SEARCH

SEARCH BY CITATION

Keywords:

  • bulk density;
  • fuel moisture;
  • fuel structure;
  • heathland;
  • redundancy analysis;
  • shrubland;
  • wind speed

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

1.Calluna-dominated heaths occur throughout Europe but are in decline across their range. There is growing interest in using prescribed burning for their management, but environmental and social change will impact future fire regimes. Understanding fire behaviour is vital for the sustainable use of fire, but no robust models exist to inform management.

2.  Shrub fuels display complex fire behaviour. This is particularly true in UK moorlands which are unusual in their fuel structure and moisture regime, being dominated by live fuel and an oceanic climate.

3.  We burnt 27 experimental fires in the Scottish uplands during the legal burning season using a replicated experimental design. Plots were assigned to one of three commonly identified growth phases. We estimated a range of prefire fuel characteristics, including heterogeneity in fuel structure. We recorded wind speed and direction and estimated rate of spread (RoS).

4.  Redundancy analysis was used to investigate the relationship between fire behaviour parameters as a whole and control variables. Fuel structure and heterogeneity, wind speed and canopy fuel moisture content were strongly related to variation in fire behaviour.

5.  Best subsets regression was used to generate models of fire spread based on wind speed, vegetation height, canopy fuel moisture and an index of fuel heterogeneity. RoS was determined largely by wind speed, but this interacted strongly with vegetation structure. Changes in fuel horizontal continuity and vertical structure reduced rates of spread in low wind speeds.

6.Synthesis and applications. Live fuel moisture and fuel heterogeneity play an important role in dampening fire behaviour, aspects of shrub fuels that have previously not been examined in detail. Careful use of fire for moorland management increases habitat diversity and creates fire-safe landscapes. Escaped prescribed fires burn large areas, homogenize landscapes and have severe impacts on ecosystem services. The complex relationship between fuel structure and wind speed implies that changes in behaviour can be rapid and unexpected. Models can be used to assess fire hazard prior to prescribed burning and to choose fuels that can be burnt safely under prevailing or forecast conditions.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Recent ‘horizon-scanning’ exercises have identified changes in fire regimes, resulting from climatic and environmental change, as one of the most significant threats facing biodiversity in the UK (Sutherland et al. 2008). There is also growing debate about the role of fire in the context of peatland carbon balance (Gray 2006; Limpens et al. 2008) and water quality (Mitchell & McDonald 1995; Clay, Worrall, & Fraser 2009). Prescribed burning can be used to achieve multiple management objectives whilst helping to protect against undesirable changes in fire regime, such as increased fire frequency and higher fire severity, (Davies et al. 2008a), but a good understanding of fire behaviour is crucial for its safe application. A thorough conception of factors affecting fuel flammability, fire spread and intensity are also central to any fire danger rating system (e.g. Canadian Forest Service undated, National Rural Fire Authority). Although there has been some fire research in the UK, much has focused on measuring fire temperatures (e.g. Kenworthy 1963; Kayll 1966; Hobbs & Gimingham 1984) and ecological effects rather than fire behaviour per se. Notable exceptions include Hamilton (2000) and Bruce & Servant (2003). Further research is urgently needed to inform management, to allow assessment of fire hazard in the context of a changing environment and to contribute to the development of a robust fire danger rating system for moorland habitats (Kitchen et al. 2006; Legg et al. 2007).

Fire behaviour in shrub fuels is acknowledged to be complex and difficult to model. Simple empirical models based on aspects of vegetation structure and wind speed exist, but it has been difficult to develop models that could be used in a fire danger rating system and that account for day-to-day variation in fire hazard (Anderson 2006). We know that important controls on shrub fire behaviour include fuel loading and structure (e.g. Fernandes 2001), fuel height (e.g. Catchpole et al. 1998; Fernandes, Catchpole, & Rego 2000), fuel moisture (e.g. Weise et al. 2005; Plucinski & Anderson 2008), the quantity of dead fuel (Baeza et al. 2002) and wind speed (e.g. Molina & Llinares 2001; Morvan & Dupuy 2004). Hobbs & Gimingham (1984) suggested that ignition line length may be important for moorland fires, although Davies (2005, p. 110) demonstrated that their rate of spread (RoS) data was somewhat questionable.

The moorlands of the UK uplands are of international conservation importance (Thompson et al. 1995) and are widely managed to maximize populations of red grouse Lagopus lagopus scoticus Latham for sport shooting. Appropriate fire regimes are a key factor determining their productivity and environmental services (Davies et al. 2008a). Elsewhere in Europe, heathlands are under threat and fire is seen as a mechanism to prevent their loss and increase landscape diversity (e.g. Sedláková & Chytry 1999; Calvo, Tarrega & Luis 2002; Vandvik et al. 2005; Ascoli et al. in press). The characteristics of fire in heathlands are poorly understood and, in the UK uplands in particular, are unique due to the particularities of the vegetation and the oceanic climate (Davies et al. 2006). Whilst most fire research has been completed on dead or seasonally desiccated fuel, ours is dominated by a green, live component with relatively high fuel moisture content (FMC). This may have important dampening effects on fire behaviour (Catchpole & Catchpole 1991; Dimitrakopoulos & Papaioannou 2001). UK moorlands are frequently dominated by a single dwarf-shrub species, Calluna vulgaris L. Hull. (hereafter Calluna), that forms a dense canopy under which often lies a deep layer of pleurocarpous mosses, lichens and litter. As stands age, gaps become more frequent (Gimingham 1988), and the ratio of woody stems to fine fuels within the canopy increases (Davies et al. 2008b). Greater heterogeneity in fuel beds with age means that relationships between increasing fuel load and fire behaviour may not be simple, but we are not aware of previous empirical studies in shrubland fuel types that have examined this in any detail.

