in economically developed regions of the temperate zone.Projections are that the input of N r will increase substantially in tropical regions such as Southeast Asia and South and Cen-tral America due to increasing agricultural and industrial use of N (Galloway et al., 2008).A recent study suggested that elevated anthropogenic N r deposition is probably already widespread in tropical forests (Hietz et al., 2011).
Elevated depositions of mineral N (ammonium (NH + 4 ) and nitrate (NO − 3 )) and N fertilization to forest ecosystems have been shown to affect CH 4 fluxes from forest soils (Steudler et al., 1989;Brumme and Borken, 1999).Several mechanisms have been proposed to explain how mineral N affects CH 4 fluxes in upland soils.Most commonly, the inhibition of CH 4 oxidation in the soil by increased NH + 4 levels is mentioned, not only in temperate soils (Steudler et al., 1989;Crill et al., 1994) but also in tropical soils (Veldkamp et al., 2001).The enzyme methane monooxygenase, which initiates the oxidation pathway of CH 4 , is also able to oxidize NH + 4 .When NH + 4 competes with CH 4 for reactive sites of methane monooxygenase, this can cause inhibition of CH 4 oxidation (Bédard and Knowles, 1989).
An osmotic effect may also contribute to the inhibition of CH 4 oxidation (Nesbit and Breitenbeck, 1992;Veldkamp et al., 2001).There is a discrepancy in published literature about the duration over which NH + 4 can inhibit CH 4 oxidation.An inhibition effect of NH 4 for 13 yr has been reported (Mosier et al., 1996), whereas in another study inhibition lasted only about four weeks (Veldkamp et al., 2001).On the other hand, increased NO − 3 levels can inhibit CH 4 production because NO − 3 is preferred as an electron acceptor over bicarbonate (Conrad, 1989), and some intermediates if NO − 3 is denitrified (NO − 2 , NO, N 2 O) can be toxic for methanogenic microorganisms (Klüber and Conrad, 1998).
Methanotrophic microorganisms also need a N source and thus could be N limited (Bender and Conrad, 1995;Bodelier et al., 2000).However, Bodelier and Laanbroek (2004) showed through a literature review that many indications for N limitation of soil CH 4 consumption have been ignored in earlier studies.Apart from the effects N limitation has on the growth and activity of CH 4 -oxidizing bacteria, they also proposed that switching from fixation of molecular N to assimilation of mineral N can cause almost instantaneous changes in CH 4 -oxidizing activity.
To date, only one N-manipulation study has been published on N effects on soil CH 4 fluxes from (sub)tropical forests, and this was conducted in China (Zhang et al., 2008;Zhang et al., 2011).In this study, CH 4 uptake decreased with increasing N application rate, whereas in the disturbed and rehabilitated forest no N-addition effect was observed.The authors concluded that the response of soil CH 4 uptake to N addition in tropical forests varied depending on the soil N status; the lack of effect from the disturbed and rehabilitated forest was explained by intense competition for N by the vegetation (Zhang et al., 2008).
Here we report the impact of chronic N additions on soil CH 4 fluxes from two species-rich, old-growth forests in Panama: a lowland, moist forest on clayey Cambisol and Nitisol soils, and a montane, wet forest on a sandy loam Andosol soil covered with an organic layer.We hypothesized that (1) in the lowland forest, with large soil N-cycling rates (Corre et al., 2010) and tree stem diameter growth (≥ 10 cm diameter trees), as well as fine litterfall that is not N limited (Wright et al., 2011), long-term N addition will inhibit CH 4 uptake; (2) in the montane forest, with comparatively small soil N-cycling rates (Corre et al., 2010) and tree stem diameter growth (10-50 cm diameter trees), as well as fine litterfall that is N limited (Adamek et al., 2009), long-term N addition will stimulate CH 4 uptake.We tested these hypotheses by comparing soil CH 4 fluxes over a period of four years (2006)(2007)(2008)(2009) in the lowland forest between control and N-addition plots during 9 to 12 yr of N additions and in the montane forest between control and N-addition plots during 1 to 4 yr of N additions.Our objectives were to (1) assess changes in soil CH 4 fluxes as a result of long-term N addition, and (2) relate these changes to soil-extractable NO − 3 , NH + 4 and soil water-filled pore space, which are factors that potentially control soil CH 4 fluxes.This is the first study to report how CH 4 fluxes change under chronic N addition in diverse, old-growth Neotropical forests.

