Greenhouse gas emissions along a peat swamp forest degradation gradient in the Peruvian Amazon: soil moisture and palm roots effects

Tropical peatlands are a globally important carbon stock of 87–350 GtC (Gumbricht et al. 2017; Page et al. 2011) and are crucial for mitigating climate change (Murdiyarso et al. 2013). In addition, tropical peatlands are among the most efficient terrestrial ecosystems for carbon sequestration (Dommain et al. 2011; Page et al. 2004), as continuous input of organic material from lowland tropical evergreen vegetation combined with anaerobic soil conditions lead to a build-up of soil organic matter (SOM) over time (Jauhiainen et al. 2012). In natural conditions, the portion of SOM that is decomposed and emitted as carbon dioxide (CO₂) or methane (CH₄) is usually outweighed by the continuous input of fresh litter and roots (Jauhiainen et al. 2005; Hergoualc’h and Verchot 2011; Hoyos-Santillan et al. 2015). Models such as the Holocene Peat Model—HPM (Frolking et al. 2010; Kurnianto et al. 2015) use this balance to predict long-term peat accumulation via vegetation-specific characteristics such as decomposition speed and primary production.

Peru is estimated to harbor one of the largest extents of tropical peatlands in the world (Gumbricht et al. 2017; Draper et al. 2014; Page et al. 2011). Mauritia flexuosa-dominated palm swamp forests are the dominant peatland ecosystem type in the Peruvian Amazon (Draper et al. 2014). Interest in the carbon pools of the Peruvian Amazon peatlands has increased in recent years. Since Lähteenoja et al. (2009a) explored their extent, research further expanded into other fields such as palaeoecology (Roucoux et al. 2013), C stocks estimates (Draper et al. 2014), M. flexuosa management (Virapongse et al. 2017), the characterization of degradation (Hergoualc’h et al. 2017), and greenhouse gas (GHG) fluxes (Teh et al. 2017). The challenges that Peruvian peatlands face are substantially different from those in Southeast Asia. Therein Southeast Asia, peatlands are under great pressure from agricultural expansion, artificial drainage, and fires, which result in considerable GHG emissions (Gaveau et al. 2014; Hergoualc’h and Verchot 2014). In contrast, anthropogenic degradation of peatland in the Peruvian Amazon is mostly related to recurrent harvesting of M. flexuosa palms from natural stands without drainage or fire. The fruits from M. flexuosa palms (locally referred to as “Aguaje”) and palm weevils (Rhynchophorus palmarum) that grow inside dead palms are highly demanded products in the regional market and are important sources of vitamins and proteins for rural communities (Pacheco Santos 2005). Even though more sustainable (climbing) techniques exist (Horn et al. 2012), fruit harvesting continues to involve cutting down entire palms. The effect of this practice substantially alters the composition and structure of the forest (Hergoualc’h et al. 2017); however, its impact on long-term peat accumulation remains unstudied.

Information about driving factors of GHG fluxes is crucial to accurately model and predict long-term changes in the soil carbon pool and exchanges of GHG with the atmosphere (Ryan and Law 2005). Hereafter and throughout the text, soil CO₂, CH₄, and nitrous oxide (N₂O) fluxes are collectively referred to as GHG fluxes. Factors that drive GHG flux dynamics include, among others, soil substrate quality, nutrient availability, and aeration. Soil water-filled pore space (WFPS) is commonly used as an indicator for soil aeration. For a wide range of soils, the WFPS has been reported to affect the emissions of CO₂ (e.g., Husen et al. 2014; Howard and Howard 1993), CH₄ (e.g., Verchot et al. 2000; Del Grosso et al. 2000), and N₂O (e.g., van Lent et al. 2015; Davidson et al. 2000). Maximum respiration rates typically occur around 60% WFPS (Linn and Doran 1984); below this level, the microbial activity is limited by water, and above it, oxygen deficiency hampers microbial respiration (Moyano et al. 2013). N₂O emissions predominantly arise from nitrification and denitrification (Davidson et al. 2000), and maximum N₂O emission reported by Van Lent et al. (2015) was around 60% WFPS for a wide range of tropical mineral soils. Methanogenesis is the anaerobic microbial decomposition of organic material, which occurs in waterlogged soils or in anaerobic microaggregates, whereas methanotrophy takes place in parts of the soil where oxygen is available. Both processes concur in soils, and—among other factors—are modulated by soil structure and WFPS% (Smith et al. 2003; Teh et al. 2005). In waterlogged soils at 100% WFPS, methanogenesis is likely the dominant process, whereas at WFPS < 100% CH₄ is solely produced in anoxic microsites and part of the CH₄ is oxidized.

