how data science can help insurers understand wildfire risk · 2019-03-06 · may 3, 2018 aicp...
Post on 15-Jul-2020
0 Views
Preview:
TRANSCRIPT
Proprietary & Confidential: Weather Analytics 2018
Wildfire AnalyticsHow Data Science Can Help Insurers Understand Wildfire Risk
May 3, 2018
AICP Conference
Proprietary & Confidential: Weather Analytics 2018
Today’s Conversation
• Why We’re Here▪ Lookback to the paradigm shift of 2017
• How Wildfire Risk is Changing▪ Ignition and propagation conditions for wildfire
▪ The role of development and climate change in wildfire
• How Data Analytics Support Smart Adaptation▪ Robust risk assessments through opportunistic use of technology
2Proprietary & Confidential: Weather Analytics 2018 2
Proprietary & Confidential: Weather Analytics 2018
Overview of the 2017 Wildfire Season
Proprietary & Confidential: Weather Analytics 2018
Proprietary & Confidential: Weather Analytics 2018
2017
Wildfire: a large, destructive fire that spreads quickly over woodland or brush
In 2017, California suffered its most destructive wildfire season in state history
• Acres burned: 1.3 million
• Structures destroyed: 10,000 +
• Insured losses: ~$14 billion
• Fatalities: 46
4Proprietary & Confidential: Weather Analytics 2018
Proprietary & Confidential: Weather Analytics 2018
More wildfires in winter months
December U.S. Wildfires (2000-2017)
5
Proprietary & Confidential: Weather Analytics 2018
More wildfires in winter months
Year-to-Date U.S. Wildfires (2000-2017)
6
Proprietary & Confidential: Weather Analytics 20187
Wildfires have significant downstream effects that are equally impactful to policyholders as damage caused by flames
• Wind-borne embers
• Near-surface smoke damage
• Business interruption
• Long-term health impacts
Proprietary & Confidential: Weather Analytics 2018 7
Proprietary & Confidential: Weather Analytics 2018
Trends Driving Wildfire Risk
Proprietary & Confidential: Weather Analytics 2018
Proprietary & Confidential: Weather Analytics 2018
What causes wildfires?
Most frequently, us.
• During 1992-2012, humans accounted for 84% of all wildfires and 44% of all acres burned
• Human-caused fire season 3x longer than lightning-caused fire season
Proprietary & Confidential: Weather Analytics 2018 9
Proprietary & Confidential: Weather Analytics 2018 Source: Anchorpoint
Historical fires
Proprietary & Confidential: Weather Analytics 2018 10
10
Proprietary & Confidential: Weather Analytics 2018
• Conditions required for ignition are simple: fuel, heat, and oxygen
• Conditions that drive wildfire growth are complex
o Topography, land cover, and weather are the major determinants of wildfire growth
• Topography: the steeper the slope, the more rapidly and intensely the fire will burn up-slope
• Land cover: the type of vegetation will affect how quickly wildfire spreads and how long the wildfire burns
• grasses fuel fast growth but short duration; old timber fuel slower growth but longer duration
• Weather: high winds, low humidity, drought all enable faster-moving wildfires
Wildfire growth
11
Proprietary & Confidential: Weather Analytics 201812
Proprietary & Confidential: Weather Analytics 2018 12
Proprietary & Confidential: Weather Analytics 2018 1 based on survey of Redfin housing and FEMA risk data
• Continued real estate development in the wildland-urban intermix—the exurban region where undeveloped land intermingles with housing
• Often this development occurs in landscapes prone to wildfire
• 7.7% of US homes—1.5 trillion USD in value—now at risk of wildfire damage1
• Climate change is driving more frequent extreme events that favor wildfire ignition and growth
• Rising temperatures dry vegetation faster
• Lightning occurs more frequently in hot weather
• Earlier springs prolong the fire season
How wildfire risk has changed
13
Proprietary & Confidential: Weather Analytics 201814
Minimal likely exposure
Intermix
Interface
Wildland
Proprietary & Confidential: Weather Analytics 2018 14
Proprietary & Confidential: Weather Analytics 2018
0
20
40
60
80
100
120
140
160
180
200
NV AZ UT CO ID TX NM WA OR MT CA WYChange
in N
um
ber
of
Hom
es
in W
UI
(%)
Real Estate Development in Wildland-Urban Interface Regions,1990-2010
Real Estate Development
15
Proprietary & Confidential: Weather Analytics 2018Proprietary & Confidential: Weather Analytics 2018
Increased frequency in rainfall extremes…
16
Proprietary & Confidential: Weather Analytics 2018Proprietary & Confidential: Weather Analytics 2018
Increased frequency in rainfall extremes…
17
Proprietary & Confidential: Weather Analytics 2018Proprietary & Confidential: Weather Analytics 2018
…and increased intensity of winds
18
Proprietary & Confidential: Weather Analytics 2018
How Data Analytics Support Smart Adaptation
Proprietary & Confidential: Weather Analytics 2018
Proprietary & Confidential: Weather Analytics 2018
• Local weather can drive drastic fluctuations in wildfire conditions
• These fluctuations often occur over the span of hours, leaving little time for organized response
• Accessing and disseminating reliable forecasts on local weather conditions is key to mitigating losses
• NOAA’s High Resolution Rapid Refresh (HRRR) produces quality, actionable information for stakeholders
o 3 km (1.9 mi) spatial resolution
o Next 18-hour outlooks updated in hourly releases
o Radar observations assimilated into hourly updates
