Functional Ecology, Institute of Biology and Environmental Sciences, Carl-von-Ossietzky-University of Oldenburg, Oldenburg, Germany
Corresponding author: F. A. Werner, Functional Ecology, Institute of Biology and Environmental Sciences. Carl-von-Ossietzky-University Oldenburg, PO Box 2503, Oldenburg, DE-26111, Germany. (firstname.lastname@example.org)
 Tank bromeliads are common epiphytic plants throughout neotropical forests that store significant amounts of water in phytotelmata (tanks) formed by highly modified leafs. Methanogenic archaea in these tanks have recently been identified as a significant source of atmospheric methane. We address the effects of environmental drivers (temperature, tank water content, sodium phosphate [P], and urea [N] addition) on methane production in anaerobically incubated bromeliad slurry and emissions from intact bromeliad tanks in montane Ecuador. N addition ≥ 1 mg g−1 had a significantly positive effect on headspace methane concentrations in incubation jars while P addition did not affect methane production at any dosage (≤ 1 mg g−1). Tank bromeliads (Tillandsia complanata) cultivated in situ showed significantly increased effluxes of methane in response to the addition of 26 mg N addition per tank but not to lower dosage of N or any dosage of P (≤ 5.2 mg plant−1). There was no significant interaction between N and P addition. The brevity of the stimulatory effect of N addition on plant methane effluxes (1–2 days) points at N competition by other microorganisms or bromeliads. Methane efflux from plants closely followed within-day temperature fluctuations over 24 h cycles, yet the dependency of temperature was not exponential as typical for terrestrial wetlands but instead linear. In simulated drought, methane emission from bromeliad tanks was maintained with minimum amounts of water and regained after a short lag phase of approximately 24 h. Our results suggest that methanogens in bromeliads are primarily limited by N and that direct effects of global change (increasing temperature and seasonality, remote fertilization) on bromeliad methane emissions are of moderate scale.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Bromeliads (Bromeliaceae) form a family of herbaceous plants occurring throughout moist and dry tropical and subtropical ecosystems in the Americas, where they grow epiphytically (nonparasitically on trees), lithophytically (on rocks), or terrestrially. In many species, highly modified leaves arranged in a dense spiral (rosetta) around a compressed erect shoot allow the external storage of up to several liters of water in phytotelmata or so-called “tanks”. Due to their high density in many tropical forests, tank bromeliads may impound as much as 50,000 L of water ha−1 [Fish, 1983]. Besides water, bromeliad tanks effectively intercept leaf litter, which decomposes and accumulates in the narrow bases of leaf axils as a fine, dense detritus. Martinson et al.  have recently shown that these “microswamps” host diverse and compositionally peculiar communities of methanogenic archaea which can produce significant quantities of methane (CH4). Analogous to paddy rice [Wang et al., 1997], bromeliads facilitate the emission of CH4 through a leaf aerenchyma that connects to the atmosphere via the plant stomata [Martinson et al., 2010]. While bromeliads may account for not even a hundredth of CH4 emissions at a global scale [Wuebbles and Hayhoe, 2002], they can affect regional budgets. In a tropical montane moist forest of South Ecuador, bromeliad tanks were estimated to emit 3.6 g CH4 ha−1 d−1, hence neutralizing the sink strength of the forest soil of 3.1 g CH4 ha−1 d−1 [Martinson et al., 2010]. Extrapolated to the area covered by neotropical forests, these emissions would approximate 1.2 Tg yr−1 and help explain the high CH4 concentration above neotropical forests that are not accounted for by current atmospheric models [Frankenberg et al., 2008; Martinson et al., 2010]. Since methane is a potent greenhouse gas globally accounting for 20–25% of radiative forcing [Shindell et al., 2009], understanding its sources is critical for the projection of methane budgets under global change. While Martinson et al.  found that methane emissions were correlated with measures of plant size (e.g., dry weight, tank water volume), emission rates were characterized by high residual variance among plants. This variability of emissions has not been addressed experimentally and is not understood. As a consequence, the magnitude of bias in ecosystem-level emissions from sampling solely during daytime is unknown. Likewise, the effects of global change (e.g., atmospheric warming, increased remote fertilization such as from biomass burning or industrialization) on methane emissions from bromeliads are unpredictable.
