Mound architecture and season affect concentrations of CO2, CH4 and N2O in nests of African fungus‐growing termites

Termites are a significant natural source of greenhouse gases (GHGs), but quantifying emissions especially from large termite mounds is problematic as they rarely fit in measurement chambers. Predicting fluxes based on internal and atmospheric concentrations could provide an indirect way to assess mound emissions, but developing such models necessitates better understanding of the concentration levels and their variance. We used gas chromatography to measure carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) concentrations within nests of two co‐occurring species of fungus‐growing termites in both dry and wet seasons in Kenya, with termite Macrotermes michaelseni building mounds with a closed and Macrotermes subhyalinus with an open ventilation system. Gas concentrations were 3–100 times higher in mounds than the global averages in atmosphere, implying that termite mounds are sources of all three GHGs. Carbon dioxide concentrations were higher in closed than in open mounds. Methane concentrations remained constant in open mounds, whereas closed mounds exhibited considerable variation between nests and across seasons. Concentrations of both CH4 and N2O correlated positively with mound volume during the wet season, whereas interactions with mound size were not observed in CO2 concentrations or during the driest sampling period. These findings underline that among fungus‐growing termites, mound size, ventilation type and precipitation affect nest gas concentrations and with this likely the magnitude of mound GHG emissions. Potential reasons behind the observed relationships are discussed, including differences in population size, biomass of fungus gardens and CH4 oxidation.


INTRODUCTION
Termites, especially those species that cultivate fungi (Termitomyces spp., Basidiomycota) to assist in plant matter degradation (i.e.fungusgrowing termites, subfamily Macrotermitinae), are important carbon and nutrient recyclers in many dry tropical ecosystems (Collins, 1981;Jones, 1990;Jouquet et al., 2011).For example, in East African arid and semi-arid savannas, the biomass of grass harvested by the fungus-growing termites corresponds roughly to that consumed by grazing mammal herbivores (Lepage, 1979(Lepage, , 1981a)), and wood litter degraded in their nests may represent even 90% of the dead wood formed in the ecosystem (Buxton, 1981).Fungus-growing termites collect dead plant matter, including litter from grasses, trees and shrubs (Boutton et al., 1983;Lepage et al., 1993;Vesala, Rikkinen, et al., 2022) and translocate the material into their climatically controlled nest chambers.There, the plant cell walls are broken down into simple sugars by actions of the Termitomyces symbionts in wellaerated fungus combs and by the intestinal bacterial flora during two subsequent passages through guts of termite workers (Badertscher et al., 1983;Nobre & Aanen, 2012;Poulsen et al., 2014;Vesala, Arppe, & Rikkinen, 2022).As a result of effective decomposition, termite nests commonly emit an order of magnitude higher amounts of carbon dioxide (CO 2 ) than the surrounding savanna soils (Konaté et al., 2003;Räsänen et al., 2023;Van Asperen et al., 2021).
Alongside producing CO 2 , termites are among the rare insects that produce methane (CH 4 ) through activities of hindgut microbes (Brauman et al., 1992;Brune, 2018).Due to the high total biomass of termites in many tropical ecosystems, their contribution to the global CH 4 budget is significant (Collins & Wood, 1984;Fraser et al., 1986;Kirschke et al., 2013;Rasmussen & Khalil, 1983;Sanderson, 1996;Zimmerman et al., 1982).Currently, contribution of termites to the global atmospheric CH 4 budget is estimated to be 3-15 Tg CH 4 y À1 , corresponding to approximately 1%-4% of CH 4 released annually from non-anthropogenic sources (Saunois et al., 2020).In addition to CO 2 and CH 4 , nests of some termites maintain slightly elevated levels of nitrous oxide (N 2 O) (Khalil et al., 1990), a greenhouse gas (GHG) with a 273-fold warming impact compared with CO 2 (Forster et al., 2021).Brümmer et al. (2008Brümmer et al. ( , 2009) ) demonstrated that mounds of a soilfeeding termite Cubitermes fungifaber (Sjöstedt 1896) were significant N 2 O sources in a West African savanna ecosystem, where the surrounding soils generally had low N 2 O emission rates.It has been suggested that the magnitude and direction of N 2 O fluxes could be linked to termite diet: under controlled laboratory conditions workers of soilfeeding and fungus-growing termites, generally consuming relatively nitrogenous food sources, emitted N 2 O, whereas wood-feeding species thriving on low-nitrogen food sources were rather net N 2 O consumers (Brauman et al., 2015).
