Biological Conservation 191 (2015) 243–250

Contents lists available at ScienceDirect

Biological Conservation

j ourna l homepage: www.e lsev ie r .com/ locate /b ioc

Vegetation burning for game management in the UK uplands isincreasing and overlaps spatially with soil carbon and protected areas

David J.T. Douglas a,⁎, Graeme M. Buchanan a, Patrick Thompson b, Arjun Amar a,1,Debbie A. Fielding c, Steve M. Redpath d, Jeremy D. Wilson a

a RSPB Centre for Conservation Science, RSPB Scotland, 2 Lochside View, Edinburgh Park, Edinburgh EH12 4DH, UKb RSPB, 1 Sirius House, Amethyst Road, Newcastle Business Park, Newcastle-upon-Tyne NE4 7YL, UKc The James Hutton Institute, Craigiebuckler, Aberdeen AB15 8QH, UKd Institute of Biological and Environmental Sciences, Zoology Building, Tillydrone Avenue, Aberdeen AB24 2TZ, UK

⁎ Corresponding author.E-mail addresses: david.douglas@rspb.org.uk (D.J.T. Dou

graeme.buchanan@rspb.org.uk (G.M. Buchanan), patrick.tho(P. Thompson), arjun.amar@uct.ac.za (A. Amar), debbie.fields.redpath@abdn.ac.uk (S.M. Redpath), jeremy.wilson@rspb.o

1 Present address: Percy FitzPatrick Institute of AfricanExcellence, University of Cape Town, Rondebosch 7701, S

http://dx.doi.org/10.1016/j.biocon.2015.06.0140006-3207/© 2015 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 March 2015Received in revised form 28 May 2015Accepted 9 June 2015Available online xxxx

Keywords:Blanket bogCalluna vulgarisDesignated siteHeather moorlandPeatlandRed grouse

Burning for habitatmanagement is globally widespread. Burning over carbon-rich soils is a global environmentalconcern due to the potential contribution to climate change. In the UK, upland heath and blanket bog, so-called'moorland', often overlies carbon-rich soils, and has internationally important conservation value, but is burnedasmanagement for gamebird shooting and to a lesser extent for livestock grazing. There is little detailed informa-tion on the spatial extent or temporal trends in burning across theUK. This hinders formulation of policies for sus-tainablemanagement, given that the practice is potentially detrimental for soil carbon storage, water quality andhabitat condition. Using remotely sensed data, we mapped burning for gamebird management acrossc45000 km2 of the UK. Burning occurred across 8551 1-km squares, a third of the burned squares in Scotlandand England were on peat ≥0.5 m in depth, and the proportion of moorland burned within squares peaked atpeat depths of 1–2 m. Burning was detected within 55% of Special Areas of Conservation and 63% of Special Pro-tection Areas that were assessed, and the proportion of moorland burned was significantly higher inside sitesthan on comparable squares outside protected areas. The annual numbers of burns increased from 2001 to2011 irrespective of peat depth. The spatial overlap of burning with peat and protected areas and the increasingnumber of burns require urgent attention, for the development of policies for sustainable management and re-versal of damage to ecosystem services in the UK uplands.

© 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Prescribed burning of natural and semi-natural vegetation is anestablished management across the world. It can reduce wildfire risk(Boer et al., 2009), manage forest succession (Sah et al., 2006), enhancegrazing pasture (Augustine and Milchunas, 2009), manage game(Kilburg et al., 2015) and deliver nature conservation by mimickingnatural fire events to maintain desired habitat conditions (Williamset al., 2012). However, across a range of systems, there is increasingevidence of negative environmental impacts of burning, including soilerosion (Cawson et al., 2012), alteration of soil processes including nu-trient cycling and soil hydrology (Neary et al., 1999), water quality

glas),mpson@rspb.org.uking@hutton.ac.uk (D.A. Fielding),rg.uk (J.D. Wilson).Ornithology, DST/NRF Centre ofouth Africa.

(Battle and Gollday, 2003) air pollution (Tian et al., 2008) and habitatcondition and biodiversity impacts (Suárez and Medina, 2001). Theseimpacts must in some cases be evaluated against purported benefitsof burning, for example wildfire risk reduction, with the benefits anddisbenefits debated (Altangerel and Kulla, 2013).

Burning in the UK uplands for game management is a case wherebenefits and disbenefits need to be better understood (Grant et al.,2012; Sotherton et al., 2009). This burning is widely practised abovethe altitudinal limits of enclosed agriculture on upland heath andblanket bog dominated by ling heather Calluna vulgaris, for the endemicred grouse Lagopus lagopus scotica. On these ‘moorlands’, burning isundertaken in small strips recommended up to 2 ha and in rotationsof 10–25 years (Defra, 2007), tomaintain afine-scalemosaic of differentages which provide nesting cover and food resources for red grouse.Together with intensive predator control and use of medication toreduce grouse parasites and disease, thismanagement supports a globallyunique form of recreational shooting which depends on achieving highpost-breeding densities of red grouse (≥200 birds km−2, MoorlandAssociation, 2006). The area of heather moorland managed principallyfor grouse shooting has previously been imprecisely estimated at

244 D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

6600–17000 km2 (Bunce and Barr, 1988; Grant et al., 2012; Hudson,1992), representing 5–15% of the UK uplands, and about 20–60% ofheather dominated moorland, although not all of this may be subjectto rotational burning (Ball et al., 1983; Brown and Bainbridge, 1995).

Grouse management is therefore a major land use in the UK, butconcerns have been raised about the environmental impacts of burningpracticed for grouse shooting (Bain et al, 2011; Glaves et al, 2013).Manymoorlands are underlain by blanket peat soils, of which the UK holds10–15% of the global resource (Dunn and Freeman, 2011; Lindsayet al, 1988;Milne and Brown, 1997), storing 3.2 billion tonnes of carbon(Bain et al, 2011) — the largest terrestrial carbon reserve in the UK(Grayson et al, 2012). There is particular concern about the effectsof burning over deep peat (usually defined as ≥0.5 m depth; Bainet al., 2011) on peat hydrology, chemistry and physical properties.Where blanket bog is damaged by burning, impacts include alowered water table, breakdown of the active peat-forming structureand resulting long-term loss of the carbon store (Lindsay, 2010;Brown et al., 2014).

