Fire Growth Modelling at Multiple Scales Kerry Anderson
Canadian Forest Service Kerry Anderson Canadian Forest Service
1
Slide 2
Fire Growth Modelling at Multiple Scales Introduction Weather
and fire activity can be thought of occurring on different time and
space scales. Forecasting at each of these scales has its
individual practical and physical limitations.Introduction Weather
and fire activity can be thought of occurring on different time and
space scales. Forecasting at each of these scales has its
individual practical and physical limitations. 2
Slide 3
Scales 1 mm1 m1 km1000 km year month week day hour min sec
Turbulence Thunderstorms Longwaves Cyclones Fronts Cumulus Clouds
Jet Stream MICRO MESO SYNOPTIC Length Time 3
Slide 4
Scales Holdover Fires 1 mm1 m1 km1000 km year month week day
hour min sec Ignitions Turbulence Detected Fires (~0.1ha) Campaign
Fires Thunderstorms Longwaves Cyclones Fronts Cumulus Clouds Jet
Stream MICRO MESO SYNOPTIC Length Time Class E Fires (> 200 ha)
Fire Whirls 4
Slide 5
Fire Growth Modelling at Multiple Scales Introduction Fire
growth modelling can fall into three scales depending on weather
forecasting ability: Short-range: 1-2 days Medium-range: 3-7 days
Long-range: 8+ daysIntroduction Fire growth modelling can fall into
three scales depending on weather forecasting ability: Short-range:
1-2 days Medium-range: 3-7 days Long-range: 8+ days 5
Slide 6
Fire Growth Modelling at Multiple Scales Introduction As the
run time increases, the model becomes less deterministic and more
probabilistic.Introduction deterministic weather-based detailed
deterministic weather-based detailed probabilistic climate -based
generalized probabilistic climate -based generalized Short-range
Medium-range Long-range 6
Slide 7
Fire Growth Modelling at Multiple Scales The Models This
presentation shows a fire growth modelling system designed to run
consecutively over three time scales. The models are designed so
that they may be run in sequence, with the results of one model
initializing the subsequent model run. The Models This presentation
shows a fire growth modelling system designed to run consecutively
over three time scales. The models are designed so that they may be
run in sequence, with the results of one model initializing the
subsequent model run. 7
Slide 8
Fire Growth Modelling at Multiple Scales The Models The
Probabilistic Fire Analysis System is a suite of models designed to
capture the transition across multiple scales. 8
Slide 9
Short-range Fire Growth 9
Slide 10
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth USFS Farsite and the Canadian Prometheus USFS Farsite and
the Canadian Prometheus are examples of operational, deterministic,
short-range fire growth models. 10
Slide 11
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth The short-range model is a deterministic, sixteen - point
fire growth model that uses hourly weather. Short-range Fire Growth
The short-range model is a deterministic, sixteen - point fire
growth model that uses hourly weather. 11
Slide 12
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth Hourly weather is obtained from numerical weather models.
12
Slide 13
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth Fire locations are determined using NOAA/AVHRR and MODIS hot
spot data. 13
Slide 14
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth Short-range predictions are produced for the next 24 to 48
hours. 14
Slide 15
Fire Growth Modelling at Multiple Scales Short-range Fire
Growth Short-range predictions are currently being presented
through the CWFIS interactive web- page and GoogleEarth. August 15,
2013 15
Slide 16
Medium-range Fire Growth 16
Slide 17
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth The medium-range model uses the same deterministic fire
growth but uses more general, daily weather information.
Medium-range Fire Growth The medium-range model uses the same
deterministic fire growth but uses more general, daily weather
information. Day Three T max : 22 T min 8 Rh noon 25 POP: 0% Day
Four T max : 21 T min 19 Rh noon 30 POP: 0% Day Five T max : 18 T
min 14 Rh noon 60 POP: 30% Day Two T max : 20 T min 10 Rh noon 35
POP: 0% 17
Slide 18
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth Daily forecasts are combined with diurnal trend models to
produce hourly weather profiles. 18
Slide 19
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth Uncertainty in the forecast can be introduced as
perturbations (e.g. Temp + 1 o C), each producing a new weather
series. Uncertainty in the forecast can be introduced as
perturbations (e.g. Temp + 1 o C), each producing a new weather
series. 19
Slide 20
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth Ensemble forecasts are produced by running the model through
a range of weather conditions creating probable fire perimeters.
Medium-range Fire Growth Ensemble forecasts are produced by running
the model through a range of weather conditions creating probable
fire perimeters. Temp + 1 o C RH + 5% WD + 10 o WS + 3 kmh 20
Slide 21
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth Probable fire perimeters can be further enhanced by
introducing the probability of precipitation. Medium-range Fire
Growth Probable fire perimeters can be further enhanced by
introducing the probability of precipitation. rainno rain 21
Slide 22
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth 11% 33% 55% 77% 100% Current hotspot Old hotspot (now
extinguished) Case studies show an improvement in fire growth
predictions when using an ensemble approach. 22
Slide 23
Fire Growth Modelling at Multiple Scales Medium-range Fire
Growth Currently we are examining CMC ensemble products for use in
medium-range fire growth predictions. 23
Slide 24
Long-range Fire Growth 24
Slide 25
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
The long-range fire growth model is probabilistic and is based upon
climatology. The long-range fire growth model is probabilistic and
is based upon climatology. Such models are based on probabilities
and thus output is represented as probable fire extents. Long-range
Fire Growth The long-range fire growth model is probabilistic and
is based upon climatology. The long-range fire growth model is
probabilistic and is based upon climatology. Such models are based
on probabilities and thus output is represented as probable fire
extents. 25
Slide 26
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Long term fire growth from one location to another within a given
time period is the probability that the fire will spread across the
distance before a fire-stopping rain event occurs. Long-range Fire
Growth Long term fire growth from one location to another within a
given time period is the probability that the fire will spread
across the distance before a fire-stopping rain event occurs. p(t)
= p (t) P (t) spread survival 26
Slide 27
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Probability of spread is calculated by assuming the rate of spread
follows an exponential distribution where 1/ = mean rate of spread
Long-range Fire Growth Probability of spread is calculated by
assuming the rate of spread follows an exponential distribution
where 1/ = mean rate of spread P spread (r > r o ) = e - r o o
27
Slide 28
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Mean rates of spread can be calculated for each fuel type, each
month, each direction. These are similar to wind roses but for the
rate of spread for each fuel type. Mean rates of spread can be
calculated for each fuel type, each month, each direction. These
are similar to wind roses but for the rate of spread for each fuel
type. Mean rates of spread in C2 fuel type for August 28
Slide 29
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Probability of survival can be modelled using the DMC. If the DMC
drops below an extinction value, the fire is assumed to go out.