Approach
We applied N fertilizer to create N-enriched conditions, which ultimately will simulate future increased atmospheric N deposition.N deposition normally enters the ecosystem at the canopy level at low N concentrations with each rain shower whereas we applied N fertilizer to the soil at high N concentration in four doses per year (see below).One of the artefacts of N fertilization is the occurrence of pronounced peaks of soil mineral N concentrations, which can affect short-term CH 4 fluxes within the first weeks following N application (Veldkamp et al., 2001).We therefore did a separate statistical analysis for CH 4 fluxes that include all measurements conducted from 1 day to 3 months after an N application and for CH 4 fluxes that were measured ≥ 6 weeks after the last N application (hereafter referred as long-term CH 4 fluxes).The long-term CH 4 fluxes are unlikely to be affected by the artificially high mineral N concentrations directly following N application.Furthermore, the type of N fertilizer (in our case urea) will be less important for the long-term CH 4 fluxes because within six weeks following urea application in our study sites, urea-N was hydrolyzed and processed in the internal soil N cycle (Koehler et al., 2009).

Site description and experimental design
The lowland forest (25 to 61 m elevation) consists of an old-growth (> 200 yr), semi-deciduous forest and is located on the Gigante Peninsula (9 • 06 N, 79 • 50 W), which is part of the Barro Colorado Nature Monument, Republic of Panama.On the nearby Barro Colorado Island, annual rainfall averages 2715 ± 139 mm (1999-2010) with a dry season from January to April.Ambient N deposition from rainfall was 9 kg N ha −1 yr −1 , measured bi-weekly in 2006-2007 at the shore of Gigante Peninsula near the study site (Corre et al., 2010).The mean annual air temperature is 27.2 ± 0.1 • C. Stem diameter growth of trees with ≥ 10 cm diameter at breast height (dbh), fine litter production, and fine-root biomass within 0-10 cm depth were not affected by 11 yr of N addition (Wright et al., 2011).The soils are Endogleyic Cambisol in the lower parts of the landscape and Acric Nitisol in the upper parts of the landscape, both with heavy clay texture.Bulk density was 0.62 g cm −3 in the top 5 cm depth of mineral soil (Koehler et al., 2009).After 8 yr of N addition, we measured significant decreases in soil pH (control = 5.1 ± 0.1, N addition = 4.8 ± 0.1) and base saturation (control = 67 ± 8%, N addition = 41 ± 7%), while exchangeable aluminium (Al) increased (control = 213 ± 39 g Al m −2 , 8 yr N addition = 297 ± 44 g Al m −2 ) in the top 50 cm of mineral soil.
The montane forest (1200-1300 m elevation) consists of an old-growth lower montane forest and is located in the Fortuna Forest Reserve in the Cordillera Central (8 • 45 N, 82 • 15 W), Chiriquí Province, Republic of Panama.Mean annual rainfall is 5461 ± 250 mm (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) with no dry season.Ambient N deposition from rainfall was 5 kg N ha −1 yr −1 , measured biweekly in 2006-2007 at a forest clearing near the study site (Corre et al., 2010).The annual mean air temperature is 20.3 ± 0.2 • C. Stem diameter growth of trees with 10-50 cm dbh and fine litter production increased during the first two years of N addition compared with the control plots (Adamek et al., 2009), whereas fine-root biomass and production (from organic layer down to 20 cm depth of the mineral soil) were not affected by N addition (Adamek et al., 2011).The soil is Aluandic Andosol with sandy loam texture and has an organic layer thickness of 10 ± 1 cm.Bulk density of the organic layer was 0.07 g cm −3 and the underlying mineral soil had a bulk density of 0.47 g cm −3 in first 5 cm depth (Koehler et al., 2009).After 3 yr of N addition no significant changes in pH (control = 4.7 ± 0.1, N addition = 4.6 ± 0.2), base saturation (control = 8 ± 3 %, N addition = 11 ± 4 %), and exchangeable Al (control = 252 ± 16 g Al m −2 , N addition = 280 ± 24 g Al m −2 ) were observed in the top 50 cm of mineral soil.
The N-addition experiment in the lowland forest was part of an on-going nutrient manipulation study established in 1998 (Wright et al., 2011).The experiment includes Naddition and control plots, among other treatments, laid out in four replicates across a 26.6 ha area in a stratified random design.In the montane forest, the experiment was set up in 2006 in a paired-plot design with four replicates of control and N-addition plots (Corre et al., 2010).At both sites, the size of the plots was 40 m × 40 m, separated by at least 40 m of buffer zone where no manipulation was done.The N-addition plots received 125 kg urea-N ha −1 yr −1 split in four applications (i.e. during the rainy season (May-December) for the lowland forest, and every quarter of the year for the montane forest).Measurements were conducted in the central 20 m × 20 m area of the plot to prevent possible edge effects (e.g.roots from trees outside the plots growing into the N-fertilized plots).

