Modelling the fire weather of Black Saturday · Hour-by-hour comparison (0.012° model run) 1800...

03/09/2012

1

The Centre for Australian Weather and Climate ResearchA partnership between CSIRO and the Bureau of Meteorology

Modelling the fire weather of Black Saturday

Robert FawcettHigh Impact Weather Research28 August 2012

www.cawcr.gov.au


Acknowledgements

• Co-authors• Will Thurston, Jeff Kepert and Kevin Tory

High-Impact Weather – Weather and Environmental PredictionThe Centre for Australian Weather and Climate Research (CAWCR)Docklands, Victoria

• Similar work• Unpublished simulations by Les Logan (2009)• Engel, Lane, Reeder and Rezny (2012)

The meteorology of Black SaturdayQuarterly Journal of the Royal Meteorological Society (available online)

• Thanks to• Bushfire CRC for partial funding of this work• Project partners for useful discussions• Colleagues in CAWCR and the UKMO for significant assistance in setting up

the ACCESS model runs

03/09/2012

2


What are we trying to accomplish?

• to model the meteorology of Black Saturday (2009-02-07) across central Victoria

• using the Australian Community Climate and Earth System Simulator (ACCESS), including the UK Met Office atmospheric model

• at high resolution (0.012° latitude/longitude spacing 1.3 × 1.1 km)• at very high resolution (0.004°latitude/longitude spacing 440 × 350 m)

• test runs also at 0.008°( 890 × 700 m) and 0.006° resolution ( 670 × 530 m)• we are performing analogous simulations for some other recent fires

• Margaret River (Nov 2011), Eyre Peninsula (Jan 2005) already done, another underway

• to see which aspects of the modelled meteorology might be important for fire behaviour and to see if aspects of the observed fire behaviour can be “explained” by the meteorology

• to see which aspects of the very-high-resolution meteorology are also seen in the lower-resolution simulations

• lower-resolution simulations are cheaper and faster to run, and therefore closer to operational implementation – e.g., resources(0.004°) 10 × resources(0.008°)


Why are we trying to do this?

• because the Bushfire Cooperative Research Centre wants the results• Fire Impact and Risk Evaluation – Decision Support Tool (FIRE-DST) project

• simulations also feeding into BCRC Understanding Complex Fire Behaviour: Modelling Investigation of Lofting Phenomena and Wind Variability project

• See associated conference posters

• meteorological gridded outputs will be fed into fire-spread models (Phoenix RapidFire)

• to facilitate FS-model upgrading from single AWS / single sounding to fully 4D meteorological inputs

• to assess the sensitivity of fire outcomes• to fire-ignition location and timing relative to “observed”

weather• to timing of significant weather aspects

(e.g., fronts, wind changes)• scenario planning for urban interface – risk assessment

• because it is interesting• because it is (now) possible

Phoenix RapidFire simulation of the Kilmore East fire domain2009/02/07

03/09/2012

3


Antecedent conditions

• Many periods ending Jan 2009 are either “lowest on record” or in “serious deficiency” (below the 10th percentile)

• Jan 2009 also very dry

Anomaly base period: 1900-2011

wet1950s

dryMD

wet

dryWWIID wet

1970s

Data: AWAP low-res analyses

Accumulated rainfall anomalies (in mm) – Victoria (1900-2011)

• 1997-2009 (Millennium drought) represents the most sustained dry period across Victoria since Federation and likely since at least 1870.

dryFedD


Synoptic meteorology

• Analyses of mean sea-level pressure

• 11 am EDT 6 Febto 5 am EDT 8 Feb(6 hours apart)

• High-pressure system in Tasman Sea

• Low-pressure system with embedded cold front in Southern Ocean

• February max. temp. records broken across much of Victoria on 7 Feb, Melbourne’s all-months record broken

• Primary feature of relevance is the pre-frontal trough (blue line) which crosses Melbourne around 5 pm EDT on 7 Feb 2009.

03/09/2012

4


Nesting details

140 142 144 146 148 150 152

-44

-42

-40

-38

-36

-34

Longitude (°)

Latit

ude

(°)

• The modelling begins with a global model run. The second level is at 0.11° resolution covering all of Australia and surrounding waters. Further levels of nesting increasingly focus on the area of interest.• Boundaries of the third (red), fourth (green) and fifth (blue)nesting levels for the Black Saturday model runs.• Model boundaries are chosen, as far as is possible, to avoid areas of elevated topography.• Comparison test runs also done at 0.008° and 0.006° (fifth level).