Moorlands are also of high international conservation value for theirvegetation, invertebrate and bird communities (Pearce-Higgins et al.,2009; Thompson et al., 1995) with large areas given legal protectionsfor nature conservation under the European Commission Habitats Di-rective (Special Areas of Conservation (SACs), European Commission,2015a) and Birds Directive (Special Protection Areas (SPAs), EuropeanCommission, 2015b). Burning alters vegetation composition and struc-ture and, where fire temperatures are high and rotation lengths areshort, it may be damaging to fire-sensitive, peat-forming species such asSphagnum spp. mosses (Brown et al, 2014). These problems are cited asa primary reason for poor condition of upland sites designated for theirconservation value, contributing to the reasons for “unfavourable” condi-tion in 53% of the total area in England (Natural England, 2008) and on87% of “unfavourable” upland bog features in Scotland (SNH, 2010).

Finally, upland areas provide N70% of the UK's water supply (Wattset al., 2001), however water colouration and the occurrence of particu-late and dissolved organic carbon (POC and DOC) from peat soils inupland areas are major factors affecting water quality, incurring highfinancial costs for treatment (Grayson et al., 2012). Growing evidencelinks burning over peat to a range of impacts on water quality includingdiscolouration (White et al., 2007), lower pH and higher DOC content(Brown et al., 2014; Clay et al., 2012). Furthermore, riverine inverte-brate diversity may be lower within burned catchments (Brown et al.,2013).

Before policies for sustainable resource management can be devel-oped, there is a need for a detailed knowledge of the extent of burningin the UK, and how this overlaps with areas of high concern (e.g. deeppeat and areas designated under EU conservation directives). Despitethe widespread use of burning for gamebird management in the UK,and increasing evidence of its wider environmental impacts, objectivereporting of the spatial extent of moorland burning, and variation inspatial and temporal patterns, is currently lacking except at coarse(10-km square, Anderson et al., 2009) resolutions or in localised studies(Amar et al., 2012; Hester and Sydes, 1992; Yallop et al., 2006). Howev-er, the characteristic pattern of strip burning is readily visible on highresolution aerial and satellite images (Fig. A1), allowing fine-scalerepeatable mapping over large areas, whilst the heat generated byactive fires is detectable by fire monitoring satellites. We use thesedata to map spatial and temporal patterns in strip burning acrossthree countries (Scotland, England and Wales) of the UK at a 1-kmresolution. We use high resolution aerial and satellite imagery (hence-forth images or imagery) to assess the extent to which burning takesplace over deep peat soils and in protected areas designated for theirconservation value. From satellite-derived thermal anomaly datawe measure trends in the annual number of burns from 2001 to2011 nationally, and in relation to peat depth.

For the conservation reasons above,we focus on strip burning for redgrouse in the UK, however we demonstrate the utility of combining

remotely sensed burning data with wide scale maps of environmentalcorrelates, which is widely applicable where land use is associatedwith environmental impacts (Willis, 2015).

2. Methods

2.1. Mapping burning extent

Areas assessed were based on Anderson et al. (2009), who visuallyestimated the extent of strip burning for red grousemanagementwithina single image of each of 2707 10-km British National Grid squares(Ordnance Survey, 2014) covering the mainland UK (Scotland, EnglandandWales).We selected all 10-kmsquareswith burning present and 60additional 10-km squares that Anderson et al. (2009) did not assess dueto cloud-obscured imagery. From the resulting 488 10-km squares, weused a single image of each constituent 1-km British National Gridfrom Getmapping (2011) (n = 46 239 squares, aerial photography) orGoogle Earth (2011) (n = 863 images, satellite images); the latterwas obtained for a separate study (Newey et al., 2012). We excludedsquares with less than 5% land or inadequate imagery (cloud cover orblurred), and assessed a total of 47102 images. For all Getmapping im-ages, and most Google Earth images, the image year was known. Forsome Google Earth images we had not recorded the exact year butcould assign images to one of two years; 2001, b0.1%, 2002, 0.1%,2003, b0.1%, 2004, 7.3%; 2005, 27.5%; 2005 or 2006, 0.1%, 2005 or2008, 0.1%, 2006, 11.0%; 2007, 47.0%; 2007 or 2008, b0.1; 2008, 6.7%;2009, 0.2%; and 2010, b0.1%.

Getmapping images were viewed in ArcGIS 9.3 (Esri, 2009).Shapefiles of 1-km squares to be viewed in Google Earth were firstopened in MapWindowGIS (2011), converted to KML files usingthe plug-in “Shape2Earth” before being opened in Google Earth. Thepercentage of each square comprising moorland and the percentage ofthe moorland with strip burning was visually assessed (Fig. A1).Between-observer variation in image scoring was minimal (AppendixA). Both measures were estimated to the nearest 5%, including an addi-tional category of “1%” for the area of moorland burned, to identifysquares with very limited burning where rounding to zero wouldhave incorrectly classified the square as “no burning”, but rounding to5% would have over-represented burning extent. The percentage ofmoorland multiplied by the percentage of moorland burned gave anestimate of the percentage of the square burned. Using the colour toneof vegetation types on images, moorland was defined as an unenclosedhabitat dominated by ericaceous dwarf shrubs (typically ling heatherbut also bell heather Erica cinerea, cross-leaved heath Erica tetralix andcrowberry Empetrum nigrum). Our assessment of moorland area willalso include some areas of grassland and bracken Pteridium aquilinumcover where these occur in mixes with dwarf shrub heath, but onlywhere dwarf shrub heath dominates. Burning of grass-dominatedmoorland also occurs for grazingmanagement, but generally compriseslarger burn patches, is less tightly rotational and is a much less signifi-cant feature of upland landscapes nationally than strip burning for redgrouse (Yallop et al., 2006). Therefore, whilst some of this burningmay be included within the areas assessed, in general we ascribe burn-ing within our study to that undertaken by gamekeepers for grousemanagement.Mechanised cutting ofmoorlandmay occur as an alterna-tive to burning, andmay also be detectable on imagery.Whilst wewereunable to quantify the relative extent of cutting and burning, cutting canonly take place where machinery can be deployed (not on wet, steep orrocky ground (Defra, 2007)), and we expect any cutting to compriseonly a very limited area relative to that burned.