 CH4 efflux from a system to the atmosphere is the positive net result of two counteracting microbial processes being CH4 production (methanogenesis) and CH4 oxidation (methanotrophy) [Satpathy et al., 1997]. Due to the variety of involved processes and organisms, environmental drivers of total emissions are difficult to predict as shown by contradictory results in numerous studies on rice paddies and other wetlands. For instance, urea application has been shown to strongly increase CH4 production [Bharati et al., 2000; Lindau, 1994], has zero net effect on productivity [Wang et al., 1992], or even inhibit CH4 production [Bodelier et al., 2000; Cai et al., 1997] due to the toxic effect of denitrification intermediates such as nitrite, NO, and N2O on methanogenesis [Kluber and Conrad, 1998]. Since a fertilization experiment at our study site indicates P limitation of soil microbes [Krashevska et al., 2010; Wullaert et al., 2010], we expected a positive effect of P but not of N on methane emissions in our system. Furthermore, we predicted a positive relationship of emission rates with temperature and relative tank water content. While methanogen metabolic rates have a strongly positive relationship with temperature [Fey and Conrad, 2000], tank desiccation due to drought should increase oxygenation and hence decrease methane production and emission from bromeliads. Due to the very small size of bromeliad tanks and their high surface:volume ratios, we anticipated the “microswamps” formed by tank bromeliads to be highly susceptible to atmospheric temperature fluctuations and desiccation.
 In this study, we address the effects of nitrogen (N) and phosphorus (P) availability, temperature, and tank water content on methane production and net emission from the tanks of epiphytic bromeliads through a series of experiments.
2 Material and Methods
2.1 Study Site and Model System
 The study was conducted at Estación Científica San Francisco (ECSF), a research station on the eastern Andean slope in Ecuador (1840 m above sea level). Mean annual precipitation is approximately 2000 mm, and air temperature averages 16°C [Moser et al., 2008] (Figure 1). The natural vegetation of this area is evergreen tropical montane moist forest [Homeier et al., 2008]. We used the C3 species Tillandsia complanata Benth. as our study system. This pollakantic, medium-sized, tank-forming canopy epiphyte occurs from Costa Rica to Bolivia where it is commonly found from lowland to upper montane forests [Gilmartin, 1977]. Architecture and anatomy of this species represent a typical compact tillandsioid tank bromeliad as they strongly dominate montane moist ecosystems. We collected midsized adult plants from the midcanopy of local forest at 1950–2050 m in a homogenous stand of old-growth forest along a steep trail (“Transect 1”). After removing them carefully from their substrate, plants were transferred to an open-sided polythene-roofed greenhouse. Because the roots of T. complanata only serve for anchorage but not for the acquisition of water and nutrients, individuals can be transplanted without affecting plant health [Benzing, 2000]. Plants were randomly designated to treatments and positions in the greenhouse, where they were allowed ≥ 14 days for acclimatization. Plants were supplied with 2 g of dry litter from litter traps at the collection site before the experiments and watered up to the maximum tank capacity every day. Plant size and water storage capacity were measured after the experiments were completed. Each plant was then carefully dismantled, separating litter (coarse organic particles > 2 mm) from sludge (fine organic particles ≤ 2 mm). Dry weight of litter, sludge, and bromeliad tissue was determined after oven-drying to constant weight at 70°C. Nitrogen concentrations in bromeliad sludge were determined with a CHN-S element analyzer (Flash EA, Thermo Electron, Milan, Italy), while P concentrations were analyzed following the protocol outlined in Chen et al. .
2.2 Plant Experiments
 We distinguished four treatments (N, P, N + P, and a control group with watering only) with six replicates per treatment. The first nutrient application consisted of 2.6 mg N and/or 0.07 mg P plant−1 dissolved in water followed by a second application of 26 mg N and 0.7 mg P plant−1 after 20 days. In a parallel run, we applied 26 mg N and 5.2 mg P under higher temporal resolution (6, 24, 48, and 72 h after fertilization) to a different set of plants (n = 6 per treatment). N was applied as urea ((NH2)2CO) and P as NaH2PO4 × 2 H2O. These nutrient levels correspond to 0.5 and 5 months of additional nutrient input in a fertilization experiment at our field site, respectively. In this replicated nutrient manipulation experiment, urea and phosphate are applied at low dosage (50 kg N, 10 kg P ha−1 a−1) to the ground of forest plots of 400 m2 [Homeier et al., 2012]. We calculated nutrient dosage to bromeliads as the surplus contents of N and P from litter and canopy throughfall in fertilized plots (i.e., relative to unfertilized plots) during the first year of fertilization of an area equivalent to the “effective” tank area (defined as the central tank area where leaves are inclined inward and therefore channel water and litter toward the plant center) of an average T. complanata tank (491 cm2). Gas measurements were consistently taken in the same sampling order at the same time of the day to minimize temperature influence.