GHG production is also influenced by the quality and quantity of fresh roots and litter that enter the soil (Updegraff et al. 1995). For instance, in the DAYCENT model, the carbon-to nitrogen-ratio (C/N ratio) of different plant parts is an input for SOM turnover rates, and consecutively influences C mineralization, which are in turn controlled by factors such as soil moisture, temperature, and texture (Parton et al. 1993, Del Grosso et al. 2001). In addition, Oktarita et al. (2017) found increased soil N₂O when soil C/N ratios decreased. We hypothesize that palm harvesting leads to changes in vegetation composition, and in turn to alteration of litter inputs by secondary regrowth and the C/N ratio of the soil substrate. We expect this change in substrate to result in increased decomposition in areas where palms are harvested as compared to undisturbed conditions, which would be reflected in higher GHG emissions in those disturbed sites.

In saturated conditions, aerating roots potentially play an important role for soil GHG fluxes. Plants need to adapt themselves to cope with anoxic conditions (Bruhn et al. 2012). Such adaptations are generally aimed at increasing the influx of oxygen to the root zone through adventitious roots, lenticels, and enlarged aerenchymous tissues (Haase and Rätsch 2010). This acquired oxygen is used for a variety of essential plant functions, such as root respiration (Colmer and Voesenek 2009). A thickened exodermis around the aerenchyma acts like a barrier and only allows for oxygen to leak near and around the root tip (van Noordwijk et al. 1998). Granville (1969) described the presence of such tissues in the aerating pneumatophores of M. flexuosa palms. These pneumatophores may also conduct N₂O or CH₄ produced in the anoxic soil layers to the atmosphere, as has been shown for aerenchymous tissues in the stalks of rice (see, e.g., Minoda and Kimura 1994) as well as for tree stem lenticels in temperate and tropical forested wetlands (Gauci et al. 2010; Pangala et al. 2013). In this way, CH₄ is directly emitted from the soil, and avoids potential oxidation in upper aerobic soil layers. Similarly, dissolved N₂O and CH₄ can be taken up by roots and subsequently emitted during root or leaf respiration (Pihlatie et al. 2005; Gauci et al. 2010). The existence and the contribution of pneumatophores to total soil GHG emissions have not been reported for palm swamp forests in the Amazon.

Current practices of palm harvesting in the Peruvian Amazon may lead to changes in soil moisture, alteration of the quality of substrate inputs, and decreases in aerating pneumatophores. However, to date, there is no assessment on how these changes could affect GHG emissions. In this context, we set as a first objective for this study to test the response of GHG production to varying levels of WFPS, across a gradient of forest degradation involving substrate alteration. Secondly, we aimed to evaluate the contribution of aerating pneumatophores to surface GHG fluxes in undisturbed and degraded palm swamp sites.

We conducted two experiments. The first one was to determine the response of soil microbial GHG emissions to variations in soil moisture and was performed in vitro with samples without live roots. In the second experiment, we estimated the potential of root pneumatophores to conduct GHGs by relating pneumatophore densities to corresponding in situ soil GHG fluxes. The effect of degradation was assessed by including three sites differing in level of degradation. We acknowledge (and further discuss) the limits of our experimental design, which include no field-level replication for each degradation level.

The topsoil at all three sites displayed minerotrophic properties, in agreement with findings by Lähteenoja et al. (2009a) at the Quistococha reserve. The sites exhibited similar levels of cations and a similar minerotrophic state. Accordingly, the lower soil C content and C/N ratio at the hD site compared to those at the other sites may indicate a higher degree of humification (Tfaily et al. 2014). The lower P content at the hD site than at the other sites may also suggest a more advanced decomposition status of the peat (Jordan et al. 2007; Könönen et al. 2015). Recurrent harvesting of M. flexuosa palms leads to shifts in forest structure and composition (Hergoualc’h et al. 2017). Bhomia et al. (unpublished) found that the I site harbored a higher density of dicot trees (~ 1500 ha⁻¹) than the degraded sites (~ 700 ha⁻¹). Furthermore, the density of palms overall and of M. flexuosa palms in particular was much lower at the hD site (53 and 16 ha⁻¹, respectively) than at the I and mD sites (~ 300 and 170 ha⁻¹, respectively). The presence of a high number of individuals of Cecropia spp.—a pioneer species observed at the hD site—is indicative of its degraded status. In addition, the gaps formed upon harvesting are colonized by small (< 2 m) fast-growing herbaceous vegetation (personal observation). Such drastic differences in forest structure can be expected to lead to lower above and belowground litter inputs at the degraded sites as compared to inputs at the I site. Differences in forest composition may also induce significant changes in litter quality, as indicated by the difference in litterfall C/N ratio between sites (Table 1). Decreased C/N ratio of litterfall with increased degradation is in agreement with observations by Feldpausch et al. (2004) along a sequence of secondary forest regeneration in Central Amazonia. During early forest succession the biomass accumulates rapidly and consists mainly of easily decomposable litter and less of wood (Gehring et al. 2005; Guariguata and Ostertag 2001). More readily decomposable litter may on the long-term lead to reduced peat accumulation at the degraded sites (Frolking et al. 2001).

