Soil Is a Net Source of Methane in Tropical African Forests

: Research Highlights: Monitoring of soil CH 4 ﬂuxes in African tropical forest conducted run for almost two years, contributing to the scant information on greenhouse gas (GHG) ﬂuxes from forests available from this region. Data showed that the forest soil acted as a net yearly source of CH 4 . Hotspots of CH 4 emissions were measured both in upland and lowland areas of the forest, and on an annual basis they overcame the soil CH 4 sink during drier periods or in well-drained areas. Background and Objectives: Atmospheric studies indicate that tropics are a strong CH 4 source. Regional budgets attribute the majority of this source to wetland ecosystems and ﬂooded lowland forests, whereas un-ﬂooded forests are considered net CH 4 sinks, although few studies in tropical forests, in particular in Africa, are available. The present work aims to contribute to this knowledge gap. Materials and Methods: Monitoring campaigns were conducted along the year in the tropical forest of the Ankasa National Park, Ghana, in two contrasting environments, uphill and downhill, using close static chambers coupled with gas chromatography. Results: The uphill area was a net weak CH 4 sink with mean daily ﬂuxes ranging from − 1.29 to 0.44 mg CH 4 m − 2 d − 1 . The downhill area was a signiﬁcant CH 4 source with mean daily ﬂuxes ranging from − 0.67 to 188.09 mg CH 4 m − 2 d − 1 and with peaks up to 1312 mg CH 4 m − 2 d − 1 in the wet season. Conclusions: The net annual soil CH 4 budget for the Ankasa Park, normalizing the proportion of downhill areas over the whole park surface, was a source of about 3.3 kg CH 4 ha − 1 yr − 1 . Overlooking such areas might lead to underestimates of the total CH 4 source strength of forested areas.


Introduction
Methane (CH 4 ) is the second most relevant biogenic greenhouse gas, after carbon dioxide, in terms of radiative forcing on the climate [1,2]. The CH 4 atmospheric concentration depends on many natural and anthropogenic sources and sinks [2][3][4] and it has been increasing, in the last decades, at variable rates, the strongest being reached in 2014 (12.5 ppb CH 4 yr −1 ) [2,4]. On a global basis, tropical areas are considered a key source area of CH 4 production [2][3][4]). However, as top-down estimates cannot disentangle sources and sinks of CH 4 when more processes occur simultaneously, global CH 4 estimates have been quantified in combination with bottom-up approaches, mainly process-based modelling and/or inventory [2][3][4]. This detailed analysis indicated that wetlands and forested, flooded or saturated Forests 2020, 11 lowlands (permanent and ephemeral) were the most prominent CH 4 source in the tropics. This source, however, is also characterized by significant uncertainty [3]. A key feature of CH 4 producing systems is the occurrence of anaerobic conditions in soils and sediments, either saturated by high water table or under flooded conditions. As a consequence, the correct quantification of the surface of inundated or saturated soils is a critical factor in determining wetland CH 4 source strength and extension and it greatly contributes to the high uncertainty associated to this CH 4 source [5]. In land surface models, the extent of flooded land is computed from remote sensing data or by using hydrological models [3][4][5]. Both approaches are characterized by high uncertainty as remote sensing signal deteriorates over dense vegetation, which typically occurs in tropical forested areas, and because the microwave signals can only detect water above or at the soil surface [6], leaving a "black hole" in the quantification of sub-saturated or saturated soil in forested areas. On the other hand, hydrogeological models must rely on complex sets of information often not available for extended areas of tropical regions [5].
