Resolving a heated debate: the utility of prescribed burning as a management tool for biodiversity on lowland heath

1. Lowland heath is a priority habitat for conservation, nowadays largely managed for biodiversity. Historically, prescribed burning has been the principal management tool, but there are increasing calls to substitute burning with cutting to improve biodiversity outcomes. However, poor understanding of potential impacts compromises decision making. 2. Our study was carried out in the New Forest National Park, the largest area of lowland heath in Europe. Using a multi-trophic approach, we compared the ecological impact of prescribed burning with two types of vegetation cutting (swiping and baling) as management tools for biodiversity outcomes for up to 20 years after


| INTRODUC TI ON
British heathlands are semi-natural landscapes that evolved on nutrient poor, usually acidic soils through the removal of nutrients and biomass, largely developing after anthropogenic Neolithic forest clearances. They are characterised by dwarf shrub communities, dominated by Calluna vulgaris, varying in species composition along an altitudinal gradient from upland moorland to lowland heath (<300 m elevation; <1000 mm annual precipitation; Elkington et al., 2013;Webb, 2008). From the medieval period, most heaths in Britain were managed as commons on which local people had the right to graze animals, gather herbage and fodder, practice turbary (turf cutting), and collect peat and wood for fuel. Small areas were periodically burnt to provide nutritious forage, and heather was cut as winter fodder for cattle. These activities maintained an open landscape dominated by dwarf ericaceous shrubs that supported unique and valuable communities of flora, invertebrates, reptiles and birds (Webb, 2008).
In the UK, as in the rest of Europe, traditional uses of heathland declined from the 1930s onwards, many heaths becoming fragmented as land was converted to arable agriculture or used for urban development. Heathland habitat became rare, thereby threatening heathland specialist species. Since 1981, lowland heath has been listed under Annex 1 of the EU Habitats Directive and is a priority habitat under the UK Biodiversity Action Plan; it is currently proactively managed to conserve the characteristic habitat as an end in itself. The UK contains an internationally significant proportion of dry heathland, supporting approximately 18% of the world total, of which 95,000 ha (11%) is lowland heath (Townshend et al., 2004).
With heathlands managed as conservation landscapes, rather than working landscapes, the manner of their management has opened to public debate. Prescribed burning, the controlled application of fire to vegetation in either their natural or modified state, under specified environmental conditions which allow the fire to be confined to a predetermined area and at the same time to produce the intensity of heat and rate of spread required to attain planned resource management objectives (FAO Wildland Fire Management Terminology, 2003), has become a contentious management tool. Although rotational burning was used traditionally, albeit varying in extent and frequency across the heathland range (Webb, 2008), concerns have been raised that burning results in poor environmental outcomes by negatively affecting water quality, carbon dynamics and habitat composition (Harper et al., 2018). There has been particular concern expressed for amphibians and reptiles, but the impact of prescribed burning on these species-especially in the UK-remains poorly understood. While extensive summer burns can directly kill large numbers through exposure to fire and loss of cover post fire, winter burning in modest patch sizes is likely to have the lowest impact, as the temperature is unlikely to penetrate the soil to the hibernating animals and the distance for habitat re-colonisation will be low (Jofré & Reading, 2012;Santos et al., 2022).
In contrast to fire-prone regions, where wildfires are key drivers for environmental outcomes, the impacts in temperate regions (where the majority of heathland burning is prescribed), are poorly researched (Newton et al., 2009). This paper aims to fill that gap in evidence-based decision making.
In the UK, decision making over the use of prescribed burning is further compromised by an increasingly polarised and heated debate. The dispute originates in upland areas, where rotational prescribed burning is traditionally used to create a mosaic of differently aged heather to benefit red-grouse (Harper et al., 2018), hence, much of the UK heathland management debate is enmeshed with conflict over grouse management. Burning management for grouse has a different objective in terms of the type of vegetation required; more frequent burning is conducted to maintain a supply of young shoots.
Burning is conducted when heather is 20-30 cm high, typically 8 years on the most fertile soils, longer on less productive ones. In contrast, in the New Forest (NF) it is to prevent scrub developing and maintain grazing for commoners' livestock and thus conducted on a 20-year rotation. Despite the difference in management targets, the debate is being driven as much by political and economic interests as ecological understanding and consequently, and crucially, it lacks nuance, even though the limited evidence available demonstrates spatial heterogeneity in burning practices and comprises examples of both good and bad practice (Davies et al., 2016). Furthermore, the majority of research into prescribed burning has been carried out on upland systems where driven grouse shooting is contested (comprising 77% of the evidence base, Harper et al., 2018). Davies et al. (2016) further suggest that the tone of the debate inhibits necessary research by discouraging land managers collaborating when "the prevailing public perception of fire is negative and managers are inclined to view scientists as having an agenda". This perception compromises research in lowland areas too. Furthermore, the common standards monitoring (CSM) protocol is restricted to results. Current heathland CSM does not adequately assess wider biodiversity on protected areas but is effectively an assessment of vegetation feature condition.
Including invertebrates in surveys provides a more nuanced assessment of heathland condition.