The changes in litter quality and the less-decomposed SOM pool as outlined above could explain the differences in C mineralization we found between sites. This is in line with Nilsson and Bohlin (1993), who found that more decomposed peat soils produced more CO₂ than less decomposed, more fibrous peat soils in temperate climate. In contrast, incubation studies by Jauhiainen et al. (2016) and Swails et al. (2017) showed that soils from degraded peatlands with secondary regrowth or cultivated with oil palm produced less CO₂ than soils from intact peat swamp forests. However, the degraded peatlands of Indonesia were drained and burnt, while sites in our study were not. The effect of such drainage and fire practices generally accelerates the peat mineralization (Hergoualc’h and Verchot 2011, 2014). An intact hydrology at our study area might prohibit the easily-decomposable litter in the degraded sites to decompose. During relative dry periods (60–70% WFPS), the in situ CO₂ fluxes in degraded peat swamp forests could show a stronger response to lower WFPS as compared to undisturbed peat swamp forests. Periods with less precipitation are therefore expected to result in increased CO₂ fluxes in situ.

The response of CO₂ production to changes in WFPS followed a cubic polynomial relationship, which is in agreement with results reported by Howard and Howard (1993) for a wide range of soils and by Husen et al. (2014) for peat soils under oil palm cultivation in Indonesia. The maximum CO₂ production at 60–70% WFPS for all three sites is in accordance with common values around 60% WFPS found in the literature (Davidson et al. 1998; Linn and Doran 1984), but slightly higher than the 50% value reported by Husen et al. (2014).

In the incubation study, CH₄ was solely produced in soil from the heavily degraded site at anoxic conditions (WFPS > 100%), as has been reported by Smith et al. (2003) and similar to 88% WFPS reported by Melling et al. (2005) for tropical peat swamp forest. The soil samples from the I and mD sites were likely not lacking in C substrate for the methanogens (Table 1), and the water addition treatment mimicking flooded conditions should have ensured anaerobic soil samples. Possibly, CH₄ production was inhibited due to high nitrate and other denitrification products, which inhibits methanogenesis in rice paddy soils (Kluber and Conrad 1998a, b). The absence of N₂O production in the hD site could indicate a lack of nitrate or other denitrification product in these soils. In addition, the soils from the I and mD produced large amounts of N₂O at WFPS% > 70%, which could indicate the inhibition of methanogenesis due to the presence of nitrate or other denitrification products. Most likely, the time for recovery to conditions favoring methanogenesis in the soil samples was inadequate. We did see an effect of increased CH₄ production toward the end of the incubation experiment (Appendix), which confirms this lag in recovery.

N₂O was predominantly produced at WFPS greater than 70% for soils from the I and mD sites. Mathieu et al. (2006) used ¹⁵N stable isotope tracers to demonstrate that in water-saturated soils denitrification accounted for 85–90% of emitted N₂O. Therefore, we expect that N₂O in our peat soils predominantly originated from denitrification instead of nitrification. In addition, we observed a reduced production with increased incubation time that could indicate a depletion of substrate for denitrification, or a further reduction to dinitrogen (N₂) (Appendix). In contrast to the soils from the I and mD sites, the soils from the heavily degraded site did not produce any significant amount of N₂O. The large CH₄ flux from the same hD soils indicates highly reducing conditions in the soil sample. Possibly denitrification in these soils predominantly resulted in N₂, which is the dominant product of denitrification in wet, anaerobic conditions (Davidson et al. 2000), while at the same it is non-toxic for methanogenesis (Roy and Conrad 1999). The maximum N₂O production rate occurred at 100 and 84% WFPS for soils from the I and mD sites (Fig. 2c), respectively, much above the 60% value presented by Van Lent et al. (2015) for a wide range of tropical mineral soils. This indicates that nitrification is lacking in these peat soils and N₂O predominantly originates from denitrification instead.