CH 4 flux measurements
Soil CH 4 fluxes were measured using vented static chambers.Four permanent chamber bases (area 0.04 m 2 , height 0.25 m, total volume with cover 11 L) were installed on each plot in a stratified random design along two perpendicular 20 m long transects that crossed in the plot's centre.During the first two years (2006)(2007), sampling frequency was from once a month to four times a month when we intensively measured following an N application.During the third and fourth year (2008)(2009), sampling was conducted at least once a month.Four gas samples (100 mL each) were removed at 2, 12, 22 and 32 min after chamber closure and stored in pre-evacuated glass containers with a teflon-coated stopcock.Gas samples were analyzed in the field station in Panama using a gas chromatograph (Shimadzu GC-14B, Columbia, MD, USA) equipped with a flame ionization detector and an autosampler (Loftfield et al., 1997).CH 4 concentrations were determined by comparison of integrated peak areas of samples with those of three to four standard gases (depending on concentrations: 250, 1499, 1996, 9900 and 20 010 ppb CH 4 ; Deuste Steininger GmbH, Mühlhausen, Germany).Gas fluxes were calculated from the concentration change in the chamber versus time and were adjusted for air temperature and atmospheric pressure measured at the time of sampling.To account for the decreasing diffusion gradient over time caused by the chamber feedback, we fitted both a linear and a quadratic regression model if CH 4 concentrations increased or decreased asymptotically (Wagner et al., 1997).We chose the statistically more adequate model based on the Akaike information criterion.The quadratic model was used in 14 % of the flux calculations in the montane forest and in 20 % of the gas flux calculations in the lowland forest.If CH 4 concentrations leveled out over time and the quadratic model was statistically inferior, we excluded the last data point and calculated the flux based on a linear model.These data screening and calculation procedures ensure that we minimized underestimations which may occur if a linear model was uncritically applied to static chamber flux data (Livingston et al., 2006).Positive fluxes indicate CH 4 emission from the soil; negative fluxes indicate CH 4 uptake by the soil.Zero fluxes were included.The annual CH 4 fluxes were approximated by applying the trapezoid rule (linear interpolation of time intervals between measured flux rates), assuming constant flux rates per day.

Soil mineral N and moisture
From earlier experience in tropical forests, we learned that short storage of disturbed soil samples can considerably alter mineral N concentrations (Arnold et al., 2008).We therefore conducted mineral N extractions in the field.Parallel to gas sampling, four samples of mineral soil (0-0.05m depth) were collected within the central 10 m × 10 m of each plot.For the montane site, we sampled the organic layer and 0-5 cm depth mineral soil separately.While in the field, samples were pooled for each plot, leaves and roots were manually removed, and a subsample (50-60 g fresh weight) was added to a prepared extraction bottle containing 150 mL of 0.5 mol L −1 K 2 SO 4 .Shaking (1 h) and filtering continued upon arrival at the field station, which was at most 6 h after field extraction.Soil extracts were stored in a freezer and kept frozen during air transport to the University of Göttingen (Germany), where NH + 4 and NO − 3 contents were analyzed using continuous flow injection colorimetry (Cenco/Skalar Instruments, Breda, Netherlands).NH + 4 was determined using the Berthelot reaction method (Skalar Method 155-000) and NO − 3 was measured using the copper-cadmium reduction method (NH 4 Cl buffer but without ethylenediamine tetraacetic acid; Skalar Method 461-000).The rest of the field-moist sample was stored in plastic bags for gravimetric moisture determination, conducted in the field station on the same sampling day.We dried 40-100 g of fresh-weight soil for 24 h at 105 • C. We expressed moisture content as WFPS using measured bulk density and particle densities of 2.65 g cm −3 for mineral soil (Linn and Doran, 1984) and 1.4 g cm −3 for organic layer (Breuer et al., 2002).