Where macroscopic evidence of water inundation fails due to the technical limits of its detection, the CH 4 budget of tropical forests is hence computed by land surface models, treating forested areas mostly as net CH 4 sinks [2][3][4]. Aerobic soils can provide moderate and transient sources of CH 4 , due to methanogenic activity generated in anoxic microsites within an aerobic soil matrix. This phenomenon has been hypothesized in order to explain the relatively small net CH 4 emissions observed from upland and aerobic soils across a range of continents and ecosystems, ranging from well-drained systems to tropical mountain grasslands [7][8][9][10][11][12][13][14][15]. However, also unaccounted for is subterranean termite activity, particularly in tropical soils, which has been proposed as a possible explanation for observed hotspots of CH 4 emissions in aerobic tropical soils [7,8,15]. Although, in some cases CH 4 emissions as high as those in wetlands have been reported [16]; the majority of the abovementioned net CH 4 emissions observed in un-flooded forests are small compared with inundated areas [2], but their relevance as a global CH 4 source is due to on the dramatic extension of these tropical and subtropical natural areas.
The strength and the extension of the CH 4 source from these areas is, however, still high uncertain [2]. Moreover, the availability of only few and short time series of in situ CH 4 measurements, makes tropical areas some of the least constrained regions for inversions using surface measurements [3].
To gather further evidence on the contribution of tropical forests to CH 4 soil-atmosphere exchange, a study was conducted in a humid equatorial tropical forest in the Ankasa National Park, Ghana. Several campaigns to quantify the CH 4 soil-atmosphere gas exchange were carried out along one year in two contrasting environments representative of the main extreme conditions for CH 4 production and consumption we might typically find in these areas, uphill and downhill, and the evaluation of the overall annual net flux of the Ankasa forest was provided.

The Study Site
The study site was a mix moist semi-deciduous and wet evergreen forest (05 • 16 11.2 N; 02 • 41 41.55 W, 104 m a.s.l.,) located in the Ankasa Conservation Area in Ghana [17]. The mean annual temperature is about 25°C and the mean annual precipitation is 1500-2000 mm, mainly concentrated from the end of March to mid-July and from September to November with a lower intensity [18]. A dry period extends from December to March and a shorter dry period occurs in August [18]. The landscape is characterized by the presence of hills (average elevation 90 m a.s.l.), alternating with narrow valleys. Considering the whole surface of the Ankasa Park, the downhill area was estimated to be around 20% of total Park surface (data derived by SRTM DEM at 90 m resolution, CGIAR-CSI free source). The soils are classified as Oxisols [19]. They developed on coherent biotite-rich granites of the Cape Coast complex to form the Ankasa association consisting of "Abenia" series, the most widespread series on top of the hills, which alternates with the "Ankasa" series, along the slopes. Soils are deeply weathered and highly acidic with a pH in the upland areas varying between 3.6 at 0-5 cm depth and 4.5 at 15-30 cm depth. Alluvial soils occupy the bottom of the fluvial valleys, presenting a pH between 4.3 (0-5 cm  [17]). Contents of sand, silt and clay along the soil profile in the two areas are reported in Castaldi et al. [17]; overall the sand was the predominant soil mineral fraction. In the top-hill, soil C was highly concentrated in the top 5 cm (average C value 9.8 ± 1.12 g kg −1 , average total N content 0.66 ± 0.21 g N kg −1 ), decreasing to 1.15 ± 0.51 g C kg −1 at 5-15 cm depth and to 0.78 ± 0.19 g C kg −1 between 30-50 cm depth. In the downhill areas, C content varied between 2.74 and 2.44 g C kg −1 in the top 15 cm, reaching 1.10 ± 0.05 g C kg −1 at 30-50 cm depth [17]. Surface litter C in the top-hill area was estimated around 15 ± 9 Mg C ha −1 while C stock in the mineral soil, to a depth of 1 m (151 ± 20 Mg C ha −1 ), was estimate to be in the same order of magnitude of the aboveground biomass C (138-170 Mg C ha −1 ) including live and dead wood [20].

