Register      Login
International Journal of Wildland Fire International Journal of Wildland Fire Society
Journal of the International Association of Wildland Fire
RESEARCH ARTICLE (Open Access)

The 1986 Annaburroo experimental grassland fires: data

James S. Gould A , Miguel G. Cruz https://orcid.org/0000-0003-3311-7582 A and Andrew L. Sullivan https://orcid.org/0000-0002-8038-8724 A *
+ Author Affiliations
- Author Affiliations

A CSIRO, GPO Box 1700, Canberra, ACT 2601, Australia.

* Correspondence to: Andrew.Sullivan@csiro.au

International Journal of Wildland Fire 33, WF23100 https://doi.org/10.1071/WF23100
Submitted: 21 June 2023  Accepted: 25 April 2024  Published: 13 May 2024

© 2024 The Author(s) (or their employer(s)). Published by CSIRO Publishing on behalf of IAWF. This is an open access article distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND)

Abstract

Background

In 1986, CSIRO conducted a large program of experimental fires in grassland at Annaburroo Station, Northern Territory, Australia, with the objective of quantifying the effect of fuel condition (load and height) on fire behaviour.

Aims

This paper provides the data collected during this program, representing a unique set of observations and measurements of large, free-burning experimental fires conducted in a multi-factor experimental design.

Methods

Data are collated by experimental burn plot, providing detailed measurements of weather (wind speed, air temperature, relative humidity), fuel state (load, height, moisture content, curing) and fire behaviour (rate of spread, flame depth, flame height, head fire width), as well as processed information (e.g. steady-state rate of spread).

Data availability

The data are made available for free download on the CSIRO Data Access Portal (https://data.csiro.au/collection/csiro:58746) and include detailed metadata descriptions of the data and their structure, also provided in this article.

Conclusions

We have made the data available for fire behaviour researchers around the world to use in their research under the Creative Commons Attributions licence. It is hoped they will analyse these data and extract new and innovative insights to help improve our understanding of wildland fires burning in grass fuels.

Keywords: Experimental fires, fire behaviour, fuel state, grassland fire.

Introduction

Our current understanding of wildland fire behaviour dynamics is underpinned by empirical evidence (e.g. Van Wagner 1977; Cheney et al. 1993; Cheney and Gould 1995a, 1995b; Fernandes et al. 2008; McCaw et al. 2012; Finney et al. 2018). Empirical data, the basis of an evidence-based scientific approach, provide insights into the physical processes and mechanisms driving a phenomenon and ensure theories and hypothesis can be objectively evaluated. As fire is a scale-dependent phenomenon (Fons et al. 1963; Byram 1966), the study of wildland fire requires field-based experiments at a scale (e.g. size, intensity) that ensures the key processes governing behaviour and spread are captured. The planning and safe execution of well-thought-out experimental fire research projects that adequately test hypotheses (e.g. Cheney et al. 1993; Stocks et al. 2004; Gould et al. 2007; McCaw et al. 2012; Finney et al. 2018; Katurji et al. 2023), particularly of high-intensity wildfire propagation, are often constrained by legal, operational and safety regulations. This largely explains why the field of wildland fire behaviour is data-poor. The open sharing of experimental fire data is a necessary component for the advancement of fire behaviour science.

The CSIRO (Commonwealth Scientific and Industrial Research Organisation) conducted a large experimental burning program during the dry season of July and August 1986 at Annaburroo Station, 100 km south-east of Darwin, in the Northern Territory, Australia, in collaboration with a significant number of Australian State and Territory rural fire and land management agencies (Cheney et al. 1989). These experiments were initially intended to evaluate the effect of grass fuel attributes (such as fuel load, fuel bed height, fuel particle size, moisture content) on the forward rate of fire spread.

The major direct outcomes of this research were the separation of the calculation of fire rate of spread from that of fire danger in the McArthur (1973) Grassland Fire Danger Meter (Cheney and Gould 1995a) and development of new fire spread models (CSIRO 1997; Cheney et al. 1998; Cheney and Sullivan 2008) to supplant that of McArthur for fires in grasslands. These models are currently the recommended models for use in grasslands in Australia for fire behaviour prediction (Cruz et al. 2015; Plucinski et al. 2017) and recently have also been implemented in the new Australian Fire Danger Rating System (Hollis et al. 2024). Other outcomes included analysis of the effect of fuel load on fire spread and flame characteristics (Cheney et al. 1993), the effect of fireline width on fire growth and acceleration (Cheney and Gould 1995b, 1997), the relevance of non-dimensional convection numbers on grassland fire dynamics (Sullivan 2007), and future grassland fire management under the impacts of climate change (Sullivan 2010). This dataset was also used to parameterise grassland fire spread models in the Canadian Fire Behaviour Prediction System (Forestry Canada Fire Danger Group 1992). A small subset of the dataset, comprising detailed fire propagation isochrones and associated burning conditions, has been made available to a number of researchers to calibrate and test physical based models, such as the Wildland–Urban Interface Fire Dynamics Simulator (Mell et al. 2007; Moinuddin et al. 2018), FIRETEC (Dupuy et al. 2011) and FIRESTAR (Morvan et al. 2009; Morvan 2011). More recently, Khanmohammadi et al. (2022) used the full dataset to train different machine learning algorithms and evaluate their suitability to predict the spread rate of grassfires.