In the UK, interest in fire behaviour is focused both on management fires during the core legal burning period of 1 October to 15 April and on wildfires that occur throughout the year but mainly during the spring (March to April) and summer (July and August) (Legg et al. 2007). Statutory controls and recommendations on the use of burning for land management are summarized in SEERAD (2001) and Natural England (2007). Weather in the UK uplands can vary rapidly from cold and wet, to warm and dry with low atmospheric humidity and periods of physiological drought when the ground is frozen (Davies 2005). Variation in fuel moisture can be large, particularly in spring, due to frequent precipitation alternating with periods of frozen ground, low humidity and over-winter cuticular damage (Davies 2005). There are usually rather few days suitable for burning. This makes understanding fire behaviour all the more important if we are to enable land managers to make efficient use of time and resources without feeling under pressure to burn unsafely in order to meet management targets. Fire controllability, of which RoS is an important component, affects the ability of managers to target burns to specific areas and to maintain the patchwork of Calluna ages associated with diverse moorland landscapes. Fires that escape control burn large areas, create homogenous landscapes, risk losing Calluna cover where fires are severe or burn large areas of older Calluna that regenerates poorly and can ignite peat leading to the destruction of carbon stores and widespread damage to a range of ecosystem services (Davies et al. 2008a). We sought to define relationships between weather and fuel conditions and the RoS of management-size fires whilst removing, as far as possible, the influence of other factors such as slope and ignition line length.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Experimental site

Experimental fires were burned at two sites using a randomized block design. Three fires, one in each of three Calluna developmental phases (Gimingham 1988), were burnt on any 1 day. Two different plot designs were used (Supporting Information, Appendix S1): 21 plots were burnt according to design one (15 m ignition line), whilst six were burnt according to design two (20 m ignition line). Design one was used at Crubenmore Estate, near Dalwhinnie on the edge of the Cairngorms National Park in NE Scotland (04°15′W, 56°57′N; Ordnance Survey Grid Reference NN 6386). Design two was used on Black Hill, Whitborough Estate, in the Pentland Hills outside Edinburgh (03°20′W, 55°51′N; OS Grid Reference NT 185625).

Design and monitoring

Our experimental fires were necessarily small to ensure safety and controllability, but plot size was within the range of recommended fire widths for moorland management (SEERAD 2001; Natural England 2007). The preburn vegetation of plots was species poor, Calluna-dominated upland heath (NVC community H12, C. vulgaris–Vaccinium myrtillus heath, Rodwell 1991). Plots contained a mixture of coarse grasses and sedges with Trichophorum cespitosum, Deschampsia flexuosa and Molinia caerulea frequent, although never forming a significant proportion of the fuel load. Most plots were underlain by continuous mats of pleurocarpous mosses. All plots were located on slopes of less than 10%.

Fires were ignited using a drip torch and burnt as head fires (with the wind). We used a line ignition along the full length of the upwind plot edge. This pattern was altered for safety reasons during the ‘Mature’ burns on Black Hill where, due to strong winds, the top or edge of a plot was lit first. This allowed a small area of the plot to be burnt as a back-fire to strengthen swiped firebreaks. In these cases observers stopped recording fire characteristics if they deemed the head fire to be influenced by the back-fire.

Prefire fuel structure and loading was surveyed using three (design 1) or four (design 2) ‘FuelRule’ transects, with individual readings taken 2 m apart. The method is based on visual obstruction of a measuring stick and has been calibrated for use in our vegetation type (Davies et al. 2008b). The FuelRule is a 2·5-cm-wide stick painted with alternating white and yellow bands. This is placed vertically into a stand and down through the moss/litter layer. The user makes a visual estimate of the percentage of each band obscured by vegetation and records the maximum height at which Calluna touches the stick. Data are entered into a program that calculates a variety of metrics to provide an index of canopy density (CDI) and estimates of total and fine fuel load (here defined as foliage and stems <2 mm in diameter). Repeated measurements of fuel structure across a plot allow standard deviations of metrics to be used as measures of fuel-bed heterogeneity. The ratio of dead-to-live fuel was estimated using a single, representative 50 × 50-cm2 fuel assessment quadrat per plot analysed as in Davies et al. (2008b).

We collected four (design 1) or five (design 2) canopy fuel moisture content (FMC) samples. These comprised live shoots that bear a somewhat variable amount of dead and live foliage in the form of tiny adpressed leaves. Detailed analysis in the laboratory of eleven live shoots collected from Site 1 demonstrated that they contained between 8% and 22% dead foliage (mean = 14·2%, standard deviation 5·1). In later experiments, to try and better understand the relative importance of dead and live FMC, we additionally collected five samples of completely dead material. We also collected samples from the top 2 cm of the moss/litter layer beneath the Calluna canopy. Samples were taken from randomly selected subplots (see Supporting Information, Appendix S1 for details of plot layout) and FMC calculated on percentage dry weight basis.

A portable meteorological station was located roughly 10–20 m upwind of each fire and used to record wind speed, temperature and humidity at 10-s intervals during the fire. For four fires equipment failure meant we had to use data from a nearby permanent weather station. Davies (2005) describes the calibration of these data for use at the experimental site.

Rate of spread was estimated by timing fire front arrival at four 2-m-high posts located 5 m apart. For design 1 we also measured RoS using two sets of thermocouples located 5 m apart in the centre of the fire (Fig. 1). The RoS used in our analyses was calculated by averaging the observational and thermocouple estimates. Where thermocouples were not used we corrected observed RoS using the relationship between the thermocouple-observed mean and observed values. Monitoring of fire behaviour was concentrated in the centre of plots where fires were likely to spread the fastest. Fire RoS from ignition to the first post was omitted from all calculations as fires were observed to be accelerating during this period (Davies 2005).

image

Figure 1.  Variation in the fuel structure for three fuel load categories, n = 9 per category. (a) total and (b) fine fuel loads (kg m−2), (c) Calluna height (cm) and (d) bulk density (kg m−3). Boxes show the first and third quartiles; the dividing line is the median. Whiskers show the general range of the data with outliers shown as an asterisk.