High resolution weather forecasts
20
Proprietary & Confidential: Weather Analytics 2018
Area under mandatory evacuation noticeAlert Issued at 12/6/17, 11:00pm
As was the case during the Thomas Fire, the difference in time between forecasts and official alerts can often be more than 12 hours
• At 4am on December 6th, 50+ mph wind gusts were forecasted for later in the day at the northern perimeter of the fire
• At 11pm on December 6th, after conditions had already deteriorated, a mandatory evacuation was issued for the highlighted area
Access to hyper-local weather forecasts extends the critical decision window for first responders, enabling alerts and evacuation notices to be issued in real-time
Smarter Alerting
Proprietary & Confidential: Weather Analytics 2018 21
Proprietary & Confidential: Weather Analytics 2018Proprietary & Confidential: Weather Analytics 2018 22
22
Proprietary & Confidential: Weather Analytics 2018
• Intelligent wildfire risk assessments require gap-free historical data
• Blending ground-based weather stations and remote sensing instruments allows for a fuller picture of past events
• Insurance carriers can use this data to assess overall risk of a location, as well as identify how risk has changed over time
• Having gap-free historical data enables analytics teams to identify important underlying patterns
o I.e., identifying changes to vegetative biomass caused from changing precipitation patterns
High quality historical data
23
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 1980-1985
1980 1985 1990 1995 2000 2005 2010 2015
24
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 1985-1990
1980 1990 1995 2000 2005 2010 20151985
25
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 1990-1995
1980 1985 1995 2000 2005 2010 20151990
26
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 1995-2000
1980 1985 1990 2005 2010 20151995 2000
27
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 2000-2005
1980 1985 1990 1995 2010 20152000 2005
28
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 2005-2010
1980 1985 1990 1995 2000 2010 20152005
29
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index, 2010-2014
1980 1985 1990 1995 2000 2005 2010 2015
30
Proprietary & Confidential: Weather Analytics 2018
Extreme Temperature Index 1980-1985 compared to 2010-2014
1980-1985 2010-2014
31
Proprietary & Confidential: Weather Analytics 2018
Extreme Precipitation Events OverviewExtreme Rain Index 1980-1985 compared to 2010-2014
1980-1984 Average 2010-2014 Average
32
Proprietary & Confidential: Weather Analytics 2018
Plant Disease Susceptibility Index 1980-1985 compared to 2010-2014
1980-1984 Average 2010-2014 Average
33
Proprietary & Confidential: Weather Analytics 2018 Source: Deloitte Unitersity Press | DUPress.com
Cloud computing trends
Computing cost-performance (1992-2012)
34
Proprietary & Confidential: Weather Analytics 2018
The past few years has seen rapid pace of innovation in data analytics
35
Cloud computing and deep learning
• “Deep learning” is the technology behind facial recognition and self-driving cars
• Deep learning algorithms require large amounts of data
• Until recently, the value of deep learning remained theoretical, constrained by high computing costs and insufficient amounts of data
• Cloud computing resources are now a commodity
• Available data only continues to grow with advances in remote sensing
35
Proprietary & Confidential: Weather Analytics 201836
Proprietary & Confidential: Weather Analytics 2018 36
Proprietary & Confidential: Weather Analytics 201837
©2018 Weather Analytics, LLC; Proprietary & Confidential
Integration of pre/post-event imagery with deep learning
provides carriers with actionable risk outlooks and data
Big Data platforms reduce the latency in post-event reporting
Carriers can quickly identify and assess areas with high volume of
impending claims
Proprietary & Confidential: Weather Analytics 2018
Situational awareness: Canyon 2 fire
37
Proprietary & Confidential: Weather Analytics 2018
Extensive historical data enable risk
profiling and trend analyses
Bringing it all together
Remote sensing innovations empower
carriers with real-time wildfire
footprints
High-resolution weather forecasts give
highly actionable outlooks for
proactive risk mitigation
Deep learning programs have the potential to enhance all of the above
38
Proprietary & Confidential: Weather Analytics 201839
Data fusion and predictive modeling provides insurance carriers with powerful decision-making and risk assessment tools for writing and renewing policy in at-risk areas
However, many industry models rely on a narrow set of variables and/or focus on exposure management, resulting in limited capability for real-time response and forecasting
Proprietary & Confidential: Weather Analytics 2018
Data fusion enables more informed underwriting
39
Proprietary & Confidential: Weather Analytics 2018
Wildfire Analytics
In the midst of catastrophe, actionable and accurate data can make all the difference for both emergency responders and insurers.
Weather Analytics is committed to working with insurance carriers to enable actionable, data-driven decision-making before, during and after wildfires.
40
Proprietary & Confidential: Weather Analytics 2018
Questions?
weatheranalytics.com
chris.skarinka@weatheranalytics.com
top related