Samples of UK peatland vegetation species were collected from the Kinder Scout region of the Peak District National Park, UK (53°23′04″N, 1°52′20″W). The species were C. vulgaris, E. vaginatum, Polytrichum juniperinum and V. myrtillus (hereafter Calluna, Eriophorum, Polytrichum and Vaccinium respectively). These samples were selected to be representative of ‘woody’ shrubs (Calluna and Vaccinium), and ‘non-woody’ grass (Eriophorum) and moss (Polytrichum) communities. A modest quantity (∼250 g) of vegetation was collected per species studied, with cuttings being taken from a 1 m² area in an effort to reduce PyC feedstock variability. Vegetation cuttings were taken with secateurs, placed in sealed plastic bags in the field then stored at 4°C in a laboratory fridge to minimise moisture losses prior to PyC production. Initial moisture contents for Calluna, Eriophorum, Polytrichum and Vaccinium were 44 ± 1, 67 ± 5, 58% ± 2% and 41% ± 2%, respectively. Whilst moisture content is an important factor with regards to mass loss in early phases of combustion (Keiluweit et al., 2010), it was considered too difficult to manipulate as an experimental factor and as such it was recorded but otherwise left unaltered.

Based on biomass loss estimates from Worrall et al. (2013), and references therein), typical biomass losses from fires in UK peatland vegetation range between 60% and 100%. We used this mass loss range to define four different severity groups: ‘low severity’, ‘moderate severity’, ‘high severity’ and ‘very high severity’ (hereafter LS, MS, HS and VHS respectively) corresponding to 60%, 70%, 80% and 90% mass loss respectively. This range allowed us to simulate a PyC continuum ranging from materials that had only undergone minimal charring to those that have undergone near complete combustion, whilst also comparing the effects of ‘short-hot’ and ‘long-cool’ burn conditions. To achieve the desired mass losses, samples were burnt for durations of 2, 5 and 10 min, and exposed to constant temperatures which ranged from 250°C–800°C (see Table 1 for further details). These temperatures and durations are representative of field conditions (Hobbs and Gimingham, 1984; Hudspith et al., 2014) and have been used in other experimental studies (e.g., Santana and Marrs, 2014; Worrall et al., 2013).

Vegetation samples, ∼1–2 cm in length, were weighed into 15 mL ceramic crucibles and placed in a Carbolite ELF 11/14 muffle furnace. Samples were not pre-dried and oxygen availability was not restricted (e.g., by wrapping fuels in foil) with samples burnt on the order of minutes rather than hours. Samples were placed inside the furnace as quickly as possible once they had reached the intended target temperature. Repeated triplicate burns were carried out until sufficient PyC material was available for analysis. Any individual sample that deviated more than 5% from the desired mass loss was discarded. Whilst muffle furnaces do not recreate the temperature curves and airflow conditions experienced in real-world wildfires, it is an approach that has been successfully implemented in previous UK peatland PyC studies (e.g., Worrall et al., 2013) and other related research (Bodí et al., 2011; Santos et al., 2016; Saputro et al., 2018). The chosen production method also allowed for observations against carefully controlled burn conditions, where open air burns provide greater realism at the cost of greater difficulty in studying specific fire scenarios (e.g., ‘short-hot’ and ‘long-cool’ burn conditions). Future research, without the aim of studying specific formation conditions, may wish to focus open-air PyC production methods (e.g., modified combustion efficiency) in order to maximise the real-world relevance of PyC formation conditions. Whilst much of the research looking at peatland wildfire phenomena assess smouldering fire conditions (i.e., limited oxygen combustion), the research presented here is part of a series of experiments evaluating above-ground fire scenarios, where vegetation is subjected to varying degrees of combustion. This research makes no attempt to study the effects of fire on peat or burning in low oxygen conditions.

The PyC samples were subject to Brunauer-Emmett-Teller (BET) analysis in bulk to determine surface area. The PyC samples were then crushed and homogenized in an agate pestle and mortar and oven dried at 105°C for 24 h for subsequent Fourier-transform infrared spectroscopy (FTIR) and CHNO elemental analysis. Triplicate analyses were carried for each combination of factors (e.g., 3x 2 min LS Calluna, 3x 2 min MS Calluna, etc.).