Statistical analyses
For CH 4 fluxes, statistical analysis was conducted on the plot means (average of 4 chambers) of each sampling day.Linear mixed effects models were used to test for the fixed effects of site (lowland vs. montane control plots) or treatment (control vs. N addition for each site) on the repeated measurements of soil CH 4 fluxes and soil factors (WFPS, soil temperature, NH + 4 and NO − 3 concentrations).The spatial replication and time (sampling days) were included as random effect.A function which allows different variances of the response variable per level of the fixed effect and/or a first-order temporal autoregressive process was included if this improved the relative goodness of the model fit based on likelihood ratio tests.The significance of the fixed effect was evaluated using analysis of variance (Crawley, 2009).If residual plots revealed non-normal distribution or non-homogenous variance, square root or logarithmic transformation was used for right-skewed data and quadratic transformation for leftskewed data, and the analysis was repeated.Effects were considered significant if P value ≤ 0.05.Pearson correlation tests were conducted on treatment means (average of 4 plots) of each sampling day to investigate the linear influences of WFPS, soil temperature, NH + 4 and NO − 3 concentrations on soil CH 4 fluxes.A few CH 4 fluxes from the N-addition plots of the montane forest were exceptionally high (21 out of 196 plot means with emissions > 60 µg CH 4 -C m −2 h −1 ), and correlation analyses were conducted both including (using logarithmic transformation) and excluding these high emissions.We also used Pearson correlation to test the influences of annual rainfall, soil clay and sand contents, organic layer thickness, and annual N deposition on annual soil CH 4 -C fluxes of tropical forests published so far.Mean values in the text are given with ±1 standard error.Analyses were conducted using R 2.15.2 (R Development Core Team, 2011).

Soil water content, temperature and mineral N
In the lowland forest, the pronounced dry season from January to April caused a strong seasonality in WFPS, which ranged from approximately 55-70 % during rainy season to 35-45 % during dry season (Fig. 1a).Mean annual soil temperature was 25.5 • C and the seasonal variation was 2.5 • C (Fig. 1c).Neither WFPS nor soil temperature differed between the control and N-addition plots (P = 0.37 to 0.95).In the montane forest, where the dry season is absent, the WFPS in the mineral soil was high (70-80 %) throughout the year.The organic layer with its low bulk density had a much lower WFPS (20-35 %; Fig. 1b).Mean annual soil temperature was 18.1 • C and the seasonal variation was 3.8 • C (Fig. 1d).Also, WFPS and soil temperature were similar between the control and N-addition plots (P = 0.31 to 0.47).
In the lowland forest, NH + 4 concentrations did not differ between the control and N-addition plots (P = 0.82) (Fig. 2a), but NO − 3 concentrations increased with N addition (P < 0.01) (Fig. 2b).In the montane forest, mineral N was dominated by NH + 4 in both organic layer and mineral soil.N addition increased NH + 4 concentrations in the mineral soil (P < 0.01) but did not show an effect on NH + 4 concentrations in the organic layer (P = 0.31) (Fig. 2c and  e).NO − 3 concentrations increased in both mineral soil (P = 0.01) and organic layer (P = 0.03) with very large increases in the fourth year of N addition (Fig. 2d and f).

CH 4 fluxes from control forest soils
CH 4 fluxes from the lowland forest control plots (−21.47 ± 1.57 µg CH 4 -C m −2 h −1 ) did not differ (P = 0.82) from the fluxes of the montane forest control plots (−3.99 ± 3.40 µg CH 4 -C m −2 h −1 ; Fig. 3, Table 1).This seemingly larger CH 4 uptake rates in this moist (2.7 m yr −1 rainfall) lowland forest soil than the wet (5.5 m yr −1 rainfall) montane forest soil was not statistically significant because of the large spatial and temporal variations (Fig. 3).Before elaborating on how soil factors influence CH 4 fluxes, we want to point out the implications of correlations: a positive correlation between CH 4 fluxes and a soil variable indicates a decrease in CH 4 uptake rates with an increase in the soil parameter values, whereas a negative correlation indicates an increase in CH 4 uptake rates with an increase in the soil parameter values.In the lowland forest, CH 4 fluxes were positively correlated with WFPS (Table 2).In the montane forest, CH 4 fluxes were negatively correlated with NH + 4 concentrations and positively correlated with NO − 3 concentrations of the organic layer and mineral soil (Table 2).These opposing correlations of CH 4 fluxes with NH + 4 and NO − 3 were because the temporal patterns of NH + 4 and NO − 3 showed the opposite trend (Fig. 2c-f).The correlation between CH 4 fluxes and total soil mineral N (NH + 4 + NO − 3 ) concentrations (organic layer R = −0.51,P = 0.01, n = 28; mineral soil R = −0.56,P = 0.00, n = 27) followed that of NH + 4 , because NH + 4 comprised the largest part of mineral N.

