The role of termite CH₄ emissions on the ecosystem scale: a case study in the Amazon rainforest

The magnitude of termite methane (CH₄) emissions is still an uncertain part of the global CH₄ budget and current emission estimates are based on limited field studies. We present in situ CH₄ emission measurements of termite mounds and termite mound subsamples performed in the Amazon rainforest. Emissions from five termite mounds of the species Neocapritermes brasiliensis were measured by use of a large flux chamber connected to a portable gas analyser measuring CH₄ and CO₂. In addition, the emissions of mound subsamples were measured, after which the termites were counted so that a termite CH₄ and CO₂ emission factor could be determined.

Mound emissions were found to range between 17.0 and 34.8 nmol mound⁻¹ s⁻¹ for CH₄ and between 1.1 and 13.0 µmol mound⁻¹ s⁻¹ for CO₂. A termite emission factor of 0.35 µmol CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ was found, which is almost twice as high as the only other reported value for the Amazon. By combining mound emission measurements with the termite emission factor, colony sizes could be estimated, which were found to range between 55–125 thousand individuals. Estimates were similar to literature values, and we therefore propose that this method can be used as a quick non-intrusive method to estimate termite colony size in the field.

The role of termites in the ecosystem's CH₄ budget was evaluated by use of two approaches. Termite mound emission values were combined with local mound density numbers, leading to an estimate of 0.15–0.71 nmol CH₄ m⁻² s⁻¹, on average, emitted by termite mounds. In addition, the termite CH₄ emission factor from this study was combined with termite biomass numbers, resulting in an estimate of termite-emitted CH₄ of ∼1.0 nmol m⁻² s⁻¹. Considering the relatively low net CH₄ emissions previously measured at this ecosystem, we expect that termites play an important role in the CH₄ budget of this terra firme ecosystem.

1 Introduction

Methane (CH₄) is one of the most important greenhouse gases, but its natural sources are still not well understood. Anaerobic decomposition processes in wetlands are expected to represent the largest natural CH₄ source, but estimates remain a large source of uncertainty (Kirschke et al., 2013; Saunois et al., 2020). Recently, alternative CH₄ production mechanisms and their possible important role on the ecosystem scale have been proposed, such as CH₄ production by living vegetation (Bruhn et al., 2012; Wang et al., 2014), CH₄ emission due to photo and thermal degradation (Lee et al., 2012), or the transport of anaerobic soil-produced CH₄ through wetland trees (Pangala et al., 2015; Rice et al., 2010). An additional known CH₄ source in tropical ecosystems is emission by termites.

Termites (Isoptera) can mostly be found between 45^∘ N and 45^∘ S and are especially abundant in warm ecosystems (Bignell, 2006; Brian and Brian, 1978; Gomati et al., 2011; Wood, 1988). They are highly socialised insects, living in large communities of up to several million individuals (Wood, 1988). Termites are considered “ecosystem engineers”: they are known for decomposing organic substances and moving and mixing organic and mineral materials, thereby enhancing humus formation, modifying soil structure, and improving soil fertility (Bignell, 2006; Brian and Brian, 1978; Bignell and Eggleton, 2000; Mishra and Sen-Sarma, 1980; De Bruyn and Conacher, 1990; Wood, 1988). In addition, they are able to modify their environment to their needs: most termite species live in complex above- or (partly) below-ground nests where temperature and moisture remain stable (Bignell, 2019; Noirot and Darlington, 2000; Wood, 1988). Recently, it was shown that termites increase their activity during droughts, resulting in, among other things, enhanced litter decomposition, elevated soil moisture, and higher seedling survival rates, thereby demonstrating a mitigating effect during droughts in tropical rainforests (Ashton et al., 2019).

Three main groups of termites can be distinguished based on their main feeding habits: soil-feeding (humiverous) termites, which can mainly be found in and on the soil, decomposing decayed organic soil material; xylophagous termites, which feed on (decomposed) wood and can also be found in living trees; and fungus-feeding termites, which live in a symbiotic relationship with fungus (Eggleton, 2000; Sanderson, 1996).

CH₄ production by termites was first described and measured by Cook (1932). Follow-up studies found that methane is produced by almost all termite species and that its production takes place in the termite gut. In higher termites (dominant in tropical forests; more evolved species with respect to diet and community complexity) CH₄ production is caused by symbiotic bacteria, and in lower termites the production is caused by flagellate protozoa (Bignell et al., 1997; Brune, 2018; Lee and Wood, 1971). In a laboratory experiment Zimmerman et al. (1982) measured the emission strength of individual termites and, by use of termite biomass numbers, presented a global termite emission estimate of 150 Tg CH₄ yr⁻¹, which was estimated to be 40 % of the global natural CH₄ emissions. Different estimates followed, resulting in lower values, such as by Seiler et al. (1984) of 2–5 Tg yr⁻¹, by Fraser et al. (1986) of 14 Tg yr⁻¹, by Khalil et al. (1990) of 12 Tg yr⁻¹, and by Martius et al. (1993) of 26 Tg yr⁻¹. More recent literature uses estimates in the range of 2–15 Tg CH₄ yr⁻¹ (Ciais et al., 2014; Kirschke et al., 2013; Sanderson, 1996; Saunois et al., 2020), which is approximately 0.5 %–4 % of the total estimated natural source of CH₄ emissions (Saunois et al., 2020). While global-scale termite emissions can be considered small in comparison to natural sources like wetland emissions (∼147 Tg yr⁻¹) or fresh water emissions (∼159 Tg yr⁻¹) (Saunois et al., 2020), the question of what their role can be in the CH₄ budget of a local tropical ecosystem remains.

Estimates of global termite CH₄ emissions are based on field and laboratory measurements. To estimate global CH₄ termite emissions, most commonly the CH₄ emission per termite (mg CH₄ termite⁻¹ h⁻¹) or termite mass (mg CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹) is measured, whereby termite mass can either be measured directly or be taken from literature (Sanderson, 1996). The disadvantage of this approach is that termites are removed from their natural environment, thereby possibly changing their emission and behaviour. Another approach is to measure termite nest CH₄ emissions in situ in the field. In this case, emissions are expressed per mound or nest (mg CH₄ mound⁻¹ h⁻¹). While this method does not disturb the natural environment, correct estimation of termite nest colony size is challenging; therefore, values are hard to convert to emission-per-termite values (Jones et al., 2005).