The aim of this paper is to make the Annaburroo experimental burn program data available to the international fire behaviour research community, and to foster its use by fire behaviour modellers. Use and analysis of this dataset should be preceded by a careful reading of Cheney et al. (1989, 1993, 1998), Cheney and Gould (1995b), Durre and Beer (1989) and Gould (1991).

The data

We present here the base environmental and fire behaviour data associated with the free-burning experimental fires carried out in open grasslands at Annaburroo Station (the Annaburroo Grassland Experimental Fire dataset). Data are provided at different spatial and temporal scales: at the original sample scale (e.g. 2-m wind speed at 5-s intervals, fuel load at the sample quadrat location), observation period (average per burning observation period within each fire) and plot (i.e. averaged for each experimental fire). Data are made available for free download from the CSIRO Data Access Portal (DAP, https://data.csiro.au/collection/csiro:58746). Fig. 1 illustrates the structure of the data files in the repository. Under the main root folder are two files that collate fire and environmental data by burn observation period (Burn_Period_Data.csv) and by fire experiment (Plot_Average_Data.csv). Burn observation period is defined by the timing of oblique aerial photography taken during each experimental fire at approximately 20–50 s intervals (see Fig. 2 for an example) – continuous observation of fire behaviour was not feasible at the time of the experiments. Plot average is the average value for all variables for each experimental fire. An additional file, AWS_Data.csv, provides the meteorological data for each burn day.

Fig. 1.

Annaburroo Experimental Grassland Fire Data file structure and naming convention. FirePlotData consists of 120 folders, one for each fire plot encoded with the fire plot ID given as one letter and three digits, i.e. annn. Each FirePlotData folder consists of wind speed measured at 2 m at each plot corner, a series of oblique aerial photographs illustrating the spread of the fire at semi-regular periods, defining the spread interval and annotated in the filename with the time of the photo in hhmmss (hour, minutes, seconds) format, a map of the fire isochrones at each spread interval, and a map of the fuel attributes as measured across each plot prior to ignition. The folder ‘Site and block maps’ contains maps of the location of the experimental site including the location of the automatic weather station (AWS), as well as detailed maps of the layout of each experimental block. AWS_Data.csv consists of meteorological data collected at the AWS for each burn day during the burn program. Burn_Period_Data.csv collates the data for each fire plot by each burn period (interval) while Plot_Average_Data.csv provides the average of these values for each fire plot. Numerical data are provided as comma-separated values in ASCII text files (csv file format) whereas images (photos and isopleth maps) are provided as JPEG format files.


WF23100_F1.gif
Fig. 2.

Example series of oblique aerial photographs (ae) showing growth of experimental fire B111 and a composite map of fire perimeter isochrones for experimental fire B111 after rectification to planar view (f). Ignition took place at 13:26:10 hours. Photograph file names indicate plot number (as annn) and time photograph was taken (as hhmmss: hour, minutes, seconds).


WF23100_F2.gif

The Site and block maps folder contains, as PNG image files, maps of the experimental site as well as individual maps of each experimental block within the site. Each experimental block was further subdivided into experimental plots.

The FirePlotData folder consists of 120 folders, one for each experimental fire encoded with the fire plot ID given as one letter and three digits (annn). Each of these folders contains a number of files with the basic data for each experimental fire, namely: (1) the annn_2mwind.csv file with the wind speed measured at 2 m at each plot corner; (2) the series of oblique aerial photographs illustrating the spread of the fire at semi-regular periods defining the burn period observation and spread interval and annotated in the filename with the time of the photo in hhmmss format (e.g. Fig. 2ae; annn_hhmmss.jpg); (3) a geo-rectified map of the fire isochrones at each spread interval (Fig. 2f; annn_FireIsochrones.jpg); and (4) a map of the transects of fuel height and load samples (annn_FuelMap.jpg) as measured in each plot prior to ignition. Each aerial photograph defines the start or the end of a burn observation period used in file Burn_Period_Data.csv. In this file, data on environmental and fire behaviour variables have been processed and averaged per identified burn observation period for each individual experimental fire. Plot-level data that aggregates each experiment burn observation period data into a single plot level average are provided in the Plot_Average_Data.csv file.