0.036° (4 km × 3.2 km)

0.012° (1.3 km × 1.1 km)

0.004°(440 m

× 350 m)


Nesting details (2)

• 50 vertical levels• 30 + 20• 60 km modelling depth

• APS1 (UKMO Unified Model version 7.5)• UK Met Office Initial Condition

• 2009-02-06 0300 UTC (2009-02-06 1400 EDT)• 0.012° resolution and finer

• no convective parameterisation• 1D planetary boundary layer scheme replaced with full 3D Smagorinsky

turbulence closure• archiving

• 5-minute surface outputs• 15-minute model-level outputs

03/09/2012

5


What do the simulations show?

7:05 am to 4:00 am (EDT), 0.012° simulation

Screen-level air temp. (°C)10-metre wind vectors

Notional instantaneousForest Fire Danger Index(Mark 5; Drought F = 10)


What do the simulations show?• many interesting things

• primary and secondary wind change• pre-frontal boundary-layer rolls over a wide area• numerous small-scale vortices along the primary wind change (trailing

wakes behind the wind change)• an undular bore

Notional instantaneousFFDI for NW Vic.36.5 - 35.5°S,143.5 - 144.5°E(0.004° resolution)

0200 UTC / 1300 EDTto1055 UTC / 2155 EDT

NSW

Vic.

Google terrain map

FFDI(DF = 10)

03/09/2012

6


What do the simulations show?

Notional instantaneousFFDI – Kinglake area0.004° simulation37.9 to 37.2°S144.9 to 145.8°E

Kinglake escarpment, Mt Sugarloaf,Murrindindi River valley

2000 UTC / 0700 EDTto0955 UTC / 2055 EDT

FFDI(DF = 10)

Google terrain map


How do we validate the simulations?

• Comparisons against surface observations• one-minute and thirty-minute automatic weather station (AWS)

observations across Victoria• one-minute AWSs mostly around Melbourne

• Comparisons against upper-air observations• radiosonde balloon-flight observations at available locations (Melbourne,

Hobart, Wagga, Sydney, Mount Gambier)• Comparisons against radar data from the Yarrawonga Doppler radar

• + others where available• Comparisons against satellite observations

• visual, infra-red imagery• None of these observational data are assimilated into the model

simulation, so the simulation can be validated against completely independent data.

03/09/2012

7


What aspects of the modelling can we verify?

• timing of the main wind change on Black Saturday• forecast screen-level maximum temperature• forecast maximum 10-metre wind speeds• upper-level temperature, dewpoint, wind speed• features observed in the radar and satellite imagery

• e.g., boundary-layer rolls, undular bores


Primary wind change on Black Saturday

Vic. Bushfires Royal Commission final reportSource: Bureau of Meteorology

03/09/2012

8


Hour-by-hour comparison (0.012° model run)

142 144 146 148 150

-40

-38

-36

-34

Location of the front: 1200 EDT

Longitude

Latit

ude

analysed model

Wind direction (°)

N E S W N

Front positions manually digitised from maps (VBRC map and 0.012°model simulation).

1200 EDT



142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model

1300 EDT

Wind direction (°)

N E S W N

03/09/2012

9



142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model

1400 EDT

Wind direction (°)

N E S W N


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


1500 EDT

Wind direction (°)

N E S W N

03/09/2012

10


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


1600 EDT

Wind direction (°)

N E S W N


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


1700 EDT

Wind direction (°)

N E S W N

03/09/2012

11


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


1800 EDT

Across central Victoria, themodel front is slightly behindthe observed front.

Wind direction (°)

N E S W N


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


1900 EDT

Wind direction (°)

N E S W N


03/09/2012

12


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


2000 EDT

Wind direction (°)

N E S W N



142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


2100 EDT

Wind direction (°)

N E S W N


03/09/2012

13


142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed model


2200 EDT

Wind direction (°)

N E S W N



2300 EDT

142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude

analysed modelIn both the observations and the model, it becomes very difficult to diagnose the precise location of the front across the mountains of the southeast.

Wind direction (°)

N E S W N

03/09/2012

14



0000 EDT

142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude


Wind direction (°)

N E S W N



0100 EDT

142 144 146 148 150

-40

-38

-36

-34


Longitude

Latit

ude


Wind direction (°)

N E S W N

03/09/2012

15


Determining the timing and max. temp. errors

1.0 1.2 1.4 1.6 1.8 2.0

010

2030

40

088164 (400m)

Tem

pera

ture

(°C

), W

ind

spee

d (m

/s)

Air t

empe

ratu

re (°

C),

win

d sp

eed

(m/s

)

0000 0448 0936 1440 1912 2400Time (EDT) on Black Saturday

change

AWS and 0.004° model data for 088164 Eildon Fire Tower

model data

AWS data

AWS data

10-metre wind direction

Screen-level airtemperature (°C)

Screen-leveldewpointtemperature (°C)

10-metre windspeed (m/s)

model data

AWS data

model data

five-minute model data

one-minuteAWS data andfive-point movingaverage

change


A well-observed aspect of the meteorology

• At 2334 EDT (1234 UTC) on Black Saturday (2009-02-07), extreme fire behaviour and significant spotting were reported including up to 1 km from the town of Myrtleford in northeast Victoria.