 To quantify within-day flux variation, we measured four bromeliads over periods of 24 consecutive hours. Methane flux was measured every 3 h (eight measurements per day) for 3 days (“runs”) with different weather conditions (ranging from warm and sunny to cool and rainy).
 By excluding rainfall, eight randomly selected plants were allowed to dry out progressively while four additional plants were watered regularly to serve as controls. After 10 days, all tanks were refilled completely with rain water. Gas samples were taken every 48 h during desiccation but more frequently (3, 6, 12, 24, 54, and 102 h) after refilling. Tank water contents were determined gravimetrically after each gas sampling.
2.3 Sludge Incubation
 We incubated sludge anaerobically, excluding methane oxidation by methanotrophic bacteria following Wang et al. . Fine detritus (sludge) was collected from basal leaf axils of epiphytic tank bromeliads (Guzmania spp.) at 1950–2000 m in the same area where bromeliad plants were collected and stored at −20°C for 6 months prior to incubation. Deep freezing is widely and effectively used to preserve soil samples for microbial studies [Rubin et al., 2013]. The sludge was homogenized thoroughly in a blender. We used each nine jars with nutrient additions (N, P, and N + P) at exponentially increasing dosage levels (P [µg g−1 sludge]: 3.91, 7.81, 15.63, 31.25, 62.5, 125, 250, 500, 1000 and N [µg g−1]: 62.5, 125, 250, 500, 1000, 2000, 4000, 8000, 16000) and four control jars (31 jars in total). Nutrient dosages were in the range of fertilization studies on paddy rice soils [Adhya et al., 1998; Mohanty et al., 2006; Rath et al., 2005]. Jars were 330 ml glass bottles closed with modified lids to allow gas sampling from the headspace through a rubber septum. Bottles were filled with 5.85 g of homogenized bromeliad sludge (equivalent to 1 g dry weight) and 50 ml of demineralized water with dissolved nutrients. Samples were stirred vigorously after sealing, flushed with 900 ml N2, and incubated at 22°C (maximum range: ± 2°C). N was applied as urea ((NH2)2CO) and P as NaH2PO4 × 2 H2O. Cumulative methane production was measured in daily intervals for 11 days and additionally at 14, 17, and 22 days after incubation start. Gas samples of 13 ml were taken from the headspace using a 30 ml polypropylen syringe and placed in evacuated Exetainer glass vials (Labco, High Wycombe, UK). After each sampling, 13 ml of N2 were injected into the jars to maintain pressure equilibrium. The experiment was run three times.
2.4 Gas Sampling and Methane Flux Calculation
 Gas sampling from bromeliads in the greenhouse was done using vented chambers following Martinson et al. . The whole plant was placed in the chamber, and 70 ml of gas were taken from the headspace 4 times in a 45 min interval and placed into preevacuated 60 ml glass vials. Ambient air samples were collected at the beginning, half way through, and at the end of each sampling campaign to align the estimated methane concentration of the first sample in each run. Ambient air temperature and tank water temperature (taken near the base of water-filled leaf axils using a GTH175 thermometer, Greisinger, Regenstauf, Germany) were measured immediately before sampling. Gas samples were stored at ambient temperature and analyzed within a week after collection. Methane concentrations were determined using a gas chromatograph (GC-14B, Shimadzu, Duisburg, Germany) equipped with a flame ionization detector as described by Wolf et al. . Check standards were injected during each run every 10 samples to verify consistency of the analysis. Methane emission rates were calculated after Bharati et al.  and converted to µg CH4 plant−1 h−1.
2.5 Statistical Analysis
 The incubation data were analyzed using linear mixed effect (LME) models (LME4 package in R). We fitted a LME model (fixed effects: N, P; random effect: run) including the interaction between N and P and tested for interactions by Markov chain Monte Carlo (MCMC; function pvals.fnc) at each nutrient dosage for the log-transformed final headspace CH4 concentrations. Since there were no significant interactions between N and P at any dosage of nutrients (all p > 0.2), we then defined a more simple, nonfactorial model. Treatment (N, P, or N + P in each nine dosage levels plus controls) and time (22 days) were defined as fixed effects and incubation runs (three replicates) as random effect. Multiple comparisons (ghlt function in R) with Tukey test was implemented as a post hoc test. Simple linear regression was used to compare differences in the effect of added nutrient concentrations on final cumulative methane production (after 22 days of incubation) following logarithmic transformation of nutrient dosage.
