In an effort to broadly understand the response of peatlands to past, present, and future environmental change, a predictive modeling approach that links fen vs. bog traits with C exchange functions is needed. Although C dynamics are spatially variable within a single peatland due to small-scale heterogeneity associated with hummock-hollow microtopographic gradients (Waddington et al., 1996; Baird et al., 2009), there is merit in distinguishing between fens and bogs for the general purpose of coupling dynamic peatland models with Earth System Models (Frolking et al., 2009). Likewise, landscape-scale mapping of peatland types would lead to better estimates of peatland-atmosphere C exchanges (Baird et al., 2009). Also, knowing the timing of critical ecosystem shifts such as the FBT, both in the past and the future, could help model changes in peatland C dynamics and their role in the global C cycle. In particular, knowing how the extent of fens and bogs has changed throughout the Holocene would provide constraints on peatland-atmosphere greenhouse gas exchange (e.g., Frolking and Roulet, 2007; Yu et al., 2013). Indeed, it has been suggested that the progressive decrease in atmospheric CH₄ concentration that occurred from 8000 to 4000 years ago might relate to the FBT across the boreal biome (MacDonald et al., 2006; Loisel et al., 2015). Likewise, Yu (2011) suggested that peat-C accumulation could have contributed to the Holocene changes in atmospheric carbon dioxide (CO₂) concentration and δ¹³CO₂ observed along Antarctic ice cores. It is also possible that, in the near future, large swaths of wet arctic tundra switch to bogs, following landscape-scale permafrost thaw and drainage (Yu et al., 2009; Hugelius et al., in press). Despite its key role in the Holocene global C cycle, little is known about the timing and mechanisms of the FBT across the mid- and high-latitude biomes.

In this study, a data synthesis of FBT timing across 90 southern Patagonian peatlands is presented. We show a regional-scale switch from fens to bogs across the region around 4200 years ago, pointing toward a regional-scale control on the FBT. We present two non-exclusive hypotheses that could explain this regional shift in peatland type. A conceptual model of the FBT is then discussed. To our knowledge, no models provide a conceptual understanding of the FBT [but see Loisel et al. (2012), where a preliminary version of this model was presented]. In this model, allogenic factors push the fen beyond its resilience capacity, inducing a regime shift to ombrotrophy. Overall, this information may be of use to improve ecological models via the identification and quantification of ecosystem functional traits between bogs and fens; it also has applications for peatland management efforts such as rehabilitation, restoration, and conservation.

The study region is southernmost South America (SSA), which in this case refers to the landmass located between ∼ 50 and 56°S. This area includes Chile’s Magallanes region and Argentina’s portion of Tierra del Fuego (Figure 1). The regional climate is heavily influenced by the southern westerly winds, which carry moisture and heat from the Pacific Ocean onto the continent, in a west-to-east direction (Garreaud, 2009). By forcing the maritime air upward, the Andes Mountains create a strong precipitation gradient across the region (Paruelo et al., 1998). Three main peatland types are found across this gradient. First, cushion bogs are established under hyper-oceanic conditions. These ecosystems are dominated by Astelia pumila and Donatia fascicularis, two species that are tolerant of windy, stormy, and cold conditions (Auer, 1958; Moore, 1983; Pisano, 1983). Second, ombrotrophic peat bogs are located under humid to dry conditions (∼ 400–1000 mm/yr); these ecosystems are dominated by Sphagnum magellanicum carpets and scarce stunted trees (Pisano, 1983; Grootjans et al., 2010). Lastly, groundwater-fed herbaceous fens are found under dry to arid conditions (<400 mm/yr) and are usually dominated by grasses, sedges (Carex spp.), rushes (Marsippospermum grandiflorum), and a few shrubs. Peat bogs today thus occur within a narrow band of land located along the lee side of the Andes (Figure 1), where mean annual temperature is around 5°C (Loisel and Yu, 2013a), the temperature seasonality (difference between the warmest and coldest months) is around 10°C (Loisel and Yu, 2013a), and frost is rare (Schneider et al., 2003; Arroyo et al., 2005). These peat bogs are embedded within the deciduous and evergreen forests of southern Patagonia; the bogs are rain-fed and are typically found in valley bottoms atop lacustrine sediments or marine clays and silts (Bentley and McCulloch, 2005; Rabassa et al., 2006).

