fires and floods: new models for disaster prediction swlgema hilton 16 9 2.pdf · fires and floods:...
TRANSCRIPT
Fires and floods: New models for disaster prediction SWLGEMA Conference, Bunbury, August 2015
DIGITAL PRODUCTIVITY FLAGSHIP
James Hilton, Raymond Cohen, Yunze Wang and Mahesh Prakash
ENIAC (1943) iPhone 6 A8 (2014) 150,000,000
times faster.. • What can we do with this new computing power?
Introduction Natural Hazard Modelling
CSIRO | SWLGEMA 2015
Introduction Natural Hazard Modelling • Ability to model natural hazards • Digitally represent natural systems • Movement of water or fire can be predicted
• Greater resolution • More physics • Not necessarily more accurate
More computing power gives:
CSIRO | SWLGEMA 2015
• Fire propagation • Arrival time • Fire intensity • Risk assessment
Introduction Natural hazard modelling
Swift - floods • Flash flooding • Coastal inundation • Mitigation modelling • Integrated hydraulic model
Spark – bushfires:
Goal: Open and reliable disaster modelling software for decision makers
CSIRO | SWLGEMA 2015
Analysis • Risk identification (Manual analysis) • Visualisation (Impact/communication) • Automated data gathering
Generation • Data conversion/generation (Clean-up) • Parameter space
Synthesis Synthesis Computation • Run computer models • Multiple simulations
Scenario • Requirements (Risk assessment, validation) • Gather data (Terrestrial, bathymetric, vegetation)
Introduction Anatomy of prediction
CSIRO | SWLGEMA 2015
Introduction Workflow Architecture
• Workflow execution system • GUI workflow editor • Freely available at: http://research.csiro.au/workspace
CSIRO Workspace
Workspace GUI editor
CSIRO | SWLGEMA 2015
• Collection of parts for modelling natural hazards • Each part is represented by a block
• These parts can simply be plugged together
CSIRO | SWLGEMA 2015
Introduction Workflow Architecture
• The parts build up a system
1. Solver 2. Execution 3. Input 4. Output
CSIRO | SWLGEMA 2015
Introduction Workflow Architecture
External linkage
CSIRO | SWLGEMA 2015
Introduction Overview • Integrated software • Multiple OS
• Can be run on a cloud server
CSIRO | SWLGEMA 2015
Introduction Overview
Bushfire Model Spark • Tracks evolution of fire perimeter • Front speed from empirical models • Uses fuel, wind, moisture parameters • Tile- based level set method
• Arrival time • Fire intensity
Front speed
CSIRO | SWLGEMA 2015
• Track fire front • Front speed from empirical models • Dependant on wind, fuel, topography… • Many different empirical models
Fire front 1 2 3
Computational level set method
• Natural for interfaces • Front represented by level set • Handles topological changes • Computationally efficient
New fire front
Front speed
CSIRO | SWLGEMA 2015
Bushfire Model Computational Model
Solver
Input layers
Log
Fire model
OpenLayers display Processor
Bushfire Model Spark
CSIRO | SWLGEMA 2015
Bushfire Model Spark
CSIRO | SWLGEMA 2015
• System can be behind a user interface
CSIRO | SWLGEMA 2015
Bushfire Model Spark framework
• Fire behaviour models can easily be added • For example, including a spotting model
Spotting demo in workspace
CSIRO | SWLGEMA 2015
Bushfire Model Spark framework
• One result isn’t useful for random conditions, such as spotting • Need to run ensemble simulations
• Ensemble/analysis functionality built into framework
Run 1 Run 2 Run 3 Run 4 1000 runs
Bushfire Model Ensemble analysis
CSIRO | SWLGEMA 2015
• Input conditions from image analysis of from real fires using OpenCV
Bushfire Model Image analysis
CSIRO | SWLGEMA 2015
• Is curvature crucial to bushfire propagation?
• Curvature is not incorporated in any conventional bushfire models
• Radiation or convection?