Besides CH₄, termite emissions of other gases have also been investigated, such as for CO₂, O₂, CO, H₂, CHCI₃, N₂O, and different hydrocarbons (Cook, 1932; Khalil et al., 1990; Zimmerman et al., 1982). In previous studies, measurements of termite CO₂ emissions were often performed alongside CH₄ emission measurements and generally a clear relationship between CH₄ and CO₂ emissions was found, of which the ratio is expected to be species dependent (Seiler et al., 1984; Jamali et al., 2013). For termite-emitted CO₂, reported global estimates are 50 Pg yr⁻¹ (Zimmerman et al., 1982), 4 Pg yr⁻¹ (Khalil et al., 1990), and 3.5 Pg yr⁻¹ (Sanderson, 1996) (1 Pg = 1000 Tg). In addition, Khalil et al. (1990) observed mound CO uptake and emissions but reported them to be irregular and small. Strong termite mound N₂O emissions have also been detected (Brümmer et al., 2009b; Brauman et al., 2015), although they were also found to be very irregular or undetectable (Khalil et al., 1990; Zimmerman et al., 1982). Brauman et al. (2015) suggested that termite mound N₂O emissions occur if nitrogen-rich organic matter is available.

Current global termite CH₄ emission estimates are based on relatively few studies, and there is still a lack of data on termite CH₄ emission rates (Brune, 2018). In addition, existing studies have mostly focused on Australian or Asian species (Eggleton et al., 1999; Fraser et al., 1986; Jamali et al., 2011a, b, 2013; Khalil et al., 1990; Macdonald et al., 1998; Sugimoto et al., 1998a, b) or African species (Brauman et al., 1992; Brümmer et al., 2009a; Macdonald et al., 1998; Rouland et al., 1993; Sawadogo et al., 2011, 2012; Seiler et al., 1984). To our knowledge, only two studies focused on CH₄ emission of termites in the Amazon (Martius et al., 1993; Queiroz, 2004) and only one study reported CH₄ emission values for Amazonian termites (Martius et al., 1993). Martius et al. (1993) performed field measurements on wood-feeding termites by semi-field and laboratory measurements, and suggested that Amazonian termites release more methane than species in other regions. In addition, for the Amazon, it is expected that most termites are soil feeding (Jones and Eggleton, 2010), a group which are expected to be the strongest emitters of CH₄ (Bignell and Eggleton, 2000; Brauman et al., 1992).

In this paper, we present a case study performed in a tropical rainforest in the Amazon, where we measured the emission of CH₄ and other gases of epigeal (above-ground) termite nests of the species Neocapritermes brasiliensis, a soil-feeding species¹ abundant in the Amazon (Constantino, 1992; Pequeno et al., 2013) and one of the most common species in the region (Dambros et al., 2016). In addition we measured the CH₄ emission of countable groups of termites. The goal of our research was twofold. Firstly, we provide the first CH₄ and other gas emission measurements of the species N. brasiliensis, thereby expanding the limited literature on CH₄ emissions from Amazonian termites. Secondly, we aim to quantify the role of termite emissions in the CH₄ budget of this specific ecosystem as part of a larger ecosystem CH₄ budget study (in preparation). In addition, we are presenting a possible quick, non-intrusive field method to estimate termite colony size in situ.

2 Material and methods

2.1 Study site

The study was conducted at the experimental field site Reserva Biológica do Cuieiras–ZF2 (2^∘36^′32.67^′′ S, 60^∘12^′33.48^′′ W, 40–110 m above sea level (a.s.l.)), which is managed by the Instituto Nacional de Pesquisas da Amazônia (INPA) and located ∼ 50 km northwest of Manaus (Brazil). Field site ZF2 consists of plateaus and valleys with typical terra firme forest with tree heights of 35–40 m on the plateaus and 20–35 m in the valleys. Soils on the plateau are clayey and can be classified as Oxisols and Ultisols. Soils in the valleys contain more sand and can be classified as Spodosols (Luizão et al., 2004; Zanchi et al., 2014). The field site has a strong seasonality, with a wet season from December to April and a dry season from June to September. Annual average temperatures range between 26–28^∘ C and annual average precipitation is around 2400 mm. More information about the field site can be found in Araújo et al. (2002), Chambers et al. (2004), Luizão et al. (2004), Quesada et al. (2010), and Zanchi et al. (2014). Measurements took place at the end of the wet season (March 2020).

2.2 Selection of termite mounds

In the study area, two main trails exist following the topography from valley to plateau, and termite nests in the vicinity of these trails were inventoried. For practical reasons, only free-standing epigeal (above-ground) nests were considered (hereafter called mounds). Twenty termite mounds were selected for further research and for each mound the termite species was determined. For flux chamber measurements, five mounds with the same termite species were selected (nos. 13, 14, 15, 16, and 19); for practical reasons, chosen mounds were in close proximity to each other and all located in the valley. As an exploratory measurement, an additional mound of a different species was selected on the plateau (no. 6). For each mound, height and perimeter were measured. Termite mound volumes were estimated by use of the following formula, as also used in Ribeiro (1997) and in Pequeno et al. (2013):

$$\begin{array}{}\text{(1)}& V={\displaystyle \frac{\mathit{\pi }HWT}{\mathrm{6}}},\end{array}$$

where V is the mound volume (cm³), H is the height (cm), W is the width (cm), and T is the thickness (cm) of the mound. The termite mound surface was estimated by mathematically considering the lower part of the mound as a column and the upper part as half a sphere. Details of each mound (dimensions, species, location) are given in Table 1.

Table 1 Termite mounds: location, dimensions, and observed species. Volume is the estimated mound volume as calculated by Eq. (1) and surface is the estimated mound surface by mathematically considering the lower part of the mound as a column and the upper part as half a sphere. In mound 1, two different termite species were found. N. bra stands for Neocapritermes brasiliensis, H. ten for Heterotermes tenuis, R. bra for Rotunditermes bracantinus, and E. neo for Enbiratermes neotenicus. The five mounds indicated in bold (mound nos. 13, 14, 15, 16, and 19) were the mounds selected for flux measurements.

2.3 Mound flux chamber setup

Collars (stainless steel, 15 cm height, 56.5 cm diameter) were placed around the five selected termite mounds a week before the start of the measurements. Collars were inserted approximately 5 cm into the soil and litter layer. A flux chamber was created by use of a 220 L slightly cone-shaped polyethylene bucket, with a diameter of 57.5 cm. A strip of closed-pore foam (1 cm × 1 cm × 57.5 cm) was attached over the whole inner perimeter so that if the bucket was placed on the collar, the foam strip would seal the part between the bucket and the collar. Two one-touch fittings ($\mathrm{1}/\mathrm{4}$ in., SMC Pneumatics) were installed on each side of the bucket. On the inside of the bucket, a four-inlet vertical sampling tube was placed so that air was sampled from different heights (∼ 10, ∼ 25, ∼ 35, and ∼ 50 cm) in the headspace (Clough et al., 2020). The setup (chamber and tubing) was tested for internal emissions of all measured gases. For CO (see Appendix), an internal emission of <0.014 nmol s⁻¹ was found; the presented CO fluxes are not corrected for this possible internal emission.