 The effect of tank water temperature as well as air temperature on methane efflux from bromeliads was tested using linear regression and a LME with air and plant temperature as fixed effect and plant identity and daytime as random effects (individual nested within daytime). We used stepwise model selection to identify the best model. P values and MCMC confidence intervals for mixed models were calculated using R's function pvals.fnc. The fit of linear regression models of CH4 emissions versus temperature was evaluated against models from the exponential family for each plant using the curveFinder function of CurveExpert 1.4.
 Because responses strongly varied nonlinearly with time, effects of nutrient addition on CH4 effluxes from entire plants were analyzed using a factorial repeated-measures analysis of variance (ANOVA) with N and P addition as two-level factors, making pairwise comparisons of fluxes 0.3–6 d after versus before nutrient addition. Data were log transformed to match parametric assumptions. Curve fitting was done through CurveExpert 1.4 [Hyams, 2010], factorial repeated-measures ANOVA using Statistica 8.0 (Statsoft Inc., Tulsa, OK, USA). Other analyses were conducted using R 2.14.2 [R Development Core Team, 2009]. Significance was set at p ≤ 0.05.
3.1 Effects of Nutrient Addition on Methane Production Under Anaerobic Incubation
 After a lag phase with low productivity of approximately 5 days, methane concentrations rose exponentially in the headspace of all incubation jars. No temporal saturation effect could be observed during the 22 days of incubation. Addition of ≥ 1 mg N g−1 dry sludge upward resulted in a twofold increase in CH4 production on day 5, and this effect remained significant until the end of the experiment (day 22, Table 1). Addition of P did not influence methane production at any dosage, with P-treated and control jars being statistically at par during all measurements. Likewise, the stimulatory effect of N + P addition did not result in significantly higher cumulative methane production than that of N alone, at any of the nine dosage levels tested (LME; all p > 0.05). The stimulatory effect of N and N + P addition initiated at a minimum N dosage of 0.5–1 mg from where it increased with dosage before leveling off at dosages over ≥ 8 mg N g−1. A sigmoid growth function (a/(1 + (b/x)c)) provided slightly better fit for N and N + P amendment (N: r2adj = 0.83, p < 0.0001; N + P: r2adj = 0.63, p < 0.0001) than linear regression (r2adj = 0.81, p < 0.0001; 0.60; p < 0.0001) (Figure 2).
Table 1. Influence of Nutrient Addition on CH4 Production on day 1, 5, 11, and 22 After Incubation Start (Nutrient Treatments: n = 3; Controls: n = 12)a
Nutrient Addition (µg g−1)
Days of Incubation
Given are mean ± SD. Asterisks indicate significant differences between treatments and controls (Tukey test).
3.2 Effects of Fertilization on Plant Methane Emissions Rates
 We found no significant interactions between addition of N and P at any of the sampling dates. Neither were there significant main effects in response to the low nutrient dosage of run 1 (2.6 mg N and/or 0.07 mg P per plant). However, 24 and 96 h after a second tenfold nutrient application (26 mg N and/or 0.7 mg P per plant), N amendment showed a significant positive effect on methane effluxes (F1, 20 = 16.2, p < 0.001 (24 h), F1, 20 = 4.5, p < 0.05 (96 h)) (Figure 3a).
 In order to examine emission response to nutrient application in higher temporal resolution, the experiment was repeated at shorter sampling intervals (6, 24, 48, and 72 h after fertilization) in a third run. As in the previous trial, N addition had a significant stimulatory effect on CH4 efflux 24 h after amendment (F1, 20 = 13.9, p < 0.01). However, subsequent measurements (48 and 72 h after nutrient addition) did not show a stimulatory effect of nutrient addition (Figure 2b). Again, P addition (5.2 mg P plant−1) did not affect CH4 emissions at any time. Nutrient concentrations in the sludge of control plants (n = 6) were 22.27 ± 7.34 mg N g−1 and 2.48 ± 0.75 mg P g−1(mean ± standard deviation (SD)).