• Current study to look into curvature effects
Tim
e
Experiment Conventional model
Curvature incorporation
Bushfire Model Fire dynamics
CSIRO | SWLGEMA 2015
Automate input: fire locations, precipitation predictions, storm surge Up-to date land classification: remote sensing data analysis
Bushfire Model Operational Systems • Input data must be as recent as possible:
• Output data must clearly quantify impact and risk
• Simulation data must be generated as fast as possible
• System fulfils all of these requirements
CSIRO | SWLGEMA 2015
• Water depth << width • 2D formulation • Predicts water height, momentum • Shallow water finite volume model
Hydrodynamic Model Swift
width
depth
• Dam breaks • Flash flooding • Coastal inundation
CSIRO | SWLGEMA 2015
Solver
Visualisation sub-workflow
Processor: Hydraulic model
Hydrodynamic Model Swift
Display
CSIRO | SWLGEMA 2015
Hydrodynamic Model Swift
CSIRO | SWLGEMA 2015
• NOAA standard validation test cases • ‘Simple beach’ test case • Wave run-up and inundation
Initial wave
Hydrodynamic Model Validation
• All NOAA standard validation cases • Historical dam breaks • Historical flash flood events
CSIRO | SWLGEMA 2015
Hydrodynamic Model Townsville storm surge
http://hardenup.org
• Risk assessment for large-scale inundation event
CSIRO | SWLGEMA 2015
Hydrodynamic Model Kakadu Saline Inundation
CSIRO | SWLGEMA 2015
Structural mitigation options: • Levees, basins and dams • Drainage networks
Elwood Swamp 1886
Hydrodynamic Model City of Port Phillip
CSIRO | SWLGEMA 2015
Drainage flow direction
Normal conditions
• Head-based network model • Assume sound speed >> gravity wave speed
= Storm tide Drainage flow direction
Drainage flow reversal
• Drainage reversal
Hydrodynamic Model City of Port Phillip
CSIRO | SWLGEMA 2015
• Extensive urban drainage network (~13k pipes) • Model effects of 1.3 m storm surge
0.1
m3/s
0
m3/s
Flo
w rate
N
Hydrodynamic Model City of Port Phillip
CSIRO | SWLGEMA 2015
• Present day sea level +80 cm
Hydrodynamic Model City of Port Phillip
CSIRO | SWLGEMA 2015
• Storm surge and flooding scenarios • Range of mitigation options • Included one-way valves
Without mitigation With mitigation
24 hrs 2 hrs Retention time
CSIRO | SWLGEMA 2015
Hydrodynamic Model City of Port Phillip
Present day sea level
Without mitigation With mitigation
24 hrs 2 hrs Retention time
CSIRO | SWLGEMA 2015
Hydrodynamic Model City of Port Phillip • Storm surge and flooding scenarios • Range of mitigation options • Included one-way valves
0.4 m SLR
CSIRO | SWLGEMA 2015
Hydrodynamic Model City of Bunbury
• Coastal region susceptible to flooding • Look at both storm surge and rainfall events • Preston River and the Leschenault Inlet areas historically worst hit
• WA National Disaster Resilience Program (WA-NDRP) project
• Study flooding for future planning and mitigation options
• Investigate extreme present and future (with SLR) flooding events
• Impact of potential mitigations including
• Retention/detention schemes
• Changes to the drainage network
• Changes to pumping stations
CSIRO | SWLGEMA 2015
Hydrodynamic Model City of Bunbury
Lidar, bathymetry, land usage Pipe networks, pumps,
retention systems
• Input data sets Rainfall
Tides
Joint probability analysis
Statistical extreme tide & rainfall scenarios
Summary Modular solver architecture
CSIRO | SWLGEMA 2015
• Efficient models for natural hazards • Open and configurable • Platform for risk modelling • Flexible, adaptable framework • Embedded data analysis • Interactive applications
DIGITAL PRODUCTIVITY FLAGSHIP
CSIRO Digital Productivity James Hilton
t +61 3 9545 8002 e [email protected] w http://www.csiro.au
CSIRO Workspace http://research.csiro.au/workspace
Thank you