Download figure to PowerPoint

Data analysis

A generalized linear model with Tukey’s comparison (minitab 14; Minitab Inc 2003) was used to investigate differences in fuel structure between growth stages (fixed factor). Redundancy Analysis (canoco 4.5; ter Braak & Šmilauer 2002) was used to explore the relationships within and between control and response variables for all fires with dead FMC excluded, and then again on just those fires for which dead FMC data were available. This technique is more normally associated with the analysis of species community composition (e.g. Mitchell et al. 2009). Here, we treat fire behaviour characteristics as ‘species’ and analyse variation in the ‘community’ of these characteristics. Fuel consumption, fireline intensity and flame length were included in the ordination as dependent variables, along with RoS, to gain an accurate overall representation of variation in fire behaviour. Davies (2005) describes how these variables were estimated. The dependent variables were centred and standardized prior to analysis. Explanatory variables (fuel and weather characteristics) were added using manual forward selection. For this analysis, we used a value of P ≤ 0·1 in order to include all predictors that are, or may prove to be with further data, functionally significant. The significance of the predictor variables was tested using a Monte Carlo permutation with 999 iterations.

Rate of spread was modelled using best subsets regression analyses (minitab). The models were selected largely on goodness of fit to the experimental data (adjusted r2), but consideration was also given to model parsimony, ease with which the selected independent variables could be used in practice and consistency with existing fire science research and theory. The use of best subsets regression not only picks the combination of predictor variables that gives the ‘best’ model (i.e. with highest r2) but also gives alternative models using different combinations of independent variables that while statistically less efficient, may be more useable in practice. As dead FMC data were available only for a subset of the fires, regression analyses were completed both on all fires (excluding dead FMC as an independent variable) and on just those fires where dead FMC data were available.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Variation in burning conditions

Total fuel load and height of the Calluna canopy increased through the developmental stages of Calluna as represented by the Building, Late-building and Mature plots (Fig. 1). Fine fuel load increased significantly between the Building and Late-building plots [F(2, 24) = 4·90, = 0·02] but was more variable in the Mature plots and did not differ with the other stages. Bulk density was significantly lower in Mature compared with Late-building plots [F(2, 24) = 4·60, = 0·027] and nearly so compared with Building plots (= 0·055).

Standard deviations of CDI were generally the greatest in shorter Calluna stands, probably relating to the mix of dense, dome-shaped shrubs and open areas where regeneration was slow following the previous fire. The standard deviation of mean height was, however, the greatest in old stands, where variation was caused by the differential growth of adjacent heather plants and, crucially, by the collapse or death of some bushes. Combining these two terms created a variable, here termed S, that combines the influence of both horizontal and vertical heterogeneity in fuel structure. A number of plots had particularly high values for this measure of heterogeneity (Table 1).

Table 1.   Experimental fires with high values of ‘S’– the product of the standard deviations of height and CDI
Fire IDSSource of heterogeneity?
  1. The mean and median values for S were 2·31 and 1·80 respectively (SD ± 1·61).

BE2 P234·43Fuel equivalent to a Mature load in one corner of the plot and an area dominated by grasses and sedges.
BE2 P103·76Very short, young Calluna with a number of bare patches and grass-dominated areas. Would not ignite despite repeated attempts.
BH73·39Degenerate/rank Calluna. Canopy very broken with prostrate stems, large moss-dominated open areas, many dead Calluna bushes and very dense patches where Calluna stems had layered.
BE2 P83·26Significant areas of dead material and prostrate Calluna stems. Canopy very gappy in places as a result.
BE2 P192·98Many grass/sedge-dominated patches with a significant amount of Trichophorum cespitosum. Fuel equivalent to a Mature load in one corner. The plot was crossed by a sheep track.
BH42·88A mixture of dense areas of layered heather between prostrate patches of overgrazed, ‘drumstick’Calluna.
BE2 P212·52A number of significant open areas in the Calluna canopy and a distinctly higher loading in one corner of the plot.

The majority (80–85%) of the fuel in the plots was live material. Particularly high proportions of dead fuel (up to 30%) were found in a small number of Building plots (comprising mostly coarse stems, 2–5 mm in diameter remaining unburnt from the previous fire) and in older stands on the Black Hill experimental site. Fuel moisture varied considerably for both canopy (55–97%) and dead (15–29%) material.

A wide range of fire behaviour was captured during the experimental burns, with RoS estimated from 0·5 to 12·6 m min−1 (Fig. 2). The variability in RoS in Mature fires was particularly noticeable. One fire (BE2 P10) had an extremely low fuel loading. Repeated ignition attempts in this plot all failed despite successful burns in other fuel loads on the same day. This plot was treated as missing in the RDA and log-linear regression modelling.

image

Figure 2.  Variation in rate of spread (m min−1) in three fuel categories (n = 9 per category). Plot layout is as for Fig. 1.

Download figure to PowerPoint

Redundancy analysis

Redundancy analysis using all fires demonstrated the influence of several variables on fire spread (Fig. 3). A number of significant relationships were evident (described in the order predictors were added to the model). The standard deviation of canopy density index (SD CDI) was highly significantly linked to slower moving fires. Increased bulk density was relatively important in dampening fire spread. Wind speed explained a significant amount of the variation in the fire behaviour data with higher speeds linked to faster spread rates. Canopy fuel moisture significantly influenced behaviour and was strongly linked to slower moving fires. The influence of mean height was only significant at the < 0·10 level.

image

Figure 3.  Redundancy analysis of all fires with dead fuel moisture content excluded from the analysis. Axes 1 and 2 explain 44% and 13% of the variation in the fire behaviour data respectively. Predictor variables are in grey and dependent, fire behaviour variables in black italics. M/L consumption is the mass of the moss and litter layer consumed by the fire; SD CDI is the standard deviation of canopy density index. The numbers in brackets are P-values from Monte Carlo permutation testing (999 iterations) and show the significance of the variable, upon entry to the model, in explaining variation in the fire behaviour data.