Post-burning re-growth of ling heather typically occurs in fourstages over subsequent years, which are distinguishable on aerialimages (Yallop et al., 2006) from newly burned (stage 1) to degeneratewith no remaining visible burn scar (stage 4). Observers scoring imagesindependently confirmed that stages 1–3 were distinguishable fromsurrounding unburned vegetation (consistent with Yallop et al., 2006)

245D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

and together constituted the estimate of burning extent in a 1-kmsquare.We therefore assume that all burning detectable in these imageswas conducted within a maximum of c25 years of the image date(Yallop et al., 2006), although post-burning regeneration rates varyaccording to environmental factors including soil wetness (Yallopet al., 2009).

2.2. Trends in burning frequency

Temporal burning patterns were extracted from MODIS ThermalAnomalies data (MODIS, 2013). Sensors on the Aqua and Terra satellitesdetect ground surface temperatures twice daily, identifying events whenthe fire strength is sufficient to be detected relative to its background (toaccount for variability of the surface temperature and reflection bysunlight). Daily MODIS data for Scotland, England and Wales weredownloaded in November 2013 for 2001–2011. After re-projection toBritish National Grid (Ordnance Survey, 2014), thermal anomalies withprobability of ≥8 (Giglio, 2013) were defined as fires, probably excludingsome fires, but reducing false detection rates of non-fire thermal anoma-lies. We included burns within the fully permissible annual period forburning (1 October–15 April in England and Scotland (Defra, 2007;Scottish Government, 2011)) and 1 October–31 March in Wales (WelshAssembly Government, 2008) and these data were summarised as theannual number of burns per 1-km square. Thermal anomalies will onlybe detected if they are sufficiently hotwhen the satellite passes overhead.Similarly, fires will not be detected on cloudy days as the ground will beobscured from the sensor. Consequently, an unknown proportion offires remain undetected, and our data are considered a minimum value.

2.3. Environmental data

Mapped peat depth data, produced using a combination of mea-sured and extrapolated peat depths, were available separately for thewhole of England and Scotland, but not Wales. We derived a value ofmean peat depth per 1-km square. The dataset for England (NaturalEngland, 2012) comprised 25 m raster data with a peat depth valueper pixel, derived from two data layers describing peat types aboveand below 150 m elevation, explaining 28.4% and 37.8% of variationsin peat depth respectively (Penny Anderson Associated Ltd, 2013).Data for Scotland were supplied by the James Hutton Institute, derivedfrom a 1:250000 scale soil map, comprising polygons of a mean area4.3 km2, each with a value of average peat depth, with overall 7.2%error on peat depth estimates (Chapman et al., 2009). The proportionalarea of each polygon within each 1-km square was calculated toproduce a weighted average of peat depth per square.

We classified squares into three categories of average peat depth; nopeat = 0 m (squares overlying non-peat soils), shallow peat of b0.5 mor deep peat of ≥0.5 m (Bain et al., 2011) (Fig. A2). Mean altitude andgradient per 1-km grid square were calculated using a digital terrainmodel with a 50-m grid of points.

2.4. Burning in protected areas

Special Areas of Conservation (SACs) comprise high-quality conser-vation sites aiming to provide increased protection to a variety of wildanimals, plants and habitats (JNCC, 2014a), whilst Special ProtectionAreas (SPAs) are considered to be of international importance for arange of rare and vulnerable bird species and regularly occurringmigra-tory species (JNCC, 2014b). Using GIS we identified all 1-km squaresassessed for burning lying wholly within one of these protected areatypes; squares overlapping site boundaries were excluded as it wasnot possible to easily define how much of the moorland or burning laywithin the site boundary. Thus, of the sites within the area assessedfor burning, we achieved coverage of 40.1% (range 10.0–75.3% persite) of the area of these SACs and 43.3% (range 10.1–65.4% per site) ofthe area of these SPAs (Tables A2 and A3). The areas burned per site

should therefore be considered minima. Where SACs and SPAs over-lapped these were treated as separate sites for clarity.

2.5. Analysis

All analyses were conducted in R 3.0.2 (R Core Team, 2013) unlessotherwise stated.

2.5.1. Spatial correlation between burning and environmental correlatesUsing the burning dataset from imagery, we examined the relation-

ship between the percentage of moorland burned per 1-km square andpeat depth, across England and Scotland (with peat data for all squares),removing squares with no moorland (yielding 32905 squares). Theresponse variable was the percentage burning measure, rescaled to aproportion and logit-transformed (thereby normalising it) after adding0.01 to all values (Warton and Hui, 2011). The key continuous explana-tory variable was peat depth, with additional covariates of altitude (m),gradient (degrees), image year (averaged where an image was fromone of two possible years), easting and northing, all fitted as linearand quadratic terms to test for potential non-linear relationships.

The proportion of moorland burned per square might show spatialautocorrelation, especially within individual estates subject to similarmanagement practices. Estate boundaries were unavailable so weexamined whether such autocorrelation occurs, and over what scale,by first fitting a non-spatial GLM with response and explanatory vari-ables as described above, normal errors and identity link. We examinedspatial autocorrelation in the residuals using Moran's I in SAS 9.4 (SAS,2012), at increasing increments of 1-km lags between squares, to iden-tify at what scale Moran's declined to 0.1, when spatial autocorrelationeffects in model residuals are unlikely to compromise inference (Ryanet al., 2004; Kraan et al., 2009). This occurred at a 20-km lag and wecreated a factor grouping squares within 20 × 20-km blocks along OSgrid lines. We fitted this as a random factor, thereby assuming spatiallynon-independent errors at a 20-km scale, in a Generalised LinearMixedModel (GLMM) with normal errors, identity link, and response variableand explanatory variables as above.