Statistical analyses were performed using IBM SPSS version 28 (IBM Corp. Released, 2021). The data were non-normally distributed necessitating the use of non-parametric Kruskal–Wallis one-way ANOVA. The study can be considered a two-factor experiment: species and severity. The species factor had four levels (Calluna, Eriophorum, Polytrichum, Vaccinium) and severity also had four levels (LS, MS, HS, VHS). All relationships were assessed at the 95% confidence level (p < 0.05). Pairwise comparisons were performed using a Kruskal–Wallis 1-way ANOVA with Bonferroni corrections. Eta-squared values (η²) were calculated to determine each factor’s effect size (Meyer et al., 2018).

Principal component analyses (PCA) were performed using Past 4 (a freeware software statistical package originally designed for palaeoenvironmental data; Hammer et al., 2001) to assess the clustering of samples in relation to different production conditions. Individual PCA were performed on each vegetation species, using the CHNO data and FTIR ratios. BET results were excluded from the PCA due to missing values in this dataset. To remove the effects of differences in scale between variables, data were normalised by subtracting the average value and then dividing by the standard deviation of that variable for all values in each variable. Components with eigenvalues >1.0 were retained (Wold et al., 1987).

The results of the PCA are shown in Figure 6. For Calluna and Eriophorum PC1 explained ∼70% variance, whilst >55% variance was explained in the case of Polytrichum and Vaccinium. PC2 explained <20% of variance for all species (Table 6).

For all species, PC1 values increase with H/C, O/C, C/N, A₃₃₄₀/A₁₆₀₀, A₂₉₂₀/A₁₆₀₀ and A₁₀₆₀/A₁₆₀₀ values, whilst A₈₀₀/A₁₆₀₀ values decline for all species except Eriophorum. Additionally, A₁₇₀₀/A₁₆₀₀ undergoes very slight declines for Polytrichum as PC1 increases. Therefore, PC1 essentially tracks moisture evolution and thermal degradation of aliphatic components, with lower values indicating a greater degree of aromatic condensation in most species. The patterns of PC2 are less consistent across the different species. Woody species see increases in A₈₀₀/A₁₆₀₀ and decreases in A₂₉₂₀/A₁₆₀₀ and A₁₇₀₀/A₁₆₀₀ as PC2 values increase, indicating combustion of aliphatic components and increases in the degree of aromatic condensation. Non-woody species undergo increases in A₃₃₄₀/A₁₆₀₀, A₁₇₀₀/A₁₆₀₀ and C/N values as PC2 increases.

With the exception of Vaccinium, the different burn severity groups produce distinct clusters on the PCA axes demonstrating that they represent comparable materials in spite of differing formation conditions (i.e., burn durations).

Clear physical and chemical changes were evident with increasing burn severity for peatland PyC, though trends differed by fuel type. These changes to PyC characteristics included increasing carbon concentration and aromaticity, and consistent shifts in elemental and spectral composition with greater burn severity. The observations discussed here allowed for a partial acceptance of both hypotheses owing to mixed results from fuel types and the analytical methods employed.

The methods used to produce the samples meant that LS samples had small quantities of uncharred material remaining in some cases. Whilst some small components of the samples remained uncharred with LS samples, reflecting very early phases of combustion that could be seen to relate to peripheral fire effects, this was no longer present by MS burn conditions. Early changes in PyC elemental composition likely reflect dehydration reactions between LS and MS, whilst shifts in H content at higher burn severities likely relate to dehydrogenation and demethylation, and reductions in O abundance through decarboxylation reactions (Hammes et al., 2006). These changes in elemental composition mirrored progressive reductions in peaks at 3400–3300 cm^-1, 3000–2800 cm^-1 and 1060–1030 cm^-1, reflecting the losses of–OH, aliphatic CH, and C-O or alcohol–OH compounds respectively (Figures 4, 5), which have been shown to lose absorption intensity with relatively low combustion temperatures (∼300°C) (Guo and Bustin, 1998; Marchessault, 1962). NMR studies of similar low severity PyC have found very low levels of carbohydrate and methoxyl C, diminishing O-aryl and alkyl C, and the emergence of aromatic aryl spectral features that are prominent in higher severity PyC (350°C–700°C) (McBeath et al., 2011). Whilst–OH and CH were largely absent in the HS samples, this pattern differed across species. For example, Calluna and Eriophorum showed marked reductions in CH molecules from LS to HS samples, whilst Polytrichum and Vaccinium showed no significant difference across this range. These compounds can be affected by low burn temperatures, but dehydroxylation has been observed to occur at relatively high temperatures (440°C–620°C) (Lammers et al., 2009). Thus, the LS, MS and HS samples presented here may still have been undergoing dehydroxylation. These molecular changes, alongside peaks at 1600 cm^-1 persisting across species and burn severities (Figures 4, 5), reflect increases in PyC aromaticity with burn severity and thus support hypothesis 2.