Download figure to PowerPoint

When the Redundancy Analysis was repeated on the 12 fires for which there was information on dead fuel moisture (Fig. 4), the only significant predictor variables were SD CDI and fuel load (entered first and second respectively). Dead fuel moisture and wind speed (entered third and fourth) were significant at the < 0·10 level. Fuel load was not strongly linked to RoS and only weakly correlated with the first two Axes of the RDA. It was, however, closely correlated with fuel consumption and this relationship dominated Axis 3 of the RDA. This explains why the Fuel Load arrow appears to be so short (Fig. 4) despite its significant P-value. RoS was still observed to be positively influenced by wind speed and negatively affected by SD CDI. Dead FMC was negatively correlated with RoS.

image

Figure 4.  Redundancy analysis (n = 12) for fires where dead fuel moisture content data were available. Axes 1 and 2 explain 46% and 16% of the variance in the fire behaviour data respectively. Grey arrows are predictor variables, whilst fire behaviour characteristics are shown in black italics. The numbers in brackets are P-values (Monte Carlo Permutation test with 999 iterations).

Download figure to PowerPoint

Rate of spread models

There were significant correlations between a number of predictor variables (Table 2). Some of the internal correlations are to be expected (e.g. relative humidity and temperature); others are less expected (e.g. the positive association between canopy FMC and air temperature). Wind speed, an important control on fire spread, had very little association with the other predictor variables.

Table 2.   Pearson correlation matrix for weather and fuel moisture variables
 Dead FMCCanopy FMCM/L FMCTempRHWind speedSD wind speed
  1. P-values are given in parentheses; significant correlations are shown in bold. There are 10 degrees of freedom for coefficients including Dead FMC, 25 degrees of freedom for all other coefficients.

Dead FMC      
Canopy FMC0·36 (0·26)     
M/L FMC0·04 (0·90)0·14 (0·48)    
Temperature0·36 (0·25)0·68 (<0·001)0·38 (0·05)   
Relative humidity0·50 (0·10)0·31 (0·11)0·31 (0·12)0·36 (0·06)  
Wind speed−0·06 (0·87)−0·15 (0·47)−0·15 (0·44)0·10 (0·61)−0·45 (0·02) 
SD wind speed−0·09 (0·77)−0·24 (0·27)0·13 (0·56)−0·06 (0·79)−0·43 (0·03)0·40 (0·05)
The role of wind speed

Initial analysis showed that RoS seemed to respond to changes in fuel structure and wind speed in a nonlinear fashion. Fire spread in taller, Mature plots responded to wind speed much more strongly than in Building or Late-building fuels (Fig. 5). The two Mature loading plots in very heterogeneous stands on the Black Hill site appeared to behave differently from other members of the Mature fuel group.

image

Figure 5.  Regression of rate of spread on wind speed for each of the fuel groups: grey triangles: Building fuel loads; dark grey squares: Late-building; black circles: Mature. The r2 values were 0·41, 0·42, 0·52 for Building, Late-building and Mature fuels respectively. The two Mature loading plots with prostrate stems (crossed circles) burnt on Black Hill behaved differently from others in the group and were excluded from the regression (see text).

Download figure to PowerPoint

Predicting rate of spread

Models were developed using a nonlinear relationship between fuel structure, wind speed and fire spread. When all fires were analysed, excluding dead FMC data, a number of equations were developed (Table 3). The simplest of these (equation 1), uses just vegetation height and wind speed, both of which could be estimated with ease by managers in the field. Equation 2 demonstrates a significant improvement in prediction (= 0·008) with the inclusion of canopy fuel moisture.

Table 3.   (a) Regression equations to predict rate of spread (R, m min−1); (b) Detailed description of model performance showing the P-value and standard error (SE) for all terms in each model
Equation no.Fuel typesCharacteristicEquationr2 (adj)N
(a)
1Allr0·791 + 7·917h2U0·5626
2Allr8·304 + 7·286h2U − 0·097Ml0·6626
3Allr9·553 + 6·622h2U − 0·083Ml − 99·020S0·7226
4Allln(r)0·381 + 1·743h2U0·4125
5Allln(r)0·592 + 5·673h2 + 0·191U − 44·23S0·6725
6Allln(r)1·550 + 2·564h + 0·197U − 0·062Md − 39·060S0·8411
EquationConstantCoefficient 1Coefficient 2Coefficient 3Coefficient 4
PSEPSEPSEPSEPSE
  1. h is the mean Calluna height (m), U is the wind speed (m s−1), Ml is the canopy fuel moisture content (% dry weight), Md is the dead fuel moisture content (% dry weight) and S is the product of the standard deviations of mean height and CDI. N shows the number of fires on which the equation is based. Significant terms (< 0·05) are shown in bold. Independent variables are described in the order in which they appear in the model.

(b)
10·2650·694<0·0011·360      
20·0052·664<0·0011·214  0·0080·034    
30·0012·491<0·0011·144  0·0140·0310·02541·420  
40·0820·210<0·0010·404      
50·1550·402  0·0011·483<0·0010·0460·00111·840  
60·0320·582  0·0080·007  0·0010·0330·0070·0160·01612·29

Equation 3 includes S, the product of the standard deviations of mean height and CDI, a measure of fuel-bed heterogeneity. A number of plots had particularly high values for the predictor S (Table 1). Equations 1–3 yielded negative predictions of RoS where canopy fuel moisture and/or the standard deviations of mean height and CDI were high. Log transformation of RoS prevented this problem, but the information contributed by moisture was replaced by S (Table 3, equations 4 and 5).