K E Y W O R D S
chronosequence, common standards monitoring, heather cutting, heathland invertebrates, heathland management, long-term monitoring, lowland heath, prescribed burning vegetation, thus encouraging a one-dimensional focus on this as the indicator of habitat quality. As a result, there is a paucity of evidence of the broader ecological effects of prescribed burning, yet there remains an urgent need for scientific evidence to inform land managers and policy makers on the management of lowland heathland, as it differs fundamentally from upland heathland/moorland. Lowland heathland occurs below 300 m on sands and gravels and has a higher floral diversity compared to upland heathland which is on shallow peat and mineral soils. Consequently, the impact of management interventions may not always be directly comparable.
Our study area was the New Forest National Park, the largest area of remaining lowland heath in Europe (McLeod et al., 2005), containing approximately 14,600 ha of heathland and similar habitats (Tubbs, 1974). The national park is an IUCN-designated Category V protected area ('Protected Landscape'; Chape et al., 2005 The aim of our research was to compare the ecological impact of prescribed burning with vegetation cutting (i.e. the principal alternatives of swiping and baling) as management tools for biodiversity outcomes, taking a multi-trophic approach. We employed standardised methods, repeatable, locally tailored but with global relevance and gathered a large quantity of data to ensure the rigour of our results. The Forestry Commission maintain records of all management that takes place in the NF. These records enabled us to identify a replicated chronosequence of sites to investigate impacts of the management techniques in both time and space.

| Sampling design
One hundred and five sites were selected in a replicated, three block design (See Figure 1) across the New Forest National Park. Each block comprised three replicate sites each of swiped and burnt areas in a chronosequence of 0, 1, 6, 10 and 20 years after management.
Because of the small number of baled sites, it was not possible to fully incorporate baling into the experimental design. In each block, we identified one replicate of each of five sites between the ages of 0 (newly baled) and 12 years. These were not included in main analyses but used as Supporting Information. Permission to carry out field work was given by the New Forest National Park as part of the grant agreement.
Sampling and sample identification was carried out by 42 volunteers, recruited and trained by staff at the Natural History Museum and the Game and Wildlife Conservation Trust.
Sites were identified using the Forestry Commission management database. Sampling took place at each site within 50 × 50 m plots, selected as close as possible to the site centre. Vegetation structure and species composition was recorded to species in six, randomly located, 2 × 2 m quadrats. In addition, the key components of the Condition Standards Monitoring target indicators (Table 1) were incorporated into the vegetation survey. Soil invertebrates were sampled from six randomly located 25 × 25 × 10 cm deep soil pits. Soil was hand sorted to remove invertebrates which were then preserved in ethanol and identified to species using a binocular microscope. Ground active invertebrates were sampled using pitfalls traps. Six large (500 mL) and six small (250 mL) plastic pitfall traps were set and left open for 1 week. Pitfall traps contained water and ethylene glycol (preservative) and a couple of drops of scentless detergent to break surface tension. Samples were subsequently preserved in 70% ethanol and identified to species. Invertebrates active in the aerial parts of the vegetation were sampled using five sweep net samples-the maximum number that fitted into the sampling area. Each sample comprised 25 sweeps taken on a random zigzag walk. Samples were initially frozen before being processed, preserved in 70% ethanol and identified primarily to family (due to the presence of many nymphs) and to species for selected groups.
Grazing is unrestricted across the Open Forest but resulting variability in grazing pressure between sampling sites was minimised by replication over a large area.

| Data preparation
Common standard monitoring (CSM) guidelines formed the basis from which to estimate the 'Heathland Condition' at each site.
Data were averaged across quadrats within each site. Allocation of average data scores was used as a quantitative variable for analysis: for each positive target condition met (Table 1), one point was allocated, for each negative indicator a point was deducted. Total points provided the 'Heathland Condition' estimate.
Vegetation data were grouped to represent the aspects of vegetation of interest to managers (i.e. heathland condition); amount of grass for grazing; bare-ground (for basking reptiles and some invertebrates) and cover of indicator species such as lichen. Variables used in analyses were: heathland condition (see above), vegetation height and % cover of bare ground, dwarf shrubs (together), heathers, gorse, graminoids (grasses, rushes, sedges), grass alone, sedges and rushes (with no grass), the three most frequently occurring grasses (bristle bent Agrostis curtisii. purple moor grass Molinia caerulea, heath grass Danthonia decumbens), bracken Pteridium aquilinum, moss and lichen.
Invertebrate data were grouped to represent aspects of biodiversity that are key indicators of good heathland management.