Numerical data are provided as comma-separated values in ASCII text files (csv file format) while images (photos and isopleth maps) are provided as JPEG image format files. Table 1 lists the variables provided in the burn observation period and plot average data files (Burn_Period_Data.csv and Plot_Average_Data.csv). Table 2 lists the variables provided in the site AWS and plot wind data files (AWS_Data.csv and annn_2mwind.csv).

Table 1.List of fuel, weather and fire behaviour variables collected during the Annaburroo grassland fire experiments and provided in burn period and plot average data files (Burn_Period_Data.csv and Plot_Average_Data.csv).

VariableDescriptionBurn period dataPlot average data
FireIDIndividual fire plot identification
Date.yymmddDate of experimental fire, yy, year; mm, month; dd, day
IgnTime.hhmmssTime of ignition, hh, hour; mm, minutes; ss, seconds
IgnLngth.mIgnition line length (m), 0 = point ignition fires
2mWind.ms2-m wind speed (m s−1); average of the four anemometers located on the corners of the experimental plot
MidFlWind.msMid-flame wind speed calculated from 2-m average wind speed and the wind speed profile at the time of the measurement (m s−1)
10mWind.msWind speed at 10 m in the open as an extrapolation of the 2-m average wind and the wind speed profile at the time of the measurement (m s−1)
NE2mWind.ms2-m wind speed (m s−1) measured at the NE corner of the experimental plot
NW2mWind.ms2-m wind speed (m s−1) measured at the NW corner of the experimental plot
SW2mWind.ms2-m wind speed (m s−1) measured at the SW corner of the experimental plot
SE2mWind.ms2-m wind speed (m s−1) measured at the SE corner of the experimental plot
Temp.CAir temperature (°C) at 1.4 m during fire from the AWS weather tower
RH.%Relative humidity (%) at 1.4 m during the fire from the AWS weather tower
CloudCov.8Cloud cover (visual okta estimates), e.g. 2/8 represents 2 oktas (two-eighths cloud cover)
FuelTypeFuel type: ER, Eriachne sp.; TH, Themeda triandra
FuelTreatFuel treatment
E1 = Eriachne sp., natural/undisturbed
E2 = Eriachne sp., grass cut at 50% of natural height and left on site
E3 = Eriachne sp., grass cut at 50% of natural height and removed from site
T1 = Themeda triandra, natural undisturbed
T2 = Themeda triandra, grass cut at 50–25% of natural height and left on site
T3 = Themeda triandra, grass cut at 50–25% of natural height and removed from site
HeadFShapeHead fire shape
1 = Pointed shape
2 = Parabolic shape
FuelHt.mFuel height (m)
FuelLoad.kgm2Fuel load dry mass (kg m−2)
SAVR.cmSurface area-to-volume ratio (cm−1)
Cure.%Degree of grass curing (%) visual estimate
DFMC.%Dead fuel moisture content (%) measured
BurnPeriodEnd.hhmmssTime of isopleth fire perimeter, hh, hour; mm, minutes; ss, seconds
CumDist.mDistance head fire travel from ignition (m)
IncDist.mIncrement distance between fire perimeter isopleths (m)
FlameDpthAir.mFlame depth interpreted from oblique aerial photographs (m)
FlameDpthGnd.mFlame depth (m) from ground observations
FlameHt.mFlame height (m) from ground observations
FlameAng.degFlame angle (°) from ground observations
FireDir.degBearing of head fire travel (°)
TimeSinceIgnFire perimeter time since ignition (s)
MidTimeMid-point time between fire spread intervals (s)
ROS.msForward rate of spread (m s−1)
EffHFWidth.mEffective head fire width (m) calculated as per Cheney and Gould (1995b)
RSS.msSteady-state rate of fire spread calculated from observed rate of fire spread and effective width of the fire front (EffHFWidth.m) (m s−1); see Cheney and Gould (1995b)

Plot average data are generally the average of all the burn period data for a given plot. In some cases, specific burn periods were omitted from the plot average value owing to missing or non-representative data.

Table 2.List of variables given in AWS_Data.csv and annn_2mwind.csv.