Experimental Set up
Fluxes were measured over an area of about half a hectare in two contrasting environments, uphill and downhill, using eight chambers (replicates) in each site per each day of sampling. The first monitoring site to be established was the top-hill area, as measurements started as part of a wider monitoring experiment aimed at quantifying the forest GHG exchange, which also included CO 2 gas exchange measurements by eddy covariance. Top-hill measurement campaigns of soil-atmosphere gas exchange were conducted from April 2009 to November 2010. The measurements in the downhill area started in May 2010 and lasted until November 2010, covering the wet season period, when CH 4 emissions were hypothesized to occur, and the short dry season of August. Campaigns to monitor gas fluxes were conducted for six consequent days, with few exceptions, each month. Chamber position was modified at each sampling event so as to improve the coverage of the spatial variability over the study. However, in order to avoid the "soil disturbance effect" due to chamber base insertion in the soil, bases were positioned in the ground the evening before sampling [17,21]. In the wet season, during the periods of most intense rainfall, ephemeral ponds (around 50 m 2 , max water depth around 20 cm) occurred in the lowest locations of the downhill area. The pond water drained within a week or two, depending on the rainfall duration and intensity. The monitoring campaigns were not continuous, but concentrated over few days per each month, therefore it was not possible to estimate the total area of forest soil which was oversaturated along the year. However, in order to have an idea of the order of magnitude of the CH 4 concentration in the water, and of the magnitude of the diffusive flux which might derive from these ponds, in one of the campaigns, water was sampled from the pond closest to the chamber sampling area over three consecutive days, while the water was slowly draining away.

Determination of Gas Exchange between Atmosphere and Soil or Water
Soil-atmosphere gas exchange was measured using closed static PVC chambers (20 cm h × 15 cm ∅) which were placed on bases inserted in the ground (7 cm h × 15 cm ∅) [17]. Gas fluxes were withdrawn at time 0, 30 and 60 min after closure and gas was stored in slightly overpressured, pre-evacuated gas-tight vials (20 mL), which were then sealed with thermal glue and shipped at the end of each six day campaign to Italy for gas chromatographic (GC) analysis. The analytical set up of the GC used for gas analyses is described in detail in Castaldi and others [17]. Flux rates were determined via linear regression of the three sampling points (0, 30, 60 min) for each chamber and by applying a temperature and pressure correction. The minimum detectable flux over a period of 60 min was 0.07 mg CH 4 m −2 d −1 . The analytical precision of the GC for standards at ambient concentration was approximately 1.5%, using one standard deviation as a measure of mean error. Any flux below these reported detection limits was assigned a value of zero. At each sampling event, soil temperature (HI93510 thermometer, Hanna Instruments Canada Inc., Laval, Quebec) and volumetric soil water content (ThetaProbe ML2, Delta-T Device Ltd., Cambridge, UK) were hand-measured close by each chamber (0-5 cm depth), 5 cm from the chamber edge [17].
Gas samples to determine the concentration of CH 4 dissolved in water were obtained using the headspace technique [22]. 150 mL of water was sampled just below the water surface of the pond, taking 10 replicates of water each day for 3 consecutive days, when the water level measured on average 10, 8 and 6 cm above soil level, respectively. The dimension of the ephemeral pond was in the order of about 50 m 2 and water samples were taken circling the pond and extending the sampling towards the center of the pond using a support, so as to avoid pressing the soil with boots and releasing the CH 4 trapped in the soil close to the sampling point. The water was gently transferred inside a 325 mL glass bottle which was immediately capped with butyl rubber septa and closed with a screw cap for air-tight closure. The cap was provided with a circular hole made with a needle, to allow gas sampling through the septa using a syringe. The water was vigorously shaken for about one minute to allow the gas phase of the headspace and water to achieve equilibrium. Then a sample from the bottle headspace was transferred into a 20 mL air tight vial using a double needle system and by replacing the displaced air volume with water. The temperature of the water was measured at the same height of the water sample just after water sampling. To have a reference concentration of the CH 4 concentration in the headspace at T 0 , before the equilibration, a sample of air was taken just at the surface of the water and stored in a 20 mL vial as described for the other gas samples. Vials were shipped to Italy for GC analysis. CH 4 partial pressure was converted into concentration in water using Henry's constant, water temperature, the measured gas partial pressures in air, and taking into account the volume of water and headspace inside the bottle [23].