2.5.2. Spatial overlap between burning and protected areasGlobally, protected areas are not randomly distributed with respect

to key environmental characteristics that might relate to theireffectiveness (Joppa and Pfaff, 2009). Therefore, before we testedwhether the percentage of moorland burned per 1-km square dif-fered inside and outside protected sites, we controlled for potentiallyconfounding effects. Covariates potentially associated with the percent-age area burned (peat depth, area of moorland, altitude, gradient, imageyear, easting and northing) differed markedly inside and outsideprotected areas (Table A1). We therefore selected squares inside andoutside protected areas in England and Scotland, withmoorland presentandwith similar distributions in the covariates, following an establishedmatchingmethod (Andam et al., 2008).We used the ‘matchit’ procedurein the R package MatchIt, selecting squares for both SPAs and SACs, usingnearest-neighbour covariate matching and a caliper of 0.5 standarddeviations of each covariate (Andam et al., 2008). We compared the per-centage of moorland burned per square within and outside SPAs andSACs, respectively, using two-sample unpaired t-tests applied tothe matched samples of squares, where the response variable waslogit-transformed to normalise (Warton and Hui, 2011).

2.5.3. National trends in burning frequency and its environmental correlatesFrom the MODIS data, we extracted all 1-km squares assessed for

burning that containedmoorland.We summed the total annual numberof burns across all squares from 2001 to 2011. These data showedextremely high counts in 2003 and 2007; a generalised ExtremeStudentized Deviate test, for detecting one or more outliers (Rosner,1983) revealed that these years were statistically significant outliers.MODIS cannot distinguish between wildfires or prescribed burns, and

246 D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

with 2003 in particular having a high incidence of UKwildfires, includingsome spring fires within the permissible burning period (McMorrow,2011), 2003 and 2007 were removed. Whilst some wildfires may beincluded in the remaining years, by restricting our study to known areasof red grouse management at the 10-km square scale, this risk isminimised. We tested for a trend in the annual number of burns, fittinga Generalised LinearModel (GLM)with quasi-Poisson errors (accountingfor overdispersion where residual deviance/residual d.f. N 2 (Lindsey,1999)) and log link, with total annual burns as the response variableand year as a continuous explanatory variable.We conducted two furthertests by modifying this model, first testing whether trends differedbetween England, Scotland andWales, using quasi-Poisson errors, fittingcountry as a three-level factor and testing a year ∗ country interaction.Secondly, we tested whether trends in annual burns differed betweensquares in the three peat depth categories, in a model using quasi-Poisson errors, using all squares in England and Scotland, fitting peatdepth as a three-level factor and testing a year ∗ peat depth interaction.Trends inside and outside of protected areas were not analysed due tovery small numbers of burns recorded byMODIS in thematched samplesof squares (b7 burn events annually, inside or outside, across all squaresper site type).

As the number of burns detected by MODIS could be influenced bycloud cover, we tested for a trend in the annual number of sunshinehours, using publicly available data from four weather stations situated≤10 km from areas assessed for burning, with complete sunshine datafrom 2001 to 2011 (Met Office, 2014). We included data from monthsoverlapping with the permissible burning period (October to Aprilinclusive), summing total annual sunshine hours across stations andfitting this as the response variable in a GLMwith quasi-Poisson errors,log link and year as a continuous explanatory variable.

Fig. 1. (a) Extent ofmoorlandburningwithin Scotland, England andWales.Map shows the percesquare obtained from 2001 to 2010; (b) extent of moorland burning within 1-km squares clasoverlying deep peat and with burning present.

3. Results

3.1. Mapping burning extent

Strip burning was detected in 8551 1-km squares, with 5245 inScotland, 3139 in England and 167 in Wales (Fig. 1a). Within squares,the percentage area of moorland burned varied from 1 to 95% (meanacross squares with burning 16.7 ± 0.2%; England 19.0 ± 0.3; Scotland15.7 ± 0.2%; Wales 5.9 ± 0.6%).

3.2. Burning and peat depth

A third (33.5%) of burned 1-km squares (43.6% in England and 27.6%in Scotland) were classified as overlying deep peat (Fig. 1b). Of the totalarea burned, approximately 466 km2 (39%) occurred in squares classi-fied as deep peat, c278 km2 (60% of burned area) in England andc188 km2 (40% of burned area) in Scotland.

The percentage area of moorland burned within squares was corre-lated with peat depth, peaking at approximately 1–2 m peat depth(Fig. 2a) and also increased with increasing area of moorland within asquare (Fig. 2b), peaking at 400–600 m altitude (Fig. 2c) and peakingin squares with a mean gradient about 10o (Fig. 2d).

3.3. Burning in protected areas

Burning was significantly greater inside protected areas than inmatched areas that were not protected (Fig. 3), and burning was wide-spread across protected areas (Fig. 4). Burning was detected in 26 of 47(55%) SACs assessed (Table A2), with the percentage of total site areaburned ranging from b0.1 to 11.5% per site. The North Pennine Moors

ntage area ofmoorland burnedwithin 1-kmgrid squares, assessed from a single imagepersified as deep peat (averaging ≥ 0.5 m) in England and Scotland. Shown are only squares

Fig. 2. Fitted GLMM relationships between the percentage areas of moorland burned, assessed from images of 1-km grid squares, and environmental predictors. There were significantquadratic relationships between the area ofmoorland burned and peat depth (χ21= 15.79, P b 0.001),moorland area (χ21= 22.33, P b 0.001), altitude (χ21= 469.25, P b 0.001), gradient(χ21 = 61.90, P b 0.001) and image year (χ21 = 39.86, P b 0.001), and non-significant relationships with easting (χ21 = 1.85, P = 0.174) and northing (χ21 = 3.30, P = 0.069).

247D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

SAC in the north of England was notable in having high relative andabsolute levels of burning, with the 118.2 km2 burned representing51.7% of all assessed SAC burning (despite accounting for only 16.5% ofassessed SAC area).

Burning was detected in 20 of 32 (63%) SPAs assessed (Table A3),with b0.1–10.8% of site areas burned. The North Pennine Moors SPA(overlapping partly with the North PennineMoors SAC) had the largestarea burned, both as a percentage of site area and also total area burned,with 158.4 km2 burned representing 59.5% of the total area of burningdetected within all SPAs assessed, but only 23.6% by total area.