This study includes a total of 90 peatlands (Supplementary Figure S1), of which 12 were sampled by our team over the course of two field seasons (Supplementary Table S1). In 2010 (PAT-10), four peat bogs were visited and a core was retrieved at each one of the following sites (unofficial names): Harberton (HB), Cerro Negro (CN), Upper Andorra Valley (UAV), and Escondido (ESC). Some results from these cores have been published elsewhere (Loisel and Yu, 2013a, b; Loisel, 2015). In 2018 (PAT-18), eight new sites were cored: Beef-Penguin (BP), Mercedes (MP), Jen (JB), Rasmussen (RAS), Ariel (AP), Cura (CP), Pat-Andy-Nat (PAN), and Flarks (FP). Some of the results from cores MP and BP are published (Bunsen and Loisel, 2020), while those from the other cores remain unpublished at the moment (but see Bunsen, 2020). Note that Ariel Peatland has also been studied by Xia et al. (2018, 2020); likewise, Harberton Bog was the object of previous studies (e.g., Markgraf, 1991; Pendall et al., 2001).

Eleven of the twelve sites are S. magellanicum peat bogs that are characterized by microtopographic gradients, from dry hummocks to wet hollows. Shrubs Empetrum rubrum and Nothofagus antarctica were identified at all sites, with the exception of Jen Bog and Mercedes Peatland, where N. antarctica was replaced by cypress tree Pilgerodendron uviferum, a conifer only found in hyper-humid habitats of Patagonia (Moore, 1983). The sedge and grass communities typically included Marsippospermum grandiflorum, Tetroncium magellanicum, Alopecurus magellanicus, Carex magellanica, and Carex curta. The 12th site (Cura Peatland) is a spring-fed fen dominated by herbaceous plants and Bryales. Cores were extracted in 50-cm increments using a Russian auger. The stratigraphy of each section was described in the field in terms of main peat types (moss, herbaceous, ligneous, etc.) and soil properties (texture, particle size, color). The cores were then wrapped in plastic and foil, and stored in PVC pipe for shipping. Cores were kept at 4°C until lab analyses. Inception age varies between 15,550 and 1450 calibrated years Before Present (cal. yr BP). All 11 peat bog cores exhibit a clear, sharp switch from herbaceous-dominated fen peat to Sphagnum-dominated bog peat (Figure 2). The FBT has thus occurred at some point during the Holocene at these sites.

Peat core chronology was constrained using radiocarbon (¹⁴C) dating (Supplementary Table S2). Hand-picked plant macrofossils were cleaned with deionized water, dried, and ¹⁴C-dated using AMS at the University of California – Irvine’s Keck AMS Carbon Cycle Lab (samples collected in 2010) and at Lawrence Livermore National Lab’s CAMS Radiocarbon Lab (samples collected in 2018). When the samples were too decomposed for the identification and cleaning of plant macrofossils, root-free bulk peat samples (63–125 μm) were used instead. As the regional tephrochronology is relatively well known (e.g., Stern, 2008), a few volcanic tephras were identified along our peat profiles on the basis of visual examination (in the field) as well as changes in bulk density (in the lab). The tephra layers that were identified are: Aguilera I, Burney I and II, Hudson I, and Reclus I, as well as cryptotephras Hudson a and b (Mansilla et al., 2016; Supplementary Table S3). While they were not used to build the chronologies, the tephras were used to confirm our age-depth models and compare core stratigraphy between sites.