Gas saturation was calculated as the ratio between the measured gas concentration and that expected if the water sample was in equilibrium with the atmosphere. The hourly flux was determined using an ensemble of three different modelling approaches, as none of the commonly used approaches for water-gas exchange are generally tested and applied in conditions such as those presented in this paper, i.e., under a forest canopy, but they refer to open wetlands and lake ecosystems. The first two models estimated the gas flux using the gas concentration in water and the gas transfer velocity K600 [24,25]. In both models, which used wind velocity as a key variable, the latter was set at a value of <0.15 m s −1 (2 m height), based on measures done in the same area below the canopy during eddy covariance campaigns (data not shown). The third model is a steady-state diffusion model based on Fick's law, which doesn't take into account the wind factor but could still be considered appropriate for the conditions of very low wind under the forest canopy of Ankasa.

Statistic Analysis
Simple linear regression and non-linear regression analysis was performed to find the relationship between independent and dependent variables. Data were tested for normality distribution (Shapiro-Wilk test) and constant variance (Sigma Stat, Jandel Scientific). Only for eight of the 98 days of monitoring was the average CH 4 flux above zero (net CH 4 source), and these days mainly occurred in the wet season 2010 ( Figure 2). The average net CH 4 source was due to a few chambers characterized by high CH 4 emissions among chambers showing almost zero flux, as can be noted by the very high standard error for these days of positive fluxes ( Figure 1). These single chamber hotspots, however, never exceeded 8 mg CH 4  The water-filled pore space (WFPS) is a key control parameter for soil gas diffusivity, which in turn strongly controls the rate of soil CH 4 uptake. In Figure 3, CH 4 fluxes measured in the top-hill area were plotted against the values of WFPS%, with each data point representing the daily average value of gas and WFPS (n = 8). Figure 3 shows that there was a trend of reduction in CH 4 uptake rates for increasing water-filled pore space values, which followed a sigmoidal shape, with the steepest reduction between 50% and 60% WFPS. This range is generally regarded as a limit above which the rate of aerobic processes slows down, due to gas diffusion limitation. The small net CH 4 emissions observed were recorded only above 50% WFPS, but overall WFPS values rarely exceeded 60%.

Statistic Analysis
Simple linear regression and non-linear regression analysis was performed to find the relationship between independent and dependent variables. Data were tested for normality distribution (Shapiro-Wilk test) and constant variance (Sigma Stat, Jandel Scientific).

Results
Fluxes of CH4 were measured uphill for almost two consecutive years between April 2009 and November 2010, during 19 campaigns for a total of 98 days of sampling. The daily mean flux ( Figure  1) varied between −1.29 and 0.44 mg CH4 m −2 d −1 .    The water-filled pore space (WFPS) is a key control parameter for soil gas diffusivity, which in turn strongly controls the rate of soil CH4 uptake. In Figure 3, CH4 fluxes measured in the top-hill area were plotted against the values of WFPS%, with each data point representing the daily average value of gas and WFPS (n = 8). Figure 3 shows that there was a trend of reduction in CH4 uptake rates for increasing water-filled pore space values, which followed a sigmoidal shape, with the steepest reduction between 50% and 60% WFPS. This range is generally regarded as a limit above which the rate of aerobic processes slows down, due to gas diffusion limitation. The small net CH4 emissions observed were recorded only above 50% WFPS, but overall WFPS values rarely exceeded 60%.  Only for eight of the 98 days of monitoring was the average CH4 flux above zero (net CH4 source), and these days mainly occurred in the wet season 2010 (Figure 2). The average net CH4 source was due to a few chambers characterized by high CH4 emissions among chambers showing almost zero flux, as can be noted by the very high standard error for these days of positive fluxes (Figure 1). These single chamber hotspots, however, never exceeded 8 mg CH4 m  The water-filled pore space (WFPS) is a key control parameter for soil gas diffusivity, which in turn strongly controls the rate of soil CH4 uptake. In Figure 3, CH4 fluxes measured in the top-hill area were plotted against the values of WFPS%, with each data point representing the daily average value of gas and WFPS (n = 8). Figure 3 shows that there was a trend of reduction in CH4 uptake rates for increasing water-filled pore space values, which followed a sigmoidal shape, with the steepest reduction between 50% and 60% WFPS. This range is generally regarded as a limit above which the rate of aerobic processes slows down, due to gas diffusion limitation. The small net CH4 emissions observed were recorded only above 50% WFPS, but overall WFPS values rarely exceeded 60%.  Figure 4 shows that the daily average CH4 fluxes were significantly higher than those observed   Taking into account all the measured fluxes over the 38 days of sampling from eight chambers, 47% of measured fluxes were above 0 mg CH4 m −2 d −1 , i.e., net CH4 emission, and about 13% of positive fluxes were above 10 mg CH4 m −2 d −1 , reaching values as high as 1312 mg CH4 m −2 d −1 (Table 1, Figure 5). The highest emissions were reached during the period of most intense rainfall, which corresponded to the months of June and October (Figure 4). These were also the months when the highest spatial variability was observed ( Figure 5).
