Prescribed burning, atmospheric pollution and grazing effects on peatland vegetation composition

1. Peatlands are valued for ecosystem services including carbon storage, water provi-sion and biodiversity. However, there are concerns about the impacts of land management and pollution on peatland vegetation and function. 2. We investigated how prescribed vegetation burning, atmospheric pollution and grazing are related to vegetation communities and cover of four key taxa ( Sphagnum spp . , Calluna vulgaris, Eriophorum vaginatum and Campylopus introflexus ) using two datasets from a total of 2,013 plots across 95 peatland sites in the UK. 3. Non- metric multidimensional scaling and permutational multivariate analysis of variance showed differences in vegetation community composition between burned and unburned plots at regional and national scales. cover and 10 years Synthesis and applications. Burning, grazing and atmospheric pollution are associated with peatland vegetation composition and cover of key species, including Sphagnum . We suggest that, to promote cover of peat- forming species, peatlands should not be routinely burned or heavily grazed. Current or historical atmospheric pollution may hinder peat- forming species, particularly on burned sites.

Due to the carbon storage and hydrological functions of Sphagnum, its re-establishment is often a focus in peatland restoration (Evans et al., 2014;Rochefort, 2000). Sedges such as Eriophorum vaginatum also form peat, and can colonise and stabilise eroding peat (Evans & Warburton, 2008). Dwarf shrubs such as Calluna vulgaris occur naturally on blanket peatlands, often as a short canopy (Averis et al., 2004), but dominance can impact peatland soil ecosystem functions such as net carbon storage (Dixon, Worrall, Rowson, & Evans, 2015), water storage and routing (Holden, 2005) and dissolved organic carbon production (Armstrong et al., 2010). Invasive species including Campylopus introflexus, a moss native to the southern hemisphere which is now widely distributed in several European countries (Equihua & Usher, 1993), may also threaten ecosystem function by competing with native species.
Fire is common in peat-rich biomes and can threaten carbon stocks (Turetsky et al., 2015). Prescribed burning occurs on many peatlands globally for purposes including wildfire prevention, forestry, land clearance and habitat management Buytaert et al., 2006). Unlike wildfires, prescribed burns are usually controlled to ignite vegetation, but not the underlying peat. In the UK, patches of up to 4000 m 2 are burned in rotations of 5-20 years to create a mosaic of C. vulgaris ages to improve forage and nesting habitat for the game bird red grouse (Lagopus lagopus scotica). Despite guidance against burning on deep peat in the UK (Defra, 2007;Scottish Government, 2011;Welsh Assembly Government, 2008), there is evidence that burns have increasingly encroached onto these areas (Douglas et al., 2015;Yallop et al., 2006). In recent years, there has been considerable debate over burning (Brown, Holden, & Palmer, 2016;Dougill et al., 2006), and further evidence of how it affects vegetation is required to inform management and policy.
Fire can affect vegetation by both combustion and alteration of local environmental conditions. For example, greater peat bulk density  and lower near-surface hydraulic conductivity  in recently burned areas may reduce water availability to Sphagnum, which relies on passive capillary transport (Thompson & Waddington, 2008). Burning can cause higher maximum and lower minimum near-surface soil temperatures in the years following a fire , which may impact Sphagnum growth negatively (Walker, Ward, Ostle, & Bardgett, 2015). Interactions between vegetation properties, fire severity and local soil conditions may explain why past evaluations of burning impacts on Sphagnum have yielded mixed conclusions (Glaves et al., 2013;Worrall, Clay, Marrs, & Reed, 2010). Calluna vulgaris shows evidence of fire adaptation and may increase after burning unless suppressed by grazing. Burning may also allow new species to colonise. C. introflexus can carpet bare soil rapidly (Southon, Green, Jones, Barker, & Power, 2012) but whether this occurs to an ecologically significant extent after burning is unknown.
One issue that prevents generalisations from being made is that the current evidence base on the effects of burning on peatland vegetation draws mainly from localised case studies, and the applicability of these findings to larger scale (e.g. national) patterns remains unestablished.
When considering peatland vegetation patterns at large scales (i.e. national to global), atmospheric pollution becomes an important consideration. Sphagnum dieback near industrial centres has been attributed to sulphur deposition (Ferguson, Lee, & Bell, 1978). Nitrogen additions can limit Sphagnum growth directly (Granath, Strengbom, & Rydin, 2012) or by increasing competition (Malmer, Albinsson, Svensson, & Wallén, 2003), and can favour C. introflexus (Field et al., 2014;Southon et al., 2012). Grazing is also widespread on many peatlands and experimental work has shown varying responses among plant species (Milligan, Rose, & Marrs, 2016). Further knowledge of the relationship between grazing and peatland vegetation on a national scale may help when weighing up economic, cultural and conservation concerns.
The UK has approximately 1.5 million ha of blanket peatland (Bain et al., 2011); around 13% of the global total. The extent to which this peatland is affected by widespread management practices (burning and grazing), atmospheric pollution, and interactions between these drivers is not clear. In the UK, prescribed vegetation burning and atmospheric pollution both intensified during the 20th century, so knowledge of their interactive effects is valuable.
In this paper, we present the first synthesis of national scale peatland vegetation survey data from 1893 plots across 85 sites in England, alongside a regional dataset with 123 plots across 10 sites.
The study aims to evaluate relationships between environmental drivers (including prescribed burning, atmospheric pollution and grazing) and blanket peatland vegetation community composition, as well as abundance of four key taxa (Sphagnum spp., C. vulgaris, E. vaginatum and C. introflexus). The findings are discussed in the context of peatland restoration and management.  (Table 1)  . Within each burned site, plots were distributed between four age classes; burned <2 years (B2), 3-4 years (B4), 5-7 years (B7) and >10 years (B10+) prior to the study. The Domin abundance of vascular plants, bryophytes and lichens within a 2 m × 2 m quadrat in each plot was recorded and transformed to an approximation of percentage cover using the Domin 2.6 transformation (Currall, 1987). Peat depth was measured up to a 118 cm probe length limit. National Vegetation Classification (NVC) types (Rodwell, 1991) present at each site were also determined.