A₈₀₀/A₁₆₀₀ values observed in the samples presented here (Figure 5) varied between fuels but generally showed increases with burn severity. In the case of Eriophorum, the lowest degree of sample condensation is at HS, with declining signals to this point likely reflecting losses of aliphatic CH (e.g., see Figure 5 in Guo and Bustin, 1998). However, ‘severity’ was not found to have a significant effect on this condensation proxy, which was not anticipated and thus deviates from the expectations of hypothesis 2.

Indicators of the degree of aromatic condensation were greatest for Eriophorum suggesting greater losses of amorphous biomass structures with more formation of graphitic sheet structures in this fuel than the other fuels investigated, potentially pointing to Eriophorum being more susceptible to molecular changes under the burn conditions used here relative to the other species investigated. This finding in conjunction with the significant effect of species on A₈₀₀/A₁₆₀₀ (Table 4) directly contradicts hypothesis 1. The non-significant effect of burn severity on A₈₀₀/A₁₆₀₀ potentially relates to the relatively short duration/low temperature conditions used here (2–10 min, 250°C–800°C), whereas increases in condensation have been noted in the range of 400°C–1000°C in previous studies (Schneider et al., 2010; Wiedemeier et al., 2015). The burn conditions used in this study may therefore have resulted in dehydration occurring up to the HS samples presented here, meaning that apparent early declines in condensation likely reflect losses of H in the range of ∼800 cm^-1 and increases in A₈₀₀/A₁₆₀₀ for HS and VHS samples are indicative of the formation of aromatic complexes. In addition to differences in response to specific burn conditions between fuels, the significant effect of ‘species’ on A₈₀₀/A₁₆₀₀ values may be the result of differing lignin content present in the ‘non-woody’ and ‘woody’ species. In previous research this has been found to result in a greater presence of aromatic complexes in ‘woody’ fuels than ‘non-woody’ fuels under higher severity burn scenarios (Wurster et al., 2013). However, this trend is not reflected in the data presented here (Figure 5), suggesting that further analysis (e.g., BCPA) may be necessary to more comprehensively understand the molecular changes occurring in United Kingdom peatland PyC under more complete combustion (Schneider et al., 2010).

For some samples there was relatively little change between MS and HS for certain measures. For example, features at 2920 cm^-1 and 1700 cm^-1, and A₂₉₂₀/A₁₆₀₀ and A₁₇₀₀/A₁₆₀₀ values, for Polytrichum and Vaccinium (Figures 4, 5) showed <0.2 change in ratio values. However, for some species (e.g., Eriophorum) this phase of the sample continuum shows much more marked changes. This reflects a transitionary phase between largely unaltered plant material, where original plant structures remain, to more amorphous char materials. Here continued dehydration and depolymerization of cellulose and lignin likely led to the formation of volatile compounds and randomly arranged aromatic structures, with pore space beginning to increase (Keiluweit et al., 2010). Spectral features at 2920 cm^-1 and 1700 cm^-1 were amongst those identified both in this study and by Uhelski et al. (2022), with 1700 cm^-1 being strongly associated with peat-bound PyC. Uhelski et al. (2022) suggest that spectral features in the range of 1700 cm^-1 in PyC may arise from field aging of PyC, indicating that it has potential in PyC field degradation studies (e.g., Kennedy-Blundell et al., 2023). However, the variable response of certain spectral features across the species studied here suggests that further investigation of PyC produced from a range of United Kingdom peatland fuels is required to more fully understand the relationship between initial biomass and resultant PyC characteristics.