As also noted in the top-hill site, fluxes started to significantly increase when the WFPS was above 40% (Figure 6). However, while the top-hill site remained on average a net sink of CH4 even at higher values of WFPS, the downhill site became a very strong CH4 source (fluxes > 100 mg CH4 m −2 d −1 ) above 60% WFPS. Moreover, when the WFPS was between 20% and 40%, a significant number of chambers had a flux between 1 and 100 mg CH4 m −2 d −1 . The probe measuring the water content was 0-7 cm deep, hence, it cannot be excluded that at lower depth a higher water content was reached compared to the topsoil centimeters, due to an increase in the water table during the rainy season. Taking into account all the measured fluxes over the 38 days of sampling from eight chambers, 47% of measured fluxes were above 0 mg CH 4 m −2 d −1 , i.e., net CH 4 emission, and about 13% of positive fluxes were above 10 mg CH 4 m −2 d −1 , reaching values as high as 1312 mg CH 4 m −2 d −1 (Table 1, Figure 5). The highest emissions were reached during the period of most intense rainfall, which corresponded to the months of June and October (Figure 4). These were also the months when the highest spatial variability was observed ( Figure 5).
As also noted in the top-hill site, fluxes started to significantly increase when the WFPS was above 40% ( Figure 6). However, while the top-hill site remained on average a net sink of CH 4 even at higher values of WFPS, the downhill site became a very strong CH 4 source (fluxes > 100 mg CH 4 m −2 d −1 ) above 60% WFPS. Moreover, when the WFPS was between 20% and 40%, a significant number of chambers had a flux between 1 and 100 mg CH 4 m −2 d −1 . The probe measuring the water content was 0-7 cm deep, hence, it cannot be excluded that at lower depth a higher water content was reached compared to the topsoil centimeters, due to an increase in the water table during the rainy season. Indeed, CH4 fluxes did not correlate with the short-term cumulative rainfall, i.e., the cumulative rainfall of the two days prior the gas flux measurement, which might mainly influence the topsoil water content. On the contrary, CH4 fluxes significantly correlated with the mid and long term cumulative rainfall, i.e., cumulative rainfall of 10 and 30 days prior gas sampling, respectively (Figure 7). The two latter relationships were fitted by non-linear curves, the best fit being the one related to the cumulative rainfall of 10 days prior gas flux measurement (p < 0.0001). Overall, the steepest increase in CH4 positive fluxes per unit of rainfall increase was found for the cumulative rainfall in the 10 days preceding the flux measurements.   Indeed, CH4 fluxes did not correlate with the short-term cumulative rainfall, i.e., the cumulative rainfall of the two days prior the gas flux measurement, which might mainly influence the topsoil water content. On the contrary, CH4 fluxes significantly correlated with the mid and long term cumulative rainfall, i.e., cumulative rainfall of 10 and 30 days prior gas sampling, respectively (Figure 7). The two latter relationships were fitted by non-linear curves, the best fit being the one related to the cumulative rainfall of 10 days prior gas flux measurement (p < 0.0001). Overall, the steepest increase in CH4 positive fluxes per unit of rainfall increase was found for the cumulative rainfall in the 10 days preceding the flux measurements.  Indeed, CH 4 fluxes did not correlate with the short-term cumulative rainfall, i.e., the cumulative rainfall of the two days prior the gas flux measurement, which might mainly influence the topsoil water content. On the contrary, CH 4 fluxes significantly correlated with the mid and long term cumulative rainfall, i.e., cumulative rainfall of 10 and 30 days prior gas sampling, respectively (Figure 7). The two latter relationships were fitted by non-linear curves, the best fit being the one related to the cumulative rainfall of 10 days prior gas flux measurement (p < 0.0001). Overall, the steepest increase in CH 4 positive fluxes per unit of rainfall increase was found for the cumulative rainfall in the 10 days preceding the flux measurements.