CH₄ and CO₂ concentrations were measured with a Los Gatos Ultraportable Greenhouse Gas Analyser. The instrument was connected in a closed loop with the flux chamber (2×2 m PTFE tubing, $\mathrm{1}/\mathrm{4}$ in.). For air circulation, the internal pump of the Los Gatos instrument was used with a flow of ∼ 0.35 L min⁻¹. The instrument measures concentrations every second; 10 s averaged concentrations were saved and used for flux calculations. For each measurement, the flux chamber was closed for 20 min, during which time concentrations were measured continuously. All five mounds were always measured on the same day and in the same order. Over one week, each mound was measured three times, each time at approximately the same hour of the day.

2.4 Flux calculations

Fluxes were calculated as follows. By use of the ideal gas law, mole fractions (mol mol⁻¹) were converted to concentrations (mol m⁻³). For chamber temperature, a standard temperature of 25 ^∘C was assumed. For chamber volume (CV), the termite mound volume (Table 1) was deducted from the bucket volume (220 L).

Fluxes could be calculated as follows:

$$\begin{array}{}\text{(2)}& F={\displaystyle \frac{\mathrm{d}C}{\mathrm{d}t}}\cdot \mathrm{CV},\end{array}$$

where F is the mound emission (mol s⁻¹), $\mathrm{d}C/\mathrm{d}t$ is the concentration change (mol m⁻³ s⁻¹), and CV the corrected chamber volume (m³). Linear regression was used to derive the concentration change and the given error bars are the propagated standard error of the linear regression slope. Concentration increases were calculated over the last 10 min of the chamber closure to avoid possible effects of the bag filling (see Appendix). If clear headspace concentration fluctuations were observed in the beginning of this time window, possibly by a remaining effect of the bag filling, the window was shortened by a maximum of 2 min (leaving a time window of 8 min). All calculated $\mathrm{d}C/\mathrm{d}t$ increases showed an R²>0.95. Unless mentioned otherwise, the given mound CO₂ emissions are corrected for the estimated contribution of soil respiration by subtracting the average valley soil emission (see Sect. 2.5). For mound no. 6, the average plateau soil emission was subtracted.

2.5 Valley and mound-adjacent soil fluxes

To quantify the CH₄ and CO₂ emissions of the soils adjacent to the termite mounds, four soil collars were installed around each mound: two soil collars were placed at 20 and 45 cm distance from the mound (distance between mound collar and middle of soil collar) and two additional soil collars were placed on the opposite side of the mound at the same distances. The soil collars were of 20 cm diameter with a height of 10 cm and were inserted 5 cm into the soil. The flux chamber height was 15 cm so that the soil chamber volume was 4.7 L. To be able to connect the Los Gatos instrument, the soil chamber had two one-touch fittings on top. The chamber and collars were created from a common PVC sewage pipe. Every mound-adjacent soil flux measurement was 4 min, and the set of 4 collar measurements was performed once per mound, with the exception of mound no. 19. For mound nos. 13 and 14, the measurements were performed on the second measurement day, for mound nos. 15 and 16, the measurements were done on the third measurement day. Mound-adjacent soil fluxes will be expressed per mound-collar area (0.25 m²) to be better comparable to mound emissions. The same chamber setup was also used in a substudy at a nearby transect (∼ 500 m from termite mounds) where, among other things, valley soil (10 collars) and plateau soil (10 collars) fluxes were measured (three repetitions). Measured soil fluxes from the valley will be shown for comparison.

2.6 Termite mound subsample emission measurements

After each last mound flux measurement, a mound sample was taken of approximately 1 L volume. From this, three small subsamples were taken (volume not determined). When selecting a piece, we tried to look for solid not crumbling pieces, so that the inside of the subsample was undisturbed. From the sample from mound no. 19, only one suitable subsample was found. Each subsample was placed in a small closed box (12.6 cm × 19.2 cm × 6.8 cm) with two one-touch fittings, functioning as a small closed flux chamber. A blank measurement was made with the small box and no internal emissions were found. Each mound subsample was measured with the Los Gatos instrument for 5 min, to determine the CH₄ and CO₂ production in the chamber over time. After each measurement, the mound sample was carefully broken open and termites were counted, so that the CH₄ and CO₂ emission per termite (the termite emission factor) could be calculated. The measurements took place next to the mound and time between sampling and measuring was always less than 15 min. To verify whether the termite emission factor was stable between seasons and mounds, additional measurements were performed. In October 2020 (dry season), the same type of measurements were performed on 15 subsamples of the same termite mounds, and in December 2020 (transition dry–wet season), measurements were performed on five subsamples of a different mound of the same species.

2.7 Termite mass measurement

Termite mass was measured in the Laboratory of Systematics and Ecology of Soil Invertebrates at INPA. A total of 480 living workers of the species N. brasiliensis were weighed in five subgroups ($\mathrm{4}\times n=\mathrm{100}$, $\mathrm{1}\times n=\mathrm{80}$) by use of a precision scale (FA2104N). Reported individual termite mass is fresh weight per termite (mg termite⁻¹).

3 Results

3.1 Mound CH₄ and CO₂ emissions

Headspace concentrations increased strongly during chamber closure, and chamber concentrations climbed up to 5750 nmol CH₄ mol⁻¹ and up to 1950 µmol CO₂ mol⁻¹. CH₄ emissions of mounds nos. 13–19 ranged between 17.0 and 34.8 nmol mound⁻¹ s⁻¹ (Fig. 1), with an average emission of 25.2 nmol mound⁻¹ s⁻¹. Additional valley measurements showed heterogeneous soil CH₄ fluxes with small uptake and emission taking place alongside, ranging between −0.1 and 2.9 nmol m⁻² s⁻¹ (med = −0.02, avg = 0.15, SD = 0.54). Mound-adjacent soil CH₄ fluxes, measured at 20 and 45 cm from the mound, ranged between 0.4 and 8.9 nmol CH₄ m⁻² s⁻¹ (avg = 2.14, SD = 2.00) and were, on average, enhanced in comparison to valley soils (Fig. 2). Soil valley CO₂ fluxes were found to range between 0.9 and 3.7 µmol m⁻² s⁻¹ (avg = 2.14, SD = 0.74) (Fig. 2) and the average plateau soil CO₂ emission was 4.03 µmol m⁻² s⁻¹ (SD = 1.36). Mound-adjacent soil CO₂ fluxes showed an average emission of 4.81 µmol CO₂ m⁻² s⁻¹ (range = 2.0–10.1, SD = 2.04), thereby being enhanced with respect to the surrounding valley soils (Fig. 2). Mound CO₂ emissions, corrected for the average valley and plateau soil respiration, ranged between 1.1 and 13.0 µmol mound⁻¹ s⁻¹, with an average emission of 8.14 µmol mound⁻¹ s⁻¹ (average of mounds nos. 13–19).