Effects of N addition on soil CH 4 fluxes
In the lowland forest, neither all CH 4 fluxes (−24.22 ± 1.64 µg C m −2 h −1 ) nor the long-term CH 4 fluxes (−26.14 ± 2.00 µg C m −2 h −1 ) from the N-addition plots differed (P = 0.55 to 0.57) from the CH 4 fluxes of the control plots (Fig. 3a and c; Table 1).The reason was the occasional CH 4 emissions from three of the four replicate plots of the control and N-addition treatment regardless of seasons (46 emission fluxes out of 373 plot-mean fluxes or 12 % of the observations, ranging from 0.4 to 210 µg C m −2 h −1 ), resulting in the large spatial and temporal variations (i.e.large SE bars; Fig. 3a and c).For all CH 4 fluxes, we detected a positive correlation with WFPS and negative correlations with soil temperatures and NO − 3 concentrations (Table 2).The same soil factors showed similar trends of correlations with the long-term CH 4 fluxes (Table 2).
In the montane forest, despite the large mean CH 4 emissions from the N-addition plots (for all CH 4 fluxes 50.94 ± 19.62 µg C m −2 h −1 ; for long-term CH 4 fluxes 62.13 ± 31.26 µg C m −2 h −1 ), neither all CH 4 fluxes nor the long-term CH 4 fluxes differed (P = 0.32 to 0.71) from those of the control plots (Fig. 3b and d; Table 1).The reason was that frequent CH 4 emissions were observed from all eight plots (83 emission fluxes out of 351 plot-mean fluxes or 24 % of the observations, ranging from 0.2 to 2575 µg C m −2 h −1 ).These CH 4 emissions were dominated by one pair of control and N-addition plots (49 emission fluxes out of 351 plotmean fluxes), causing the large spatial and temporal variations (i.e.large SE bars; Fig. 3b and d).If we exclude this one pair of control and N-addition plots from the statistical analysis, there remained no difference between the N-addition and control plots (P = 0.28 to 0.82), but the mean CH 4 fluxes showed net uptake instead of net emission (Table 1    remained negatively correlated with NH + 4 concentrations of the organic layer (Table 2).Considering only the long-term CH 4 fluxes, we observed also a negative correlation with NH + 4 concentrations of the organic layer both including and excluding the large emissions (Table 2).

CH 4 fluxes from control forests in comparison with published values
The mean annual CH 4 uptake rate in the control plots of the lowland forest was within the range of published values from (sub)tropical forests below 800 m elevation (Table 3).The few published CH 4 uptake rates that were lower than our lowland forest soil were mainly from Amazon forest soils with low sand or high clay contents, while those with larger CH 4 uptake rates were mostly at sites with low clay content (Steudler et al., 1996;Sousa Neto et al., 2011).Indeed, from studies compiled (Table 3), the only significant correlation between annual CH 4 fluxes and site factors for the tropical forests below 800 m elevation was a positive correlation between annual soil CH 4 fluxes and clay contents (R = 0.58, P = 0.02, n = 16).A high content of clay decreases the contribution of coarse pores to the total porosity (Hillel, 1998).
As coarse pores are especially important for gas diffusive transport, soil texture may be a good proxy variable for gas diffusion control on CH 4 uptake.Consistent with this correlation pattern, earlier studies have shown that CH 4 uptake is often limited by gas diffusion in the soil (Keller and Reiners, 1994).Also, the seasonal changes in CH 4 uptake of our lowland forest soil (Fig. 3a, c) were best explained by gas  diffusion, as was illustrated by the correlation of CH 4 fluxes with WFPS (Table 2); during the wet season when WFPS was high, CH 4 uptake was low because CH 4 diffusion from the atmosphere to this site's clayey soils was probably slowed down by the high soil water contents.
The mean annual CH 4 uptake rate in the control plots of the montane forest was the lowest published so far for tropical forests above 800 m elevation (Table 3).This was caused by the frequent CH 4 emissions from our wet, montane forest soil (Fig. 3b and d).From tropical forests above 800 m elevation (Table 3), we detected a positive correlation between annual CH 4 fluxes and rainfall (R = 0.78, P = 0.04, n = 7), which is in line with the gas diffusion control on soil CH 4 uptake as discussed above.Rainfall influences gas diffusion through its effects on soil moisture content.However, in contrast to the forests below 800 m elevation, we detected a negative correlation with clay contents (R = −0.68,P = 0.04, n = 9).This can probably be explained by the occurrence of thick organic layers (Table 3) at the surface of some of these soils, which may interfere with gas exchange between soil and atmosphere.From an earlier study we conducted in montane forests of Ecuador, we found that, contrary to common belief, the deeper part of such organic layers can contribute to the CH 4 -oxidation capacity of soils (Wolf et al., 2012).The thickness, bulk density and CH 4 -oxidation capacity of these organic layers may influence CH 4 uptake stronger than the soil texture of the underlying mineral soil.We also detected a positive correlation between annual CH 4 fluxes and annual N deposition rates (R = 0.96, P < 0.00, n = 6) of tropical forests above 800 m elevation.This may suggest that CH 4 uptake is lower at sites with higher N deposition.However, this correlation is based on six sites that had N deposition rates of only ≤ 5.0 kg N ha −1 yr −1 .At such low rates of N deposition, we think that inhibition of CH 4 oxidation by NH + 4 is unlikely.Instead, we think that such correlation is only circumstantial because in these six sites annual N deposition was positively correlated with annual rainfall (R = 0.89, P = 0.02, n = 6), signifying that low CH 4 uptake was reported for sites with high rainfall and high N deposition.Thus, we think it is more likely that soil water content (which controls gas diffusion) as influenced by rainfall was the reason behind the observed correlation between annual CH 4 fluxes and annual N deposition rates.
For the control plots of the montane forest, we interpret the negative correlations of CH 4 fluxes with NH + 4 and total mineral N concentrations as evidence that CH 4 consumption was N limited.We had similar findings of negative correlation between CH 4 fluxes and total mineral N concentrations in montane forest soils in Ecuador, suggesting N limitation on methanotrophic activity (Wolf et al., 2012).While the positive correlation of NO − 3 with CH 4 fluxes may indicate inhibitory effects of nitrification on CH 4 consumption, we think that this is very unlikely since the NO − 3 concentrations were one to two orders of magnitude smaller than the NH + 4 concentrations (compare Fig. 2c and d and 2e and f, and note the different scales on the y axis).Furthermore, measurements of gross and net nitrification rates showed very low nitrification activity (Koehler et al., 2009;Corre et al., 2010).Although Bodelier and Laanbroek (2004) suggest that N limitation of methanotrophic bacteria is less likely at (sub-)atmospheric CH 4 concentrations in the soil, we had ancillary measurements of the soil-air CH 4 concentrations in our montane forest soil that showed CH 4 concentrations in this forest soil were occasionally high.These measurements were conducted monthly from October 2008 to January 2010 in three control plots and three N-addition plots for various layers: 0.10 m above the soil surface, at the interface of the organic layer and mineral soil, at 0.05, 0.20, 0.40, 0.75 and 1.25 m depths in the mineral soil; we employed the same gas sampling methods described in our earlier study (Koehler et al., 2012).We found that 34 % of 421 observations had CH 4 concentrations in the mineral soil higher than the concentration at 0.10 m above the soil surface of 2.0 ± 0.1 ppm CH 4 -C, particularly during periods of high rainfall and thus high soil water contents.Such high CH 4 concentrations in our montane forest soil air may allow for population increases of methanotrophic bacteria which, in turn, may lead to N limitation on their activity (Bodelier and Laanbroek, 2004).