3.4. National trends in burning frequency

The annual number of burns from 2001 to 2011 increased signifi-cantly overall across Scotland, England and Wales at a rate of 11% peryear (0.106 ± 0.042, χ21 = 64.77, P = 0.011), with an acceleratingtrend in more recent years (Fig. 5). This was unlikely to be related todeclining cloud cover and improved burn detection, as therewas no sig-nificant trend in sunshine hours over this period (−0.012 ± 0.011,χ21 = 31.91, P = 0.285). There was no significant difference in annualburn trends between the three countries (χ22 = 0.77, P = 0.950). InEngland and Scotland, where we had country-wide peat depth data,there was a significant overall increase in annual burn trends(0.105 ± 0.029, χ21 = 62.64, P b 0.001) but the rate of increase didnot differ between squares overlying zero peat, shallow peat or deeppeat (χ21 = 0.75, P = 0.926).

Inside Outside

Mea

n ar

ea o

f moo

rland

bur

ned

(%)

0

2

4

6

8

10

Fig. 3. Area of moorland burning per 1-km square inside and outside of two categories ofprotected areas, Special Areas of Conservation (SACs; black bars) and Special ProtectionAreas (SPAs; grey bars). Analyses compared samples of squares inside and outside sites,matched by underlying environmental and topographical covariates (SAC, n = 2095 in-side and outside, SPA n=2238 inside and outside). Differences in burning inside and out-side sites; SACs, t4025 = 8.64, P b 0.001; and SPAs, t4370 = 8.85, P b 0.001).

4. Discussion

This study provides the first fine-scale assessment of the extent,trends and environmental correlates of burning for recreational shoot-ing of red grouse in the UK uplands. Burning occurred across 8551 1-km squares and based on typical vegetation regeneration rates weassume that this burning took place within the last c25 years (Yallopet al., 2006). This estimate of the extent of moorland burning is withinthe range of previous estimates of the area under management for redgrouse of 6600–17000 km2 (Bunce and Barr, 1988; Grant et al., 2012;Hudson, 1992), albeit towards the lower end; this may be becauseother activities such as predator control may extend beyond the areaburned and may be included in estimates of the total area under grousemanagement.

The annual number of burning events has increased significantlyfrom 2001 to 2011 at a rate of c11% per annum. These national trendsin burning are consistent with smaller scale studies such as in northernEngland where Yallop et al. (2006) reported markedly increasing burnevents,with the percentage of the areanewly burned (b4 years) doublingfrom 15 to 30% between the 1970s and 2000, and declining repeat burn-ing times to squares from 20 to 16 years. Our results suggest that in-creases in the annual number of burns are widespread across theuplands of themainlandUK. These increasesmay be due to general inten-sification of management for red grouse; for example, the numbers ofgamekeepers in employment, and the potential number of ‘shootingdays’ that could be sold, increased in northern England from 2001 to2009 (Natural England, 2009).

Set against annual increases in burning, aerial and satellite imageryused to estimate burning may rapidly require updating, or the trueextent of burning may be underestimated. In fact, our study is likely tobe a conservative estimate since N90% of data used to map burningare derived from images pre-dating 2008, and therefore before muchof the recent large increase in the number of burns was detected.

This increase occurred irrespective of peat depth, so that burning isnow widespread over deep (≥0.5 m) peat in England and Scotland,especially in England, where 44% of all squares with burning overliedeep peat, and where 60% of the measured area of burning acrossthese two countries over deep peat occurred. This result is set in thecontext of increasing global concern surrounding burning on carbon-rich soils, and its impacts on carbon storage and emissions, where burn-ing occurs across temperate-boreal and tropical peatlands (Goldammer,2013; Hoscilo et al., 2011; Tansey et al., 2008).

A growing evidence base highlights negative impacts of burning ondeep peat for a range of ecosystem services including drinking waterquality (Glaves et al., 2013; White et al., 2007), for example throughelevated DOC contents (Clay et al., 2012; Holden et al., 2012). There isalso evidence of both negative (Garnett et al., 2000; Worrall et al,2010) and positive (Clay et al., 2010) impacts of burning on carbon stor-age and further research is needed on the impacts on overall carbon

Fig. 4. Percentage area of site burned within two categories of protected areas; (a) Special Areas of Conservation (SACs); and (b) Special Protection Areas (SPAs).

248 D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

budgets. Much of this evidence has been obtained from studies at alimited number of sites, but our study provides a broader picture ofwhere burning is practised over deep peat, notably northern Englandand the central Highlands and far north of Scotland (Fig. 1b). Theimpacts of burning on carbon budgets may not have previously beenaccounted for in national emission statistics (UK Government, 2015).

The significantly greater percentage area of burning inside protectedareas than matched areas outside these sites is set in the context ofincreasing concern surrounding the condition of upland protected

2000 2002 2004 2006Year

2008 2010 2012

Num

ber

of b

urns

0

20

40

60

80

100

120

Fig. 5. Trend in the annual number of moorland burns recorded in Scotland, England andWales, using MODIS satellite recording of thermal events. Plot shows raw data with fittedline from GLM.

areas. Special Areas of Conservation and Special Protection Areasare designated for the conservation value of their habitats andbirds, however in the uplands many of these sites are failing toreach standards of condition required as part of their designation(Adaptation Sub-Committee, 2013). Inappropriate burning is amajor contributor to the poor (“unfavourable”) condition of blanketbog in upland protected areas in England (Natural England, 2008)and Scotland (SNH, 2010) and our results provide strong evidencethat a greater percentage of moorland is burned within these sitesthan outside sites.

The UK government has agreed to effectively protect more than 17%of its land by 2020 (Convention on Biological Diversity, 2010). Theimpact of burning within SPAs and SACs, together with the evidenceof an increase in burning, suggests that current moorland managementis not making a sufficient contribution to this target; indeed it may bedetracting from it. Our results highlight the high overlap of this inten-sive burning management with environmental concerns such as soilcarbon and protected areas. The semi-natural moorlands of the UKuplands are of high international biodiversity conservation value(Pearce-Higgins et al., 2009; Thompson et al., 1995) and store importantcarbon resources (Bain et al., 2011). Whilst burning may in somecircumstances enhance plant species diversity (Harris et al, 2011) orsupport carbon storage (Clay et al, 2010), a wide range of negativeenvironmental impacts of burning are now increasinglywell document-ed (Brown et al., 2014). Policies to reverse these damaging effects, forexample over deep peat, must be implemented as a matter of urgency.These should includemanagement agreements between landmanagersand statutory agencies to halt damaging burning practices, and tighterenforcement of these agreements.