Macrotermitinae termites construct intricate air circulation systems to maintain ideal thermal and climatic conditions for their fungal symbionts and ensure sufficient oxygen for the decomposition process within the fungus gardens (Korb, 2003;Korb & Linsenmair, 1998;Noirot & Darlington, 2000).Large termite mounds built by the genus Macrotermes serve as the most prominent examples of these complex systems (Korb, 2011;Noirot & Darlington, 2000).Their mounds may include termite populations of up to a few million individuals and contain several kilograms of fungus combs (Darlington, 1986) consisting of decomposing plant matter and Termitomyces mycelium.Mature colonies of Macrotermes jeanneli (Grassé 1937) in Kenya were measured to produce 800-1500 L of CO 2 and 0.5-1.3L of CH 4 per day (Darlington et al., 1997).Such production rates of metabolic gases within a relatively small mound interior space necessitate efficient gas exchange between the nest chambers and the ambient air.
Despite the potential relevance of termites for savanna GHG balances, current estimates of termite-derived GHG emissions contain high levels of uncertainty due to lack of information about termite biomasses in many ecosystems and inadequate flux measurements considering the high taxonomic and functional diversity within the insect group.Direct flux measurements are, however, challenging, especially in case of large Macrotermes mounds with their height and width often exceeding 2 and 5 m, respectively.In case of several mound-building termite species, the concentrations of CO 2 and CH 4 inside the nest have been found to correlate positively with the respective fluxes from nests to the atmosphere (Jamali et al., 2013;Khalil et al., 1990) and could thus potentially be used to model GHG emissions.Developing such models, however, necessitates a better understanding of baseline levels and variance of the internal gas concentrations in different types of termite mounds.
Closed mounds rely on within-mound air circulation driven by solar heat induced convection flows in the air conduits near the mound surface, and the gas exchange between internal and ambient air takes place through porous outer walls of the mound (Korb, 2011;Ocko et al., 2017;Turner, 2000Turner, , 2001) ) (Figure 1a).Airflow in an individual cavity may change during a day depending on the position of the sun in relation to the mound, until after sunset cooling of the mound surface reverses the circulation (Ocko et al., 2017).In open mounds, ventilation is facilitated by wind-induced currents of fresh air that pass through the underground hive via several air channels opening to the mound surface at different elevations (Darlington, 1984;Korb, 2011;Weir, 1973) (Figure 1b).Unlike in closed mounds, gas exchange with fresh air takes place at several underground locations through soil walls that separate the actual nest and fungus comb chambers from the ventilation channels (Darlington, 1984;Noirot & Darlington, 2000).
Although the physical basis of gas exchange is reasonably well known in both closed and open systems (King et al., 2015;Korb & Linsenmair, 2000;Ocko et al., 2017;Turner, 2001;Weir, 1973), the concentrations of different gases inside the nests have not been studied systematically.Instead, most reported observations are derived from sporadic measurements of individual mounds, typically representing the closed ventilation type, and as far as we know, N 2 O concentrations have not been registered from any species of fungus-growing termites.Lüscher (1956Lüscher ( , 1961) ) was the first to document CO 2 concentrations in different parts of a closed mound built by Macrotermes bellicosus (Smeathman 1781) in the Ivory Coast.The highest CO 2 concentrations in the nest concavities were 2.7% (27,000 ppm), exceeding atmospheric levels by two orders of magnitude.Even higher values were reported by Ocko et al. (2017), who found that CO 2 concentrations within the hive of a Namibian M. michaelseni colony remained at a constant level of ca.50,000 ppm throughout the day.In addition to CO 2 , approximately tenfold CH 4 concentrations (15.8-21.8ppm) compared to ambient air (1.8 ppm) have been reported from Macrotermes mounds (Noirot & Darlington, 2000).Given that the two mound ventilation types most likely differ in their efficiency to facilitate mound-atmosphere gas exchange, concentrations and mutual proportions of GHGs might differ systematically between the open and closed mounds.Correspondingly, nest internal gas concentrations might show a relationship with the mound volume, as the efficiency of gas exchange is commonly thought to increase with the increasing mound dimensions.
Here, we used gas chromatography to explore mound internal gas (1) the two mound ventilation types differ in their gas exchange efficiency leading to systematic differences in gas concentrations between the open and the closed mounds, and (2) as a result of enhanced ventilation, the nest internal gas concentrations would decrease with the increasing mound volume.Furthermore, we hypothesised that (3) nest internal gas concentrations show seasonal variance generated, for example, by changes in precipitation and food availability.

Studied termite mounds
A total of 30 Macrotermes mounds were studied in Taita-Taveta County, Kenya, during three field campaigns in June 2019, November 2019 and June 2021.The June 2019 sampling trip was shortly after the 'long rains' (March-May), representing a period of good availability of fresh plant matter and moderate soil humidity (Figure S1).In F I G U R E 1 Simplified cross sections and main principles of nest ventilation in mounds built by two Macrotermes species in southern Kenya (based on Darlington, 1984Darlington, , 1985;;Korb, 2011;Noirot & Darlington, 2000;Ocko et al., 2017;Turner, 2000Turner, , 2001;;and Weir, 1973).Air currents drawn in the closed mound type (a) represent situation in midday when sun heats both sides of the mound equally creating a temperature gradient between the narrow surface cavities (warmer) and the nest chambers (cooler).During different times of the day, direction of air currents may further change depending on which side of the mound is faced towards the sun.At night, internal air circulation is powered by the metabolic heat generated within the nest and fungal galleries and therefore the air circulation turns to the opposite direction (temperature gradient from high temperature inside to lower temperature outside).In mounds with an open ventilation system (b), ambient air flows through several air channels piercing the nest, and gas exchange takes place in multiple locations within the hive through soil walls that separate nest and fungus chambers from the air channels.Air movement in the ventilation channels of open mounds is generated by wind, which typically draws out air from passages on the mound top while replacement air is sucked in from the passages near the mound base.