| Data analysis
Data were tested to determine whether they satisfied the assumption of homoscedasticity by inspection of residuals versus fitted values and F I G U R E 1 Position of the New Forest, Hampshire, in Southern England. Each dot represents a sampling location. Experimental blocks 1, 2 and 3 are coloured pink, brown and blue, respectively. Within each block are: three replicate chronosequences (0, 1, 6, 10 and 20 years) burnt and swiped sites; one replicate of 0, 1, 6, 10 and 12 years baled sites.
QQ plots; where this was not the case, percentage data were arcsine transformed and count data were log 10 (X + 1) transformed and tested again. All data presented here satisfy the assumptions of the test.
ANOVA models assessed the impact of management on vegetation and invertebrates (GENSTAT v15.1) testing for the effect of management type, age of plot and controlling for the block effect. Because insufficient replication of baled areas was available, we split the analyses to minimise use of unbalanced analytical designs. Primary analyses were conducted on the full chronosequence of 0-20 years, comparing burned and swiped sites. Secondary analyses included all three treatments (burned, baled, swiped) excluding 20-year plots.
In order to determine potential for management outcomes to affect heathland birds, we explored the relationship between vegetation components and the food items available to Dartford warblers and Nightjars. First, a partial correlation matrix was created using the PARTI ALC ORR ELA TIONS procedure in GENSTAT (v15.1), which calculates a symmetric matrix of partial correlations from a set of variates. The matrix contains the correlation between each pair of variates after adjusting for all the other variates in the set. We calculated correlations between each pair of: percentage cover of graminoids (which included grasses, sedges and rushes), grass, Ulex, heather, bare ground and vegetation height. The resulting correlation coefficient of 0.97 between grass and graminoids indicated that the graminoid group was dominated by grasses, therefore 'graminoids' was excluded from analysis. Two general linear models were then run, with cover of grass, Ulex, heather, bare ground and vegetation height as predictive parameters and Dartford warbler and Nightjar food groups as a response variable (in two separate analyses).

| Species assemblages
In order to understand how treatments affected the species assemblage of communities arising, and how this changed across years, we used Canonical Correspondence Analysis (CCA) in the R package Vegan, a multivariate method that examines patterns of species occurrences across samples and relates them to measured explanatory variables. This allows us to understand if the balance of species abundances, as well as the species identity, varies according to the variables of interest. In this case, the explanatory variables were: (1) treatment (burned, swiped and baled); (2) age of each plot. The three blocks were used as co-variables and their effects partialled out before analyses of treatment effects.
Vegetation quadrats within plots were treated as split plots within the whole plots, so as not to overestimate the p-values of the analyses. Invertebrate samples were pooled within plots for the soil pit, small and large pitfall, and sweep net data.
The association of treatments and species present in the plots was tested using a permutation test for each explanatory variable using indicator species analysis (R package, indval). These tests showed which variables had a significant association with particular species in the plots. This approach has long been identified as useful for identifying indicator organisms in the monitoring of protected areas (Kremen, 1992). Analyses were carried out in R version R.2.14.0, using the packages Vegan (CCA) and indicspecies (indicator species analysis).

| Vegetation
Comparing burnt versus swiped sites ( Variables unaffected by management included vegetation height and percentage cover of bristle bent Agrostis curtisii (widespread over all sites), grasses, gorse and lichen (see Table S2).
Dwarf shrub cover, heather and lichen all increased over time, although by 20 years after management, there were signs of these decreasing: bare ground decreased over time (Table 4). No other variables responded to time since management significantly (Table S2).
Baled sites had a greater cover of dwarf shrubs including heathers (mean cover 45.9%) than either the burnt (39.8%) or swiped (24.9%) sites. Cover of heath grass, purple moor grass, sedges and rushes, forbs and moss on baled sites was also similar to that of the burnt sites. The only exception was bracken; mean cover was intermediate (6.1%) between burnt (3.5%) and swiped sites (12.9%).
Sixteen broadleaved plant species were found across all types of sites, of which heath milkwort, heath bedstraw, tormentil and sheep's sorrel were the most common. All are typical of mildly acidic heaths. Bramble was also widely distributed. Just seven species were exclusively recorded on burnt plots, including some of the typically wet heath species such as bog myrtle and oblong-leaved sundew. Round-leaved sundew was found on both burnt and swiped sites-which shared five additional species not occurring on baled sites: hawkweed, honeysuckle, lemon balm, common sorrel and field speedwell. Overall swiped sites were the most diverse with 39 species occurring exclusively, including many species typical of grassland or associated with waste ground. Only four species occurred exclusively on baled sites, while baled and swiped sites shared a further 10 species. Full details in Table S1.