Response of soil CH 4 fluxes to N addition in the lowland and montane forests
In contrast to the findings from temperate forest soils (Steudler et al., 1989;Brumme and Borken, 1999), tropical pasture soil (Veldkamp et al., 2001) and subtropical forest soil (Zhang et al., 2008), CH 4 uptake in our lowland forest soil was not inhibited by chronic N addition.Instead, the negative correlation of CH 4 fluxes with NO − 3 concentrations in the N-addition plots suggests that increased NO − 3 levels in these plots had stimulated CH 4 consumption (Bodelier and Laanbroek, 2004) and/or had inhibited CH 4 production (Conrad, 1989).The latter is however unlikely because our ancillary measurements of CH 4 concentrations at various depths of the mineral soil (0.05, 0.20, 0.40, 0.75, 1.25 and 2 m depth) in this lowland forest during the same study years (May 2006-January 2009) showed that 11 % of the observations had higher soil-air CH 4 concentrations than the average soil-air CH 4 concentrations at a specific depth.These high soil-air CH 4 concentrations occurred in all depths of both N-addition and control plots regardless of season, indicating www.biogeosciences.net/10/5367/2013/Biogeosciences, 10, 5367-5379, 2013 that inhibition by high NO − 3 levels in N addition plots on CH 4 production was unlikely (Koehler et al., 2012).Instead, there were other supporting indications that methanotrophic activity was N limited aside from the negative correlation of soil CH 4 fluxes with NO − 3 concentrations: soil-air CH 4 concentrations and contents (or the total amount of CH 4 in a soil-air volume) down to 0.4 m depth were 30 % lower in N-addition than in control plots, and the minimum CH 4 concentration of 552 ± 42 ppb was reached at shallower depth (already at 0.40 m) in N-addition than in control plots (only at 1.25 m depth) (Koehler et al., 2012).It should be noted that these patterns were not influenced by WFPS because there were no differences in WFPS between control and N-addition plots at all depths.The reason why we did not detect significant differences in soil CH 4 fluxes despite stimulated CH 4 uptake by chronic N addition is first due to the large spatial and temporal variations of CH 4 fluxes (Fig. 3a and c).Similar large variability was reported for tropical lowland forest soils and was attributed to production of CH 4 by termites or in microsites of anaerobic conditions, and to temporal patterns of rainfall and soil moisture contents (Verchot et al., 2000;Davidson et al., 2004;Koehler et al., 2012).Second, CH 4 consumption was also largely limited by gas diffusion as shown by the positive correlation of CH 4 fluxes with WFPS (Table 2).Even if N addition stimulated methanotrophic activity, the supply of CH 4 as substrate from the atmosphere to the soil through diffusion did not change, and thus chronic N addition did not necessarily result in a larger CH 4 uptake rate.Stimulation of methanotrophic activity may be explained by a shift in N nutrition of type II methanotrophic bacteria from energy-demanding N 2 fixation to assimilation of soil mineral N (Bodelier and Laanbroek, 2004;Koehler et al., 2012), of which the NO − 3 concentrations had increased under chronic N addition (Fig. 2b).
In the montane forest soil, there was also an indication that methanotrophic activity was stimulated by chronic N addition as shown by the negative correlations between CH 4 fluxes and NH + 4 concentrations of the organic layer.However, this N-stimulated methanotrophic activity was masked by the frequent CH 4 emissions.The frequent CH 4 emissions in this wet montane forest soil indicated the regulation of WFPS of the mineral soil on CH 4 fluxes, as was shown by their positive correlation when all CH 4 fluxes are included in the statistical analysis (Table 2).WFPS did not only regulate CH 4 fluxes through the diffusive limitation of CH 4 as substrate for methanotrophs but also through the occurrence of anaerobic conditions for CH 4 production.Indeed, the WFPS of this montane forest was high throughout the year (Fig. 1b), and our ancillary measurements of WFPS at various depths in the mineral soil of these plots, conducted monthly during October 2008 to January 2010, showed WFPS between 96 ± 1 % and 88 ± 1 % from 0.20 m down to 1.25 m depth.Such high WFPS may have favoured CH 4 production and thus the frequent CH 4 emissions from all eight plots.This was probably the principal reason why we were not able to detect potential differences in CH 4 uptake rates between control and N-addition plots despite an indication of N limitation on CH 4 consumption.Exclusion of one pair of control and N-addition plots that strongly dominated the CH 4 emissions during our four-year measurements did not change the statistical trend even though the mean CH 4 uptake rates in the N-addition plots were seemingly larger than the control plots in all years (Table 1).