November 2019, samplings took place in the middle of the 'short rains' (November-December), when soil was much softer and moister than during the first trip in June 2019 (Figure S1).Conditions during the third trip in June 2021 were much drier than that in June 2019, as the amount of precipitation received during the preceding rainy season (in March-May 2021) had been below average (Figure S1).This also caused mounds and the adjacent soils to be dry and hard com- Twenty-one of the studied mounds were closed ventilation (M.subhyalinus), and nine were open ventilation type (M.michaelseni).In order to increase the number of independent observations, different termite colonies were targeted during each of the three sampling trips, with the exception of six mounds that were sampled repeatedly during consecutive field trips.See Table S1 for the summary of the sampled mounds during each field trip.
To compare nest internal gas concentrations with the mound size, the height and width of each mound were measured in two directions (South-West and North-East).Based on these measured parameters, the volume of the aboveground mound was estimated using the equation (Equation 1) defined in Räsänen et al. (2023): where R is the mound radius (mean of the two directions) and H is the height from the estimated ground level (beyond outwash pediment if present) to the mound top.

Sampling and gas chromatography
A small hole (dimeter 20 mm) was drilled at the mound base to the fungus chambers using a long custom-made hand drill (Figure S2).
The location of sample collection was aimed to an area with active fungus combs with the drilled depth ranging from 55 to 110 cm from the mound surface.In seven mounds, samples were collected separately from two opposite sides of the mound to assess within-nest variation and to evaluate representativeness of our sampling.As a pre-screening step, colony activity and a direct connection of the sampling hole to fungus chambers were confirmed by visual examination using an endoscope camera (ATP Instrumentation, VB-END5).If fresh fungus combs and living termites were detected, an aluminium pipe (20 mm Â 1200 mm) with several 6 mm holes near the apex (Figure S2) was inserted into the hole, and a multi-use transfer pump connected to a small custom chamber (Nalgene HDPE 250 mL flask) was attached to the end of the pipe.As an initial check of metabolic gas production, the CO 2 concentration of nest internal air, drawn out with the pump system, was measured with an infrared CO 2 probe (Testo 440 CO 2 kit) inserted into the chamber.If CO 2 concentrations of 5000 ppm or more were observed, sampling for gas chromatography proceeded.Prior to sampling, the contact point between the aluminium pipe and mound was sealed using moistened soil, ambient air was removed and the pipe was flushed with nest air using the pump.
Subsequently the external end of the aluminium pipe was closed with a double layer of duct tape, with a probe wire of a digital thermometer and a 2-5 m long gas sampling tube (PTFE, 6/4 mm) inserted through the tape and pushed to the internal end of the pipe (Figure S2).
Three parallel gas samples were taken in sequence into 20 mL evacuated glass vials equipped with aluminium caps and septa using a 60 mL luer-lock syringe and a needle connected to the PTFE sampling tube with a 3-way stopcock.To flush the vials with nest air, a volume of 40 mL of the gas sample was pressed through the vial using a release needle, after which the remaining 20 mL of the sample was pressed into the vial to create over-pressure to decrease risk of contamination with ambient air during sample storage (Arias-Navarro et al., 2013).Before taking samples and between parallel samples, the sampling PTFE tube was flushed by pumping 3 Â 60 mL nest air through the tube.After installing the aluminium pipe, gas sampling was finished in less than 10 min.The brevity of the sampling time and the multiple perforations drilled to the aluminium pipe ensured free flow of gas from the nest interior space to the sampling tube without significant hindrance caused by the generally swift onset of repair activities by termite workers.At the end, the depth of the sampling point (from mound surface) was measured, and sampling time was registered.Gas collections were performed only during daytime (from 10:00 AM to 18:00 PM).

Data analysis
Based on preliminary data evaluation, the first samples of each threesample sequence collected during the last sampling campaign in June

Gas concentrations in open and closed mounds
Concentrations of all three gases were several times higher in termite nests compared with the global averages in atmosphere (Figure 2).for individual field campaigns (Figure 3).Principal components 1 and 2 explained 52% and 33% of the total variance, respectively.Concentrations of CH 4 and CO 2 were the main contributors to the PC1 with loading values 0.70 and 0.68, respectively, whereas the main driver for PC2 was N 2 O concentration with a loading value À0.96.