| Data sources
The second source was a habitat condition monitoring dataset (henceforth CM data). This was derived from a representative sample of mapped habitat polygons in Natural England's Blanket Bog Priority Habitat Inventory. Percentage cover values for indicator species (Table S1) were recorded using 2 m × 2 m quadrats. Peat depth was measured up to a 150 cm probe length limit and the presence of livestock droppings was recorded. Data from 1,893 plots (distributed randomly within 85 sites) with peat depth ≥30 cm were used in this analysis; 30 cm representing the minimum depth on which blanket peatland vegetation normally occurs (Lindsay, 2010). Satellite imagery from Google Earth™ and Bing Maps™ was used to determine whether plots had been recently burned (788, 41.6%) or not (1105) based on the visibility of burn patches. This is an effective method for determining the burn status of areas of peatland (Allen, Denelle, Ruiz, Santana, & Marrs, 2016;Yallop et al., 2006). However, in this instance, it was not possible to determine the age of individual burn patches, so the 'burned' category encompassed a range of burn ages. The time taken for visual recovery after burning varies according to vegetation regeneration, but is usually at least 25 years where dwarf shrubs are abundant (Yallop et al., 2006). Detection of burning on grass dominated moorland from imagery is problematic due to a short-lived burn signature (Yallop et al., 2006), so 303 plots with grassland vegetation types (14% of 2,196 total plots) were excluded from the analysis.

| Vegetation community composition
Non-metric multidimensional scaling of the EMBER data showed separation of burned and unburned plots (Figure 2), but no clear pattern within burned plots related to time since burning. Permutational multivariate analysis of variance indicated significant effects of site (df = 9,106, R 2 = 0.40, p < .001) and burn status (all levels included, df = 4,106, R 2 =0.14, p < .001) on species composition. When only burned plots were compared, the effect of burn age was not significant (p = .06, df = 3,112). The distribution of sites in the NMDS plot reflects both location and NVC type (Table 1). Within the unburned sites, two groups are apparent (Figure 2a, Figure S1); were associated with species composition (Figure 2).
In the NMDS ordination of the CM data, unburned and burned plots occupy overlapping areas in the ordination space (Figure 3)

