Top Banner
Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University of Alaska Fairbanks 1 Geophysical Institute 2 College of Natural Sciences, and Mathematics
29

Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Dec 29, 2015

Download

Documents

Harold Ryan
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Development of a Tool for Downscaling of Operational Climate Forecasts to

Regional and Local Fire Indices

1,2Nicole Mölders & 1Gerhard Kramm

University of Alaska Fairbanks1Geophysical Institute

2College of Natural Sciences, and Mathematics

Page 2: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Motivation

Wildfires reduce visibility => affect land and air trafficReleased aerosols and trace gases reduce air quality PM2.5 may affect healthDestruction of propertiesNatural treat with high occurrence nearly worldwide…

Agencies need data for planning fire management several months ahead a fire seasonA tool to provide suitable data for support in decision making and planning is required

Page 3: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Wildfires occur worldwide

Page 4: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Alaska has long wildfire history

Page 5: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

High temporal variability in wildfire frequency

and area burned

Modified after Mölders and Kramm (2006)

Page 6: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Fire weather assessment difficult ahead of a season or when data are sparse

Operational climate model predictions are too coarseHigher resolution non-linearly slows down the CPU and turnaround time Only 7 first class sites in Interior Alaska21 additional sites of unknown data quality run by volunteers for limited amount of quantitiesClouds or smoke from existing fire may affect remote sensing

None of the above methods permits 3 months ahead assessment of fire risk as required by regional fire agencies for planning

Page 7: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Moisture, temperature and wind dependent Fosberg Fire Weather Index typically used for fire risk assessment

Equilibrium moisture content

Fosberg Fire Weather Index

(FFWI)

Wind factor

Moisture damping coefficient

Wind speed

Temperature

Relative humidity

after Goodrick (2002)

Page 8: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Fosberg Fire Weather Index obtained from climate simulations not very helpful

Arctic Ocean

Alaska Canada

FFWI (-.-)

Page 9: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Modified Fosberg Fire Weather Index (mFFWI) by inclusion of Keetch-Byram Drought Index (KBDI) and fuel availability

),T,y,x,t(f)T,P,P,y,x,t(fKBDI cfdailymax,dailymax,annualdaily

Spatially varyingfuel availability factor

FFWI Modified FFWI

observations model

Page 10: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Heterogeneous precipitation distribution requires spatial fuel availability factor

30-average observed precipitation according to the GPCC (color)and as simulated by CCSM (solid lines)

Page 11: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Dryness yields to high fuel availability

Arctic Ocean

Arctic Ocean

Page 12: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

mFFWI obtained from climate simulations lacks regional details

Improvement compared to FFWI, but too coarse

modified FFWI (-.-)

FFWI (-.-)

Arctic Ocean

Arctic Ocean

Page 13: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Schematic to bridge operational climate forecasts to regional and local mFFWI

operationalclimate

forecasts

WRF

WRFnested runs in endangered

areas

weeklymFFWI maps

for quick-looks

6-hourlyforcing data

data base of climate model data plus derived mFFWI,regional model data plus derived mFFWI,

observations

statistical analysis toderive relationships, evaluation,

improvement, etc.

observations

dailymFFWI maps foranalysis

dailymFFWI

local maps

if required

Page 14: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Schematic of time staggering for mFFWI ensembles

operational climate forecasts

WRF simulations every x days for 5d

time

loca

l mF

FW

I

mFFWI form various WRFruns started at different times envelop

of mFFWI

start procedure 3 month ahead a fire season for guesstimates

repeat procedure until end of the fire season if a “real” month is over

Page 15: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Application of WRF for 4 km x 4 km grid increments

6-27-2005 1400 AST

Page 16: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Great horizontal variability of fuel availability factorwithin the Interior

Page 17: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Importance of consideration of fuel availability within the region

6-27-2005 1400 AST 6-27-2005 1400 AST

Page 18: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Modified FFWI derived from scare meteorological observations often shows lower fire risk than FFWI

Page 19: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Fires in June 2005

2005/06/14 2215 UT Fires in eastern Alaska Aquahttp://rapidfire.sci.gsfc.nasa.gov/gallery/?search=alaska&date

2005/06/26 2050 UT Fires in Alaska and Yukon Territory Terra http://rapidfire.sci.gsfc.nasa.gov/gallery/?search=alaska&date

Page 20: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Evaluation of simulated (m)FFWI difficult

Not every area with high (m)FFWI burns

Network density (4 sites in Interior Alaska)

Complex terrain, low representativeness

Random errors due to initial and boundary conditions or observations

Systematic errors from consistent misrepresentation of geometrical, physical, or numerical factors

Error propagation in measurements/simulations

Actual fires and burned areas affect temperature, humidity, precipitation

Overall evaluation recommended

Page 21: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Small uncertainty of (m)FFWI from propagation of measurement errors

2

2n

1i ii

Where stands for FFWI or mFFWI and is RH, T, v, P

Assume: (RH)= 5% (T)= 0.5K (v)= 0.5m/s (P)= 0.01inch

Note that uncertainty of FFWI and mFFWI will only differ if precipitation occurs

Page 22: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Note that McGrath and Northway airports are outside the WRF model domain!

WRF well provides modified FFWI

Page 23: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Forecast errors propagate in mFFWISimulation started 6-11-2005 0600 UTvalid for 6-11-2005 2300 UT (1400 AST)

Simulation started 6-7-2005 0600 UTvalid for 6-11-2005 2300 UT (1400 AST)

Errors in P yield errors in soil moisture, FAF, KBDI, mFFWIErrors in T yield errors in RH, FFWI, mFFWIErrors in RH yield errors in FFWI, mFFWIErrors in v yield errors in FFWI, mFFWI

=> Estimate uncertainty of mFFWI with Gaussian Error Propagation (GEP) principles

Page 24: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Suitability of WRF for fire risk assessment

Simulation length mattersMean FFWI and mFFWI obtained by WRF at the 4 observational sites are the same as for observationsErrors between (m)FFWI derived from WRF and observations are within the range of observational uncertaintyRMSE are lower for mFFWI than FFWIDetermination of (m)FFWI more difficult in mountainous than relatively flat terrainMeans of (m)FFWI derivded from WRF and observations do not differ significantly (95% confidence) according to a t-test, but variance does according to f-testWRF is suitable for fire weather forecast

Page 25: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Acknowledgements

We thank

Edward O'Lenic and the organizing team for inviting us

Zhao Li, Debasish Pai Mazumder, Ted Fathauer for collaboration

You for your attention

Page 26: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Released aerosols and trace gases reduce air quality

Courtesy: J. Connor

Page 27: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Fires, smoke, and burn scars in Alaska and Yukon Territory in August 2005

2005/08/10 2210 UT(false color) Aqua http://rapidfire.sci.gsfc.nasa.gov/gallery/?search=alaska&date

Page 28: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Large fire scars increase of upward transport and development of a non-classical mesoscale circulation

0500 AST 1100 AST

2100 AST1400 AST

From Mölders & Kramm 2006

Page 29: Development of a Tool for Downscaling of Operational Climate Forecasts to Regional and Local Fire Indices 1,2 Nicole Mölders & 1 Gerhard Kramm University.

Cloud formation changes due to wildfire scars

0500 AST 1100 AST

2100 AST1400 ASTFrom Mölders & Kramm 2006