The effect of mound volume
The linear models for each season-specific dataset indicated that nest internal CO 2 concentrations did not have any relationship with the mound volume (Figure 4a-c).This result was the same for datasets collected during all three field campaigns and regardless of whether the mound volume was included in the models either alone or together with the mound type.In contrast, mound internal CH 4 concentrations correlated significantly with mound volume in the datasets collected during the first two field campaigns (June 2019 and November 2019), representing wetter conditions compared to the third campaign (June 2021).
In June 2019, besides mound volume, also mound type and the interaction of the two variables had significant (p < 0.05, Table S2) effects on nest CH 4 concentrations.As shown in Figure 4d, this result could be interpreted as a positive correlation between CH 4 concentrations and mound volume in closed but not in open mounds.However, we note that the sample size for the closed type for this correlation was only four, and thus, this interpretation should be treated with added caution.
In the dataset for June 2021 (the driest sampling period), mound volume did not correlate with CH 4 concentrations when the potential effect of mound ventilation type was included in the model.Instead, for this dataset, even very small mounds had higher CH 4 concentrations compared with most mounds sampled during the two previous campaigns (Figure 4f).S2) but only during the most humid field campaign in November 2019 (Figure 4h).During that period, the N 2 O levels also tended to be generally at a higher level than during the two other campaigns.

Relationships between CO 2 , CH 4 and N 2 O
Despite of the high variation associated with CH 4 concentrations within closed mounds, there was a significant (p < 0.01) correlation between the nest internal CH 4 and CO 2 concentrations, indicating linkage between the two gases (Figure 5a).This linkage appeared stronger in open than in closed mounds, which was also reflected in the CH 4 :CO 2 concentration ratios that varied significantly less (F-test: F 24,11 = 4.69, p = 0.01) in open (Figure 5b) than in closed (Figure 5c) mounds.In closed mounds, the CH 4 :CO 2 concentration ratios tended to be higher in June 2021 compared with the two earlier campaigns, but based on the applied mixed effects model, this difference was not significant.
Concentrations of N 2 O did not have any relationships with either CO 2 or CH 4 concentrations in either closed or open mounds (Figure S5).

CO 2 and CH 4 closed and open mounds
Mound ventilation type was an important factor affecting nest internal CO 2 and CH 4 levels, with open mounds having significantly lower CO 2 and less variable CH 4 concentrations than closed mounds.Most likely, both differences can be explained by the fundamental differences in the nest ventilation mechanism between the two mound types (Figure 1).In closed M. michaelseni mounds, nest internal air is in direct contact with the fungus gardens, royal chamber and the nursery galleries, and circulates throughout the mound interior until gases are exchanged through the porous outer surface at the mound top (Darlington, 1984;Korb, 2011;Noirot & Darlington, 2000;Ocko et al., 2017;Turner, 2001).
In contrast, in open M. subhyalinus mounds, gas exchange with the ambient air takes place in numerous locations within the underground part of the mound, where the open ventilation channels are in contact with different nest compartments (Darlington, 1985;Korb, 2011;Noirot & Darlington, 2000;Weir, 1973), thus creating multiple relatively separated gas spaces within a single mound.
Compared to the internal gas cycling and centralised gas exchange of closed mounds, the dispersed ventilation of open mounds seems to provide a more efficient way to reduce CO 2 levels within the nest interior.Compared to the CO 2 levels found in the present study, Korb andLinsenmair (1999, 2000) reported clearly lower CO 2 concentrations in West African M. bellicosus mounds, having a closed ventilation system largely resembling that of M. michaelseni.The CO 2 levels of somewhat less than 20,000 ppm on average, registered from dome-shaped M. bellicosus mounds in a cool forest habitat, were considered supraoptimal for the termite colonies due to the impairing effect of elevated CO 2 concentrations on fungal metabolism and, thus, on colony food supply (Korb & Linsenmair, 1999) 2b) and in relation to CO 2 (Figure 5).This difference between the two mound types could potentially arise from several reasons.First, in open mounds, gas exchange between the nest and the ambient air takes place close to the sites where the gases are produced (Figure 1b).Thus, the gas mixture in the fungal chambers mainly originates from Termitomyces fungus combs and from the termite workers maintaining the fungal cultivations.Unlike termite workers and soldiers, fungus combs are not known to produce notable amounts of CH 4 (Darlington et al., 1997) 5b) implies that within fungal chambers that are relatively well separated from the rest of the nest interior in that mound type, the biomass ratio between fungus combs and the nursing termites remains more or less constant.
The situation could be quite different in closed mounds, where gas spaces within the fungal gardens are less insulated from the rest of the nest interior (Figure 1a).Hence, a mixture of gases drawn from a closed mound possibly represents more a total outcome of a whole colony's metabolism, including different termite castes and age groups and the fungus gardens, than it does in open mounds.