Figure 1 CH₄ and CO₂ emissions of mounds nos. 13–19 (in valley) and of mound no. 6 (on plateau) expressed in nmol and µmol mound⁻¹ s⁻¹, which represents a collar area of 0.25 m². All mounds (except mound no. 6) were measured three times during one week and each series no. (#) was measured on the same day and in the same order. Error bars are propagated standard errors of the linear regression slope, as described in Sect. 2.4.

Figure 2 Measured mound emissions and mound-adjacent soil fluxes for CH₄ (a) and CO₂ (b) for mound nos. 13, 14, 15, and 16 expressed in nmol 0.25 m⁻² s⁻¹ for CH₄ and µmol 0.25 m⁻² s⁻¹ for CO₂ (collar area is 0.25 m²). Note that for CO₂ the total mound emissions per collar area not corrected for soil respiration are shown and stated. The centrally placed markers are the measured mound emissions (also for mound no. 19); the larger marker indicates the day-specific mound emission when mound-adjacent soil fluxes were measured. The grey bar indicates the range of additionally measured soil valley fluxes. The range and average flux for each group of measurements are given in the table. On average, measured mound CH₄ and CO₂ fluxes were a factor of 630 and 16 higher in comparison to the surrounding soil valley fluxes.

During chamber closure, the concentration changes in CH₄ and CO₂ were strongly correlated (R²>0.95 for each chamber closure). The ratio between the mound CH₄ and CO₂ emission (${\mathrm{CH}}_{\mathrm{4}}/{\mathrm{CO}}_{\mathrm{2}}$) ranged between 2.1 and $\mathrm{17.1}\times {\mathrm{10}}^{-\mathrm{3}}$ (Fig. 3) and showed a constant ratio when data from mound no. 19 (furthest away from other mounds) and mound no. 6 (different species) were excluded (average ratio: $\mathrm{2.8}\times {\mathrm{10}}^{-\mathrm{3}}$, SD = 0.4). The smallest mound (no. 19) clearly showed smaller-than-average emissions, but in general no strong correlation was found between mound CH₄ emissions and mound height (R²=0.07) or volume (R²=0.08), and a small correlation was found between mound CO₂ emissions and mound height (R²=0.43) and mound volume (R²=0.44).

Figure 3 Mound CO₂ emissions (µmol mound⁻¹ s⁻¹) vs. mound CH₄ emissions (nmol mound⁻¹ s⁻¹). Dotted lines indicate the different $\mathrm{d}{\mathrm{CH}}_{\mathrm{4}}/\mathrm{d}{\mathrm{CO}}_{\mathrm{2}}$ emission ratios.

3.2 Termite weight, individual termite emission, and colony size estimation

The average weight of five subsets of living workers of the species N. brasiliensis was determined and was found to range between 2.83 and 3.33 mg with an average weight of 3.07 mg (SD = 0.18), which is similar to what was found by Pequeno et al. (2013), who reported 3.0 mg (SD = 0.4). Since the species N. brasiliensis has a relatively low soldiers : workers ratio of 1 : 100 (Krishna and Araujo, 1968), we will use the worker weight 3.07 mg (SD = 0.18) as an average termite weight for the species N. brasiliensis.

CH₄ and CO₂ emissions of 13 mound subsamples were measured. For each subsample, the measured gas production was plotted over the counted termites (Fig. 4). The fitted line has a forced intercept at y=0. For CH₄, an emission of 0.0002985 nmol termite⁻¹ s⁻¹ was found (se = $\mathrm{1.77}\times {\mathrm{10}}^{-\mathrm{5}}$) and fitted with an R² of 0.95 (n=13). The set of additional measurements resulted in similar termite CH₄ emission factors, namely 0.0002976 nmol termite⁻¹ s⁻¹ (se = $\mathrm{1.32}\times {\mathrm{10}}^{-\mathrm{5}}$) and 0.0003043 nmol termite⁻¹ s⁻¹ (se = $\mathrm{1.41}\times {\mathrm{10}}^{-\mathrm{5}}$) for the measurements of October and December 2020 respectively. Given estimates in this paper will be based on the termite emission factor of 0.0002985 nmol CH₄ termite⁻¹ s⁻¹. For CO₂, an emission of 0.1316 nmol termite⁻¹ s⁻¹ was found (se = $\mathrm{2.59}\times {\mathrm{10}}^{-\mathrm{2}}$) with an R² of 0.68 (n=13). Excluding the outliers (32, 14.9 nmol s⁻¹ and 313, 80.9 nmol s⁻¹) gave an R² of 0.88 (n=11) with a CO₂ emission of 0.074 nmol termite⁻¹ s⁻¹ (se = $\mathrm{8.5}\times {\mathrm{10}}^{-\mathrm{3}}$). Converting the emission rates from termite to termite mass (fresh weight) and from seconds to hourly rates gives a termite emission factor of 0.35 µmol g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ (se = 0.02) for CH₄ and of 86.8 µmol g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ (se = 10.0) for CO₂ (Table 2).

Figure 4 CH₄ production (left axis, green triangles) and CO₂ production (right axis, blue circles) over counted termites. The lines (green solid for CH₄, blue dashed for CO₂) represent a linear regression fit with forced intercept at y=0. For CH₄, a production of 0.0002985 nmol termite⁻¹ s⁻¹ (se = $\mathrm{1.77}\times {\mathrm{10}}^{-\mathrm{5}}$, R²=0.95) was found and, for CO₂, a production of 0.1316 nmol termite⁻¹ s⁻¹ (se = $\mathrm{2.59}\times {\mathrm{10}}^{-\mathrm{2}}$, R²=0.68) was found. Excluding the outliers (32, 14.9 nmol s⁻¹ & 313, 80.9 nmol s⁻¹) gives an R² of 0.88 (n=11) with a CO₂ emission of 0.074 nmol termite⁻¹ s⁻¹ (se = $\mathrm{8.5}\times {\mathrm{10}}^{-\mathrm{3}}$). For comparison, two sets of additional subsample CH₄ emission measurements are shown. The first additional measurements (AM1, light grey triangles) resulted in a termite emission factor of 0.0002976 nmol termite⁻¹ s⁻¹ (se = $\mathrm{1.32}\times {\mathrm{10}}^{-\mathrm{5}}$) (one measurement point (599 termites, 0.165 nmol s⁻¹) is not shown in this figure). The second set (AM2, dark grey triangles) gave a termite emission factor of 0.0003043 nmol termite⁻¹ s⁻¹ (se = $\mathrm{1.41}\times {\mathrm{10}}^{-\mathrm{5}}$).