Analysis of the 12 fires for which dead FMC data were available produced different results (equation 6). Dead and canopy FMC did not correlate closely (Table 2) and dead, rather than canopy, FMC was included in the model with the latter never appearing as significant. The inclusion of dead FMC did not therefore account for the variance in RoS caused by canopy FMC. This suggests, contrary to the previous analysis, that canopy FMC did not play a major role in fire behaviour.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Relatively few studies of fire behaviour have been completed on Calluna-dominated heathlands. Those that do exist have been primarily concerned with documenting and modelling fire temperatures in the Calluna canopy (e.g. Whittaker 1961; Kayll 1966; Hobbs & Gimingham 1984), which are not of particular management or ecological interest (Van Wagner & Methven 1978). Hobbs & Gimingham (1984) conducted a more extensive study, but our rates of spread cover a much greater range. An artificial ceiling of 2 m min−1 seems to exist in their spread data; Davies (2005) demonstrated that this was probably due to problems with the way they used thermocouples to measure RoS. Our spread rates compared favourably with those recorded by Molina & Llinares (2001), Fernandes (2001) and De Luis et al. (2004) for experimental fires in Mediterranean shrubland.

The role of fuel structure

The fuel categories used here represent three of the easily recognized phases in the development of a Calluna stand described by Gimingham (1988). The changes we observe in fire behaviour between stand stages (Fig. 1) support a simple model of fuel development that builds on the work of Gimingham (1988) and Davies et al. (2008b). Fine fuel is all that is consumed in most management fires with the thicker stem material remaining unburnt. In spring and autumn, the moss and litter layer is usually too wet to contribute to flaming combustion. The fine fuel layer develops to form a complete canopy by the end of the Pioneer phase, although gaps may remain in the Building phase if regeneration is not uniform. The canopy of fine fuel increases with depth and loading through the Building phase, but its density changes little. The Building phase has a uniform, dense canopy of fine stems down to the ground level. In Late-building phase fuels the growth of coarse stems begins to lift the canopy off the ground and cause a reduction in total bulk density increasing flammability (Figs 2 and 3). This process continues in the mature phase, but stems also begin to die or fall prostrate and the fine fuel becomes more horizontally discontinuous. Average canopy height eventually declines and much of the fine fuel, including an increased proportion of dead fuel, lies close to the ground where it is less well aerated. The two ‘Mature’ plots on Blackhill (BH4 and BH7) represent this final stage of development, being extremely heterogeneous and with most of their fine fuel near to the ground. They are the Degenerate Calluna stands described by Gimingham (1988) and it is noticeable that they behave rather differently from Mature fires (Fig. 5). This suggests the existence of three key fuel types – Building, Mature and Degenerate Calluna. Further data are required to support this and to adequately model fire behaviour in all three fuel types.

Vegetation height recurred as an important predictor of fire behaviour (Table 2, Fig. 3) because it is closely related to changes in canopy structure. Height is a surrogate variable for total and fine fuel loading and the degree of aeration as the fine fuel is lifted above the ground. Taller stands with heterogeneous canopies also have greater surface roughness leading to turbulent air flow and greater penetration of air into the canopy (Stull 1988). In fires in older canopies there was, therefore, not just a continuous oxygen supply but also greater convective heat available to evaporate water before pyrolysis and combustion began. Although Rothermel (1972) described shrubland fires as surface fires, several authors (e.g. Alexander & Sando 1989; Fernandes et al. 2000) suggest that they have more in common with crown fires where litter is not required for sustained fire propagation and the fuel is aerated from below. Using the square of mean height captured the mechanism producing reduced fire spread rates in the collapsed mature stands on Blackhill as well as the effect of low fuel load, high bulk density and small amounts of dry, dead material in younger stands.

Rate of spread was negatively correlated with bulk density and this is reflected in the RDA (Fig. 3). Both Thomas (1971) and Rothermel (1972) identified bulk density as having an important dampening effect on fire behaviour, with increasing bulk density essentially starving the combustion process of oxygen. In the present study decreases in bulk density were largely associated with the creation of spaces beneath the main fuel mass of the canopy. This probably increased RoS through increased ventilation of the fire front, although only up to the point where significant gaps began to disrupt the fire’s flow. Large gaps were observed to break up the head fire into a number of smaller fire fronts that occasionally self-extinguished. The importance of gaps in the fuel is demonstrated by the inclusion of the SD CDI in the RDA and by the importance of S, the product of the standard deviations of height and CDI, in the regression models. Future analyses should explore the potential of a qualitative index of canopy heterogeneity that would be more practical than the rather cumbersome S term.

The role of wind speed

The effect of wind speed on RoS is well documented and understood (e.g. Pyne, Andrews, & Laven 1996). The greater variability in behaviour between fires in Mature fuel plots can be explained by the increased sensitivity to wind in more heterogeneous stands. Higher wind speeds flatten flames and increase reception of radiative and convective energy by the fuel bed ahead of the fire front allowing the fire to bridge gaps in the canopy of older Calluna stands. This increases the depth of the combustion zone and RoS through what is, generally speaking, a better aerated, drier fuel bed than that of Building phase stands. This effect appears in the regression models as the interaction term between height and wind speed (h2U).

The influence of wind speed on Building and Late-building fuels was reduced compared to the Mature load plots (Fig. 5). This may be partly associated with the increased bulk density of fuels closer to the ground, with reduced aeration of the fuel, and due to the exponential decline in wind speeds closer to the ground.

The role of fuel moisture

Previous work (Davies 2005; Davies et al. 2006) failed to find a significant relationship between fuel moisture and fire behaviour, although combustion theory tells us that there must be some effect. The results of the RDA (Fig. 3) and regression modelling (equations 2 and 3) when all fires were included (but dead FMC data were excluded) suggested that canopy fuel moisture had a significant negative effect on RoS and intensity. However, where dead FMC data were available, these became the more significant predictor of fire RoS and intensity (equation 7), albeit for a significantly reduced data set. Dead fuels probably play an important role in governing initial flammability and fire behaviour due to their significantly lower moisture content (Legg et al. 2007). They ignite first and preheat, dry and raise live fuel to combustion temperature. The importance of dead moisture content for an established fire is likely to vary depending on the status of live fuel components and vice versa. The balance between the relative amounts of dead and live material and their FMC is likely to be crucial.