Age-depth models for the four PAT-10 cores (HB, CN, UAV, and ESC) have been published in Loisel and Yu (2013a) but have been updated for this study, in light of more recent calibration curves being available. Age-depth models for the eight PAT-18 cores (BP, MP, JB, AP, CP, FP PAN, and RAS) are available in Bunsen (2020). All raw and calibrated ¹⁴C data, as well as our final age-depth models, are included in the Supplements (Supplementary Table S2 and Supplementary Figure S2), along with tephra ages obtained from the literature (Supplementary Table S3) and bulk density profiles from all 12 cores to show the stratigraphic location of those tephras (Supplementary Figure S3). We used the SHCal13 (Hogg et al., 2013; Reimer et al., 2013) and SHZ 1-2 calibration curves (Hua and Barbetti, 2004), along with Bacon version 2.3.5 (Blaauw and Christen, 2011) to constrain peat chronologies.

Geochemical measurements were performed at high resolution along all 12 cores (Supplementary Figure S3). Along the PAT-10 cores, contiguous sub-samples (1 cm³) were analyzed every cm; the sample interval was increased to 2-cm increments along the PAT-18 cores, though sub-samples (2 cm³) were still measured contiguously. Loss-on-ignition (LOI) was performed to determine peat water content, organic matter content, dry bulk density, and organic matter density on each sub-sample (total = 3854 samples); standard procedures were followed (Dean, 1974). Changes in peat water content and organic matter density were used to infer stratigraphic changes in peat type along the cores for which plant macrofossil analysis has not yet been performed. This inference is based on a statistical analysis that was performed on 438 Patagonian peat samples for which plant fossil analysis as well as geochemical measurements were performed (see Loisel and Yu, 2013a). Results indicate that low water content and high density are associated with herbaceous peat, and high water content and low density are correlated with Sphagnum peat (Loisel and Yu, 2013a). Lastly, peat-carbon accumulation rates (PCAR) were obtained by multiplying the organic matter density of each depth increment (in g OM/cm³) by the interpolated deposition rate of each sample (from the age-depth models, in cm/yr) and by an assumed 50% C content (Loisel and Yu, 2013a).

A compilation of 78 peat core stratigraphies from southernmost Patagonia was used to infer the timing of the FBT across the region. The quality of data varies across the different sources, from visual descriptions of peat stratigraphic changes (Auer, 1965) to semi-quantitative plant macrofossil-based reconstructions (De Vleeschouwer et al., 2014). Radiocarbon-based age-depth models were unavailable for Auer’s (1965) cores; instead, peat stratigraphic changes are described in relation with three tephra layers of known ages (Stern, 2008): Reclus I (15,780–14,040 cal. yr BP; median age = 14,900 cal. yr BP), Hudson I (7960–7420 cal. yr BP; median age = 7700 cal. yr BP), and Burney II (5290–3230 cal. yr BP; median age = 4200 cal. yr BP). We used these three tephra layers to compare changes in peat type, total peat mass (gOM/cm²), and PCAR (gC/m²/yr) between all 78 cores from the literature as well as our 12 new cores. Lastly, peat depths were converted to organic matter density, peat mass, and PCAR using the following values: 0.05 g OM cm³ for Sphagnum peat and 0.07 g OM cm³ for herbaceous (i.e., Carex spp., Marsippospermum sp.) and other non-Sphagnum (i.e., Bryales) peat types (Loisel and Yu, 2013a).

Altogether, 58 out of the 90 surveyed peatlands are bogs today (Figure 3 and Supplementary Figure S4). Out of these 58 Sphagnum-dominated bogs, 35 transitioned from their fen state at 4200 cal. yr BP (Burney II eruption), six switched to bogs at 7700 cal. yr BP (Hudson I eruption), and 13 underwent the FBT at other times (Figure 4). As for the sites that did not respond to those eruptions and switched to bogs at later stages (e.g., JB, UAV, ONA, and RR), it is possible that the eruption (or its tephra) was not sufficient to destabilize the fen conditions. As for the sites that switched prior to those eruptions (e.g., ESC, BP, and HAB), we do not have sufficient evidence to suggest alternative triggering mechanisms at this time, though drying conditions during the mid-Holocene might have facilitated the colonization by Sphagnum (e.g., Hughes, 2000; Hughes and Barber, 2004). Of the remaining 32 sites, 27 have persisted as fens for their entire history while the other five are Donatia-dominated cushion bogs today (Supplementary Figure S4). The temporal pattern offered by the FBT is striking, with a majority of sites simultaneously switching to bogs across SSA (Figure 4).