3.3 Within-Day Variation of Methane Emissions
 Tank water temperatures ranged from 13.9°C to 22.6°C, closely following air temperature (13.4–28.3°C). Methane emission rates from all four plants showed pronounced within-day variation throughout the three sampling days (Figure 4). Highest methane emission rates (15.8 ± 3.8 µg CH4 plant−1 h−1) were observed in the late afternoon at 14:00 or 17:00 local time. Up to 50% lower emission rates occurred between 2:00 and 8:00 (7.4 ± 2.6 µg CH4 plant−1 h−1). While we observed pronounced variability in absolute methane fluxes among individuals, the magnitude of within-day variation was similar for all plants and consistent among sampling runs.
 Within-day variation of methane emission was positively related to both air and tank water temperature (Figure 4). Linear regressions of CH4 emissions for individual plants yielded R2 values of 0.71 ± 0.18 against tank water temperature (all p < 0.0001) and 0.48 ± 0.09 SD against air temperature (all p < 0.001). The linear nature of the relationship between temperature and CH4 emissions (Figure 5) was confirmed through curve fitting. The best positively exponential functions (r = 0.84 ± 0.11; r = 0.73 ± 0.11) did not improve fit over simple linear regression models for the relationship between CH4 emission and temperature water and air, respectively (r = 0.87 ± 0.08; r = 0.74 ± 0.08). In LME modeling, air temperature was excluded through stepwise model selection since it did not improve the model significantly. Water temperature had a significant positive effect on CH4 emission rates (p < 0.01); LME estimates indicate a mean increase of 0.67 µg CH4 plant−1 h−1 per °C increase in temperature.
3.4 Variation of Methane Emissions With Tank Water Content
 The effect of tank water content on methane emissions was examined through simulated drought. The highest methane emission from desiccating plants was found after 4 days of desiccation, when maximal tank water content of treated bromeliads was already decreased by approximately 50% (Figure 6). In the following, emission rates decreased and had ceased on day 10, with the exception of two large plants (2.3 and 1.8 µg CH4 plant−1 h−1, respectively) that presumably had retained some tank water. Six hours after refill, emission rates slowly began to recover and had reached half of their initial (i.e., full tank) values 4 days after refilling (Figure 7). Variation in emission rates from control plants likely reflect pronounced temperature differences on sampling dates.
4.1 Within-Day Fluctuation of Methane Emissions
 As expected, we found a strong, significant positive correlation of both air and tank water temperature with methane emission rates. Methane emission consistently showed highest values during the afternoon when air and tank water temperatures were highest, followed by a decline to a minimum in the early morning, concurrent with lowest temperatures. However, considering the high amplitude of temperature fluctuations we found in tank water, within-day fluctuation of CH4 efflux was remarkably low compared to terrestrial wetlands [see e.g., Sass et al., 1991]. The Arrhenius equation predicts that methane production as a chemical reaction should increase exponentially with temperature
with k being rate constant, T temperature (Kelvin), A a preexponential factor, Ea activation energy, and kB the Boltzmann constant. In terrestrial wetlands such as rice paddy fields, both production and emission of methane indeed follow the Arrhenius dependence and increase exponentially with temperature [Rath et al., 2002; Sass et al., 1991; Schutz et al., 1990]. Instead, we found a linear relationship between CH4 efflux and water temperature within the considerable sampling range of 14–23°C water temperature. This result suggests the regulatory influence of factors other than reaction kinetics, such as plant stomatal conductance [Martinson et al., 2010] or methane oxidation [Mohanty et al., 2007].
 Our results further show that within-day fluctuations need to be taken into account when assessing ecosystem-level CH4 emissions. Based on measurements during daytime only, Martinson et al.  projected methane release from bromeliads to be 3.6 g CH4 d−1 ha−1for our study site. Our measurements revealed pronounced within-day variation (lower fluxes during nighttime), suggesting that these projections overestimate the true total daily emissions by approximately 20–30%.