During the periods of most intense rainfall, ephemeral ponds occurred. The lifetime of these ponds was quite variable, from a few days to a couple of weeks, depending on the rainfall regime. The estimates of the CH 4 concentration in the water and gas saturation %, determined as described in the methods section, are reported in Table 2, together with the estimates of the average daily diffusive flux from the ephemeral ponds, calculated with three different modelling approaches. Results showed that the concentration of CH 4 and its percentage saturation in water were very high at day one of sampling, increased the next day and started to slightly decrease on the third day of sampling, while the water level continuously decreased over the observation period. The estimated diffusive fluxes, considering the single replicates, ranged from a minimum of 5.34 mg CH 4 m −2 d −1 (model 1) to a maximum of 127.36 mg CH 4 m −2 d −1 (Fick's model). The average of the ensemble of the three models varied between a minimum of 9.73 mg CH 4 m −2 d −1 to a maximum of 77.13 mg CH 4 m −2 d −1 . Considering these sole diffusive fluxes, the CH 4 source strength of the analyzed pond was comparable  (Table 1) for days 2 and 3, and above the 50th percentile for day 1. During the periods of most intense rainfall, ephemeral ponds occurred. The lifetime of these ponds was quite variable, from a few days to a couple of weeks, depending on the rainfall regime. The estimates of the CH4 concentration in the water and gas saturation %, determined as described in the methods section, are reported in Table 2, together with the estimates of the average daily diffusive flux from the ephemeral ponds, calculated with three different modelling approaches. Results showed that the concentration of CH4 and its percentage saturation in water were very high at day one of sampling, increased the next day and started to slightly decrease on the third day of sampling, while the water level continuously decreased over the observation period. The estimated diffusive fluxes, considering the single replicates, ranged from a minimum of 5.34 mg CH4 m −2 d −1 (model 1) to a maximum of 127.36 mg CH4 m −2 d −1 (Fick's model). The average of the ensemble of the three models varied between a minimum of 9.73 mg CH4 m −2 d −1 to a maximum of 77.13 mg CH4 m −2 d −1 . Considering these sole diffusive fluxes, the CH4 source strength of the analyzed pond was comparable to the 90th percentile distribution of the fluxes measured with soil chambers (Table 1) for days 2 and 3, and above the 50th percentile for day 1.