Consequences of chronic N deposition on soil CH 4 fluxes from tropical forests
Nine to twelve years of N addition to a lowland forest and one to four years of N addition to a montane forest did not affect soil CH 4 fluxes, although we found indications that CH 4 consumption may have been N limited at both sites.We proposed the following reasons why such N-stimulated CH 4 consumption did not lead to statistically larger CH 4 uptake: (1) for the moist, lowland forest soil, this was caused by limitation of CH 4 diffusion from the atmosphere into the clayey soils, particularly during the wet season when WFPS was high; (2) for the wet, montane forest soil, this was due to the high WFPS in the mineral soil throughout the year, which may not only limit CH 4 diffusion from the atmosphere into the soil but also favours CH 4 production; and (3) both forest soils showed large spatial and temporal variations of CH 4 fluxes.The lowland forest soil showed occasional but low CH 4 emissions whereas the montane forest soil showed more frequent CH 4 emissions with a few exceptionally large emissions (Fig. 3).Accordingly, such high CH 4 concentrations in the soil provide large amounts of substrate for methanotrophy and favour N limitation on methanotrophic bacteria (Bodelier and Laanbroek, 2004).
Our results contrast with the only published study about Naddition effects on soil CH 4 fluxes from (sub)tropical forests, which was conducted in China, where increasing N addition rates resulted in decreasing CH 4 uptake rates.These results were attributed to several possible causes: high N status, low pH values and Al toxicity (Zhang et al., 2008;Zhang et al., 2011).Although our lowland forest soil also had a high N status (Corre et al., 2010) and our montane forest soil also had low pH and high exchangeable Al (see Sect. 2.2), the differences in site conditions between our sites and this forest in China are that the Chinese site had suffered decades of high N deposition (Table 3) leading to soil pH values below 4.0, exchangeable Al of > 400 mg Al kg −1 even in the control plots, and never emitted CH 4 during the first year of measurements.Sub-atmospheric soil CH 4 concentrations are possibly prevalent in this Chinese site, and in such conditions methanotrophic activity is less likely to be N limited (Bodelier and Laanbroek, 2004).
If our explanation for the contrasting effects of N additions between our study sites and that of Zhang et al. (2008) holds up throughout the tropics, it is unlikely that elevated N deposition on tropical forests will lead to a rapid reduction in CH 4 uptake.We expect that in tropical montane forests, which typically have low N availability, N deposition may stimulate CH 4 oxidation at sites where occasional CH 4 emissions occur or will cause no change in CH 4 uptake at sites where no CH 4 emissions occur.In tropical lowland forests, which often have a high N availability, N deposition only appears to inhibit CH 4 uptake if soil pH values have become so low that considerable Al toxicity occurs.In other situations, it seems more likely that N deposition will not affect CH 4 fluxes or may even stimulate CH 4 uptake.Whether N additions to tropical forests with N-limited methanotrophic activity can indeed stimulate soil CH 4 uptake remains to be seen.The most likely time when CH 4 uptake may be stimulated by N additions is during dry periods/seasons when CH 4 supply from the atmosphere is not or less limited by gas diffusion.The most likely place where CH 4 uptake may be stimulated by N addition is in forests with a strong seasonal rainfall where occasional CH 4 emissions occur during the rainy season and strong uptake occurs during the dry season.
for their thorough and helpful reviews.M. D. Corre acknowledges funding from the Robert Bosch Foundation (Germany) for her independent research group NITROF, and from the Deutsche Forschungsgemeinschaft (Co 749/1-1).B. Koehler acknowledges further funding by the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning and from Erwin Zehe at the Karlsruhe Institute of Technology, Germany.This Open Access Publication is funded by the University of Göttingen.