| Relationships between key taxa and environmental variables
Results from the GLMM models revealed that several environmental and management variables were correlated with cover of key taxa ( Figure 4, The CM data indicated less Sphagnum on burned plots with a mean cover of 8.2% compared to 10.5% on unburned plots, and also on plots with livestock droppings with a mean cover of 6.7% compared to 10.7% on those without ( Figure 5,

| Interactions between burning and atmospheric pollution
At EMBER sites, an interaction between burn age and nitrogen deposition impacted cover of E. vaginatum, with a significantly more negative relationship on B2 plots than unburned plots ( Figure 6, Table S5).
In the CM data, nitrogen deposition had a significantly more negative relationship with Sphagnum cover on burned plots than unburned plots ( Figure 7, Table S5).

| DISCUSSION
Analysis of the two datasets revealed that blanket peat vegetation community composition is associated with burning status, grazing, atmospheric pollution and site physical attributes including northing and elevation. The four taxa of interest in this study were associated with several of these variables. Model R 2 values indicate that our predictors explained more variation in the regional EMBER data than the national CM data (Figures 4-7), which is likely due to less variation in other environmental factors at the regional scale. The unexplained variation may be due to factors including management history and hydrological variables not captured by our predictors (e.g. drainage). Although effect size and significance differed between the two datasets, the direction of effects was generally consistent. The results suggest that burning, grazing and atmospheric pollution have the capacity to shift vegetation community composition on peatlands, and that burning and atmospheric pollution have an interactive effect on the abundance of key taxa in some cases.

| Drivers of community composition
Analyses spanning regional to national scales both suggested that burning is associated with differences in vegetation community composition, as previously observed in local (Hobbs, 1984; and regional (Harris et al., 2011) Brown et al., 2015;Holden et al., 2015).

The vegetation of the unburned sites shows a clear divide between
North and South Pennine sites in the EMBER NMDS ordination ( Figure 2) in line with the two mire NVC types they supported (Table 1). However, this north-south divide is not apparent in the burned plots, and three of the five burned sites supported heath vegetation types associated with burning, grazing and atmospheric pollution (Elkington et al., 2001). This suggests that geographically  (Table S1). These taxa represent the majority of cover and the most ecologically important species on UK blanket peatland sites (Averis et al., 2004); however, it is possible that further differences in the vegetation composition of plots would be observed with the inclusion of rarer species.

| Variation in Sphagnum cover
Analysis of the CM data, compiled from 1893 plots at 85 sites across Our results suggest that current burning practices may leave Sphagnum vulnerable to decline in some cases. This could reflect fire damage to Sphagnum similar to that described by Lindsay and Ross (1994), competitive effects or an indirect impact via peat physical and hydrological properties Clay, Worrall, Clark, & Fraser, 2009;Holden et al., 2014Holden et al., , 2015.
Data from the 10 EMBER catchments showed that differences in Sphagnum cover between unburned plots and burned plots of any age were not significant, although effect directions were negative with the exception of B4 plots. These findings contrast with the significant relationship in the larger CM dataset. It is possible that changes in Sphagnum abundance after burning were asynchronous across sites, or that the number of EMBER plots (3 per burn age × 5 sites) did not control adequately for other factors affecting Sphagnum abundance (e. g. burn severity, microtopography and livestock access).
Nitrogen deposition had a negative relationship with Sphagnum cover in both datasets. Both nitrogen and sulphur (deposition of which were correlated) can have direct physiological impacts on Sphagnum (Ferguson et al., 1978;Granath et al., 2012), and nitrogen can increase competition from vascular plants (Limpens et al., 2011), promoting graminoid cover (Field et al., 2014). Sulphur deposition has declined faster than nitrogen deposition in recent decades (Curtis & Simpson, 2014), but legacy impacts on vegetation are possible.
The significantly more negative relationship between nitrogen deposition and Sphagnum cover on burned plots compared to unburned plots in the CM data may indicate that atmospheric pollution affects Sphagnum re-establishment by limiting growth or propagule availability after burning. Sphagnum peatlands are vulnerable to N deposition (Granath, Limpens, Posch, Mücher, & de Vries, 2014), and this interaction suggests that vulnerability may intensify if burning continues to increase. The positive relationship between Sphagnum and northing is difficult to attribute to a single influence, but may be associated with geographic variation in rainfall (Nijp et al., 2014), temperature (and evapotranspiration), geology or land management.
The lower Sphagnum cover on plots with livestock droppings may be due to nutrient inputs, physical damage or hydrological impacts of trampling such as peat compaction leading to less water availability (Meyles, Williams, Ternan, Anderson, & Dowd, 2006). Finally, the relationship between elevation and Sphagnum was positive in the EMBER dataset and negative in the CM dataset. This difference may be a result of the larger range of elevations in the CM dataset (Table S2), or regional variation in the effect of elevation according to factors such as climate or abundance of competitors such as C. vulgaris.