Detailed remote mapping of the extent of widespread land usessuch as burning has other uses. It could be used to improve our under-standing of the impacts of burning on the distribution of species associ-ated with these upland areas; game management, including burning, is

249D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

known to be associated with the distribution and abundance of uplandbirds (e.g. Tharme et al., 2001; Douglas et al., 2014). National mappingwill allow for a greater study of these effects, updating previous coarseresolution assessments of not just burning (Anderson et al., 2009) butalso the interaction between grousemanagement and bird distributions(Gibbons et al., 1995).

Acknowledgements

We thank Maider Guiu, Trevor Smith, Tessa Cole, Jonathan Groomand Emma Teuten for data collection and map production. We thankNatural England (especially Matthew Shepherd) and James HuttonInstitute (especially Steve Chapman) for provision of peat depth dataand advice on interpretation. Data collection was funded by RSPB andJames Hutton Institute, the latter supported by the European Commis-sion under the 7th Framework Programme for Research and Technolog-ical Development through project HUNT (212160, FP7-ENV-2007-1).Analyses were funded by RSPB.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.biocon.2015.06.014.

References

Adaptation Sub-Committee, 2013. Managing the land in a changing climate. AdaptationSub-Committee Progress Report 2013. Committee on Climate Change, London.

Altangerel, K., Kulla, C.A., 2013. The prescribed burning debate in Australia: conflicts andcompatibilities. J. Environ. Plan. Manag. 56, 103–120.

Amar, A., Court, I.R., Davison, M., Downing, S., Grimshaw, T., Pickford, T., Raw, D., 2012.Linking nest histories, remotely sensed land use data and wildlife crime records toexplore the impact of grouse moor management on peregrine falcon populations.Biol. Conserv. 145, 86–94.

Andam, K.S., Ferraro, P.J., Pfaff, A., Sanchez-Azofeifa, G.A., Robalino, J.A., 2008. Measuringthe effectiveness of protected area networks in reducing deforestation. Proc. Natl.Acad. Sci. 105, 16089–16094.

Anderson, B.J., Arroyo, B.E., Collingham, Y.C., Etheridge, B., Fernandez-De-Simon, J.,Gillings, S., Gregory, R.D., Leckie, F.M., Sim, I.M.W., Thomas, C.D., Travis, J., Redpath,S.M., 2009. Using distribution models to test alternative hypotheses about a species'environmental limits and recovery prospects. Biol. Conserv. 142, 488–499.

Augustine, D.J., Milchunas, D.G., 2009. Vegetation responses to prescribed burning ofgrazed shortgrass steppe. Rangel. Ecol. Manag. 62, 89–97.

Bain, C.G., Bonn, A., Stoneman, R., Chapman, S., Coupar, A., Evans, M., Gearey, B., Howat,M., Joosten, H., Keenleyside, C., Labadz, J., Lindsay, R., Littlewood, N., Lunt, P., Miller,C.J., Moxey, A., Orr, H., Reed, M., Smith, P., Swales, V., Thompson, D.B.A., Thompson,P.S., Van de Noort, R., Wilson, J.D., Worrall, F., 2011. IUCN UK Commission of Inquiryon Peatlands. IUCN UK Peatland Programme, Edinburgh.

Ball, D.F., Radford, G.L., Williams, W.M., 1983. A land characteristic data bank for GreatBritain. Institute for Ecology Occasional Paper 13. Institute for Ecology, Bangor.

Battle, J., Gollday, S.W., 2003. Prescribed fire's impact on water quality of depressionalwetlands in southwestern Georgia. Am. Midl. Nat. 150, 15–25.

Boer, M.M., Sadler, R.J., Wittkuhn, R., McCaw, L., Grierson, P.F., 2009. Long-term impacts ofprescribed burning on regional extent and incidence of wildfires — evidence fromfifty years of active fire management in SW Australian forests. For. Ecol. Manag.259, 132–142.

Brown, A.F., Bainbridge, I.P., 1995. Grouse moors and upland breeding birds. In:Thompson, D.B.A., Hester, A.J., Usher, M.B. (Eds.), Heaths and Moorlands: CulturalLandscapes. HMSO, Edinburgh, pp. 51–66.

Brown, L.E., Johnston, K., Palmer, S.M., Aspray, K.L., Holden, J., 2013. River ecosystemresponse to prescribed vegetation burning on blanket peatland. PLoS ONE 8, e81023.http://dx.doi.org/10.1371/journal.pone.0081023.

Brown, L.E., Holden, J., Palmer, S.M., 2014. Effects of Moorland Burning on theEcohydrology of River Basins. Key Findings From the EMBER Project. University ofLeeds, Leeds.

Bunce, R.G.H., Barr, C.J., 1988. The extent of land under different management regimes inthe uplands and the potential for change. In: Usher, M.B., Thompson, D.B.A. (Eds.),Ecological Change in the Uplands. Blackwell, Oxford.

Cawson, J.G., Sheridan, G.J., Smith, H.G., Lane, P.N.J., 2012. Surface runoff and erosion afterprescribed burning and the effect of different fire regimes in forests and shrublands: areview. Int. J. Wildland Fire 21, 857–872.

Chapman, S.J., Bell, J., Donnelly, D., Lilly, A., 2009. Carbon stocks in Scottish peatlands. SoilUse Manag. 25, 105–112.

Clay, G.D., Worrall, F. And, Rose, R., 2010. Carbon budgets of an upland blanket bogmanaged by prescribed fire. J. Geophys. Res. 115, G04037. http://dx.doi.org/10.1029/2010JG001331.

Clay, G.D., Worrall, F., Aebischer, N.J., 2012. Does prescribed burning on peat soilsinfluence DOC concentrations in soil and runoff waters? Results from a 10 yearchronosequence. J. Hydrol. 448, 139–148.

Convention on Biological Diversity, 2010. 2010 Biodiversity Target. http://www.cbd.int/2010-target (Last accessed 7/3/15).

Defra, 2007. The Heather and Grass Burning Code: 2007 Version. Department forEnvironment, Food and Rural Affairs, London.

Douglas, D.J.T., Bellamy, P.E., Stephen, L.S., Pearce-Higgins, J.W., Wilson, J.D., Grant, M.C.,2014. Upland land use predicts population decline in a globally near-threatenedwader. J. Appl. Ecol. 51, 194–203.