29
Slide 30
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Probability of survival is solved using Markov Chains and DMC. As
time progresses, the probability of survival drops to zero. 30
Slide 31
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
The model calculates the probabilities of spread and of survival
and combines them spatially to produce a probable fire extents map.
Long-range Fire Growth The model calculates the probabilities of
spread and of survival and combines them spatially to produce a
probable fire extents map. Spread Survival Extents 50% 60% 40% 50%
31
Slide 32
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
Comparisons with historical fires indicate the model produces
realistic results Long-range Fire Growth Comparisons with
historical fires indicate the model produces realistic results
050100150200250 km 50% 30% 10% Historic Fire Perimeter but is
difficult to validate this way. 32
Slide 33
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
The long-range model predictions were compared with distributions
of fire perimeters predicted by repeated simulations using the
hourly-based, deterministic fire-growth model. Long-range Fire
Growth The long-range model predictions were compared with
distributions of fire perimeters predicted by repeated simulations
using the hourly-based, deterministic fire-growth model. The study
showed a close agreement between the long-range model and the
deterministic model. a) 40 simulated firesb) 40 simulation
probability c) Long-range model probabiltiy 10% 30% 50% 10% 30% 50%
33
Slide 34
Combined Predictions 34
Slide 35
Fire Growth Modelling at Multiple Scales Combined Prediction
The models are designed so that they may be run in sequence, with
the results of one model initializing the subsequent model run.
Short-range: day 1 to day 2 Medium-range: day 3 to day 7
Long-range: day 8 to day 21 Combined Prediction The models are
designed so that they may be run in sequence, with the results of
one model initializing the subsequent model run. Short-range: day 1
to day 2 Medium-range: day 3 to day 7 Long-range: day 8 to day 21
35
Slide 36
Fire Growth Modelling at Multiple Scales Combined Prediction
The results of one model are used to initialize the subsequent
model run. Combined Prediction The results of one model are used to
initialize the subsequent model run. 2 days 7 days 21 days Short
Medium Long 36
Slide 37
20 Fires 2009-2010 37
Slide 38
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
The long-range model PFAS was used in BC during the 2009 and 2010
fire seasons. Long-range Fire Growth The long-range model PFAS was
used in BC during the 2009 and 2010 fire seasons. Predictions were
used to assess modified response decisions. Terrain Fuels 15 day
Fire Growth Projection (Prescribed Fire Analysis System) _____
Current fire perimeter (approx 76 000 ha) _____ 15 day area at risk
(approx 180 000 ha) (50 % probability) _____ 30 day area at risk
(approx 317 000 ha) (50 % probability) 30 day 38
Slide 39
Fire Growth Modelling at Multiple Scales Long-range Fire Growth
In 2009, PFAS predictions were used to assess modified response
decisions made on 37 fires. In 2010, this was done on xxx fires. In
all cases, PFAS was used to supplement the decision support
decision made by fire suppression officers. In a few cases, the
predictions triggered a reassessment of the suppression response
plan. Long-range Fire Growth In 2009, PFAS predictions were used to
assess modified response decisions made on 37 fires. In 2010, this
was done on xxx fires. In all cases, PFAS was used to supplement
the decision support decision made by fire suppression officers. In
a few cases, the predictions triggered a reassessment of the
suppression response plan. 39
Slide 40
The Good 40
Slide 41
41
Slide 42
42
Slide 43
43
Slide 44
44
Slide 45
45
Slide 46
46
Slide 47
47
Slide 48
48
Slide 49
The Bad 49
Slide 50
50
Slide 51
51
Slide 52
52
Slide 53
53
Slide 54
54
Slide 55
The Ugly 55
Slide 56
56
Slide 57
57
Slide 58
58
Slide 59
59
Slide 60
60
Slide 61
61
Slide 62
62
Slide 63
63
Slide 64
Fire Growth Modelling at Multiple Scales Conclusions As forest
protection agencies re-evaluate the role of fire in the landscape
and its ecological benefits, as they face the prospect of excessive
fuel build up resulting from years of fire exclusion policies, and
as they contend with fiscal constraints, these agencies must start
looking at possible fire growth over extended periods. The methods
presented through these models may serve as a foundation for such
evaluation.Conclusions As forest protection agencies re-evaluate
the role of fire in the landscape and its ecological benefits, as
they face the prospect of excessive fuel build up resulting from
years of fire exclusion policies, and as they contend with fiscal
constraints, these agencies must start looking at possible fire
growth over extended periods. The methods presented through these
models may serve as a foundation for such evaluation. 64