The role of fuel moisture in affecting fire behaviour is thus still uncertain. Although fires will not burn at all if the fuel is too wet, and the RDA showed a negative relationship between fire behaviour and canopy FMC, the relationship was weak compared to the effects of fuel structure. However, the moisture content of the heather canopy varies considerably between seasons (Legg et al. 2007; Davies et al. 2008b). During management fires in the legal burning season, such as those reported here, the moss and litter is too wet to burn in all but the most exceptional of weather conditions, but canopy moisture can be very low. During the summer months, however, the fresh green foliage has high moisture content, while the moss layer is frequently dry and may sustain a fire beneath the heather canopy. Further fire tests where both canopy and dead FMC data are available, and where the amount of dead material present in the ‘live’ canopy samples is quantified, are urgently needed. In summer wildfire conditions, the relationships between fire behaviour and weather conditions may be very different and the relationships reported here may not be applicable.

Management and research implications

Globally, heathlands are a rare and threatened habitat type, declining in many areas of Europe (Thompson et al. 1995). Prescribed burning has been shown to have a number of potential biodiversity benefits such as providing habitat for upland wading birds (Tharme et al. 2001) and increasing the diversity of lichen communities (Davies & Legg 2008). Other benefits may be in controlling fuels and the hazard and impact of severe wildfires (Davies et al. 2008a), removing excess nutrients from atmospheric pollution (Barker et al. 2004) and aiding tree and forest regeneration (Hancock et al. 2005). Although the judicial use of fire is increasingly common across the global distribution of heathlands, concern has been expressed about the potential detrimental effects of inappropriate burning practices on carbon budgets, erosion, water quality and biodiversity (Holden et al. 2007; Davies et al. 2008a). Davies et al. (2008a) and Ascoli et al. (in press) discuss the need for a coordinated, ecological approach to fire management in Calluna heathlands. This should consider management objectives, potential fire behaviour and fire severity and set fire prescriptions accordingly. The safe and sustainable use of fire demands prior knowledge of probable behaviour so that fires that cause environmental damage by escaping control or igniting underlying organic soils can be avoided.

Further work is required to model and understand relationships between RoS, flame length and fire intensity. Flame length is an important visual representation of the rate of combustion and can be used by managers to quickly gauge how easy a fire will be to control. More data are required in Degenerate fuels and from a wider range of seasons and FMC scenarios. Such work will allow us to develop guidelines and simple tools, such as nomograms (Dimitrakopoulos, Mitsopoulos & Raptis 2007) and photo guides to fuels and fire behaviour, which will be useful for training and fire hazard assessment.

Agencies and managers are encouraged to use our increasing scientific understanding of fire behaviour to carefully consider the impacts of changes in practice and policy for landscape fire hazard. The relationships presented here can be used, with caution, to provide an indication of fire behaviour when planning fire management activities. Our personal experience suggests that head fires with spread rates greater than 6–8 m min−1 necessitate careful planning for adequate control. Heathland managers should expect significantly different fire behaviour in different Calluna growth phases and bear in mind the significant nonlinear relationships in fire behaviour we have observed: on a given day behaviour in one growth phase is not necessarily a simple guide to that in another. Learning to recognize the structural changes in fuel described earlier is crucial to safe burning practice.