4.2 Seasonal Fluctuation of Methane Emissions Due to Tank Water Content
 Seasonal fluctuations of temperature are moderate in evergreen tropical forests and foliar litter falls throughout the year. Especially in montane evergreen forests, bromeliad tanks usually appear saturated with fresh litter throughout the year (F. A. Werner, personal observation, 2009). Precipitation is therefore likely to be the greatest single predictor of seasonal fluctuations in methane emissions from bromeliads in such forests. Subjected to simulated drought, methane emissions from desiccating tanks decreased with reduced tank water content, and recovery began after a short lag phase of approximately 24 h following refilling (Figure 7). This is consistent with results from flooded rice fields and fen peat incubations where methane emissions decrease dramatically even after short drainage of 2 days to recover after a lag phases of 2–30 days [Estop-Aragones and Blodau, 2012; Conrad, 2002; Lindau et al., 1991; Wang et al., 1992; Yao et al., 1999]. The comparably short lag phase we observed in bromeliads may be due to a high ratio of organic matter to inorganic oxidants in bromeliad sludge (compare Yao et al. ). Rainless periods of one to few weeks, typically coupled with high vapor pressure deficits, can cause complete loss of tank water through evapotranspiration [Laube and Zotz, 2003] and hence lead to the cessation of methane production in tanks. However, our results indicate that methane production is surprisingly robust to rainfall seasonality, declining sharply only at very low water content (< 20% of holding capacity) and beginning to recover quickly after refilling. Moreover, even pronounced drought events may not directly affect annual methane effluxes from bromeliads if microorganism productivity is ultimately limited by substrate availability.
4.3 Effect of N and P Fertilization on Methane Production and Emissions
 Contrary to our expectations, we found a strong, positive effect of urea-N application on methane production in anaerobically incubated bromeliad sludge. Applied at any dosage ≥ 1 mg N g−1, headspace methane concentrations were strongly elevated after 11 days of incubation. Corresponding to < 5% of N concentration in unaltered bromeliad slurry, this threshold is remarkably low. CH4 productivity increased with dosage but leveled off at > 4 mg N g−1 which suggests saturation (Figure 2). This conclusion is further supported by the superior fit of a sigmoid function. In contrast, P amendment alone did not stimulate CH4 production nor were there significant interactions between the two nutrients, N + P addition having similar effects on CH4 production as N addition alone.
 Unlike anaerobic incubation, in vivo fertilization of entire plants allows for interactions (substrate and nutrient competition, methane oxidation) with methanotrophs and other aerobic microorganisms such as nitrate reducers, as well as nutrient competition by the bromeliad itself. In rice paddy soils, CH4 production is frequently inhibited by the toxic effect of denitrification intermediates such as nitrite, NO and N2O [Kluber and Conrad, 1998], and methanogenic archaea can be outcompeted by methanotrophs or other microorganisms [Yuan and Lu, 2009]. Our results strongly suggest that methane production in bromeliad tanks is not inhibited by nitrite toxicity and that methanotrophs do not profit disproportionally by N or P fertilization in the short term. Considering that even 21 days of incubation revealed no sign of decreasing N-fertilizing effect methanogenesis, the brevity of the stimulatory effect of N addition on plant methane effluxes (1–2 days, Figure 3) points at N competition by other microorganisms or bromeliads, which are known for their highly effective uptake of N (and P) from tank water [Winkler and Zotz, 2009; Zotz and Richter, 2006].
 Our results show that in our study system, methanogen production in bromeliad tanks is strongly limited by N but not by P availability. This outcome is remarkable since a plot-based fertilization experiment at the study site shows that plant growth and soil microbial biomass is limited primarily by P or both N and P [Krashevska et al., 2010; Wullaert et al., 2010; Homeier et al., 2012]. Unlike in a number of studies from terrestrial wetlands (paddy rice fields), a stimulating effect of N fertilization on net methane emissions was also found in bromeliads grown under field conditions, indicating that bromeliads methanogens were not affected by toxicity of denitrification intermediates. The brevity of the N fertilization effect points at effective competition for N by aerobic organisms. Methane emissions from bromeliad tanks appeared remarkably insensitive to fluctuations of tank water levels. While methane emissions closely followed within-day temperature fluctuations, they did not show an exponential dependency of temperature predicted by the Arrhenius dependence. Instead, the linear relationship found between temperature and emission rates points at strong coregulation of emissions by additional factors such as methane oxidation or gas transportation via plant tissue. Based on our measurements of within-day variation, we estimate that restriction to daytime sampling by Martinson et al.  may have led to a moderate overestimation of the true daily emissions of approximately 20–30%. Our results further suggest that direct effects of global change (increasing temperature and seasonality, remote fertilization) on bromeliad methane emissions are likely to be of moderate scale.
 We thank Amanda Matson, Anke Müller, Silvia Parra, and Edzo Veldkamp for their collaboration and Glenda Mendieta Leiva for statistical advice. Rütger Rollenbeck and Thorsten Peters provided climate data. Funding was provided by the German Research Foundation (DFG) through a research grant to FAW (WE 4551/1-2) and by the German Academic Exchange Service (DAAD) through a travel grant to MMK. This is publication no. 391 of the Yanayacu Natural History Research Group.