The absolute flux rates from the incubated soils are of limited use; however, as presented above, we were mainly interested in the relative flux rates that are useful for interpretation and modeling of field-based fluxes. Nevertheless, the average peat CO₂ and CH₄ flux rates measured in vitro were in the same order of magnitude as in situ rates from the literature. A hypothetical 30-cm peat profile with the overall average bulk density of 0.1 g cm⁻³ and the average maximum in vitro production rate of 2.7 μg C-CO₂ g⁻¹ d.w. h⁻¹ (Fig. 2a, 60–70% WFPS) would result in CO₂ surface fluxes of 81 mg C-CO₂ m⁻² h⁻¹. This compares well to the average value of 79 mg C-CO₂ m⁻² h⁻¹ reported by a review of soil surface CO₂ from tropical peat swamp forests in Southeast Asia (Hergoualc’h and Verchot 2011). The 30-cm depth is expected to represent the soil surface flux since deep peat layers contribute little to total GHG efflux (Moore and Dalva 1997; Jauhiainen et al. 2016). The average maximum in vitro CH₄ production rate of the water-saturated soils (WFPS > 95%) from the hD site was 5.9 ng C-CH₄ g⁻¹ d.w. h⁻¹ or 0.18 mg C-CH₄ m⁻² h⁻¹ which is similar to the rate of 0.33 mg C-CH₄ m⁻² h⁻¹ reported by Hergoualc’h and Verchot (2014), but an order of magnitude lower than the average of 1.50 mg C-CH₄ m⁻² h⁻¹ reported from similar M. flexuosa palm swamp forests (Teh et al. 2017). In contrast to CO₂ and CH₄, N₂O fluxes measured in vitro were much higher than rates measured in situ. The average maximum N₂O production rate from in vitro incubated soils from the I and mD sites were 105 and 10 ng N-N₂O g⁻¹ d.w. h⁻¹ (respectively, 84 and 100% WFPS), or 3.2 and 0.3 mg N-N₂O m⁻² h⁻¹, respectively. For the soils from the intact site, this was an order of magnitude above the average of 0.74 mg N-N₂O m⁻² h⁻¹ reported by Van Lent et al. (2015) for tropical peat soils. Differences between fluxes measured in situ and in vitro may originate from soil disturbance during sampling or incubation preparation, differences in active peat layers producing or consuming N₂O and environmental conditions, or lack of a plant sink for mineralized N (Butterbach-Bahl et al. 2013). Lastly, an artifact could have been created due to an up-build of GHGs in the soil and water suspension between sampling days, which then diffused into the headspace after venting at the onset of the measurements. However, we took special care to minimize this effect by leaving the flasks open between sampling days.

Cumulative in situ CH₄ emissions were correlated with pneumatophore density (Fig. 3), indicating a potential role for aerating roots to conduct CH₄ to the atmosphere in these peat swamp forests. Recently, vegetation-based CH₄ fluxes have been synthesized to represent 5–22% of the total global CH₄ budget, of which 58–78% is conducted by a variety of plant structures and 22–42% is thought to be produced by plants themselves (Carmichael et al. 2014; Schlesinger and Bernhardt 2013). Pangala et al. (2013) roughly estimated that CH₄ emissions from woody tree stems alone represented 62–81% of the total ecosystem flux in an Indonesian peat swamp forest, while the contribution of pneumatophores was thought to be negligible (< 2%). The specific contribution of roots to conduct CH₄ has been further studied in Indian (Purvaja et al. 2004) and Australian (Kreuzwieser et al. 2003) mangroves; however, to our knowledge, this study is the first to do so for a tropical palm-dominated swamp forest. Our setup was unable to quantify the contribution of these roots to the total ecosystem flux, but it highlights the need for additional research in order to elucidate the role of pneumatophores for CH₄ emissions in these ecosystems, and to investigate to what extent this observation holds with different vegetation compositions and soil types throughout the Amazon. Long-term GHG monitoring studies should therefore include aerating roots as potential sources in their experimental design.

The body of knowledge on tropical M. flexuosa-dominated peatlands in the Peruvian Amazon steadily increased in recent years, even though research in this remote region has its practical limitations. This study should be seen as a first exploration of soil moisture and roots as controlling factors for GHG fluxes along a degradation gradient in this region. Despite no site replication, we were able to find enough variation of the underlying drivers of the processes of interest to detect significant relationship. To be able to make inferences about the effects of degradation on biogeochemical cycles and GHG emissions, we would need to undertake a more extensive and replicated experiment. However, this study does show the importance of various site-specific factors that drive the spatial and temporal variations of GHG emissions, and that recurrent harvesting of M. flexuosa could alter the GHG balance on the long-term. Sustainable management of the M. flexuosa-dominated peat swamp forests is needed and should aim at providing a continuous source of income for many families, while conserving the peat and its role for climate change mitigation at the same time.