GEOS9016 2021

Fire model

AIM

Develop a simple fire hazard model for the Smiths Lake field station.

In terms of the major report, this is a key component of the site allocation process.
You need to identify sites where a fire will be of a low enough energy that it can be
controlled.

In terms of GIS more broadly, this model is implemented using map algebra 1. This is
the most common way of combining rasters in modelling processes.

Overview

The fire model is based on the McArthur Forest Fire Danger Meter Mk5. There are
several parameters we are assuming are spatially constant across the study region,
for example temperature, but that you might wish to experiment with for your own
project.

The parameters are based on the work of Noble et al. (1980) and Sirakoff (1985).
Copies of these are available through the library online resources for the course, with
bibliographic details at the end of this document. Note the discussion in Noble et al.
(1980) concerning the utility of the equations and the way in which they were derived
from paper graphs. You should also consider the general applicability of the methods
(see for example Perry, 1998 and citing documents). Derivations for the parameters
used in the model (which you need to read) are given after the instructions.

The model itself can be divided into three components that you need to calculate (see
flowchart in Figure 1). The first is the rate of spread across flat ground (based on
vegetation type). The second is the modified rate of spread to take into account slope
gradient (using the ground slope calculated in degrees from your DEM). The third is
the fuel load at each cell (based on the fuel load surveys collected in the field). These
are then combined to form your fire model.

NOTE: We are using 2018 fuel loads. If you are unfamiliar with fire behaviour, I
recommend you browse through the website
http://learnline.cdu.edu.au/units/env207/fundamentals/behaviour.html. It includes
relevant information to understand this Lab.

The DEM you are using is the one you wrote the metadata for (although you might
wish to re-interpolate it to improve its quality).

1
Map algebra is also sometimes referred to as raster algebra.
http://desktop.arcgis.com/en/arcmap/10.6/extensions/spatial-analyst/map-algebra/what-is-map-algebra.htm .
Map algebra and cartographic modelling are covered in the Lectures.

1
GEOS9016 2021

Figure 1. There are three main components to the process, and these are ultimately combined using multiplication. RoS is the Rate
of Spread component. Diamonds represent operations, other shapes represent data sets.

2
GEOS9016 2021

PROCEDURE

1. Using ArcCatalog, copy the fuel loads folder.

GEOS9016\fuel_loads_2018.

2. The remainder of the instructions are in ArcMap. Ensure the Spatial

Analyst extension is enabled. This is done under Customize ->
Extensions. Make sure Spatial Analyst has the check mark and then
close the window.
3. Open the ArcToolbox environments using the menu GeoProcessing ->
Environments (you can also access this using the right click menu from
the uppermost icon in the ArcToolbox panel).

4. Before you begin analysing the data, you need to set your toolbox
Processing Extent and Raster Analysis cell size to be the same as your
DEM. The extent can be easily set by making it the same as the DEM
(you can also enter it manually if you ever need to specify an extent
without an existing data set).

5. Make sure the Current Workspace is your GIS folder on your H: drive,
and the Scratch Workspace is C:\Data\ArcGIS\z1234567\temp (where
the last part is your user ID). You will need to create this temp folder if it
does not exist.

It should look something like this:

3
GEOS9016 2021

6. Rate of Spread component

We will calculate a raster to represent the rate of fire spread across flat
ground, which in these instructions is the raster called fire_spread. This
is the most involved part of the model process.

There are two fire spread models, one for grasslands and one for forests.
The parameters we are using are for one climatic scenario. For
variations, see the calculations at the end of the Procedure section.

If land cover is grassland, then the rate of spread is 0.801 m/s

If land cover is forest, then the rate of spread is 0.0802 m/s.

These parameters can be added to your vegetation shapefile in a new

field (you need to create it and give it a meaningful name like rate_sprd).
Remember to make the field type double so it stores decimal values.
Select those fields where the KEITH_FORM field (in the vegetation.shp
file) is blank (KEITH_FORM = ‘’) and assign the grassland value to them
(but note that many of these areas are roads and water bodies – this can
be fixed later).

The other entries can be assigned the forest value for the moment (you
will note that in reality vegetation types such as wetlands should not burn
anywhere near as well as a eucalypt forest, but I leave that to you to sort
out as an innovative step). To select the non-cleared area, just use
Switch Selection , as it is all locations not classed as cleared. It is
available under several menus.

What the above means is that you are treating all cleared areas as
grassland, and all other areas as forest. This will over-allocate cleared
land, for example it includes all roads and water bodies, but this is
something you can fix later.