Considering the potentially important role of termite population size and biomass of fungal gardens in determining the gas concentrations and their mutual relationships within nests of fungus-growing termites, it is regretful that our non-invasive study approach precluded the quantification of these parameters from the studied mounds.Thus, the discussion related to these variables is largely speculative and draws on previously published findings to provide some interpretations of the observed patterns.Based on earlier long-term studies on Kenyan M. michaelseni colonies (Darlington, 1986;Lepage, 1981a), the amount of fungus comb material per unit of termite biomass is known to vary between different seasons.Following the logic that CH 4 originates solely from termites whereas CO 2 is produced also by the Termitomyces fungi, such variation should directly reflect to the mutual abundances of CH 4 and CO 2 (i.e., CH 4 :CO 2 concentration ratios) within the mounds.On average, the highest comb biomasses in relation to termites are found in May-June, when nests do not yet have a brood of developing alates (nymphs), and the total biomass of fungus combs is high due to active foraging (Lepage, 1981a(Lepage, , 1981b)).During that period, total biomass of fungus combs per unit of termite biomass can be twofold compared to the end of the long dry season (October-November) when food reserves have been largely consumed and the sexual brood reaches its maximum size (Darlington, 1986;Lepage, 1981a).Although mass-specific CH 4 production rates of nymphs and alates are typically much lower than those of worker termites (Darlington et al., 1997;Sugimoto, Inoue, Tayasu, et al., 1998), the seasonally high biomass of the reproductive castes (Darlington, 1986) could increase the CH 4 production of a termite colony during times when the sexual brood is at its largest.2018) who also demonstrated that, depending on termite species, CH 4 oxidation could take place either in thick and porous wall structures within the aboveground mounds, or in soil surrounding the underground nest chambers.Whether corresponding methanotrophic activity could also take place in mounds of fungus-growing termites is currently a controversial question and the results originating from the two Macrotermes species studied so far are in sharp contrast (Sugimoto, Inoue, Kirtibutr, & Abe, 1998;Tyler et al., 1988).Sugimoto et al. (2000)

Mound volume and seasonal effects
In addition to ventilation type, also the size of the aboveground mound might affect nest internal gas concentrations.In case of both termite species studied here, a positive correlation has been demonstrated between the external dimensions of the mound and the size of the termite population (Darlington, 1990;Darlington & Dransfield, 1987).Assuming that the nest internal volume increases equally with the increased production of metabolic gases of a larger termite population and their fungus gardens the CO 2 levels within the nest interior should remain relatively constant apart from the mound dimensions.At the same time, a mound with higher chimneys or a larger surface area is most likely more efficiently ventilated than a small mound with modest aboveground structures, which, in turn, could decrease the gas concentrations within the nest interior.To constantly maintain optimal CO 2 concentrations within their nests, termites most likely are able to adjust dimensions of their aboveground mounds to specifically correspond the production of metabolic gases by the colony.This hypothesis, also referred to as social homeostasis (Turner, 2001), is strongly supported by our results, as the CO 2 concentrations did not increase with the mound volume in either closed or open Macrotermes mounds.
Instead, there was a significant positive correlation between mound volume and nest internal CH 4 concentration (Figure 4).This production rate.In contrast to the two first field campaigns, we did not find any correlation between mound size and internal CH 4 levels during the third campaign in June 2021.Instead, during that time, even very small mounds had higher CH 4 concentrations than most mounds during the two earlier campaigns (Figure 4f).Consistently, the CH 4 to CO 2 concentration ratios of closed mounds also tended to be higher in June 2021 compared to the previous sampling periods (Figure 5b).The most obvious differences between the three field campaigns were related to precipitation conditions and soil moisture.
In year 2019, both sampling trips took place either right after (June) or during (November) the rainy spells with relatively high soil humidity (Figure S1).Instead, in June 2021, soil humidity levels were much lower than 2 years earlier (Figure S1).Dry soil conditions in June

CONCLUSIONS
The results of this study demonstrate that the concentrations of the concentrations and mutual relationships of CO 2 , CH 4 and N 2 O in closed and open Macrotermes mounds in the Tsavo Ecosystem in southern Kenya.Samples were collected during three field trips in 2019 and 2021 representing a continuum of different precipitation conditions ranging from short rainy season with abundant precipitation (November 2019) to drought (June 2021).Our aim was to provide robust, field-based estimates of typical concentration ranges of the three important GHGs within the mounds of fungusgrowing termites during different seasons.We hypothesised that: pared with the two earlier sampling times.Rainfall and soil moisture data for the entire study period were obtained from the Maktau weather station of the Taita Research Station located at a distance of 15-45 km from the studied termite mounds.Half (n = 15) of the mounds were in a dense Acacia-Commiphora woodland (Kasigau Road) intensively grazed by cattle, and the other half (n = 15) in an open savanna grassland within a nearby nature reserve that experienced mainly wildlife grazing (Taita Hills Wildlife Sanctuary).