Fig. 1 .
Fig. 1.Mean (±SE, n = 4) soil water-filled pore space (WFPS) and temperature at 0-0.05 m mineral soil in the control ( ) and N-addition (•) plots of the lowland forest (a and c) with 9-12 yr of treatment and of the montane forest (b and d) with 1-4 yr of treatment.For WFPS in the montane forest, the upper and lower values are for the 0-0.05 m mineral soil and organic layer, respectively.Grey shadings in (a) and (c) mark the dry seasons.The first two years were previously reported by Koehler et al. (2009).

Fig. 2 .
Fig. 2. Mean (±SE, n = 4) soil-extractable ammonium (NH + 4 , left panels) and nitrate (NO − 3 , right panels) at 0-0.05 m mineral soil in the control ( ) and N-addition (•) plots of the lowland forest (a and b) and montane forest (c and d for organic layer, e and f for 0-0.05 m mineral soil).The black vertical lines indicate dates of N addition during 9-12 yr of treatment in the lowland forest and 1-4 yr of treatment in the montane forest.Grey shadings in (a) and (b) mark the dry seasons.The first two years were previously reported by Koehler et al. (2009).

Fig. 3 .
Fig. 3. Mean (±SE, n = 4) soil CH 4 -C fluxes from the control ( ) and N-addition (•) plots of the lowland forest (a and c) and montane forest (b and d).The black vertical lines indicate dates of N addition during 9 to 12 yr of treatment in the lowland forest and 1 to 4 yr of treatment in the montane forest; the grey horizontal lines mark the zero flux.The upper panels include all fluxes whereas the lower panels show only the long-term fluxes, which were measured at least six weeks after a N addition.Grey shadings in (a) and (c) mark the dry seasons.

Table 1 .
).Also, a few CH 4 emissions from N-addition plots were exceptionally high (21 out of 196 plot means with emissions Annual soil CH 4 -C fluxes (kg C ha −1 yr −1 , mean ± SE, n = 4) from the control and N-addition plots, separated into all and longterm fluxes, with the latter including only the fluxes measured at least six weeks after a N application.For the montane forest, values in parentheses are estimates that excluded one pair of plots (control and N addition) which dominated CH 4 emissions (49 emission fluxes out of 351 plot-mean fluxes).

Table 2 .
Pearson correlation coefficients between soil CH 4 -C fluxes (µg C m −2 h −1 ) and soil variables, using the mean values of each treatment on each sampling day, measured from 2006 to 2009.For the montane forest N-addition plots, coefficients in parentheses are from analyses that include the few events of large CH 4 emissions (please see Sect.2.5).

Table 3 .
Compilation of CH 4 -C fluxes (kg CH 4 -C ha −1 yr −1 ) from soils of old-growth (sub)tropical forests, sorted from smallest to largest uptake rates within each elevation category.Percentages of clay and sand were estimated from the reported soil texture class.If no organic layer was mentioned, we assumed that it was absent (i.e.thickness of 0 cm). *