VariableDescriptionAWS_Data.csvannn_2mwind.csv
AWSDate.yymmddDate, yy, year; mm, month; dd, day
AWSTime.hhmmssTime, hh, hour; mm, minutes; ss, seconds
2mAWSWind.msAutomatic weather station 2-m wind speed (m s−1)
10mAWSWind.msAutomatic weather station 10-m wind speed (m s−1)
10mAWSDir.degAutomatic weather station 10-m wind direction (°)
Temp.CAir temperature (°C) at 1.4 m during fire from the AWS weather tower
10mTemp.CAir temperature (°C) measured at the AWS at 10-m  height
RH.%Relative humidity (%) at 1.4 m during the fire from the AWS weather tower
Insol.kWInsolation (kW m−2) measured at the AWS at 1.4 m height
Time.hhmmssTime, hh, hour; mm, minutes; ss, seconds
ExpPeriodTiming of measurement relative to experiment. Pre, measurement prior to ignition; fire, measurement during the experiment; post, measurement after experiment concluded
NW2mWind.ms2-m wind speed (m s−1) measured at the NW corner of the experimental plot
NE2mWind.ms2-m wind speed (m s−1) measured at the NE corner of the experimental plot
SE2mWind.ms2-m wind speed (m s−1) measured at the SE corner of the experimental plot
SW2mWind.ms2-m wind speed (m s−1) measured at the SW corner of the experimental plot

Weather data in AWS_Data.csv were recorded at 1-min intervals. Wind speed data in annn_2mwind.csv were recorded at 5-s intervals. Wind speed data are the average value recorded over the sampling period.

Experimental design

The experiments took place at Annaburroo station, NT, Australia (12°54′25″S, 131°40′02″E, Fig. 3). The site is a broad flat area in the flood plain of the Mary River. The area was divided in 10 main blocks (A–J in Fig. 3), and then subdivided into 170 plots. Of these, 121 fitted the requirements of the study (plots with trees or inadequate fuel distribution were excluded). The experimental plots had two sizes: 100 × 100 m and 200 × 200 m. Fuel breaks of bare mineral earth 1.5 or 3 m wide separated the plots. The experiments were carried out during July and August 1986. This period is characteristically the middle of the dry season at the location. Grasses are normally fully cured, and prevailing weather patterns ensure a predominance of warm and dry easterly winds over the area. Rain events during this period are infrequent. A wide buffer area was burnt around the perimeter of the experimental area to reduce the risk of escape of experimental fires even under heightened fire danger conditions. Table 3 summarises the means and ranges of weather, fuel and fire behaviour data collected during the experimental program. Table A1 lists all the experimental fires by date.

Fig. 3.

Map of the layout of the experimental blocks (A–J) and 200 × 200 m experimental plots within each block at Annaburroo Station. Some plots were further divided into 100 × 100 m plots. Each main plot was defined by a 3-m wide graded bare-earth break. Inset map indicates the location (star) of the experimental site in the Northern Territory, Australia.


WF23100_F3.gif
Table 3.Mean and range weather, fuel and fire behaviour data in the burn observation period data file (Burn_Period_Data.csv).

VariableMeanMinimumMaximum
Weather
 Air temperature (°C)30.223.038.0
 Relative humidity (%)29.713.055.0
 NE 2-m wind speed (m s−1)3.800.08.3
 NW 2-m wind speed (m s−1)4.150.310.0
 SW 2-m wind speed (m s−1)4.140.011.0
 SE 2-m wind speed (m s−1)3.700.07.7
Fuel
 Degree curing (%)91.580100
 Fuel bed height (m)0.290.070.59
 Fuel Load (kg m−2)0.350.120.61
 Surface area-to-volume ratioER = 97.7 cm−1
TH = 122.4 cm−1
 Dead fuel moisture content (%)6.802.712.1
Fire behaviour
 Ignition line length (m)65.90200
 Rate of fire spread (m s−1)0.870.024.88
 Flame height (m)2.20.36.0
 Flame angle (°)47.910120
 Flame depth (m)5.20.112.0

For a detailed breakdown of fuel bed variables with fuel treatment, please see Table 3 in Cheney et al. (1993).

Fuel condition and sampling

The selected site contained two distinct grass species, Eriachne burkittii Jansen (commonly kerosene grass) and Themeda triandra Forrsk. (previously known as Themeda australis R.Br, commonly kangaroo grass) that provided fuel beds of dissimilar characteristics. Further manipulation of the fuel bed was achieved by controlling grass height (by cutting at 50 and 25% of the undisturbed height) and fuel quantity (by removing or leaving the cut fuel). This yielded six different fuel bed treatments (see Treatment in Table 1). Treatments were randomly assigned to each plot.

Surface-area-to-volume ratio for the two major species was measured on selected plants. The diameter of stalks and thickness of leaves were measured with a micrometer and converted to a surface-area-to-volume ratio following Brown (1970). The mean surface-area-to-volume of a plant was estimated as the average value of leaves and stalk component, weighted by their length (Gould 1991).