subsequent parallel samples and, thus, these outlying observations were omitted.The likely source of error was identified as ambient air contamination caused by insufficient flushing prior to sampling, related to the longer PTFE sampling tube used during the third field trip compared to the two earlier trips.For mound TR408 sampled in November 2019, one of the three replicate samples showed systematically lower concentrations for all three gases, indicating leakage or contamination during sampling, and these values were omitted from the dataset.For all other mounds, means and standard deviations were calculated from three replicates for each studied chamber.The standard deviations were mostly <1000 and <1.0 ppm for CO 2 and CH 4 , respectively, and <50 ppb for N 2 O, with a few exceptions that showed higher variation (FigureS3).In cases where samples were drawn from two parallel chambers from the opposite sides of the mound (n = 7), the concentrations in two chambers were compared to evaluate within-mound variation, and subsequently, the values from parallel chambers were averaged to obtain mean values for each studied mound for further analysis.The variance within the gas mixtures of different mounds was first characterised using principal component analysis (PCA).Concentrations of each studied gas (CO 2 , CH 4 , and N 2 O) were included as variables in the PCA after normalisation and centralisation of the data values.Variances of internal CO 2 , CH 4 and N 2 O concentrations and the CH 4 :CO 2 concentration ratios were compared between the closed and open mounds using a two-tailed F-test.To compare GHG concentrations and the CH 4 :CO 2 concentration ratios in closed and open mounds, a linear mixed modelling approach was adopted as six colonies were sampled repeatedly during the study.Four separate models were run using each of the three gases or the CH 4 :CO 2 concentration ratio as a dependent variable.Mound identity was used as a random effect in all linear mixed effects models that were fitted by restricted maximum likelihood (REML) using nlme package (Pinheiro et al., 2021).Mound ventilation type (open/closed), season (field campaign), sampling time (10-12, 12-15 or 15-18 o'clock), nest temperature, vegetation type (woodland/grassland) and the depth and cardinal direction of the sampling hole were included as fixed factors in the models one by one to find potential variables affecting the gas concentrations.Based on preliminary data exploration that revealed considerable differences in gas concentrations in relation to mound volume between the different field trips, the effect of mound volume on nest internal gas concentrations was studied separately for each of the three field campaigns using general linear models.In addition to mound volume, mound ventilation type and interaction of the two variables were included in the models to see whether mound size affected gas F I G U R E 2 Concentrations of CO 2 (a), CH 4 (b) and N 2 O (c) in mounds with closed or open ventilation system.Dashed horizontal line in each panel indicates the mean atmospheric concentration of the gas in question.Boxes and whiskers show variation (median, 1st and 3rd quartiles, and 1.5 IQR) of the mean values obtained from individual mounds.Numbers within the boxes indicate the number of observations.Significant p value shown in panel A originates from linear mixed model analysis indicating a difference in CO 2 concentrations between the two mound types.Mean concentrations of the two other gases did not differ between closed and open mounds but closed mounds had significantly higher variance in nest internal CH 4 levels.F I G U R E 3 Principal component analysis demonstrating systematic differences in CO 2 , CH 4 and N 2 O concentrations between closed and open mound types, especially when separated by the three measurement campaigns.Filled and open symbols refer to closed and open mounds, respectively.Each convex hull delimits observations from closed or open mounds during one field trip.Note that one exceptionally large closed mound with anomalously high internal N 2 O level sampled in June 2021 (discussed in the text) was left outside of its convex hull to improve clarity.concentrations differently in closed and open mounds.All statistical analyses were performed in R version 4.1.0(R Core Team, 2021).