Discussion
Atmospheric studies have demonstrated high atmospheric CH 4 concentrations over tropical areas [26]. The main source of this CH 4 has been attributed to permanent and ephemeral wetlands, combining top-down and bottom-up approaches [2][3][4]. Methanogenesis is the final step of a chain of anaerobic reactions, where microbial groups that utilize a variety of inorganic terminal electron acceptors (TEAs) for their respiration process [27,28] compete for organic substrates, which leads to CH 4 production. Strong reducing conditions are needed for CH 4 production until TEAs have been consumed; drops in water table levels might quickly re-oxidize reduced forms of TEAs, reducing methanogenic activity [29,30]. For this reason, sediments of permanently flooded areas are considered as the most important environment contributing to CH 4 emissions in natural ecosystems. For the same reason, CH 4 fluxes scaled up to regional level for forest areas classified as "upland forests" or "unflooded forests" are computed as net CH 4 sinks, assuming that aerobic conditions and methanotrophic activity dominates the net soil-atmosphere CH 4 exchange [3,4]. To what extent this assumption can be considered valid over the whole area of tropical forests is not clear, as a number of studies have shown small to moderate, and rarely high, temporal hotspots of CH 4 emissions from unsaturated tropical soils. A mechanism similar to N 2 O production in aerobic soils at unsaturated conditions [31] has been hypothesised to occur in anoxic microsites of CH 4 production [7][8][9][10][11][12][13][14][15]. The geomorphology of many tropical forested areas presents a very variable landscape, made by a mosaic of hills and downhill areas, where, during the period of most intense rainfall, a transient strong accumulation of water in the soil system occurs, leading to soil profile saturation and short-lived (hours to a few days) waterlogged areas. While the latter might be highly ephemeral in the dry season, disappearing within few days, during the wet season the continuous and intense rainfall might saturate a significant part of the soil profile in the downhill areas, making the water table quite superficial. This condition of saturation of the sub-surface soil layers might co-occur with significant amounts of carbon, which typically characterizes the soil profile of the tropical soil, thanks to the continuous organic inputs of the highly productive tropical forests. In the studied downhill area of the Ankasa Park, the analysis of the organic C distribution along the soil profile [17] showed a quite uniform C concentration down to 30 cm depth, varying from 2.77 g C kg −1 soil at 0-5 cm depth to 2.32 g C kg −1 soil at 15-30 cm. Between 30-50 cm depth, the C concentration of soil decreased but it was still around 1.1 g C kg −1 . The combination of good availability of electron donors, soil saturation, and high rates of soil respiration typical of tropical forests [17,32], might create a favorable environment for the complex sequence of microbial syntrophic interactions between fermentative bacteria and methanogenic archaea, which leads to significant rates of CH 4 production, as observed in wetland ecosystems [27,28]. Isotopic studies have demonstrated that the fraction of labile C transferred through methanogenesis is a better predictor of net CH 4 fluxes than the simple plant productivity rates [14,33]. Moreover, 14 C-labeling studies suggested that, in soil, carbon produced as root exudates might act as a direct substrate for CH 4 production [34,35], or indirectly influence CH 4 production by stimulating the decomposition of recalcitrant soil organic matter [36]. Topographic differences in CH 4 fluxes, linked to local differences in hydrologic and redox conditions at different landscape positions, have been reported in previous studies [11,[37][38][39]. Transitional states, where soil shifts from being a CH 4 sink to a CH 4 source, in function of rainfall and topography, could be very rapid, thus requiring a quite complex model to incorporate an explicit simulation of transient water content increase, inhibition of oxygen diffusion, and switch of CH 4 production [2,39].
Our observations showed that in the top-hill area, there was indeed a moderate number of cases when the soil acted as a net CH 4 source, and these cases occurred mostly when the WFPS was above 40% in accordance with previous observations [7][8][9][10][11][12][13][14][15]40]. Considering that the top-hill soil was well drained and at a height of about 20 m above the downhill level, it might be assumed that the observed small net CH 4 emissions might be generated in anaerobic hotspots of microbial activity, within an aerobic soil, as previously reported for other upland soils in tropical areas during the wet season. These hotspots, even if not sufficient to transform the upland areas of Ankasa into a net CH 4 source, might contribute to reducing the strength of the net CH 4 sink over the year. This, together with the strong and transient net CH 4 emissions observed in the downhill area in the wettest months, might result in the Anakasa forest soil being a net CH 4 source over a whole year cycle. CH 4 fluxes in the order of 100 mg CH 4 m −2 d −1 , and above, were observed in the downhill area when the topsoil was not saturated (WFPS just above 60%), and when WFPS was only between 20% and 40% a significant number of fluxes were still between 1 and 10 mg CH 4 m −2 d −1 . These observations, and the correlation between CH 4 positive fluxes and rainfall intensity in the 10 days preceding the gas flux measurements, might suggest a sub-surface CH 4 source. Among ephemeral flooded areas, studies in rice paddies showed that several days of flooded conditions were needed for the onset of methanogenesis to take place, i.e., when available oxygen and TEAs are reduced, usually at a redox potential more negative than about 150 to 200 mV [41,42]. The transient nature of the water table was also shown by the small ponds, which after reaching their maximum size drained quite quickly after the rain stopped, accompanied by a rapid decline in the CH 4 flux emissions and water CH 4 concentration.