Fuel bed structure at an experimental plot level was sampled through a 16-point grid located within the plot (Fig. 4). At each sample point, the mean height of the surrounding grass sward was measured and a destructive sample (0.3 × 0.6 m quadrat) of all standing and matted fuel collected. Each fuel sample was oven-dried at a nominal temperature of 104°C for 24 h to determine its dry weight, expressed as the mass per unit area in units of kilograms per square metre. Plots with large fuel heterogeneity or discontinuous fuel cover were excluded from burning or from analysis.

Fig. 4.

Experimental fire B111 fuel plot grid laid over the fire isochrone map.


WF23100_F4.gif

Weather data

An automatic meteorological station (AWS) was located at a fixed site within the experimental area for the duration of the experimental program (see Fig. 3). At this AWS, measurements of wind speed were made at 10 and 2 m above ground. Wind speed measurements used UNIDATA anemometers that were calibrated in the CSIRO Division of Atmospheric Research wind tunnel in Aspendale, Victoria. Wind direction was measured at the 10-m height using a UNIDATA wind vane. Wind speed and direction data were recorded as 1-min averages.

Air temperature, relative humidity and solar radiation at the AWS were measured at 1.4-m height. Air temperature was measured using a solid-state temperature probe (UNIDATA AD537) that was shielded and aspirated with an encapsulated UNIDATA 2B57 pre-amplifier. Relative humidity was measured with a Philips Humidicap sensor. Solar radiation was measured with a pyranometer (make and model unknown), calibrated using the maximum recorded value on cloudless days. The temperature probe was calibrated in a temperature-controlled oil bath over the range 20–40°C. The relative humidity sensor was calibrated against field readings using an Assman psychrometer. Air temperature, relative humidity and solar radiation were measured and recorded in UNIDATA dataloggers at 1-min intervals.

On selected days, vertical temperature measurements were taken every 150 m from the ground up to ~1400 m using a helicopter with either an Assman psychrometer or the helicopter’s temperature probe. On most days, the data revealed a super adiabatic layer near the ground up to 150 m, a well-mixed layer aloft to ~800 m and an inversion above that (Durre and Beer 1989).

For each experimental fire, wind speed at 2-m height was measured at each plot corner (NW, NE, SE, SW). Purpose-built three-cup sensitive anemometers (Bradley 1969) with a start-up speed of 0.05 m s−1 were used for these measurements, with the anemometers being calibrated in the wind tunnel as described above. These wind speeds were recorded on pre-synchronised UNIDATA data loggers at 5-s intervals, with the recorded value being the average of the 5-s period. Fig. 5 illustrates the variability of the wind speed measured before, during and after an experimental fire.

Fig. 5.

Wind speed variability as measured at the four anemometer locations prior to, during and after experiment B111.


WF23100_F5.gif

Fuel moisture content and curing level

Four samples of fully cured grasses were collected in, or within a few metres of, each experimental plot for determination of fuel moisture content representative of the conditions of the experimental fire. Samples (approximately 30 g each) were collected in an undisturbed area where boundary effects, such as the existence of a fire break, were not present. Two samples were collected immediately prior to ignition and two after the experimental fire was concluded. The samples were stored in a hermetically sealed container to ensure no moisture loss occurred prior to determination of wet weight. Wet weight was determined within hours of sample collection using a sensitive balance, and then the sample was oven-dried at 104°C for 24 h and reweighed. Fuel moisture was calculated as a fraction of oven-dried weight (see Matthews 2010). The average of the four measurements expressed the dead fuel moisture content of each experimental fire.

Curing level was initially not a variable in the experimental burn program as it was conducted during the dry season in uniformly fully cured grasses. However, an unseasonal rain event just prior to commencement of the burning experiments caused germination of green shoots underneath the fully cured grass midway through the burning experiment. Subsequently, curing level for each experimental plot was visually estimated by the two senior scientists on site based on guidelines developed from previous experience in determination of grass curing through destructive sampling methods. It was felt by observers that the new growth under the taller fully cured grass had minimal effect on fire behaviour as the fire propagated through the taller dead fuels.

Fire behaviour data

Ignition

The database comprises experimental fires of two types of ignitions: lines (n = 121 experimental fires) and points (n = 24). Line ignitions aimed to rapidly produce a fire spreading at a quasi-steady-state rate of spread for the prevailing conditions. Point source ignitions aimed to create fires where the dynamics of fire acceleration could be studied. A line ignition was created by two igniters with drip torches who started at the mid-point of the line on the upwind edge of the plot and moved to opposite ends of the line at a speed of ~2 m s−1. As the line was oriented at right angles to the prevailing wind, the technique allowed the rapid development of a wide and deep headfire region (Fig. 2a). The length of ignition line varied with plot area. The 100 × 100 m plots generally had an ignition line of 50 m and took 15–20 s to complete, whereas the 200 × 200 m plots generally had an ignition line of 100 m and took 30–40 s. Differences in plot characteristics, fuel distribution and occasional operator error sometimes resulted in small variations in completed ignition line length. It was these variations in ignition line length that later aided identification of the importance of head fire width in fire spread (Cheney et al. 1993).