Some peatland-rich areas of southernmost Patagonia are found outside the Sphagnum moss domain. As such, the five cushion bogs cataloged in this study (T40, T41, T42, T60, and SKY-2) are located off the western Chilean coast and along the Strait of Magellan, in areas of either very high precipitation, high wind, or storminess (Supplementary Figure S4; see Supplementary Figure S1 for detailed location of these sites). Though tephra layers were identified along these core profiles, Sphagnum moss is not typically encountered in these two regions today, due to climatic limitations (Pisano, 1983). That said, the co-occurrence of Donatia and Sphagnum has been observed in a few peatlands located near the current limit of Sphagnum bogs (near sites PAT-18-MP, between sites PAT-10-HB and T60, and at site SKY). In this area, to our knowledge, at least one peat core harbors switches between cushion- and Sphagnum-dominated assemblages (Mathijssen et al., 2019). But for the sake of this study, which focuses on the FBT, we will not investigate this topic further.

Another outstanding spatial pattern from our analysis is the location of 27 “persistent fens.” These peatlands have not undergone the FBT despite having been influenced by volcanic eruptions (Figure 3 and Supplementary Figure S4). Of these persistent fens, 13 sites are clearly found in the steppe ecoregion (T36, T37, T51, T64, T65, T66, T67, T69, P57, P58, P60, P61, and P65), making it unlikely for Sphagnum to colonize these areas due to local hydroclimatic limitations (Supplementary Figure S4; see Supplementary Figure S1 for detailed locations). That said, several sites that are either located near the modern-day boundary between the steppe and forest ecoregions (T7, T19, T30, T31, T32, T63, and P63) or within the Sphagnum bog domain (PAT-18-CP, T23, T26, T58, T61, P53, and P56) also persisted as fens. In those cases, a number of local conditions that would prohibit oligotrophy, including insufficient peat thickness, the presence of mineral-rich springs or surface runoff from surrounding hillslopes, and other conditions such as inadequate parent material or topography, could have prevented the switch to bog conditions. In particular, many of these peatlands are found in complex moraine basins that are probably strongly influenced by groundwater (Auer, 1965; Coronato et al., 2006; Rabassa et al., 2006).

A comparative analysis of Sphagnum vs. non-Sphagnum peat mass revealed that Sphagnum-dominated peatlands are characterized by greater peat stores, and thus larger soil C stocks, than their non-Sphagnum counterparts (Tukey’s HSD: p < 0.001, except for the period prior to 14,900 cal. yr BP, where the difference between the two peat types was not statistically significant). This finding is consistent through the different time periods (Figure 5). We observed relatively consistent peat mass values for both Sphagnum (median ∼ 12 g/cm²) and non-Sphagnum (median ∼4 g/cm²) peat types through time, though these figures should not be directly compared, as each time period has a different duration. To alleviate this issue, PCAR was calculated to standardize changes in peat mass over time.

There are two outstanding patterns that come out of the PCAR data (Figure 6). First, Sphagnum peat is consistently accumulating at greater rates than its non-Sphagnum counterpart (Tukey’s HSD: p < 0.001, except for the period between 14,900 and 7700 cal. yr BP, where the difference between the two peat types was not statistically significant), with median apparent rates of C accumulation that progressively increase through time, from 10 gC/m²/yr between Reclus I and Hudson I, to 16 gC/m²/yr between Hudson I and Burney II, and to 20 gC/m²/yr from 4200 cal. yr BP to modern times (Figure 6). Second, non-Sphagnum PCARs remain uniform throughout all three time periods, with a median of 7 gC/m²/yr (Figure 6). To normalize PCAR values for each peatland developmental history, time-weighted PCARs were calculated for the Sphagnum and non-Sphagnum peatland sites (Figure 7). This calculation multiplies the PCAR from each time period by its duration to ensure that each period is adequately represented in the “whole core PCAR.” We find, once again, that Sphagnum sites have been characterized by an overall greater C-sink capacity than non-Sphagnum sites (Tukey’s HSD: p < 0.00001), with median time-weighted PCAR values of 17 gC/m²/yr (min = 4, max = 49) for Sphagnum and 8 gC/m²/yr (min = 2, max = 33) for non-Sphagnum. Note that, amongst the non-Sphagnum sites, time-weighted PCAR ranged from 7 to 10 gC/m²/yr for cushion bogs and from 2 to 33 gC/m²/yr for persistent fens.

