Near realtime wildfire simulation using big data platforms
Download as PPTX, PDF1 like222 views
The document discusses developing a near real-time wildfire simulation model using big data platforms. It aims to integrate real-time weather and fuel moisture data to dynamically model wildfire spread, conduct Monte Carlo simulations to estimate fire risk, and provide better informed decision making. Current progress includes developing surface fire models integrating National Weather Service data. Future plans are to refine live fuel moisture estimation and testing the model during active wildfires for near real-time predictions. The model seeks to fill gaps in real-time wildfire simulation while leveraging freely available datasets.
Near realtime wildfire simulation using big data platforms
- 1. Near Real-time Wildfire Simulation using big data platforms Bishrant Adhikari Department of Geography, University of Wyoming badhikar@uwyo.edu
- 2. Introduction and overview • Wildfire scenario in United states • Need for near real time wildfire simulation • Possibility of using existing simulation models • Real time inputs in wildfire spread models • Monte carlo simulation for calculation of probability of wildfire • Work in progress and future plans
- 3. Progression of Wildfires in the US Source: http://www.circleofblue.org/2015/world/animation-progression-wildfires-united-states/. Year Fire Name State Total Acres 2004 Taylor Complex AK 1,305,592 2006 East Amarillo Complex TX 907,245 2007 Murphy Complex ID 652,016 2009 Railbelt Complex AK 636,224 2004 Eagle Complex AK 614,974 1997 Inowak AK 610,000 2012 Long Draw OR 557,628 2004 Solstice Complex AK 547,505 2011 Wallow AZ 538,049 2004 Boundary Fire AK 537,098 2009 Minto Flats South AK 517,078 2005 Southern Nevada Complex NV 508,751 2002 Biscuit (formerly Florence) OR 500,068 Source: National Interagency Coordination Center and the National Fire and Aviation Management Web Applications.
- 4. Source: National Interagency Coordination Center (https://www.nifc.gov/fireInfo/fireInfo_documents/SuppCosts.pdf)
- 7. Problems in Wildfire simulation • Often limited to lab models and fire labs • Steep learning curve and cumbersome data preparation process • Complex installation and setup process (some even cost $$$) • Not using freely available datasets • No near real time solution
- 9. Near real time datasets • National Weather service provides near real time weather and wind data • Data is published every 3 hours and are publicly available • Even has forecasts stored in GRIB2(Gridded Binary) format as separate bands • Resolution ~ 5Km Data Unit Temporal Resolution Relative Humidity Percentage Every 3 hrs(8 times a day) Temperature Degrees Kelvin Every 3 hrs(8 times a day) Wind Speed m/s Every 3 hrs(8 times a day) Wind Direction Degrees Every 3 hrs(8 times a day)
- 10. Live fuel moisture • One of the most important parameter in determining wildfire risk and burnability • MODIS Terra and Aqua images are used to estimate moisture content • Relative greenness and water indices • Revisit time approx. 2 days • Moderate spatial resolution (~ 500m)
- 12. Rothermel & Huygen’s Wavelet Principle Source: Rothermel(1972), http://guillermo-rein.blogspot.com/2014/06/forecasting-wildfires-and-natural.html
- 13. Pros- • Semi-empirical equation widely used across several notable softwares (FARSITE,BEHAVEPlus, FlamMap, FSPRO, WiFIRE and so on) • based upon physical equations and has strong theoretical grounds Cons- • Not all the inputs(moisture damping coefficient, Flux ratio could be measured/calculated in real world) • Have implications while used in varied landscape at a larger scale
- 14. Monte carlo simulations & Historical Fire Extent • Quantification of exact coefficients of fuel moisture, terrain, weather and wind is difficult • Changing those parameters randomly • Fire risk/Probability of fire estimated based on proportion of monte-carlo simulation runs that burned a particular area • Time stamped fire perimeters • Used for validation of modelling results (degree of agreement)
- 15. Current progress • Development of surface fire model using Rothermel(1972) and Albini(1976) and Scott and Burgan(2005) models • Integration of real time weather and wind data to dynamically calculate wildfire extent • Automatic estimation of live fuel moisture using MODIS images • Monte carlo simulation with modification of rate of spread equation coefficients
- 16. Work in progress
- 18. Future Plan • Calculation/estimation of live fuel moisture of extinction • Big data computation and parallel processing • Geotrigger based notification • Using near real time captured sensor data such as wind, humidity and precipitation • Testing with currently burning wildfire in near real time • Using updated fire extent and regularly fed field data about fire behaviour Adapted from: http://slideplayer.com/slide/10254426/
- 19. Potential • Fills current void in fire simulation domain • Better informed decision making • Better evacuation planning and resource management • Uses freely available datasets • Uses Open source softwares/Libraries • Use of big data technique to scientifically predict fires in real time
- 20. Limitations • Resolution of datasets (~ 5Km) 2 days for live fuel moisture • Higher degree of uncertainty possible between ignition and first fire extent data • Limitations on the models used • Proper integration of big data computation platform
- 21. Acknowledgements • Klaenhammer Excellence Fund • Paul Stocks Foundation Arts and Science Scholarship • University of Wyoming, Department of Geography • Dean of School of Arts and Sciences • Prof. Dr. William Gribb • Dr. Chen Xu, Dr. Paddington Hodza, Dr. Thomas Minckley
- 22. Near Real-time Wildfire Simulation using big data platforms Bishrant Adhikari Department of Geography, University of Wyoming badhikar@uwyo.edu
Editor's Notes
- #5: The Department of Interior agencies include: Bureau of Indian Affairs, Bureau of Land Management; National Park Service; and U.S. Fish and Wildlife Service. • The U.S. Forest Service is an agency of the Department of Agriculture.
- #7: Public Viewer: Designed to let users zoom to a place of interest Explore map themes Identify wildfire risk for a specific location (basically “What’s your risk?”) Allows users to observe wildfire threat and expected flame lengths for their point of interest Professional Viewer: Mainly for the planners and advanced users (govt officials, hazard mitigation planners, wildland fire professionals) To support community wildfire protection planning needs Advanced functionality and additional map themes available Function to define area of interest and generate detailed summary report. Export and download wildfire risk GIS data Usually requires valid user account and additional permission granted by administrator
- #12: The fire modeling pentagon illustrates the five major influences on fire behavior modeling simulations. Fuelbed structure and slope characteristics are timeconstant influences since those factors do not change during any single fire simulation (which typically lasts no more than a few weeks). Fuel moisture and wind characteristics are time-varying influences because those factors can vary by the minute, hour, day, and week, and thus affect all temporal fire growth simulations. Relative spread direction—heading, flanking, backing—has considerable effect on fire behavior