By combining the termite CH₄ emission factor with the termite mound CH₄ emissions, colony sizes were estimated. Colony size estimates were based on the highest measured emissions and were found to range between 55–125 thousand individuals (Table 3). Colony size can also be estimated by use of mound volume or mound external surface. Table 3 shows the colony size estimates based on values as given by Lepage and Darlington (2000) for termites in general and also shows the estimates based the “mound volume–termite biomass” relation found by Pequeno et al. (2013), specifically for the species N. brasiliensis.

Brauman et al. (1992)Eggleton et al. (1999)Fraser et al. (1986)Konaté et al. (2003)Martius et al. (1993)Rouland et al. (1993)Sawadogo et al. (2011)Sugimoto et al. (1998a)Khalil et al. (1990)Macdonald et al. (1999)Martius et al. (1993)Seiler et al. (1984)Sugimoto et al. (1998a)Brümmer et al. (2009a)Jamali et al. (2013)Queiroz (2004)

Table 2 Overview of literature values for CH₄ and CO₂ emissions of termites per weight (upper part), emissions per termite mound (middle part), and emissions per area (lower part). Values from this study are indicated in bold. If reported, the average and SD are given, otherwise a range is indicated. If multiple values were found, measurements from higher soil-feeding termite species were selected. For each study, the graph or table where the data were found is given. The ${\mathrm{CH}}_{\mathrm{4}}/{\mathrm{CO}}_{\mathrm{2}}$ is given in molar ratio (10⁻³).

^a Sawadogo et al. (2011) reported emissions per dry weight mass. To convert to fresh weight, a formula as reported by Pequeno et al. (2017) was used. With an assumed dry weight of 0.5 mg, a fresh weight of 1.57 mg is deducted. ^b Mound emissions are divided by a collar area of 0.25 m². ^c Calculated based on average values in this table. ^d Neocapritermes brasiliensis. ^e Crenetermes albotarsalis, Cubitermes fungifaber, Cubitermes speciosus, Noditermes sp., Procubitermes sp., Thoracotermes macrothorax; ^f Dicuspiditermes santschii, Dicuspiditermes nemorosus, Pericapritermes semarangi, Procapritermes nr. sandakanensis, Homallotermes eleanorae, Proaciculitermes sp. A, Pericapritermes nitobei. ^g Coptotermes lacteus. ^h Ancistrotermes cavithorax, Odontotermes n. pauperans. ⁱ Nasutitermes macrocephalus, Nasutitermes corniger, Nasutitermes surinamensis, Nasutitermes sp., Nasutitermes ephratae, Nasutitermes araujoi. ^j Noditermes sp., Crenetermes albotarsalis, Cubitermes speciosus, Thoracotermes macrothorax, Astratotermes sp. ^k Macrotermes bellicosus. ^l Microcerotermes sp., Globitermes suplhureus, Termes sp., Dicuspiditermes sp. ^m Drepanotermes perniger, Nasutitermes magnus, Nasutitermes triodiae, Tumulitermes pastinator, Amitermes laurensis, Coptotermes lacteus. ⁿ Bulbitermes sp. C, Dicuspiditermes nemorosus, Dicuspiditermes santschii. ^o Macrotermes and Odontotermes (Macrotermitinae), Trinervitermes (Nasutitermitinae), Amitermes and Cubitermes (Termitinae), Hodotermes (lower termite). ^p Cubitermes fungifaber. ^q Microcerotermes nervosus, Turnulitermes pastinator, Turnulitermes hastilis, Amitermes meridionalis.

Table 3 Colony size estimates (CSEs) based on different methods; values given per thousand (×10³). “Mound volume” is the estimated mound volume as given in Table 1 and “Mound emission” is the highest measured emission per individual mound.

^a CSE based on the highest measured mound CH₄ emission and combined with an emission factor of 0.0002985 nmol CH₄ termite⁻¹ s⁻¹ (se = $\mathrm{1.77}\times {\mathrm{10}}^{-\mathrm{5}}$). ^b CSE based on mound volume by use of mound termite density values (0.2–5.6 termite cm⁻³; Lepage and Darlington, 2000). ^c CSE based on mound surface area (given in Table 1) by use of mound termite surface values (5.6–16.7 termite cm⁻²; Lepage and Darlington, 2000). ^d CSE based on mound volume by the species-specific volume–population equation $y=\mathrm{47.94}\cdot x^{\mathrm{0.47}}$ (x is mound volume (L) and y is colony biomass (g)) as given by Pequeno et al. (2013); for termite weight, 3.07 mg (SD = 0.18) was used. Since mound no. 6 was of a different species, it is not included in this table.

4 Discussion

4.1 CH₄ and CO₂ emissions

Measured mound CH₄ emissions were of similar magnitude to emissions found by previous studies (Table 2, middle and lower part). The termite emission factor, determined for the soil-feeding species N. brasiliensis, was found to be 0.35 µmol g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ (SD = 0.02), which is similar to values found for other species (Table 2, upper part) but almost two times higher than the average value reported by Martius et al. (1993) for a wood-feeding species in the Amazon (0.19 µmol g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹). Our emission rate is within the reported range of 0.1–0.4 µmol g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ for soil feeders (Sugimoto et al., 2000). Mound CO₂ emissions and the termite CO₂ emission factor were similar to or a little higher than the few values found in literature (Table 2). Nevertheless, since mound material and termites were measured together, the contribution of indirect termite emissions, i.e. mound respiration, cannot be quantified, so that the direct termite-produced CO₂ emission is presumably lower.

There is a large variety in type of termite mounds (shape and size are dependent on, among other things, species, ecosystem, and climate; Noirot and Darlington, 2000), explaining the wide range of reported termite mound CH₄ emissions (Table 2, middle and lower part). In situ measurement of termite mounds gives information about the net CH₄ emission under natural conditions but is unable to distinguish sources and sinks inside the mound. One known CH₄ sink in termite mounds is the uptake by methanotrophic bacteria, which are also responsible for the CH₄ uptake in aerobic soils. The presence and magnitude of this process have been discussed and reviewed by different studies (Ho et al., 2013; Khalil et al., 1990; Macdonald et al., 1998; Nauer et al., 2018; Seiler et al., 1984; Sugimoto et al., 1998a; Pester et al., 2007; Reuß et al., 2015). The role of possible mound CH₄ uptake should also be acknowledged for the measurement of individual termite emissions (Table 2, upper part); most literature values, including values from this study, are based on termite incubation in the presence of mound material, with ongoing CH₄ uptake; therefore, actual termite CH₄ emission values might be higher.