The spectra of the HS samples, particularly in the woody fuels, showed clear losses of C-O rich molecules (1060–1030 cm^-1) (Figure 4), characteristic of cellulose and hemicellulose (Yang et al., 2007). These spectral changes suggest a relative increase in lignin abundance, which persisted across a wider temperature range than the less thermally resistant cellulose (Mackay and Roberts, 1982). However, Calluna and Eriophorum showed reduced absorbance at 1700 cm^-1 showing losses of C=O bonds, whilst C=C bonds remained relatively abundant (Figures 4, 5). This changing ratio of C=C to C=O molecules likely reflects emissions of CO and CO₂, with gas phase FTIR studies showing these to be the primary products of lignin combustion and aromatization (Cao et al., 2013). These changes, alongside O/C and H/C shifts towards 0,0 in the Van Krevelen plots (Figure 2), indicate more complete combustion of samples, with increasing shifts away from the properties of the initial biomass towards ‘composite chars’ as discussed by Keiluweit et al. (2010). These composite chars will be partially composed of aliphatic and O containing compounds, which are preserved alongside aromatic components, with these labile components being largely absent at higher burn severities PyC (Keiluweit et al., 2010). These shifts in elemental composition (i.e., declining O/C and H/C values) alongside molecular changes support hypothesis 2.

C/N values differed across the vegetation species and fuel types (i.e., woody vs. non-woody). Additionally, C/N values at different burn severities varied by fuel type but generally decreased by HS and VHS (Figure 3), potentially reflecting thermal degradation of amino acids and amide groups under higher severity combustion. These biomass components were found to remain relatively stable in simulated fires at 350°C for up to 90 s (Knicker et al., 1996), though these conditions do not equate to those observed for United Kingdom peatland settings or the conditions under which PyC is typically formed. Pyle et al. (2015) found that %N responded to charring intensity thresholds (rather than changing linearly), with the greater losses for non-woody than woody fuels. They proposed that N is transformed into heterocyclic and aromatic compounds before being decomposed at higher severities. The chemical transformations noted by Pyle et al. (2015) were similar to the trends observed for non-woody C/N values in this study (Figure 3); however, the trends varied by fuel suggesting that charring thresholds may have differed by initial fuel composition.

Changes in PyC physical characteristics were limited in low to high severity samples, particularly with non-woody fuels, where surface area values were generally <10 m²/g. Higher burn severities consistently produced samples with the greatest surface area (Figure 1) suggesting that physical changes, such as the generation of fine ash particles and soot microstructures (e.g., Hammes et al., 2008), are strongly linked to burn severity. Furthermore, as Keiluweit et al. (2010) state, the surface area of low burn severity chars may have been limited by lack of atomic pore space and blocking of pores by recondensation of volatile matter. Additionally, high burn severity (particularly VHS samples) PyC samples may have experienced an increased lateral growth in graphene-like sheets and development of nanopores (Keiluweit et al., 2010), potentially explaining the surface area and A₈₀₀/A₁₆₀₀ increases between the HS and VHS Calluna and Eriophorum samples presented in this study.

Whilst surface area underwent marked increases between HS and VHS, large standard deviations relating to differing burn characteristics were observed at HS and VHS, particularly for Calluna. Brewer et al. (2014) highlighted the importance of pyrolysis conditions on PyC physical characteristics, which can be seen in our results: Calluna and Vaccinium show the greatest surface area with high temperature 2 min burns, whilst Eriophorum and Polytrichum surface area values were greatest with lower temperature 10 min burns (see Supplementary data). Increasing surface area values have been observed with greater burn temperatures (e.g., Keiluweit et al., 2010; Sigmund et al., 2017) and in some cases burn duration (e.g., Sun and Zhou, 2008). However, the finding from this study, that PyC produced from differing fuel types (i.e., woody vs. non-woody fuels) may exhibit divergent physical characteristics from comparable burn conditions, is a novel observation and may warrant further investigation. In spite of this interesting trend, there were no significant differences in BET values produced by the four different species used in this study. As such, despite woody shrubs returning higher overall BET values and responding differently to burn temperature/duration conditions, surface area values only diverged from those of the non-woody fuels under the highest severity burn conditions (Figure 1). Therefore, shifts in PyC physical characteristics respond to complex interactions between burn conditions and initial fuel.