Dunn, C., Freeman, C., 2011. Peatlands: our greatest source of carbon credits? CarbonManage. 2, 289–301.

Esri, 2009. ArcGIS 9.3. Esri, Redlands, California, United States.European Commission, 2015a. Habitats Directive. http://ec.europa.eu/environment/

nature/legislation/habitatsdirective/index_en.htm (Last accessed 7/3/15).European Commission, 2015b. Birds Directive. http://ec.europa.eu/environment/nature/

legislation/birdsdirective/index_en.htm (Last accessed 7/3/15).Garnett, M.H., Ineson, P., Stevenson, A.C., 2000. Effects of burning and grazing on carbon

sequestration in a Pennine blanket bog, UK. The Holocene 10, 729–736.Getmapping, 2011. Getmapping. Hartley Wintney, Hampshire, UK.Gibbons, D., Gates, S., Green, R., Fuller, R.J., Fuller, R.M., 1995. Buzzard Buteo buteo and

ravens Corvus corax in the uplands of Britain: limits to distribution and abundance.Ibis 137, S75–S84.

Giglio, l., 2013. MODIS Collection 5 Active Fire Product User's Guide, version 2.5. http://modis-fire.umd.edu/Documents/MODIS_Fire_Users_Guide_2.5.pdf.

Glaves, D.J., Morecroft, M., Fitzgibbon, C., Lepitt, P., Owen, M., Phillips, S., 2013. The Ef-fects of Managed Burning on Upland Peatland Biodiversity, Carbon and Water.Nat. Engl. Evid. Rev. (Number 004).

Goldammer, J.G. (Ed.), 2013. Prescribed Burning in Russia and Neighbouring Temperate-Boreal Eurasia. Kessel Publishing House, Remagen-Oberwinter, Germany.

Google Earth, 2011. www.google.com/earth.Grant, M.C., Mallord, J., Stephen, L., Thompson, P.S., 2012. The costs and benefits of grouse

moor management to biodiversity and aspects of the wider environment: a review.RSPB Res. Rep. 43.

Grayson, R., Kay, P., Foulger, M., Gledhill, S., 2012. A GIS based MCE model for identifyingwater colour generation potential in UK upland drinking water supply catchments.J. Hydrol. 420–421, 37–45.

Harris, M.P.K., Allen, K.A., McAllister, H.A., Eyre, G., Le Duc, M.G., Marrs, R.H., 2011. Factorsaffecting moorland plant communities and component species in relation to prescribedburning. J. Appl. Ecol. 48, 1411–1421.

Hester, A.J., Sydes, C., 1992. Changes in burning of Scottish heather moorland since the1940s from aerial photographs. Biol. Conserv. 60, 25–30.

Holden, J., Chapman, P.J., Palmer, S.M., Kay, P., Grayson, R., 2012. The impacts of prescribedmoorland burning on water colour and dissolved organic carbon: a critical synthesis.J. Environ. Manag. 101, 92–103.

Hoscilo, A., Page, S.E., Tansey, K.J., Rieley, J.O., 2011. Effect of repeated fires on land-coverchange on peatland in southern Central Kalimantan, Indonesia, from 1973 to 2005.Int. J. Wildland Fire 20, 578–588.

Hudson, P.J., 1992. Grouse in Space and Time. Game Conservancy Ltd., Fordingbridge.JNCC, 2014a. Special Areas of Conservation (SACs). http://jncc.defra.gov.uk/page-23

(Last accessed 7/3/15).JNCC, 2014b. Special Protection Areas (SPAs). http://jncc.defra.gov.uk/page-162

(Last accessed 7/3/15).Joppa, L.N., Pfaff, A., 2009. High and far: biases in the location of protected areas. PLoS ONE

4, e8273.Kilburg, E.L., Moorman, C.E., DePerno, C.S., Cobb, D., Harper, C.A., 2015. Wild turkey

prenesting-resource selection in a landscape managed with frequent prescribedfire. Southeast. Nat. 14, 137–146.

Kraan, C., van der Meer, J., Dekinga, A., Piersma, T., 2009. Patchiness of macrobenthicinvertebrates in homogenized intertidal habitats: hidden spatial structure at a landscapescale. Mar. Ecol. Prog. Ser. 383, 211–224.

Lindsay, R. (2010) Peatbogs and carbon: a critical synthesis. Unpublished report to RSPBScotland.

Lindsay, R.M., Charman, D.J., Everingham, F., O'Reilly, R.M., Palmer, M.A., Rowell, T.A.,Stroud, D.A., 1988. The flow country. The Peatlands of Caithness & Sutherland. NatureConservancy Council, Peterborough, UK.

Lindsey, J.K., 1999. On the use of corrections for overdispersion. J. R. Stat. Soc.: Ser. C: Appl.Stat. 48, 553–561.

MapWindowGIS, 2011. www.mapwindow.org (Last accessed 11/3/15).McMorrow, J., 2011. Wildfire in the United Kingdom: status and key issues. In:

McCaffrey, S., LeBlanc Fisher, C. (Eds.), Second International Association of Wild-land Fire Conference on Human Dimensions of Wildland Fire 27 April 2010–29April 2010, 44–56. International Association of Wildland Fire, San Antonio, Texas(eScholarID:96512).

Met Office, 2014. http://www.metoffice.gov.uk/public/weather/climate-historic/#?tab=climateHistoric (Last accessed 11/3/15).

Milne, R., Brown, T.A., 1997. Carbon in the vegetation and soils of Great Britain. J. Environ.Manag. 49, 413–433.

MODIS, 2013. http://modis.gsfc.nasa.gov/ (Last accessed 11/3/15).Moorland Association, 2006. http://www.moorlandassociation.org/economics3.asp

(Last accessed 7/3/15).Natural England, 2008. State of the Natural Environment 2008. http://publications.

naturalengland.org.uk/publication/31043?category=118044 (Last accessed 7/3/15).

Natural England, 2009. Mapping values: the vital nature of our uplands. http://publications.naturalengland.org.uk/publication/47001 (Last accessed 7/3/15).