A comparison of the Sphagnum vs. non-Sphagnum portions of the cores reveals that, since the Reclus I eruption, approximately 42% of the total peat mass accumulated at these 90 sites is made of Sphagnum peat (Figure 3). This figure considers organic matter density differences between Sphagnum (0.05 gOM/cm³) and non-Sphagnum peat (0.07 gOM/cm³). For comparison, Sphagnum peat has dominated the regional peatland landscapes during approximately 30% of the entire peat history, as reconstructed from the cores. In other words, though Sphagnum peat is less dense than its non-Sphagnum counterpart, it has built a peat deposit that is greater (42%) than the total amount of time (30%) these peatlands have been Sphagnum-dominated bogs. Likewise, core analysis reveals a median peat C mass (or median soil C density) of 119 kgC/m² (Figure 8 and Supplementary Table S4), with modern-day Sphagnum bogs having significantly greater C storage (141 kgC/m²) than their non-Sphagnum counterparts (56 kgC/m²; Tukey’s HSD: p < 0.00001). A regional map of peatland C density, which also reports whether each site is a Sphagnum or non-Sphagnum peatland, suggests that the greatest peat C stocks might be found along the eastern and southern portions of Isla Grande, in Tierra del Fuego (Figure 8).

Given the high proportion of sites that exhibit their FBT at 4200 cal. yr BP, we specifically analyzed peat mass and PCAR differences between sites that switched to bogs after the Burney II eruption vs. those that did not (Figure 9). We find that, on average, sites that switched to bogs accumulated significantly more peat (and stored more C) than sites that remained fens or cushion bogs (Tukey’s HSD: p < 0.00001). These figures amount to a median peat mass of 15 g/cm², equivalent to 19 gC/m²/yr for the 58 bogs, vs. a median peat mass of 4 g/cm² or 5 gC/m²/yr for the other 32 sites (Figure 9).

In light of tipping point theory, we argue that destabilizing forces are responsible for the FBT. These positive feedbacks consist of autogenic processes that can be triggered by a perturbation that helps “tip” a fen system into a bog state, by either dampening or amplifying a number of peatland processes. We propose two mechanisms that could explain the surge of Sphagnum during the late-Holocene: (1) volcanic impact, and (2) change in hydroclimate.

The stabilizing properties of bogs occur over multiple spatial scales, from microforms (10⁰–10¹ m) to mesoforms (10¹–10² m) and whole ecosystems (10²–10³ m). These stabilizing forces include a number of negative feedback mechanisms that relate to the presence of Sphagnum, changes in peat water storage, vegetation mosaics, and the rate of peat formation, which are briefly described below.

While it can be said that a number of negative feedbacks promote the persistence and homeostasis of the bog state (see above), the situation is different for the fen state. In the literature, it has been suggested that peatlands in general, and bogs in particular, are complex adaptive systems (Belyea and Baird, 2006). But not much has been said about the persistence capabilities of fens. Indeed, fens tend to be topographically constrained, that is, they develop in wet depressions and can “swamp” low slopes as they expand horizontally (e.g., Korhola et al., 2010; Loisel et al., 2013). It has also been suggested that they tend to be more sensitive to climate and hydrological changes then bogs (e.g., Hilbert et al., 2000), perhaps as a result of their incapacity to self-regulate the way bogs do. From this perspective, the stabilizing forces that promote a fen persistence are more difficult to identify. In an attempt to fill this knowledge gap, we propose that lateral expansion and intensified peat decay are potential properties that lead to persistence, albeit being “passive” negative feedbacks, as both these processes contribute to maintaining high surface wetness, particularly near the center of the fen.