[VBRC report on the Beechworth-Mudgegonga fire]

• What happened meteorologically speaking and how well was it modelled?Fire perimeter map (VBRC).

03/09/2012

16


In the Yarrawonga Doppler radar …

1202 UTC2302 EDT

1322 UTC0022 EDT

† †

† †

• Yarrawonga radar is northwest of the Beechworth-Mudgegonga fire (‡) .• A disturbance (†) passes across the radar view at around midnight, travelling from the southwest towards the northeast.• As it passes across the B-M fire, the fire is seen in the radar data to intensify considerably.• Radar images available at 20-minute intervals.

‡ ‡

‡ ‡

300 km


In the satellite images …

1230 UTC2330 EDT

1330 UTC0030 EDT

1430 UTC0130 EDT

†

†

†

• The disturbance (†) is also visible in the infra-red satellite imagery, available at hourly intervals.• As it passes across the Beechworth-Mudgegonga fire (‡) , the smoke plume grows and remains larger for at least one hour.• A much larger smoke plume (§) is also visible in these images.

‡

‡

‡

§

03/09/2012

17


On the ground …

1.2 1.4 1.6 1.8 2.0 2.2

984

986

988

990

992

994

072160 (400m)

Days since Midnight 2009/02/06

Pre

ssur

e (h

Pa)

AWS and 0.004° model data for 072160 Albury Airport AWS

Bar

omet

ric p

ress

ure

(hP

a)

10-metre wind direction

AWS

model

AWS

model

• The disturbance (†) is visible in the wind direction and the barometric pressure data.• AWS wdir < 280°from 0101 EDT to 0108 EDT.• Disturbance also seen in the Yarrawonga AWS.

†

†

0448 0936 1440 1912 2400 0448 Time (EDT) on Black Saturday and the day after

N NW


What does the model say is happening …

1100 UTC2200 EDT

1300 UTC0000 EDT

• The disturbance () is initially just the leading edge of the wind change, but the wind change stalls, leaving an undular bore (†) to run on ahead towards the northeast.• The undular bore is characterised at the surface by a temporary change in wind direction and (to a lesser extent) speed.• It is harder to see its progress across the mountainous areas of the southeast.• around 4 to 5 km (8 to 10 min) wide• also visible in the 0.006°, 0.008°, 0.012°and 0.036° simulations

†

Wind direction (°)N NE E SE S SW W NW N

NWSW

S

SE

WSW

N

S

0.004° simulation

03/09/2012

18


50

100

150

200

250

300

350

9.00 10.00 11.00 12.00 13.00 14.00 15.00 16.00

Hour (UTC)

Dis

tanc

e tra

velle

d (k

m)

modelradar

Verification

• Positions along the SW-NE radar transect were estimated from the Yarrawonga radar and model output (0.004°-resolution) images• Model speed (43 km/hr) along transect about 30% faster than observed speed (33 km/hr).

Time (EDT) on Black Saturday and the day after21:00 22:00 23:00 00:00 01:00 02:00

†

Dis

tanc

e tra

velle

d (k

m)

43 km/hr

33 km/hr


Implications for the fire …

Hei

ght a

bove

sea

leve

l (m

)

• The modelling suggests the undular bore (†) generates updrafts strong enough to further loft pieces of bark raised to about 1000 metres by the fire.

stalled wind change †

†

transect line

Beechworth-Mudgegonga fire

cool air

warm air 0015 EDT

03/09/2012

19


Following the bore

-20 0 20 40 60

1000

500

200

100

Temperature (°C), Wind speed (m/s)

Pre

ssur

e (h

Pa)

120 °C -100 °C -80 °C -60 °C -40 °C Wagga1100 UTC2200 EDT(0.012°)

0.004°-res.orography

indicative ofstable layer

pot. temp.

heig

ht a

bove

sea

leve

l (m

)


Concluding remarks …• The simulation verifies well against the available

observational data• Modelling of maximum temperatures is very good• The wind change is modelled well, with timing errors of only

20 to 50 minutes.• Peak surface wind speeds are typically underestimated,

• The dramatic blow-up in the B-M fire behaviour in the north east can be explained by an undular bore running ahead of a stalled wind change

• this meteorological feature is well supported by the available observational evidence

• The simulation shows lots of interesting small-scale detail• Simulation data now being fed into Phoenix RapidFire• Not all fire weather cases will be as well modelled as this

one …• synoptic forcing here is very strong

Modelling the fire weather of Black Saturday · Hour-by-hour comparison (0.012° model run) 1800...

Documents

Transcript of Modelling the fire weather of Black Saturday · Hour-by-hour comparison (0.012° model run) 1800...