Small variation in mound emission magnitudes was observed between measurement days. This can be caused by a variation in colony size (due to foraging activities) or termite activity driven by fluctuations in temperature or radiation (Jamali et al., 2011a; Ohiagu and Wood, 1976; Sands, 1965; Seiler et al., 1984). However, as our termite mounds are in a tropical forest with relatively constant temperatures and only indirect daylight, strong diurnal temperature and radiation patterns are not expected. Small variation can also be caused by minimal air transport below the soil collar through the porous upper soil layer; during preliminary tests without a collar, we observed that even a light forest breeze can cause chamber headspace variations. In case our setup was subject to minor air transport below the collar, the given mound emission estimates will be slightly underestimated with respect to the actual mound emissions. Another possible underestimation is caused by the estimated corrected chamber volume CV, as used in Eq. (2). In this study, we considered the mound volume as a solid body. A previous study considered the solid nest volume as 10 % of the actual mound volume (Martius et al., 1993), leading to a larger corrected chamber volume and therefore to larger calculated mound emissions. By use of this approach, average calculated mound emissions would increase by almost 30 % to be 32.7 instead of 25.2 nmol CH₄ mound⁻¹ s⁻¹.

The mound emission ${\mathrm{CH}}_{\mathrm{4}}/{\mathrm{CO}}_{\mathrm{2}}$ ratio was found to be relatively constant over four of the five mounds. While values in literature indicate a wide range of ${\mathrm{CH}}_{\mathrm{4}}/{\mathrm{CO}}_{\mathrm{2}}$ ratios (Table 2), both Seiler et al. (1984) and Jamali et al. (2013) found little variation between mounds of the same species and concluded that the ratio is species-specific. Our average variation of a factor of ∼4 between mounds of the same species is of the same magnitude as what was observed in earlier studies (Seiler et al., 1984; Jamali et al., 2013).

4.2 Colony size estimate

To estimate colony sizes of (epigeal) nest building termites, different methods exist. One method is by fumigation of the nest (to prevent colony evacuation) followed by excavation, after which termites can be removed from the nest debris by flotation in water. This process is labour intensive and can take five persons up to three weeks to finish one nest (Darlington, 1984; Jones et al., 2005). A faster method is by subsampling known volumes of the mound, counting the termites in the subsample, and extrapolating this to the total mound volume. Termite mounds can have irregular shapes; therefore, volume estimates strongly depend on which volume estimation approach is used (Jones et al., 2005).

The population estimation method we tested combined CH₄ mound emissions with a termite emission factor measured in situ at the field site. We estimated colony sizes ranging between 57.6 and 124.0 thousand termites per mound. For all mounds, our population estimate was in the estimated range based on mound volume or external surface area, as taken from literature equations (Table 3). Comparison to estimates based on a N. brasiliensis species-specific equation shows an average difference of 20 % (Pequeno et al., 2013). It should be noted that the relation found between mound volume and colony biomass by Pequeno et al. (2013) was quite weak (R²=0.41), and our estimates would fit in the general spread they observed in their data. Interestingly, Pequeno et al. (2013) concluded that mound volume is a weak indicator for population size for nests of the species N. brasiliensis, as also indicated by the weak correlation we found between mound volume and mound CH₄ emissions.

The influence of mound CH₄ uptake on our population estimate method should be considered: mound methanotrophic CH₄ uptake probably decreases the net mound CH₄ emission, resulting in an underestimation of the colony size when linking it to termite emission factors, as also suggested by Nauer et al. (2018). However, our termite emission factor was determined inside small pieces of undisturbed mound material, so that the material's CH₄ uptake rate was presumably only mildly affected. It is therefore likely that our termite emission factor is underestimated to the same degree as our mound emissions; therefore, both values can still be combined.

Overall, our colony size estimation approach can be considered as a test case for a quick population estimation method. The combination of one mound flux measurement (15 min) in combination with five subsample measurements (5×5 min) can be performed within 1 h, thereby being faster than the original methods. Also, the method is applicable to epigeal mounds of all species, independent of internal mound structure (Josens and Soki, 2010) or species characteristics (Pequeno et al., 2013). In addition, the method is not strongly dependent on a correct mound volume estimate, which remains a source of uncertainty (Jones et al., 2005) and which has been shown to be a weak indicator of population size for some species (Pequeno et al., 2013; Josens and Soki, 2010). Moreover, mounds can also be measured several times in a row before the subsample measurement, so that colony size dynamics over time can be studied non-invasively. A disadvantage of this method is that it is only applicable to freestanding epigeal mounds, at least with the current type of chamber setup. For a possible follow-up study, we propose a setup wherein the different methods are compared.

4.3 Role of termites on the ecosystem scale

Valley soil CH₄ and CO₂ fluxes were similar to what was found by earlier studies (Souza, 2005; Moura, 2012; Chambers et al., 2004; Zanchi et al., 2014). On average, mound-adjacent soil CH₄ and CO₂ fluxes were enhanced with respect to valley soils, although differences were small and no clear emission pattern with “distance to mound” was observed. While mound-adjacent soil fluxes are possibly enhanced, we preferred to avoid overestimation and decided to treat termite mounds as very local hot spots, with measured fluxes only representative for the collar area of 0.25 m². On average, CH₄ and CO₂ fluxes per collar area were found to be a factor ∼630 and ∼16 higher when an active termite mound was present.

To estimate the role of termites on the ecosystem scale, one approach is to combine mound emission values with termite mound density numbers. A local study reported a density value of 21.6 mound ha⁻¹ for the species N. brasiliensis specifically (Pequeno, 2014), which would lead to an average CH₄ emission of 0.05 nmol m⁻² s⁻¹ caused by mounds of this species alone. Non-species-specific mound densities are known to vary strongly between and within ecosystems (Ackerman, 2006, Appendix B8). We found five local studies reporting mound (epigeal nest) density values, which were ∼ 100 mound ha⁻¹ (Queiroz, 2004), 193 mound ha⁻¹ (Oliveira, 2016), 250 mound ha⁻¹ (Dambros et al., 2016), 60 and 280 mound ha⁻¹ (de Souza and Brown, 1994), and even 760 mound ha⁻¹ (Ackerman et al., 2007). When excluding the strong outlier of 760 mound ha⁻¹, the emission of termite mounds on the ecosystem scale was estimated to range between 0.15–0.71 nmol m⁻² s⁻¹ for CH₄ and between 0.05–0.23 µmol m⁻² s⁻¹ for CO₂. Since (epigeal) mounds only represent a part of the total termite community, and not the termites located in the subsoil, in dead wood, or on trees (arboreal nests), this emission value underestimates the actual role of termites on the ecosystem scale. To our knowledge, only Bandeira and Torres (1985) (as given in Martius et al., 1996) assessed the ratio between nest-building vs. total termite biomass and estimated it to be ∼ 0.16. Considering the limited literature on this subject, we prefer to not further extrapolate our mound emission measurements.

