Three Parks Savanna Fire-effects Plot Network: Plot-based Fire Severity Associations with Season and Frequency, Kakadu, Litchfield and Nitmiluk National Parks, Northern Territory, Australia, 1994

Plot set-up The over-all plot sizes are 40 x 20 metre. Each plot is permanently located by a central metal star picket and GPS waypoint. Plots relocated by helicopter have a round large white lid on top to assist sighting the plot from air. A permanent photo picket with number plate is placed 10 metre inside the plot parallel with the plot sides. The corners and inner transects are marked by metal droppers. The inner 40 x 10 metre plot is delineated with measuring tapes. Two 50 metre tapes are laid out 5 metre either side of central photo picket, parallel to the longer central axis, starting at zero on the left hand side heading to the back of the plot, and again at zero on the right hand side heading to the top of the plot. NB: some plots are not strictly 40 metre due to plot landscape constraints.

Metal star pickets GPS 50 metre measuring tapes

Record of fire occurrence Fire occurrence is recorded based on on-ground plot visits, and from aerial ground-truthing assessments of each park’s satellite-based fire mapping programme. An assessment is made as to whether the plot was unburnt, patchily burnt (ground cover at least 20% unburnt), or burnt, in three seasonal periods as follows: Wet season (w) - from onset of rains (typically November) until their cessation (typically March - April); Early dry season (e) - from end of wet season until end of July; Late dry season (L) - from August until start of rains.

Ground-based photos Photos are supposed to be taken biannually, post wet season and late dry season. Often parks staff only manage one photo per year.

Digital camera

Fire severity score raw Fire severity is scored for each recorded fire event from plot photos. Plot photos are scored by a member of the research team and fire managers from the park. Two seasonal photos per plot in any 1 year are not always available. However, in nearly all instances it is possible to confidently attribute fires to (1) season, given recourse to comprehensive plot records; and (2) severity class, by close examination of differences between foliage condition and vegetation structure in photo temporal sequences. Note: This method may not detect all fires of very low severity or high patchiness occurring especially in wet or early dry season periods given rapid post-fire regrowth under high soil moisture conditions. Fire severity is recorded using one or more the following categories: • 0 = Unburnt • 1 = Low • 2 = Moderate • 3 = High • P = Patchily burnt • U = Unburnt • B = Burnt. Photo records missing. • wet = Burn that occurred in the wet season If no assessment is done the fire severity is recorded as NA.

Russell-Smith J, Edwards AC (2006) Seasonality and fire severity in savanna landscapes of monsoonal northern Australia. International Journal of Wildland Fire 15, 541–550. Williams RJ, Gill AM, Moore PHR (2003) Fire behaviour. In ‘Fire in tropical savannas: the Kapalga experiment’. (Eds AN Andersen, GD Cook, RJ Williams) pp. 33–46. (Springer-Verlag: New York).

Fire severity score All raw fire severity scores are mapped to • 0 = Unburnt • 1 = Low • 2 = Moderate • 3 = High U->0 B->1 P->1 NA->0 wet-> removed

Time since last fire The time since last fire occurrence is calculated in decimal years. This is NA from 1994 until the first recorded burn, unless there was no recorded burn before 1997, in which case 1994 wet season is set to be the first burn occurrence.