Overall, on the basis of the data collected in the sampling campaigns, the top-hill area was a weak sink of about −1.5 kg CH 4 ha −1 yr −1 (daily average −0.42 mg CH 4 m −2 day −1 ). The downhill area would provide about 23.5 kg CH 4 ha −1 (on average 12.8 mg CH 4 m −2 d −1 ) during the period with the most intense rainfall, between June and November. The analysis of CH 4 fluxes measured during the campaign in May indicated a daily average of −0.42 mg CH 4 m −2 day −1 , hence in line with fluxes in the drained upland site. Assuming, therefore, that during the drier season the downhill area would have a net CH 4 exchange similar to the upland, the net annual budget of the downhill area would be around 22.7 kg CH 4 ha −1 yr −1 . Over the year, the Ankasa forest would be a mosaic of CH 4 sources and sinks. We hence estimated the overall soil CH 4 flux in the whole Park surface, considering the surface covered by the bottom hill areas and the observed temporal dynamics of fluxes during the year. The flux obtained was 3.3 kg CH 4 ha −1 yr −1 , i.e., the forest acted as a net source of CH 4 . This source might even be underestimated. In fact, in the ephemeral ponds of downhill areas, bubbling might also contribute to CH 4 emissions into the atmosphere, but our diffusive approach could not capture this source. Another source of CH 4 might be represented by diffusion through plant tissues. Scientists have discovered the significant role of tropical trees as preferential escape routes for CH 4 trapped in waterlogged soils of the Amazon basin [43]. They found that many tree species were provided with lenticels for gas exchanges, similar to those present in rice plants. We have cross-checked the list of species reported by Pangala and others [43] with the species reported in the study area, as part of a parallel study on biodiversity (data not shown) but no more than a couple of genera have been found in common. This does not exclude the fact that some of the African tree species, having roots frequently exposed to soil saturation, might present specific gas exchange routes, but the issue will need further investigation. If this "tree route" of CH 4 degassing would also occur in our study site, in particular in the downhill areas, CH 4 emissions from the forest might be even higher than calculated on the sole basis of soil fluxes.

Conclusions
Studies of CH 4 fluxes in tropical forests combining observations in plateaus and valleys areas, using different techniques, have estimated that the forest was a net CH 4 source. The reported average daily source strength varied from 2 to 20 mg CH 4 m −2 d −1 depending on the study site and time of the year, with the wetter sites showing the highest fluxes [44][45][46]. The general consensus was that tropical "unflooded" forests might be an underestimated CH 4 source. We confirmed in this study that considering both the hotspots of CH 4 emissions in the upland areas which significantly reduce the annual upland soil CH 4 sink, and the strong emissions measured during wettest periods in downhill areas, the Ankasa forest could be considered a net CH 4 source. Such a CH 4 source is much less strong than what was measured or modeled for wetlands ecosystems and flooded forested areas [2], but considering the vast area covered by tropical forests receiving a significant amount of rainfall during the wet season and throughout the year in equatorial regions, the underestimated CH 4 source of such ecosystems might deserve a more in-depth analysis in the global CH 4 budget.
Author Contributions: S.C. and R.V. conceived the research and designed the study, T.B. and G.N. contributed to collecting and analyzing data, C.S. wrote the paper with feedback from all authors. All authors have read and agreed to the published version of the manuscript.