If you later try different scenarios, for example differing rates of spread
for each forest type or different rates for forest and cleared land, then
just add a new field for each scenario and then use those values in the
following steps.

Convert the vegetation polygon to a raster layer using Conversion Tools

-> To Raster -> Polygon to Raster. Make sure you use the rate of spread
field you just calculated (rate_sprd) for the values. This is referred to as
fire_spread in the rest of this document. Make sure you also have the
correct cell size set in the tool, as this tool can ignore the raster analysis
settings. (This is because it is not part of the Spatial Analyst toolbox,
which is not actually a good reason as far as I am concerned). Accept all
other default parameters.

7. Slope factor

4
GEOS9016 2021

Now we need to adjust the rate of spread for the slope gradient (in units
of m/s). This is done using the following equation (note that the slope is
in degrees)

rate_slopespd = fire_spread * exp (0.069 * slope)

This function can be broken into several steps, working from the
innermost calculation outwards (the order of operations applies to spatial
analysis as well as normal maths).

First, you will need to calculate the slope raster from your DEM using the
Spatial Analyst -> Surface -> Slope tool. Make sure the units are
degrees.

Next, multiply slope by 0.069 using the Spatial Analyst -> Math -> Times
tool. This rescales the values.The result of this is then fed into the Exp tool
(exponent, from the same toolbox as Times).

Finally, you multiply this result by your fire_spread layer of step 6.

Delete the temporary grids you have created. It will also be faster to
generate the partial process grids (like the exp grid) in your directory on
the H: drive. Slope is a good one to keep, as we will be using it in most
of the models.

8. Fuel component

The fuel_load layer (t/ha) is calculated by summing the three fuel

component layers from the fuel loads 2018 folder you copied in step 1.
(GEOS9016\fuel_loads_2018). Look for raster layers f_surf_2018,
f_bark_2018, f_elev_2018.

There are several ways of summing the rasters. The Plus tool only lets
you sum two rasters at a time so needs to be called twice to use all three;
the Raster Calculator tool is an alternative because it lets you process
multiple rasters in one go (Spatial Analyst Tools > Map Algebra > Raster
Calculator). You can also use the Cell Statistics tool and calculate the
Sum – this has the advantage that you can choose to ignore nodata when
it is appropriate.

These fuel load rasters have been developed by correlating the field
survey values with topographic, ecosystem and satellite derived data,
using a classification tree as the classification tool. The accuracy has not
been assessed (something you could try outside of the GIS after
extracting the data for the sample locations using the Sample or Extract
Multi-values to Point tools). f bark_2018 is the bark component, f_surf_2018
is the surface fuel component, and f_elev_2018 is the elevated fuel component.
All values are based on the Overall Fuel Hazard guide (McCarthy et al.
2009).

5
GEOS9016 2021

9. To calculate the fire intensity (kW/m) of fire front for each cell for one
climate scenario, the following formula is used:

Fire_intens = 18600 * fuel_load * rate_slopespd * 0.27777 * 0.1

Where the last two terms convert from km/hr to m/s (0.27777) and t/ha
to kg/m2 (0.1). The value 18600 is the approximate energy that a
eucalypt fire will burn with, in joules 2.

Use the Spatial Analyst toolbox as before. Note the multiplicative

relationship, which means you can pre-calculate the product of the
numeric constants using a calculator and include them as one value in
the GIS. To avoid having to run the Times tool several times, you could
again use Raster Calculator to process all the rasters at once (the Cell
Statistics tool does not calculate the Product).

When plotting your fire intensity, a value of 4000 kW/m of fire front is
about the limit of fire control (see link in footnote 2), so plot the data using
a minimum-maximum stretch with a range of values between 1000 and
6000 (for example). Otherwise the small number of extreme values
makes the important details difficult to see.

And so, after all those steps, you have a fire intensity model which gives
an estimate of how much energy a fire will burn with at each location,
and therefore an indication of where not to build. Note, however, that we
have made some crude approximations and that we do not consider
changes in the direction of the wind, nor the effect of ridge tops. It is left
to you to identify what is missing and how you might improve the model
(this is the innovation part of the assessment criteria). I also encourage
you to assess different scenarios.

Note also that you could do all this processing using the ModelBuilder.
Doing so will make it much is easier to later re-run models with different
parameters.

THIS IS THE END OF THE LAB INSTRUCTIONS.