A more comprehensive approach is to use termite biomass estimates and combine them with termite emission factors, a method which is commonly used for global CH₄ budget studies (Kirschke et al., 2013; Saunois et al., 2020). For active tropical ecosystems, a termite biomass of ∼ 11 g termite m⁻² is generally assumed (Bignell and Eggleton, 2000; Kirschke et al., 2013; Sanderson, 1996; Saunois et al., 2020). Considering the previously found value of 0.19 µmol CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ for wood-feeding termites in the Amazon (Martius et al., 1993) and our newly found termite emission factor of 0.35 µmol CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ for a soil-feeding termite, a termite-derived ecosystem CH₄ emission range of 0.6–1.1 nmol m⁻² s⁻¹ can be calculated. For CO₂, our termite emission factor of 86.8 µmol CO₂ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ leads to a termite-induced ecosystem CO₂ emission of ∼ 0.27 µmol CO₂ m⁻² s⁻¹.

An overview of the different estimates is given in Table 4. Each of these estimates are based on measurements from mounds and termites found in the valley, which were only measured during the wet season. Nevertheless, an exploratory measurement of a small mound of a different species on the plateau (mound no. 6) indicated CH₄ fluxes of a similar magnitude in comparison to a similar-sized mound in the valley (mound no. 19). Furthermore, exploratory dry season measurements of the same mounds showed emissions of similar magnitude (not shown) and additional dry season mound subsample measurements revealed very consistent termite CH₄ emission factors (Fig. 4). We therefore do not expect that mound CH₄ emissions are only of importance in the valleys or only present in the wet season.

Table 4 Overview of termite-induced CH₄ and CO₂ emissions based on two different approaches. For comparison, the lowest row shows the total (not termite-specific) ecosystem CH₄ and CO₂ flux values, measured at the same field site by previous studies.

^a Querino et al. (2011) performed above-canopy Eddy Covariance CH₄ flux measurements and reported an average CH₄ flux of ∼ 2 nmol m⁻² s⁻¹. ^b Chambers et al. (2004) quantified different respiratory CO₂ sources in this ecosystem and estimated the total ecosystem respiration to be 7.8 µmol CO₂ m⁻² s⁻¹.

To put the estimates in perspective, non-termite-specific ecosystem CH₄ and CO₂ fluxes measured at this field site during earlier studies are given. Ecosystem termite CO₂ emissions were estimated to range between 0.05–0.27 µmol m⁻² s⁻¹, which is approximately ∼ 1 %–3 % of the estimated total ecosystem respiration (7.8 µmol m⁻² s⁻¹; Chambers et al., 2004). Nevertheless, since the “emission per mound” as well as the “termite emission factor” are both affected by indirect effects of termite activity (mound respiration), the contribution of direct termite-emitted CO₂ into the ecosystem is presumably smaller. For CH₄, we rather expect an underestimation than an overestimation of our termite and mound emission values, therefore we expect that these ecosystem estimates are lower bound. For CH₄, it is difficult to judge the role on the ecosystem scale since the earlier measured CH₄ flux (above canopy EC measurements, ∼ 2.0 nmol m⁻² s⁻¹; Querino et al., 2011) is a net flux of uptake and emission processes with relatively unknown individual magnitudes. Nevertheless, considering the magnitude of our estimated termite-emitted CH₄ emissions (0.15–1.1 nmol m⁻² s⁻¹), it is expected that termites play a significant role in this terra firme ecosystem.

Termites contribution to tropical South America CH₄ budget

In current CH₄ budget studies, a termite emission factor of 2.8 µg CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ is used for “Tropical ecosystems and Mediterranean shrub lands” (Kirschke et al., 2013; Saunois et al., 2020),² which is mainly based on field studies in Africa and Australia (Brümmer et al., 2009a; Jamali et al., 2011a, b; Macdonald et al., 1998, 1999; Sanderson, 1996). The only termite emission factor measured for the Amazon rainforest is by Martius et al. (1993) (3.0 µg g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹) for a wood-feeding species, which are expected to emit less CH₄ than soil-feeding species (Bignell and Eggleton, 2000; Brauman et al., 1992). As a back-of-the-envelope calculation based on Kirschke et al. (2013): 36 % of global termite emission (11 Tg) is expected to come from the region of “tropical South America” ($\mathrm{0.36}\cdot \mathrm{11}=\mathrm{3.96}$ Tg). Substituting the emission factor of 2.8 with the newly found 5.6 µg CH₄ g${}_{\text{termite}}^{-\mathrm{1}}$ h⁻¹ would increase this regions estimate to 7.92 Tg and thereby the global estimate to 14.96 Tg.

Our study points out that termite emissions are still an uncertain source in the CH₄ budget and are especially poorly quantified for the Amazon rainforest. Measurement of CH₄ emissions from different termite species, preferably covering species of different feeding or nesting habits in combination with more precise termite distribution and abundance data, would allow more precise estimates and a better understanding of the role of termites in the CH₄ budget.

5 Conclusions

In situ measurement of termite mound CH₄ and CO₂ emissions confirmed that mounds are important local hot spots, playing a considerable role on the ecosystem scale. Measured mound emissions of the species N. brasiliensis were of similar magnitude to observed emissions for different soil- and wood-feeding species, and mounds showed a relatively constant ${\mathrm{CH}}_{\mathrm{4}}/{\mathrm{CO}}_{\mathrm{2}}$ emission ratio. By performing emission measurements on small groups of termites, we derived a termite CH₄ emission factor, so far only the second value reported for the Amazon rainforest. The newly found termite emission factor, measured for a soil-feeding species, is almost twice as high as the previously reported average value for the Amazon, which was determined for a wood-feeding species. By combining mound emissions and termite emission factors, mound colony sizes were estimated and values were similar to estimates based on a literature review. Considering the quick, widely applicable, and non-destructive nature of this approach, we propose that it can be used as an alternative to the traditional methods that are intrusive and time-consuming.

Assessment of the magnitude of termite-emitted CH₄ on the ecosystem scale was attempted by two approaches. Mound emission values were combined with mound density numbers, leading to an estimate of 0.15–0.71 nmol CH₄ m⁻² s⁻¹ emitted by mounds, on average; since this estimate neglects emission from termite activity outside mounds, the number is likely an underestimation. The CH₄ termite emission factor from this study and from the only other Amazon field study were combined with termite biomass numbers, resulting in an estimate of termite-emitted CH₄ of 0.6–1.1 nmol m⁻² s⁻¹. Considering the relatively low CH₄ emissions previously measured at this ecosystem, we expect that termites play an important role in the CH₄ budget of this terra firme ecosystem.