| Invertebrates
Only the small heather weevil Micrelus ericae and Cicindela campestris (the green tiger beetle) were recorded in significantly higher numbers on burnt sites. M. ericae is a heather specialist and C. campestris is a characteristic heathland species (Table 5).
In general, insects were more abundant on swiped sites, in-  Table S2).
Although abundance of heathland specialists (as a group), money spiders, M. ericae and U. reticulata were rather low, all increased significantly as time elapsed after management. (Table 6).
The invertebrate prey items of two heathland specialist birds, the Dartford warbler and the Nightjar, were found on all sites of each management type and in all years of the chronosequence. The abundances of invertebrates making up Dartford warbler and Nightjar food group were found to be significantly higher on sites managed by cutting than those managed by burning. They also increased with time as it elapsed after management (Tables 5 and 6).

| The relationship between vegetation and invertebrate food items of Dartford warblers and Nightjars
Invertebrate prey associated with grassy areas avoided bare ground.

| Species assemblages
In all cases the management treatments significantly affected the species assemblages of both vegetation and invertebrates. While there was a consistent difference between burnt and swiped plots, the baled plots tended to have a more variable response to management revealed by the different sampling method (see Table 7). For taxa collected in soil pit samples and vegetation samples, the baled sites were significantly different in composition from both burnt and swiped sites but invertebrates collected from baled sites in small pitfalls and sweep nets were more similar to the burned sites than the swiped sites. Furthermore, invertebrates collected in large pitfall traps showed no difference between management (Table 7).
Time elapsed since management influenced species composition but revealed a different response across the sampling methods: above-ground (vegetation and sweep net samples) showed a significant association with time but below-ground or ground-level (pitfall trap) sampling did not (Table 7).
In all treatments, species which were most strongly associated with either burning or cutting were generally heathland specialists or grassland/arable specialists respectively (Table 8). In soil pits these were mostly earthworms (Table 8), while in the pitfall traps, they were predominantly ground beetles (Carabidae; Table 8).
Three of the beetle species found to be influential in the ordinations  (Joubert et al., 2016).
Baling heather appeared to lead to an intermediate position, but our confidence is reduced by the low replication in the study.
Furthermore, heather is baled with the aim of producing high quality material for restoration projects and paths around the forest-consequently the baled sites are selectively chosen for high heather cover and were probably in better habitat condition at the outset.
Currently 10% of the land that is managed annually in the NF is cut rather than burnt, and our work shows that while some cutting is beneficial for biodiversity, the impact of substantially increasing this could be negative for biodiversity. However, the CSM-through relying purely on vegetation characteristics to indicate condition and ignoring invertebrates and birds-does not reveal the important resources that cut habitats provide for heathland species at higher trophic levels.

TA B L E 6
Results of analyses of variance comparing invertebrate abundance assessed on a chronosequence of sites between 0 and 20 years since heathland management event in the New Forest. Abbreviations: Pitfalls (L), large pitfalls; Pitfalls (S), small pitfalls; ns = not significant; Sweeps = sweep net samples.

TA B L E 8
Invertebrate species from different sampling methods associated with management treatments. Nb, notable b species (national scarce species found in between 31 and 100 hectads. A hectad is an ordnance survey square of 1 ha).