2
A summary with references is described in a short note by Dr Malcom Gill, then of CSIRO. The original
source is no longer available, but the text can be accessed from
http://members.iinet.net.au/~barling/reslet2.html

6
GEOS9016 2021

Calculating the rate of spread

The following are the calculations from which the rate of spread parameters
used above are derived. Use these as a guide to implementing your own
scenarios. They are derived from Fletcher Rayner’s MSciTech (GIS) major
project report, which in turn used the references list at the end of the document.

Keep in mind that this is only one scenario, and it is used as a guide (how often
does one see a humidity of 15% at the coast?). Feel free to try your own set of
parameters, deriving the relevant factors to plug into the fire model above. The
calculations are given for grasslands first (Grassland Fire Danger Meter Mk5),
and then for forests (Forest Fire Danger meter Mk5). These are also for the rate
of spread over flat (level) ground. They need to be modified to allow for the
effect of slope (see the calculations for the rate_slopespd layer above).

Scenario 1
Environmental conditions:

McArthur Drought Index 10

◦
Air Temperature 37.2 C
Relative humidity 15%
Wind speed 40km/hr

Grassland Fire Danger Meter Mk5

Fuel Moisture Content in

M = %
H = Relative Humidity 15
T = Air Temperature 37.2
C = Degree of curing 80%

Equation 1

97.7 + 4.06 × H 3000

M = − 0.00854 × H + − 30
T + 6.0 C

97.7 + 4.06 × 15 3000

M = − 0.00854 × 15 + − 30
37.2 + 6.0 80

M = 11.04%. Therefore the fuel moisture content in the

grasslands is 11.04%.

7
GEOS9016 2021

If M < 18.8% the following equation is used to calculate the fire danger index (F)
(check Noble et al. 1980 for what to do when it exceeds this value).

F = 3.35 × W × exp [−0.0897×M + 0.0403×V ]

Equation 2 ……..…..

Where:

F = Fire Danger Index

W = Fuel Weight (tonnes/ha) Geographic
data set,
see Step 4
Average Wind Velocity at 10m
V = height 40 km/hr

F = 3.35 × W × exp [ −0.0897×11.04+ 0.0403×40 ]

therefore

F = W × 6.23

The rate of spread (R) is calculated as a function of the Fire Danger index (F)
times 0.13. Since F is just W × 6.23 for this scenario, we substitute those values.

R = 0.13 × F

R = 0.13 × 6.23 × W

= 0.801 × W

Therefore, the rate of spread for grassland over level ground for this scenario
is 0.811 km/hr times W. W is derived from the fuel load surveys we conducted
on the field trip.

8
GEOS9016 2021

Forest Fire Danger Meter Mk5

The Forest Fire Danger Meter Mk5 is

T −H 
 30 + 0.0234 ×V 
F = 1.25 × D ×e  

Where:

F = Fire danger index

D = Drought Factor 10
T = Air Temperature (Dry) 37.2
H = Relative Humidity 15%
V = Average wind velocity at 10m height 40 km/hr

 37.2 −15 
 30 + 0.0234×40 
F = 1.25 × 10 ×e  

= 66.801

Equation 6 …… R = 0.0012 × F × W

Where:

R = Rate of forward spread on level ground

W = Fuel weight (tonnes/ha)

R = 0.0012 × 66.801 × W

= 0.0802 × W

Therefore, the rate of spread for fire on level ground in a forest for Climatic
Scenario 1 is 0.0802 × W . As with the Grassland model, we have derived the W
layer already, and use it in Step 4.

9
GEOS9016 2021

References
McCarthy, G.J., Tolhurst, K.G. & Chatto, K. (2009) Overall Fuel Hazard Guide,
3rd edition. Victorian Department of Natural Resources and Environment,
Fire Management Research Report 47. (Under ‘other resources for the
Major Project’ in Moodle.

Noble I. R., Bary G. A. V. & Gill A. M. (1980) McArthur’s fire-danger meters

expressed as equations. Australian Journal of Ecology, 5, 201-203.

Perry G. L. W. (1998) Current approaches to modelling the spread of wildland

fire: a review. Progress in Physical Geography, 22, 222-245.

Sirakoff C. (1985) A correction to the equations describing the McArthur forest

fire danger meter. Australian Journal of Ecology, 10, 481.

10
GEOS9016 2021

Left: Vegetation map (.shp) Red values= different Keith Forms. Blue: no value. Right: Satellite image of Smith Lakes area. What are
‘areas cleared of forest’?

11