Point ignition experiments were conducted in conjunction with the line ignition experiments in the larger plots, typically when plot width was ~200 m, and utilised an unburnt area of these plots. Point source ignitions were created using a common matchstick.

Fire behaviour monitoring

Fire behaviour observations were recorded at 1-min intervals by experienced ground observers accompanying the fire progression. These observers identified features such as flame dimensions (depth, height and angle; see Fig. 6), occurrence of spot fire ignitions ahead of the main fire, wind changes and up- and down-draughts. Low-altitude, oblique aerial photographs were taken from a helicopter at 15–30 s intervals (Fig. 2ae). The photographs were taken from a position that produced a depression angle of approximately 45° to the principal point (approximately the centre of the plot) and 60° to the base of the plot.

Fig. 6.

Cross-section of stylised wind-driven flame front depicting the three flame characteristics measured. Flame angles >90° indicate flame is leaning towards burnt ground.


WF23100_F6.gif
Fire progression mapping and rate of spread calculation

The oblique photographs were interpreted for actual ignition line length, location of the leading edge and rear of the flame front, fire progression, flame depth and identification of ground control points (e.g. centre points of the track intersections, other selected features) for later image rectification and analysis. A custom computer program generated a geometric transformation based on the apparent ground control points (the corners of each plot) using the measured distances between each point. This transformation was then applied in the program to each oblique fire perimeter map to convert it into a planar map. A composite map of time isopleths of fire perimeter (Fig. 2f) and flame depth was then plotted for each experimental fire. The effective head fire width was calculated as the width of the fire front that influenced the next period of head fire spread (Cheney and Gould 1995b). Rate of forward spread was taken as the maximum distance that the head fire advanced between successive fire isochrones. The original fire isochrone maps were digitally scanned and reference scale and isopleths time stamps added (e.g. Fig. 2f). Interested users can utilise state-of-the-art methods to extract relative flame front positions from the oblique aerial imagery for further analysis and perhaps generate new fire progression data.

Data interpretation and intermediate calculations

The process of interpreting fire propagation and averaging its behaviour over a burn observation period and for a single experimental fire requires some explanation.

A burn observation period was defined as the interval between two sequential oblique aerial photographs depicting the location of the flame front. From this, measurements of rate of fire spread and effective head fire width were taken. The 2-m wind speed measurements made at each of the four plot corners were then averaged over this observation period. Careful examination of the anemometer data was conducted to ensure they were representative of the wind field driving the fire. Data from individual anemometers were removed if the location of the anemometer was influenced by external factors such as the presence of trees, firefighting vehicles, or the approaching flame front. A vertical wind profile and estimate of atmospheric stability were calculated at the AWS for each experimental fire (see Durre and Beer (1989) for details). This wind profile, along with the measured flame characteristics and 2-m wind speeds, were used to calculate a representative 10-m open wind speed and a mid-flame height wind speed for each experiment.

The fuel sample grid was overlayed on each of the fire isochrone maps. The mean fuel load and fuel height for each fire plot was calculated from the sample points within the burnt area (Fig. 4).

Conclusions

Nearly 40 years ago, CSIRO bushfire researchers led by Phil Cheney and Jim Gould undertook the most comprehensive study of behaviour of free-burning fires in grasslands. Although the data collected during these experiments had immediate impact, namely through the development of new grassland fire spread models that now are incorporated into Australia’s fire danger rating system and fire behaviour prediction tools, the potential of this dataset for further exploring the behaviour of grassland fires has been barely touched. It is hoped that with the publication of this dataset under a Creative Commons Attribution licence, fire researchers around the world will be able to apply creative and innovative analysis techniques to generate new insights about the behaviour and spread of one of the world’s fastest burning fuel types.

Data availability

The data presented in this publication are available from: https://data.csiro.au/collection/csiro:58746.

Conflicts of interest

Andrew Sullivan is an Associate Editor of the International Journal of Wildland Fire and to mitigate this potential conflict of interest he had no editor-level access to this manuscript during peer review. The authors have no further conflicts of interest to declare.

Declaration of funding

The collation and preparation of these data for publication was funded by the CSIRO.