Measured CO 2
concentrations ranged from 7382 to 39,233 ppm, being significantly (p < 0.001) higher in closed than in open mounds (Figure2a).Concentrations of the two other gases ranged from 2.9 to 67.9 ppm for CH 4 and from 300 to 5551 ppb for N 2 O. Concentrations of either CH 4 or N 2 O did not differ between the two mound types, but between-mound variation of CH 4 concentrations was significantly higher (F-test: F 24,11 = 11.174,p < 0.001) in closed than in open mounds (Figure 2b).Methane concentrations in open mounds ranged from 10 to 30 ppm, whereas variation in closed mounds was by far larger with the concentrations differing even more than 20-fold F I G U R E 4 Nest internal concentrations of CO 2 (a-c), CH 4 (d-f) and N 2 O (g-i) in relation to mound volume during three different sampling campaigns (June 2019, November 2019, June 2021).Closed and open symbols refer to closed and open mounds, respectively.Regression lines are shown in those cases when mound volume had a significant effect on internal gas concentrations when included in the linear model either alone (e,h) or together with mound type and interaction of the two variables (d).See Table S2 for the regression equations and summary of results.between different mounds.Parallel fungus comb chambers studied from the opposite sides of the same mounds differed from each other up to 45% (CO 2 ), 47% (CH 4 ) and 28% (N 2 O), with the two mounds of the open ventilation type (TS351 and TS353) showing higher within-nest variation in CO 2 and CH 4 concentrations than any of the closed mounds (Figure S3).Whether the mounds were located in open grassland or in woodland did not affect concentrations of any of the studied gases.Likewise, sampling depth or side (N, S, E, W), nest temperature and time of the day were insignificant in all models when added as fixed variables.PCA combining information from all three gases indicated that gas mixtures of the closed and open mounds were systematically distinct from each other.This pattern was especially clear (non-overlapping convex hulls) In the June 2021 dataset, sample size for open mounds was too small (N = 2) to detect any potential differences between the closed and the open mound type.The nest internal N 2 O concentrations correlated positively with the mound volume (p < 0.05, Table

F
I G U R E 5 (a) Relationship between CH 4 and CO 2 concentrations in closed and open mounds (filled and open symbols, respectively).(b and c) Variation (median, 1st and 3rd quartiles and 1.5 IQR) in CH 4 :CO 2 concentration ratios in closed and open mounds during three different sampling trips.harbourseveral times higher levels of CO 2 and CH 4 compared to atmospheric concentrations (Figure2).We found that CO 2 concentrations were roughly 20-100 and CH 4 concentrations up to 38 times higher than the mean global atmospheric concentrations.In addition, N 2 O concentrations were on average three and occasionally more than 10 times higher in termite mounds than in atmosphere.Although we currently do not have data on N 2 O fluxes from any Macrotermes mounds, the elevated N 2 O concentrations suggest that, in addition to CO 2 and CH 4 , they may act as locally significant sources of N 2 O emissions especially during the rainy season, similar to what has been pre- and, thus, any major change in the biomass of fungus comb material in relation to termite biomass should alter the CH 4 :CO 2 production ratios, with an increased biomass of termites in relation to fungus combs leading to a relatively larger proportion of CH 4 , and vice versa.The overall consistency of CH 4 :CO 2 concentration ratios in open mounds (Figure Another possible explanation for the observed high variance in CH 4 concentrations within the closed mounds could be related to CH 4 oxidation.Based on differences in the carbon stable isotope composition of CH 4 initially produced by the termites versus CH 4 emitted from the mounds, Sugimoto, Inoue, Tayasu, et al. (1998) calculated that for a variety of termite species studied in Thailand, more than half of the CH 4 was oxidised in mound soil structures before reaching the atmosphere.Similar results were published from Australian moundbuilding termites by Nauer et al. ( effect was, however, detected only during the two field campaigns (June and November 2019, Figure 4d,e) representing relatively moist soil conditions.Interactions found in June 2019 (Figure 4d) also suggest that the correlation between the CH 4 concentration and mound volume could be restricted mainly to the closed mound type, but due to the relatively low sample size (N = 4 for closed mounds), this interpretation should be treated with some caution.As corresponding relationships were not found in CO 2 concentrations of the same mounds during the same time periods (Figure 4a,b), the observed positive correlation between CH 4 concentrations and mound volume might be related to CH 4 oxidation.The observation of higher CH 4 concentrations in larger mounds could, for example, be associated with CH 4 oxidising capacity of the mound methanotrophic microbiota that might become saturated in large mounds with high termite biomass and CH 4 2021 may have reduced methanotrophic activity, which could potentially explain the higher CH 4 levels observed in the closed mounds during that time compared to the much moister period in June 2019.Currently, it is not well understood how different environmental variables affect CH 4 oxidation within termite mounds or (semi)arid soils, but, as a crucial factor for any microbiota, sufficient soil water content is likely among the most important factors controlling activity of the methanotrophic bacteria.Variables related to N 2 ORegarding N 2 O, the highest concentrations were generally detected during the November 2019 field campaign, which took place in the midst of the short rains.During that time, N 2 O concentrations also correlated positively with the mound volume (Figure4h), but at that point, reasons for this correlation remain unclear.Interestingly, one closed mound with a very large volume had exceptionally high N 2 O concentration (>5000 ppb) during the driest campaign in June 2021 when all the other mounds showed low N 2 O levels (Figure4i).Neither CO 2 nor CH 4 concentrations within this anomalous mound were particularly high (Figure S5), which, together with the lack of any relationships between N 2 O and the two other gases, suggests that mechanisms affecting the nest internal N 2 O levels can operate quite independently from those controlling the concentrations of CO 2 and CH 4 .In general, the N 2 O concentrations measured in this study are in line with previous observations from West African savannas, where N 2 O fluxes from mounds of a soil-feeding termite Cubitermes fungifaber were much higher than those from the surrounding soils, especially during rainy season (Brümmer et al., 2009).Consistently with that, Wachiye et al. (2020) found a slight increase in soil N 2 O emission rates at the onset of both rainy seasons in a range of different landuse types in southern Kenya, although the fluxes were generally very low.Such differences in soil N 2 O production might partially explain the differences in average N 2 O concentrations within termite nests detected in the present study.However, termite gut-associated bacteria have also been shown to produce N 2 O (Brauman et al., 2015) and, thus, it remains to be resolved what proportion of nest N 2 O originates from soil microbes and how much of it is derived from the termites.Additionally, fungus combs might harbour nitrification and denitrification bacteria and could thus potentially contribute to nest N 2 O production.Brauman et al. (2015) found that workers of termite species relying on relatively nitrogen-rich diets emitted more N 2 O than those having N poor diets.Thus, also dietary differences, which can significantly alter fungus comb N content and C:N stoichiometry(Vesala,   Rikkinen, et al., 2022), could potentially induce both spatial and intraannual variations in N 2 O concentrations within Macrotermes mounds.However, more information is needed about sources and sinks of N 2 O in termite nest ecosystems to draw more definite conclusions about nest internal N 2 O levels.