Appendix A: Termite mounds: N₂O, CO, and δ¹³C of CO₂

A1 Methodology

In addition to the direct mound CH₄ and CO₂ emission measurements (performed with the Los Gatos instrument), mound N₂O and CO fluxes and the δ¹³C of the mound CO₂ flux were determined by the following method. Three bags (5 L inert foil, Sigma-Aldrich) were sampled consecutively during chamber closure. The bags were measured on the same or the consecutive day with a Spectronus FTIR analyser, which can quantify concentrations of CO₂, CH₄, N₂O, and CO, and can determine the δ¹³C of CO₂. The N₂O and the δ¹³C of CO₂ measurements of the FTIR analyser have a cross sensitivity for CO₂ concentrations, which is well quantified for the CO₂ range 380–800 ppm (Hammer et al., 2013). In order to sample air with CO₂ concentrations <800 µmol mol⁻¹, air samples were taken in the first minutes after chamber closure (2, 5, 8 min). Out of the 45 sample bags taken, two bag samples could not be used.

Before measurement of the bag sample, sample lines were flushed with bag sample air. Air samples were dried by a Nafion dryer and by a column of magnesium perchlorate. Measurements were corrected for pressure and temperature variations as well as for cross sensitivities (Hammer et al., 2013). For more information on this instrument, please refer to Griffith et al. (2012). For calibration of the instrument, two calibration gases were used: gas 1 with values 381.8 µmol CO₂ mol⁻¹, 2494.9 nmol CH₄ mol⁻¹, 336.6 nmol N₂O mol⁻¹, 431.0 nmol CO mol⁻¹, and a δ¹³C of CO₂ of −7.95 ‰, and gas 2 with values 501.6 µmol CO₂ mol⁻¹, 2127.0 nmol CH₄ mol⁻¹, 327.8 nmol N₂O mol⁻¹, 256.7 nmol CO mol⁻¹, and a δ¹³C of CO₂ of −14.41 ‰.

To calculate the fluxes of N₂O and CO, FTIR-measured bag concentrations of N₂O, CO and CO₂ were used. For each chamber closure, the $\frac{\mathrm{d}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}{\mathrm{d}t}$, $\frac{\mathrm{dCO}}{\mathrm{d}t}$, and $\frac{\mathrm{d}{\mathrm{CO}}_{\mathrm{2}}}{\mathrm{d}t}$ were calculated so that the ratios $\frac{\mathrm{d}{\mathrm{N}}_{\mathrm{2}}\mathrm{O}}{\mathrm{d}{\mathrm{CO}}_{\mathrm{2}}}$, and $\frac{\mathrm{dCO}}{\mathrm{d}{\mathrm{CO}}_{\mathrm{2}}}$ could be derived. To calculate the fluxes of N₂O and CO, the ratios were combined with the in situ determined mound CO₂ flux, as measured by the Los Gatos instrument. This approach was chosen because the intended 3 min bag sampling interval was not always accomplished, so that an exact Δt could not be assumed with certainty. To determine the δ¹³C of the CO₂ emitted by the termite mounds, Keeling plots were used (Pataki et al., 2003).

Figure A1 CO emissions of valley mound nos. 13–19, expressed in nmol mound⁻¹ s⁻¹, which represents a collar area of 0.25 m². All mounds were measured three times during one week and each series no. (#) was measured on the same day and in the same order.

Figure A2 δ¹³C of CO₂ emitted by mounds nos. 13–19, derived by use of Keeling plots. Error bars represent the standard error of the linear regression intercept. Red squares indicate intercepts based on linear regression fits with R²<0.99 or based on linear regression with only two instead of three sample points. All mounds were measured three times during one week, and each series no. was measured on the same day and in the same order. Averages were calculated for each mound, which were −38.1 ‰ (mound no. 13, se = 0.9), −36.2 ‰ (mound no. 14, se = 1.0), −35.7 ‰ (mound no. 15, se = 0.1), −34.7 ‰ (mound no. 16, se = 1.4), and −34.7 ‰ (mound no. 19, se = 1.3). For calculation of these averages, values with a linear regression of R²<0.99 or values based on a linear regression of only two measurements (indicated as dark red squares) were excluded.

A2 Mound N₂O and CO fluxes

Gas samples (three samples per chamber closure) revealed stable N₂O concentrations and headspace concentrations ranged between 333.7 and 342.4 nmol mol⁻¹ over the different chamber closures. Since headspace CO₂ concentrations sometimes exceeded 800 µmol mol⁻¹ and N₂O–CO₂ cross sensitivity becomes uncertain at higher CO₂ concentrations (Hammer et al., 2013), not all three headspace samples per chamber closure could be used; therefore, qualitative N₂O flux estimates cannot be reported. As a back-of-the-envelope calculation, N₂O fluxes were calculated if two headspace samples were with CO₂ <800 µmol mol⁻¹ and if a minimum N₂O concentration difference of 0.18 nmol mol⁻¹ was found (FTIR precision (σ) for 5 min spectra is 0.09 nmol mol⁻¹), which gave us three mound flux estimates ranging between 0.03 and 0.11 nmol N₂O mound⁻¹ s⁻¹. Similarly low fluxes were found during additionally performed soil flux measurements, performed as part of a substudy, which showed valley soil fluxes ranging between 0.008–0.106 nmol m⁻² s⁻¹. The low mound fluxes are in agreement with a previous study suggesting that termite mound N₂O emissions are dependent on the nitrogen content of the termites diet (Brauman et al., 2015), which is expected to be low in the valleys of this ecosystem (Quesada et al., 2010).

Chamber CO concentrations ranged between 120 and 220 nmol mol⁻¹ and showed a clear uptake on all days and for all mounds, ranging between −0.04 to −0.78 nmol mound⁻¹ s⁻¹ (Fig. A1). Termite mound uptake has been observed before by Khalil et al. (1990). We expect that the observed uptake is caused by aerobic CO-oxidising bacteria in the mound, which are also responsible for the CO uptake in (tropical) soils (Conrad, 1996; Kisselle et al., 2002; Liu et al., 2018; Potter et al., 1996; Whalen and Reeburgh, 2001; Yonemura et al., 2000a). Soil CO uptake is dependent on atmospheric CO and therefore often limited by low soil diffusivity (Sun et al., 2018; Yonemura et al., 2000b). The dry porous mound material (Martius et al., 1993) is therefore a suitable place for CO uptake.

A3 δ¹³C of the mound-emitted CO₂

For each chamber measurement, a mound-specific δ¹³C value of the CO₂ flux was determined. Figure A2 shows the Keeling plot intercepts, wherein error bars represent the standard errors of the intercept. In general, the values were more depleted than values found by de Araújo et al. (2008), who found a δ¹³C of −30.1 ‰ for valley litter during the dry season (August 2004). However, for our measurements, at least one sample bag per chamber closure was with CO₂ > 800 µmol mol⁻¹, so that the CO₂ cross sensitivity correction for these samples was less certain. Intercepts based on only the first two concentrations points, which were generally lower (or around) 800 µmol mol⁻¹, resulted, on average, in less depleted (∼ 1 ‰) δ¹³C values. To investigate if these values are representative for other mounds and to investigate whether an isotopic difference exists between mound- and soil-emitted CO₂, more measurements would be needed.