Acknowledgements

The publication of this dataset could not have happened without the huge efforts of all those who contributed to the execution of the experiments and helped collect the data: Phil Cheney, Peter Hutchings, Ian Knight, John Baxter, Tom Beer, Matt Dando, Mark Durré, Edwin Mak, Peter Vis, Andrew Wilson, Marion Fallon, Rob Birtles (CSIRO), Mark Dawson (South Australia Country Fire Service), Ian Smith (Tasmanian Fire Service), Bruce Ward (Western Australia Department of Conservation and Land Management), Fiona Hamilton (Victorian Department of Conservation, Forests and Lands), Chris Trevitt (Australian National University), L. Jayawardene (Sri Lanka), and the extraordinary efforts of the staff of the Bush Fires Council NT who undertook preparation of the experimental area and fire suppression, Tony Carmody (helicopter pilot) and Jim Sullivan (fuel treatment harvester).

References

Bradley EF (1969) Small sensitive anemometer system for agricultural meteorology. Agricultural Meteorology 6, 185-193.
| Crossref | Google Scholar |

Byram GM (1966) Scaling laws for modeling mass fires. Pyrodynamics 4, 271-284.
| Google Scholar |

Brown JK (1970) Ratios of surface area to volume for common fine fuels. Forest Science 16, 101-105.
| Google Scholar |

Cheney NP, Gould JS (1995a) Separating fire spread prediction and fire danger rating. CALMScience Supplement 4, 3-8.
| Google Scholar |

Cheney NP, Gould JS (1995b) Fire growth in grassland fuels. International Journal of Wildland Fire 5, 237-247.
| Crossref | Google Scholar |

Cheney NP, Gould JS (1997) Fire growth and acceleration. International Journal of Wildland Fire 7, 1-5.
| Crossref | Google Scholar |

Cheney NP, Sullivan AL (2008) ‘Grasslands: Fuel, weather and fire behaviour.’ 2nd edn. p. 150. (CSIRO Publishing: Melbourne, Vic., Australia)

Cheney NP, Gould JS, Hutchings PT (1989) ‘Prediction of fire spread in grassland.’ (CSIRO: Canberra, ACT, Australia)

Cheney NP, Gould JS, Catchpole WR (1993) The influence of fuel, weather and fire shape variables on fire-spread in grasslands. International Journal of Wildland Fire 3, 31-44.
| Crossref | Google Scholar |

Cheney NP, Gould JS, Catchpole WR (1998) Prediction of fire spread in grassland. International Journal of Wildland Fire 8, 1-13.
| Crossref | Google Scholar |

Cruz MG, Gould JS, Alexander ME, Sullivan AL, McCaw WL, Matthews S (2015) Empirical-based models for predicting head-fire rate of spread in Australian fuel types. Australian Forestry 78, 118-158.
| Crossref | Google Scholar |

CSIRO (1997) ‘CSIRO grassland fire spread meter.’ (Styrox Pty Ltd.)

Dupuy JL, Linn RR, Konovalov V, Pimont F, Vega JA, Jiménez E (2011) Exploring three-dimensional coupled fire–atmosphere interactions downwind of wind-driven surface fires and their influence on backfires using the HIGRAD-FIRETEC model. International Journal of Wildland Fire 20, 734-750.
| Crossref | Google Scholar |

Durre AM, Beer T (1989) ‘Wind information prediction study: Annaburroo meteorological data analysis’. Research Technical Paper 17. (CSIRO Atmospheric Research)

Fernandes PM, Botelho H, Rego F, Loureiro C (2008) Using fuel and weather variables to predict the sustainability of surface fire spread in maritime pine stands. Canadian Journal of Forest Research 38, 190-201.
| Crossref | Google Scholar |

Finney M, Pearce G, Strand T, Katurji M, Clements C (2018) New Zealand prescribed fire experiments to test convective heat transfer in wildland fires. In ‘Advances in Forest Fire Research 2018. Proceedings of the VII International Conference on Forest Fire Research’. pp. 10–16. (Universidade de Coimbra: Portugal)

Fons WL, Clements HB, George PM (1963) Scale effects on propagation rate of laboratory crib fires. Symposium (International) on Combustion 9, 860-866.
| Crossref | Google Scholar |

Forestry Canada Fire Danger Group (1992) Development and structure of the Canadian Forest Fire Behavior Prediction System. Information Report ST-X-3. (Forestry Canada, Science and Sustainable Development Directorate: Ottawa, ON)

Gould JS (1991) Validation of the Rothermel fire spread model and related fuel parameters in grassland fuels. In ‘Proceedings of Conference on Bushfire Modelling and Fire Danger Rating Systems’, Yarralumla, ACT. 11–12 July 1988, pp. 51–64. (Eds NP Cheney, AM Gill) (CSIRO Division of Forestry: Canberra, ACT)