We have shown that fire behaviour in Calluna-dominated moorland vegetation is determined largely by wind speed, but that this interacts strongly with vegetation structure. While height was the simplest surrogate for structure in our experimental plots, changes to the horizontal continuity and vertical structure of the fuel as the stand passes from the building to mature phase of development have a large effect in reducing RoS where wind speeds are low. Managers should be aware of the potential for rapid changes in fire behaviour when fires move across stand boundaries as well as with changes in wind speed, particularly in older stands. Careful forward planning, identification or creation of firebreaks and appropriate ignition strategies can combat such problems. To promote consistent fire behaviour, we encourage limiting individual fires to stands of a broadly homogenous age and burning when wind speeds are stable. Managers should avoid burning mature Calluna when wind speeds are high or gusty. Well-managed fires, lit in appropriate conditions, can provide a variety of benefits that include increased floral and faunal diversity, improved productivity, fire hazard reduction and a range of other environmental services (Davies et al. 2008a). Ensuring managers understand fire behaviour will help minimize the risk of escapes and wildfires that lead to large, severe burns, soil damage, poor vegetation regeneration, carbon loss and, contrary to the core aim of traditional management burning, reduced landscape diversity.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Research was funded by the Scottish Government through the Scottish Wildfire Forum, Scottish Natural Heritage, The Game and Wildlife Conservation Trust and the Natural Environment Research Council. We are extremely grateful to Ralia Enterprises and Whitborough Estate for allowing us to use their land. Scott Newey, David Howarth, Alan Kirby, Harry Robertson, Carol Smithard, Isla Graham, Ellie Watts, Ella Steele, Bill Higham, Teresa Valor Ivars and Elaine Boyd helped with the practical work. We thank Wendy Anderson for many detailed and helpful comments. This work forms part of the FireBeaters project (http://firebeaters.org.uk) and we thank Michael Bruce, Chairman of the Steering Committee, and the rest of the group for their advice.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Alexander, M.E. & Sando, R.W. (1989) Fire behavior and effects in aspen-northern hardwood stands. Proceedings of the 10th Conference on Fire and Forest Meteorology, 17–21 April, Ottawa, Canada (Eds D.C.MacIver, H.Auld & R.Whitewood), pp. 263274. Environment Canada & Forestry Canada, Ottawa, Ontario.
  • Anderson, S.A.J. (2006) Future options for fire behaviour modelling and fire danger rating in New Zealand. Proceedings of the Bushfire Conference 2006: Life in a Fire-prone Environment: Translating Science into Practice, Brisbane, 6–9 June. Griffith University, Brisbane.
  • Ascoli, D., Beghin, R., Ceccato, R., Gorlier, A., Lombardi, G., Lonati, M., Marzano, R., Bovio, G. & Cavallero, A. (in press) Developing an adaptive management approach to prescribed burning: a long-term heathland conservation experiment in north-west Italy. International Journal of Wildland Fire.
  • Baeza, M., De Luis, M., Raventos, J. & Escarre, A. (2002) Factors influencing fire behaviour in shrublands of different stand ages and the implications for using prescribed burning to reduce wildfire risk. Journal of Environmental Management, 65, 199208.
  • Barker, C.G., Power, S.A., Bella, J.N.B. & Ormea, C.D.L. (2004) Effects of habitat management on heathland response to atmospheric nitrogen deposition. Biological Conservation, 120, 4152.
  • Ter Braak, C.J.F. & Šmilauer, P. (2002) Canoco Reference Manual and CanocoDraw for Windows User’s Guide. Software for Canonical Community Ordination (version 4.5). Microcomputer Power, Ithaca, NY.
  • Bruce, M.A. & Servant, G. (2003) Fire and pinewood ecology in Scotland: a summary of recent research at Glen Tanar Estate, Aberdeenshire. Scottish Forestry, 57, 3338.
  • Calvo, L., Tarrega, R. & Luis, E. (2002) Regeneration patterns in a Calluna vulgaris heathland in the Cantabrian mountains (NW Spain): effects of burning, cutting and ploughing. Acta Oecologia, 23, 8190.
  • Canadian Forest Service (undated) Canadian Wildland Fire Information System. Available at: http://cwfis.cfs.nrcan.gc.ca; last accessed 31 May 2009.
  • Catchpole, E.A. & Catchpole, W.R. (1991) Modelling moisture damping for fire spread in a mixture of live and dead fuels. International Journal of Wildland Fire, 1, 101106.
  • Catchpole, W., Bradstock, R., Choate, J., Fogarty, L., Gellie, N., McCarthy, G., McCaw, L., Marsden-Smedley, J. & Pearce, G. (1998) Cooperative development of equations for heathland fire behaviour. Proceedings of the III International Conference on Forest Fire Research and the 14th Conference on Fire and Forest Meteorology, Vol. I, Luso, Coimbra, Portugal 16–20 November (ed. D.X.Viegas), pp. 631645. University of Coimbra, Coimbra.
  • Clay, G.D., Worrall, F. & Fraser, E.D.G. (2009) Effects of managed burning upon dissolved organic carbon (DOC) in soil water and runoff water following a managed burn of a UK blanket bog. Journal of Hydrology, 367, 4151.
  • Davies, G.M. (2005) Fire behaviour and impact on heather moorlands. PhD thesis, University of Edinburgh. Available at: http://hdl.handle.net/1842/2609; last accessed 31 May 2009.
  • Davies, G.M. & Legg, C.J. (2008) The effect of traditional management burning on lichen diversity. Applied Vegetation Science, 11, 529538.
  • Davies, G.M., Legg, C.J., Smith, A. & MacDonald, A. (2006) Developing shrub fire behaviour models in an oceanic climate: burning in the British uplands. Forest Ecology and Management, 234 (Suppl. 1), S107. Full paper available at: http://hdl.handle.net/1842/2700; last accessed 31 May 2009.
  • Davies, G.M., Gray, A., Hamilton, A. & Legg, C.J. (2008a) The future of fire management in the British uplands. International Journal of Biodiversity Science and Management, 4, 127147.
  • Davies, G.M., Hamilton, A., Smith, A. & Legg, C.J. (2008b) Using visual obstruction to estimate fuel load and structure: the ‘FuelRule’ technique. International Journal of Wildland Fire, 17, 380389.
  • De Luis, M., Baeza, M.J., Raventos, J. & Gonzalez-Hidalgo, J.C.G. (2004) Fuel characteristics and fire behaviour in mature Mediterranean gorse shrublands. International Journal of Wildland Fire, 13, 7987.
  • Dimitrakopoulos, A.P. & Papaioannou, K.K. (2001) Flammability assessment of Mediterranean forest fuels. Fire Technology, 37, 143152.