Throughout the analyses presented here the woody shrubs (Calluna and Vaccinium) showed no significant difference to each other (Table 5). The non-woody species produced a wider range of CHNO values than the woody species (Table 3), with higher H/C and O/C values (Figure 2), suggesting a lower degree of aromatisation and dehydration, particularly in the lower severity non-woody samples (Knicker, 2009). This range in CHNO values between PyC fuels likely related to differing characteristics of the unburnt fuels, where the greater lignin content of the woody fuels will have resulted in thermal decomposed over a higher temperature range then the non-woody fuels (Yang et al., 2007). Consequently, the woody shrubs generally required higher treatment temperatures (325°C–800°C) than the non-woody species (250°C–700°C), in combination with the set burn durations, to achieve the desired mass loss percentages. This highlights the importance of considering starting fuel characteristics, burn temperature and burn duration when discussing lab produced PyC samples and their characteristics (e.g., Gibson et al., 2018; Pyle et al., 2015; Santín et al., 2016). Finally, our results show that varying combinations of burn temperature and duration can result in broadly comparable PyC residues for each severity class, but combined physical, elemental and molecular techniques may provide a strong basis for identifying specific PyC formation conditions.

Whilst post-fire burn severity estimates allow for insights into fire behaviour and estimates of biomass loss, they provide limited insight into the key details of burn characteristics. The coupling of PyC mass loss estimates with tools like the composite burn index (CBI) (e.g., Davies et al., 2016), may allow for greater understanding of the burn severity experienced in a given study area after a fire (Key and Benson, 2006). For samples with unknown formation conditions, it may be possible to infer PyC characteristics (e.g., degree of aromaticity) from simple mass loss estimates. These inferences may be strengthened by comparisons to PyC produced under known conditions (i.e., burn temperature and burn duration). It is important to note though that, whilst the general observations made here are comparable to other studies (e.g., Guo and Bustin, 1998; Hammes et al., 2006; McBeath et al., 2011; Uhelski et al., 2022), clear differences can occur between lab and field burn conditions, meaning that caution should be taken when comparing between PyC properties generated in these two settings.

By coupling fire severity estimates with an improved understanding of how PyC changes over time, it may be possible to predict the longevity of PyC and understand the longer-term impacts of fire on C cycling in United Kingdom peatlands. Kennedy-Blundell et al. (2023) provide one of the few studies that has looked at in-field PyC degradation in United Kingdom peatlands, but this was limited to a short term (≤1 year) field campaign. Long term studies (i.e., multi-year) are needed to gain a greater understanding of PyC degradation prior to incorporation into peat sequences or their mobilisation to other parts of the landscape. Such research is key as the landscape the PyC was produced in, including nearby aquatic systems, constitutes the main depositional settings for PyC over short to moderate timescales (e.g., Abney et al., 2019; Smith and Dragovich, 2008). Studies that have analysed PyC-soil interactions (e.g., DeCiucies et al., 2018; Liang et al., 2010) have observed decreases and increases in native soil carbon due to the presence of PyC (e.g., Maestrini et al., 2015; Zimmerman et al., 2011), with PyC also being noted as an important contributor to aquatic DOC (Wagner et al., 2018).

The role of PyC in carbon cycling in United Kingdom peatlands has been relatively understudied (with some notable exceptions, e.g., Clay and Worrall, 2011; Heinemeyer et al., 2018; Worrall et al., 2013), meaning that fundamental C cycling mechanisms may have been overlooked in these settings (e.g., early PyC derived C fluxes, PyC incorporation into peat, transport to the aquatic systems). Therefore, this study has added valuable new insights in this research lacuna.