Gould JS, McCaw WL, Cheney NP, Ellis PE, Knight IK, Sullivan AL (2007) ‘Project Vesta – fire in dry eucalypt forest: fuel structure, fuel dynamics, and fire behaviour.’ (Ensis-CSIRO: Canberra, ACT and Department of Environment and Conservation: Perth, WA)

Hollis JJ, Matthews S, Fox-Hughes P, Grootemaat S, Heemstra S, Kenny BJ, Sauvage S (2024) Introduction to the Australian Fire Danger Rating System. International Journal of Wildland Fire 33(3),.
| Crossref | Google Scholar |

Katurji M, Noonan B, Zhang J, Valencia A, Schumacher B, Kerr J, Strand T, Pearce G, Zawar-Reza P (2023) Atmospheric turbulent structures and fire sweeps during shrub fires and implications for flaming zone behaviour. International Journal of Wildland Fire 32, 43-55.
| Crossref | Google Scholar |

Khanmohammadi S, Arashpour M, Golafshani EM, Cruz MG, Rajabifard A, Bai Y (2022) Prediction of wildfire rate of spread in grasslands using machine learning methods. Environmental Modelling & Software 156, 105507.
| Crossref | Google Scholar |

Matthews S (2010) Effect of drying temperature on fuel moisture content measurements. International Journal of Wildland Fire 19, 800-802.
| Crossref | Google Scholar |

McArthur AG (1973) ‘Grassland Fire Danger Meter Mk IV.’ (Commonwealth of Australia, Forestry and Timber Bureau: Canberra, ACT)

McCaw WL, Gould JS, Cheney NP, Ellis PFM, Anderson WR (2012) Changes in behaviour of fire in dry eucalypt forest as fuel increases with age. Forest Ecology and Management 271, 170-181.
| Crossref | Google Scholar |

Mell W, Jenkins MA, Gould J, Cheney P (2007) A physics-based approach to modelling grassland fires. International Journal of Wildland Fire 16, 1-22.
| Crossref | Google Scholar |

Moinuddin KAM, Sutherland D, Mell W (2018) Simulation study of grass fire using a physics-based model: striving towards numerical rigour and the effect of grass height on the rate of spread. International Journal of Wildland Fire 27, 800-814.
| Crossref | Google Scholar |

Morvan D (2011) Physical phenomena and length scales governing the behaviour of wildfires: a case for physical modelling. Fire Technology 47, 437-460.
| Crossref | Google Scholar |

Morvan D, Méradji S, Accary G (2009) Physical modelling of fire spread in grasslands. Fire Safety Journal 44, 50-61.
| Crossref | Google Scholar |

Plucinski MP, Sullivan AL, Rucinski CJ, Prakash M (2017) Improving the reliability and utility of operational bushfire behaviour predictions in Australian vegetation. Environmental Modelling & Software 91, 1-12.
| Crossref | Google Scholar |

Stocks BJ, Alexander ME, Lanoville RA (2004) Overview of the International Crown Fire Modelling Experiment (ICFME). Canadian Journal of Forest Research 34, 1543-1547.
| Crossref | Google Scholar |

Sullivan AL (2007) Convective Froude number and Byram’s energy criterion of Australian experimental grassland fires. Proceedings of the Combustion Institute 31, 2557-2564.
| Crossref | Google Scholar |

Sullivan AL (2010) Grassland fire management in future climate. Advances in Agronomy 106, 173-208.
| Crossref | Google Scholar |

Van Wagner CE (1977) Conditions for the start and spread of crown fire. Canadian Journal of Forest Research 7, 23-34.
| Crossref | Google Scholar |

Appendix

Table A1.List of experimental fires by date.

Burn dateFire plot identifier (fire.id)
30/7/1986A011, A012, A022, A052, A053, A061, A066
31/7/1986A023, A031, A034
1/8/1986A121
4/8/1986A032, A033, A041, A071, A072, A073, A074, A081, A082, A083, A084, A091, A092, A141, A142, A143, A144, A151, A154
5/8/1986A152, A153, B032, B033, B041, B042, B043, B044
6/8/1986B091, B092, B093, B094, B101, B102, B103, B104, B111, B114
14/8/1986E14
15/8/1986B173, B184, B201, B202, B211, B214
16/8/1986B212, B213, B221, B224, E04, E10
17/8/1986E03, E05, E09, E11, E15, E16, E17, E18, E21, E22
18/8/1986B241, B242, B243, B244, B301, B302, B303, B034, C011, C012, C013
C014, C061, C062, C063, C064, C111, C112
19/8/1986C071, C081, E06, E12, F011, F02, F061, F071, F11, F16, F211, F212, F221, F222
20/8/1986F03, F04, F081, F13, F14, F18, F19, F231, F232
21/8/1986C04, C05, C09, C10, F05, F10, F12, F15, F17