250 D.J.T. Douglas et al. / Biological Conservation 191 (2015) 243–250

Natural England, 2012. Obtained under the Open Government Licence for public sectorinformation. http://www.nationalarchives.gov.uk/doc/open-government-licence/version/2/ (Last accessed 7/3/15).

Neary, D.G., Klopatek, C.C., DeBano, L.F., Ffolliott, P.F., 1999. Fire effects on belowgroundsustainability: a review and synthesis. For. Ecol. Manag. 122, 51–71.

Newey, S., Mustin, K., Bryce, R., Fielding, D., Redpath, S., Irvine, J., 2012. The same butdifferent: upland management and bird diversity in the Scottish Highlands. In:Irvine, J. (Ed.), HUNTing for Sustainability' — A Summary of Research Findings Fromthe Scottish Case Study, pp. 13–15 (http://fp7hunt.net/. Last accessed 11/3/15).

Ordnance Survey, 2014. http://www.ordnancesurvey.co.uk/resources/maps-and-geographic-resources/the-national-grid.html (Last accessed 7/3/15).

Pearce-Higgins, J.W., Grant, M.C., Beale, C.M., Buchanan, G.M., Sim, I.M.W., 2009. Interna-tional importance and drivers of change of upland bird populations. In: Bonn, A.,Allott, T. (Eds.), Drivers of Environmental Change in Uplands. Routledge, Abingdon,pp. 209–227.

Penny Anderson Associated Ltd, 2013. Peat depthmapping and carbon storage estimationreport. North Pennines AONB and Natural England National Peat Depth Mapping andCarbon Storage Project.

R Core Team, 2013. R: A Language and Environment for Statistical Computing. R Foundationfor Statistical Computing, Vienna, Austria (http://www.R-project.org/).

Rosner, B., May 1983. Percentage Points for a generalized ESD many-outlier procedure.Technometrics 25, 165–172.

Ryan, P.A., Lyons, S.A., Alsemgeest, D., Thomas, P., Kay, B.H., 2004. Spatial statistical analysisof adult mosquito (Diptera: Culicidae) counts: an example using light trap data, inRedland Shire, Southeastern Queensland, Australia. J. Med. Entomol. 41, 1143–1156.

Sah, J.P., Ross, M.S., Snyder, J.R., Koptur, S., Cooley, H.C., 2006. Fuel loads, fire regimes, andpost-fire fuel dynamics in Florida Keys pine forests. Int. J. Wildland Fire 15, 463–478.

SAS, 2012. SAS 9.4. SAS Institute Inc., Cary, NC, USA.Scottish Government, 2011. The Muirburn Code. Scottish Government, Edinburgh.Scottish Natural Heritage, 2010. Condition of Designated Sites. http://www.snh.gov.uk/

docs/B686627.pdf (Last accessed 7/3/15).Sotherton, N., May, R., Ewald, J., Fletcher, K., Newborn, D., 2009. Managing uplands for

game and sporting interests: an industry perspective. In: Bonn, A., Allott, T. (Eds.),Drivers of Environmental Change in Uplands. Routledge, Abingdon, pp. 241–260.

Suárez, E.R., Medina, G., 2001. Vegetation structure and soil properties in EcuadorianPáramo Grasslands with different histories of burning and grazing. Arct. Antarct.Alp. Res. 33, 158–164.

Tansey, K., Beston, J., Hoscilo, A., Page, S.E., Paredes Hernández, C.U., 2008. Relationshipbetween MODIS fire hot spot count and burned area in a degraded tropical peatswamp forest in Central Kalimantan, Indonesia. J. Geophys. Res. 113. http://dx.doi.org/10.1029/2008JD010717.

Tharme, A.P., Green, R.E., Baines, D., Bainbridge, I.P., O'Brien, M., 2001. The effect ofmanagement for red grouse shooting on the population density of breeding birdson heather-dominated moorland. J. Appl. Ecol. 38, 439–457.

Thompson, D.B.A., Macdonald, A.J., Marsden, J.H., Galbraith, C.A., 1995. Upland heathermoorland in Great Britain: a review of international importance, vegetation changeand some objectives for nature conservation. Biol. Conserv. 71, 163–178.

Tian, D., Wang, Y., Bergin, M., Hu, Y., Liu, Y., Russell, A.G., 2008. Air quality impacts fromprescribed forest fires under different management practices. Environ. Sci. Technol.42, 2767–2772.

UK Government, 2015. UK greenhouse gas emissions statistics. https://www.gov.uk/government/collections/uk-greenhouse-gas-emissions-statistics (Last accessed11/3/15).

Warton, D.I., Hui, F.K.C., 2011. The arcsine is asinine: the analysis of proportions in ecology.Ecology 92, 3–10.

Watts, C.D., Naden, P.S., Machell, J., Banks, J., 2001. Long term variation in water colourfrom Yorkshire catchments. Sci. Total Environ. 278, 57–72.

Welsh Assembly Government, 2008. The Heather and Grass Burning Code for Wales2008. Welsh Assembly Government, Cardiff.

White, S., Yallop, A.R., Thacker, J.I., Clutterbuck, B., 2007. Control of water quality fromupland catchments. Report to Yorkshire Water.

Williams, D.R., Pople, R.G., Showler, D.A., Dicks, L.V., Child, M.F., zu Ermgassen, E.K.H.J.,Sutherland, W.J., 2012. Bird Conservation: Global Evidence for the Effects of Interven-tions. Pelagic Publishing, Exeter.

Willis, K.S., 2015. Remote sensing change detection for ecological monitoring in UnitedStates protected areas. Biol. Conserv. http://dx.doi.org/10.1016/j.biocon.2014.12.006.

Worrall, F., Bell, M.J., Bhogal, A., 2010. Assessing the probability of carbon and greenhousegas benefit from the management of peat soils. Sci. Total Environ. 408, 2657–2666.

Yallop, A.R., Thacker, J.I., Thomas, G., Stephens, M., Clutterbuck, B., Brewer, T., Sannier,C.A.D., 2006. The extent and intensity of management burning in the English uplands.J. Appl. Ecol. 43, 1138–1148.

Yallop, A.R., Clutterbuck, B., Thacker, J.I., 2009. The history and ecology of managed firesin the uplands. In: Bonn, A., Allott, T. (Eds.), Drivers of Environmental Change inUplands. Routledge, Abingdon, pp. 171–185.