| Variation in Calluna vulgaris cover
The greater C. vulgaris cover on burned sites in the CM data was expected, as burning is often practised to regenerate heather. It could be suggested that dwarf-shrub dominated vegetation is more likely to be selected for burning, but all plots were on deep peat (>30 cm), which suggests that this vegetation type is itself a legacy of management. On intact, Sphagnum-dominated blanket peatland, C. vulgaris is thought to regenerate naturally through layering of branches as Sphagnum grows up through the stems (Forrest, 1971;Macdonald, Kirkpatrick, Hester, & Sydes, 1995). Burning may disrupt this process and enhance C. vulgaris regeneration from seeds to roots (Forrest & Smith, 1975). The processes behind the increase in C. vulgaris may also be hydrological, as burning can result in deeper water-tables and less water availability , and C.
vulgaris is primarily a heath species which can tolerate drier conditions than many other peatland plants.
The CM data also indicated a positive correlation between C. vulgaris cover and both elevation and northing value, which could be due to geographic variation of multiple influences as discussed for Sphagnum. Plots with livestock droppings had less C. vulgaris cover than those without, perhaps indicating a negative impact of grazing in accordance with the findings of Hulme et al. (2002). However, it is also possible that dense heather limits livestock access or other areas are grazed preferentially.

| Variation in Eriophorum vaginatum cover
The lower cover of E. vaginatum on plots burned 2, 7 and 10+years ago compared to plots at unburned EMBER sites was unexpected given the dominance after fire observed in past studies (Hobbs, 1984).
However, in the CM data, there was no significant difference in E.
vaginatum cover between burned and unburned plots. The ability of E.
vaginatum to survive burning may depend on fire intensity, which was not quantified in this study, and the extent to which fire penetrates tussocks, damaging growing buds. Rapid regeneration of C. vulgaris after fire may also limit the opportunity for E. vaginatum to proliferate.
The positive relationship between E. vaginatum and both northing and elevation in the CM data may indicate enhanced competitive ability in cooler, wetter conditions. The interactive relationship of burning and nitrogen deposition with E. vaginatum observed in the EMBER data could indicate that after damage by burning, E. vaginatum is more susceptible to nutrient-driven competition (Wein & Bliss, 1973).

| Variation in Campylopus introflexus cover
Results from the EMBER data suggested that C. introflexus colonises rapidly after burning and can sustain increased cover for at least 7 years post burn. This is consistent with reports of C. introflexus colonising bare and disturbed peat (Equihua & Usher, 1993) and suggests that disturbance including burning may have assisted the spread of this non-native species in Europe. Cover was less at higher elevations, which could indicate climatic preferences or limited dispersal.
Nitrogen deposition had a positive relationship with C. introflexus, consistent with reports of a positive response of the species to nitrogen (Southon et al., 2012), which suggests it may be able to occupy a niche vacated by pollution sensitive native species such as Sphagnum.

| CONCLUSIONS
Our results suggest that burning, atmospheric

DATA ACCESSIBILITY
The data used for the analyses in this paper are archived in the Research Leeds Data Repository. DOI: https://doi.org/10.5518/230 (Noble et al., 2017).