| Management impact on heathland condition and vegetation
Heathland condition was strongly influenced by management practice. Prescribed burning delivered habitat more closely matching the criteria for good condition lowland heath according to CSM assessment. This is in agreement with early studies that investigated short-term regeneration of heath (Sedláková & Chytrý, 1999).
Burning encouraged dwarf shrubs, especially heathers, and also resulted in a marginally more open habitat. The CSM criteria indicate minimum 25% dwarf shrub cover as a target threshold. Burnt sites comfortably exceeded this (at 40%) whereas cut sites narrowly met it at 25%. However, neither burnt nor cut sites reached the 50%-75% required to meet conservation objectives. Bracken dominates in poor heathland condition and is a problem on lowland heathland. Swiping increased bracken cover, which, on average, crossed the 10% threshold at which the CSM considers it negative, whereas on burnt sites cover was maintained at an average of 3.5%. Grass cover, as a whole, did not differ between the management types but species responded differently; burning encouraged purple moor grass and cutting encouraged heath grass, both characteristic species of heathland habitats. However while heath grass cover was low (<5% irrespective of management), purple moor grass approached an average of 20% on burnt sites, potentially due to the post burn release of nutrients (Shelswell et al., 2011). Twenty percent is within the target guidelines. There is evidence that grazing could reduce this further and encourage greater ericoid cover (Newton et al., 2009). Appropriate grazing can also introduce additional vegetation structural diversity (Lake et al., 2001;Tallowin et al., 2005), although grazing impact is determined by stocking rates, species, breed and periods of grazing (Rosa García et al., 2013).
In this study, free ranging cattle, horses and deer had access to the areas throughout the year but their numbers were not recorded.
Swiping resulted in higher diversity and cover of broadleaved species including species usually associated with grassland and waste ground, not typical of heathland habitats and absent from the burnt sites. Overall, using vegetative indicators, the results suggest that, on balance, burning delivers better CSM habitat condition than vegetation swiping.

| Management impact on invertebrates
Invertebrate species assemblage composition differed between management treatments and, reflecting the vegetation data, baled sites were intermediate between burnt and swiped sites, with a tendency to be more similar to burnt sites. When examined individually, above ground, characteristic heathland invertebrate species were largely unaffected by different management techniques, although where differences were found, it was the swiped sites that supported greater abundance and generally invertebrate abundance was positively correlated with grass cover. It is worth noting that heath grasshopper, rare in the UK and understudied everywhere (Haes & Harding, 1997) was more abundant on the swiped sites.
The below ground invertebrates reveal a potential early warning that the heathland areas are in danger of deteriorating. Firstly, endogeic earthworms-horizontally burrowing species found only in areas with well-developed soil structure, such as pasture, arable land and neutral to base-rich woodlands-in the heathland plots suggests the presence of grassy patches that do not sustain good heathland.
Previous work has suggested that grassy areas are likely to be grazed heavily resulting in enrichment by dung which further improves conditions for earthworms (Carpenter et al., 2012). These processes are likely to encourage non-heathland species to grow which reflects what we observed in the swiped plots.

| Management impact on birds
We could not assess the extent to which birds were directly benefited by managed habitats, as birds operate at a larger spatial scale than the managed plots. Instead, we inferred the value of the different management techniques by calculating the abundance of the different species eaten by two insectivorous heathland specialists: the diurnal Dartford warbler and the crepuscular Nightjar. We found that the swiped sites provided a more abundant food source than burnt areas but both species nest on heathland, the Nightjar nesting on open ground (Langston et al., 2007) and the Dartford warbler nesting in tall heather or gorse bushes, on which it is known to be dependent (Tubbs, 1963;van den Berg et al., 2001), supporting the need for a mosaic of burnt and swiped patches.  (Burton et al., 2022). The beetles spend their larval stages in the soil and so soil conditions affect the numbers emerging. However, they may then undergo some redistribution influenced by above ground factors. Most of the invertebrates collected at or below ground were decomposers or predators of decomposers, with few herbivores thus explaining the low differences between treatments. In contrast, most sweep net species were herbivores or predators of herbivores-many with narrow food plant ranges. These species were strongly affected by vegetation change.

| Comparison of above-and below-ground response to management
These differences should be considered when assessing the conservation impact of changes in management policy because they may affect species of conservation concern directly or indirectly by impacting on their food supplies (as may occur with the Dartford warbler).
Management may also influence ecosystem functioning as indicated by the presence of endogeic earthworms in some of the heathland plots which may suggest more long-term changes are occurring.
Our results suggest that the current 20 year management cycle in the NF is appropriate. While heathland condition on burned sites was still good at 20 years, it had begun to decline on sites that were cut. The community assemblage results also suggest that the vegetation community as a whole shifts over time. The cover of ericaceous species (which are the dominant species on heathland) was declining 20 years after management (although this did not affect the aboveground invertebrates which continued to increase in abundance).
While our results are in agreement with the current rotation cycle, further work looking at longer time-frames would be useful. The size of management patches (from between <1-approximately 10 ha) was driven largely by pragmatic decisions in the NF (often proximity to local infrastructure and buildings). Although patch size was not investigated, given the good condition we observed on plots of all sizes, we would suggest that areas of up to 10 ha are acceptable and are likely to be rapidly recolonised by moderately mobile species.
The impact of patch size is worth further investigation.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information can be found online in the Supporting Information section at the end of this article.