  • Dimitrakopoulos, A.P., Mitsopoulos, I.D. & Raptis, D.I. (2007) Nomographs for predicting crown fire initiation in Aleppo pine (Pinus halepensis Mill.) forests. European Journal of Forest Research, 126, 555561.
  • Fernandes, P.M. (2001) Fire spread prediction in shrub fuels in Portugal. Forest Ecology and Management, 144, 6774.
  • Fernandes, P.M., Catchpole, W.R. & Rego, F.C. (2000) Shrubland fire behaviour modelling with microplot data. Canadian Journal of Forest Research, 30, 889899.
  • Gimingham, C.H. (1988) A reappraisal of cyclical processes in Calluna heath. Vegetatio, 77, 6164.
  • Gray, A. (2006) The influence of management on the vegetation and carbon fluxes of blanket bog. PhD thesis, University of Edinburgh.
  • Hamilton, A. (2000) The characteristics and effects of management fire on blanket-bog vegetation in north-west Scotland. PhD thesis, University of Edinburgh.
  • Hancock, M., Egan, S., Summers, R., Cowie, N., Amphlett, A., Rao, S. & Hamilton, A. (2005) The effect of experimental prescribed fire on the establishment of Scots pine Pinus sylvestris seedlings on heather Calluna vulgaris moorland. Forest Ecology and Management, 212, 199213.
  • Hobbs, R.J. & Gimingham, C.H. (1984) Studies on fire in Scottish heathland communities. I. Fire characteristics. Journal of Ecology, 72, 223240.
  • Holden, J., Shotbolt, L., Bonn, A., Burt, T.P., Chapman, P.J., Dougill, A.J., Fraser, E.D.G., Hubacek, K., Irvine, B., Kirkby, M.J., Reed, M.S., Prell, C., Stagl, S., Stringer, L.C., Turner, A. & Worrall, F. (2007) Environmental change in moorland landscapes. Earth-Science Reviews, 82, 75100.
  • Kayll, A.J. (1966) Some characteristics of heath fires in north-east Scotland. Journal of Applied Ecology, 3, 2940.
  • Kenworthy, J.B. (1963) Temperatures in heather burning. Nature, 200, 1226.
  • Kitchen, K., Marno, P., Legg, C., Bruce, M. & Davies, G.M. (2006) Developing a fire danger rating system for the United Kingdom. Forest Ecology and Management, 234 (Suppl. 1), S21.
  • Legg, C.J., Davies, G.M., Marno, P. & Kitchen, K. (2007). Developing a Fire Danger Rating System for the UK: FireBeaters Phase I final report. Report to the Scottish Wildfire Forum. The University of Edinburgh and The Met Office, Edinburgh.
  • Limpens, J., Berendse, F., Blodau, C., Canadell, J.G., Freeman, C., Holden, J., Roulet, N., Rydin, H. & Schaepman-Strub, G. (2008) Peatlands and the carbon cycle: from local processes to global implications – a synthesis. Biogeosciences Discussions, 5, 13791419.
  • Minitab Inc. (2003) minitab, Release 14 for Windows. Minitab Inc., State College, PA.
  • Mitchell, G. & McDonald, A.T. (1995) Catchment characterization as a tool for upland water quality management. Journal of Environmental Management, 44, 8395.
  • Mitchell, R.J., Simonson, W., Flegg, L.A., Santos, P. & Hall, J. (2009) A comparison of the resilience of four habitats to fire, and the implications of changes in community composition for conservation: a case study from the Serra de Monchique, Portugal. Plant Ecology & Diversity, 2, 4556.
  • Molina, M.J. & Llinares, J.V. (2001) Temperature-time curves at the soil surface in Maquis summer fires. International Journal of Wildland Fire, 10, 4552.
  • Morvan, D. & Dupuny, J.C. (2004) Modeling the propagation of a wildfire through a Mediterranean Shrub using a multiphase formulation. Combustion & Flame, 138, 199210.
  • National Rural Fire Authority (undated) Fire Weather. National Rural Fire Authority, New Zealand. Available at: http://nrfa.fire.org.nz/FireNet/Regions/Rural; last accessed 31 May 2009.
  • Natural England (2007) The Heather and Grass Burning Code. DEFRA, London.
  • Plucinski, M.P. & Anderson, W.R. (2008) Laboratory determination of factors influencing successful point ignition in the litter layer of shrubland vegetation. International Journal of Wildland Fire, 17, 628637.
  • Pyne, S.J., Andrews, P.L. & Laven, R.D. (1996) Introduction to Wildland Fire. John Wiley & Sons Inc., Chichester.
  • Rodwell, J.S. (ed.) (1991) British Plant Communities. Volume 2: Mires and Heaths. Cambridge University Press, Cambridge.
  • Rothermel, R.C. (1972) A Mathematical Model for Predicting Fire Spread in Wildland Fuels. USDA Forest Service Research Paper INT-115. Intermountain Forest and Range Experiment Station, Ogden, UT.
  • Sedláková, I. & Chytry, M. (1999) Regeneration patterns in a Central European dry heathland: effects of burning, sod-cutting and cutting. Plant Ecology, 143, 7787.
  • SEERAD (2001) The Muirburn Code. Scottish Executive, Edinburgh.
  • Stull, R.B. (1988) An Introduction to Boundary Layer Meteorology. Kluwer Academic, Boston, MA.
  • Sutherland, W.J., Bailey, M.J., Bainbridge, I.P., Brereton, T., Dick, J.A.T., Drewitt, J., Dulvy, N.K., Dusic, N.R., Freckleton, R.P., Gaston, K.J., Gilder, P.M., Green, R.E., Heathwaite, A.L., Johnson, S.M., Macdonald, D.W., Mitchell, R., Osborn, D., Owen, R.P., Pretty, J., Prior, S.V., Prosser, H., Pullin, A.S., Rose, P., Stott, A., Tew, T., Thomas, C.D., Thompson, D.B.A., Vickery, J.A., Walker, M., Walmsley, C., Warrington, S., Watkinson, A.R., Williams, R.J., Woodroffe, R. & Woodroof, H.J. (2008) Future novel threats and opportunities facing UK biodiversity identified by horizon scanning. Journal of Applied Ecology, 45, 821833.
  • Tharme, A.P., Green, R.E., Baines, D., Bainbridge, I.P. & O’Brien, M. (2001) The effect of management for red grouse shooting on the population density of breeding birds on heather-moorland. Journal of Applied Ecology, 38, 439457.
  • Thomas, P.H. (1971) Rates of spread of some wind-driven fires. Forestry, 44, 155175.
  • Thompson, D.B.A., MacDonald, A.J., Marsden, J.H. & Galbraith, C.A. (1995) Upland heather moorland in Great Britain: a review of international importance, vegetation change and some objectives for nature conservation. Biological Conservation, 71, 163178.
  • Van Wagner, C.E. & Methven, I.R. (1978) Discussion: Two recent articles on fire ecology. Canadian Journal of Forest Research, 8, 491492.
  • Vandvik, V., Heegaard, E., Måren, I.E. & Aarrestad, P.A. (2005) Managing heterogeneity: the importance of grazing and environmental variation on post-fire succession in heathlands. Journal of Applied Ecology, 42, 139149.
  • Weise, D.R., Zhou, X., Sun, L. & Mahalingam, S. (2005) Fire spread in chaparral –‘go or no go?’ International Journal of Wildland Fire, 14, 99106.
  • Whittaker, E. (1961) Temperatures in heath fires. Journal of Ecology, 49, 709715.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Detailed diagram of experimental plot layout

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
JPE_1681_sm_AppendixS1.doc31KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.