three major greenhouse gases (CO 2 , CH 4 , and N 2 O) show high levels of variation within mounds of fungus-growing termites depending, for example, on the nest architectural features and season.Mounds with the open ventilation system had significantly lower internal CO 2 concentrations than those with closed ventilation, suggesting that the open system provides the termite colonies with a more efficient way to remove metabolic gases than the closed system.Mean concentrations of CH 4 or N 2 O were not affected by the nest ventilation type but, instead, showed season-specific correlations with the mound volume.As linkages between the nest internal gas concentrations and the emitted fluxes have been demonstrated in various moundbuilding termite species, the high variance observed in nest gas concentrations imply that also the GHG fluxes from mounds to the atmosphere could vary seasonally and between architecturally different mounds.These observations thus highlight the need to recognise relevant variables including mound ventilation type and size, as well as seasonal variables, when assessing the role of termites in GHG budgets.While aiming to model termite mound GHG fluxes based on internal concentrations, our results point out some key parameters that should be included.Most importantly, the different baseline levels of CO 2 observed between the closed and the open mounds need to be considered while assessing CO 2 emissions from mounds with different ventilation systems.Correspondingly, interactions of CH 4 and N 2 O concentrations with the mound volume and the seasonal effects should not be ignored when aiming to model CH 4 and N 2 O emissions.The root causes behind most relationships observed in the present study are currently obscure and deserve more research.For example, the high variation associated with the CH 4 concentrations and the interaction patterns of CH 4 and CO 2 suggests that methanotrophic bacteria may have an important role in regulating net CH 4 emission of African Macrotermitinae mounds, which should be addressed in future studies.Likewise, very little is currently known about the interactions related to N 2 O levels within termite mounds, which seem to vary remarkably.Long-term year-round monitoring of nest gas concentrations together with key environmental variables, possibly combined with flux measurements in small mounds applicable to chamber measurements, could provide a viable approach to achieve a more comprehensive understanding of the seasonal variation, its drivers and the linkages between nest internal concentrations and emissions of different GHGs.Our results also demonstrate that mounds relying on closed ventilation should be preferentially used for such studies as, due to the mixing effect of the within-mound air circulation, gas samples collected from closed mounds more reliably represent the output at the whole colony level than those drawn from the open mounds with a more compartmentalised nest structure.AUTHOR CONTRIBUTIONS Risto Vesala: Conceptualization; investigation; writingoriginal draft; methodology; validation; visualization; writingreview and editing; formal analysis; data curation.Matti Räsänen: Writingreview and editing; visualization; data curation; validation.Sonja Leitner: Investigation; writingreview and editing.Daniel Girma Mulat: Investigation; writingreview and editing.Lucas Mwangala: Writingreview and editing; resources; project administration.Jouko Rikkinen: Conceptualization; investigation; validation; writingreview and editing; project administration; resources.Laura Arppe: Conceptualization; investigation; funding acquisition; methodology; validation; writingreview and editing; project administration; resources; supervision.ACKNOWLEDGEMENTS This research was funded by the Academy of Finland (decision 333868, Laura Arppe).Studied termite mounds were located within Taita Hills Wildlife Sanctuary and Mgeno Ranch.These two landowners are acknowledged for their cooperation during the study.Professor Petri Pellikka and the Taita Research Station of University of Helsinki are acknowledged for the research facilities.We thank field assistants Peter Mwasi and Darius Kimuzi for their important contribution to sample collection.Sonja Leitner and Daniel G. Mulat would like to thank all funders who contributed to the CGIAR initiatives Livestock, Climate and System Resilience (LCSR) and Low Emission Food Systems (MITIGATE+) and to the CGIAR Trust Fund.Research was done under the research authorization from National Commission for Science, Technology and Innovation of Kenya (NACOSTI/P/21/10199).