Presentation about our probabilistic coupled inland flood, hurricane wind and storm surge model. We discuss Harvey losses, SpatialKat runtimes, and show for the first time industry wide EP curves for all perils combined and by themselves. Highlights also include our climate sensitivity and climate variability modeling, another industry first. Comments welcome.
1. US Coupled Inland Flood,
Storm Surge and Wind Modeling
Embracing Volatility
RAA, Orlando, February 2018
KatRisk LLC
752 Gilman St.
Berkeley, CA 94710
510-984-0056
www.KatRisk.com
KatRisk Deutschland GmbH
Wilhelmstr. 6
79098 Freiburg, Germany
0761-5146-7600
2. ARGO / Ariel Re US Model Intercomparison
Compared all commercially available US flood models in November 2017
– KatRisk
– Impact Forecasting
– CoreLogic
– AIR
Large range of modeled AAL
– Factor 3x between models
Large differences in EP
– Number of events
– Correlations between regions
Some models don’t model TCs explicitly
Special thanks Federico Waisman (Ariel Re)
https://www.argolimited.com/flood-model-showcase/
3. 2017 KatRisk US Inland Flood, Storm Surge, Hurricane Model
Summary Highlights and some Cat Model Industry Firsts
Fully correlated multi-peril 50k year event set (up to 50 Million years
sampled) with TC and non-TC flood events
Groundbreaking low run-times from laptop to server to cloud
2-d hydraulic modeling everywhere (storm surge and inland flood) with
user defined inland flood defenses (with KatRisk defaults)
Actuarial coherent view of risk computations with repeatable location
aware correlated uncertainty sampling (allows buildings as footprints)
Global correlations through teleconnections and climate change sensitivity
Transparent financial model with multi-peril contracts
Expose key model sensitivities to user (flood defences, correlation, etc.)
4. KatRisk Simplified Global Workflow
Probabilistic Deterministic Expensive Financial + Analysis + API
Probabilistic VARMA based
Ocean SST model
Probabilistic Tropical Cyclone Track Model Probabilistic Precipitation and
Temperature Model, Surface Meteorology
TC Precipitation Model
Land Surface Model
River Routing Model
Hydraulic 2-d Flood Model
Tropical Cyclone Wind Model
Storm Surge Model,
Tidal Model
2-d hydraulic
Exposure and GU Loss Model (API for third party data integration)
Insured Loss Model (Policies, Treaties)
Statistics, Analysis, Maps (WMS), Web Interface (GUI) and Web Service (API)
Probabilistically sampled vulnerability with correlated severity distributions
Global Teleconnections
Climate Change SLR
Peril-Peril Correlation
Spatial - Temporal
Uncertainty Correlation
C
C
C C
C
5. KatRisk Hurricane and Storm Surge Model
A climate conditioned hurricane track set developed for the Atlantic Basin (1km resolution,
10k * 5 years of events)
Combined with roughness, windfield, and vulnerability models, full wind loss modeling
capabilities
Sample Tracks 100 Year Windspeed Map
6. Storm Surge and Inland Flood
Storm surge (SS) has been simulated for the
entire 50k year track set and output on a 10m
resolution grid with parametric wave model
Inland flood simulation with TC and non-TC
rainfall. 50k years of continuous simulation of
pluvial and fluvial flooding (KatRisk US Flood
Model 2017)
Correlated Wind fields, storm
surge, and TC precipitation
7. Storm Surge Modeling
Storm surge has been analyzed for 50,000 years of hurricane tracks
Houston
Chesapeake Bay
New Orleans
New York
Images show KatRisk Score (1-10)
9. Model Run-Times [1.2 Million Locations]
KatRisk SpatialKat runs on most Windows or Linux computers with minimal
resources with 100% scalability with #samples
– 50MB / sec hard drive read speed per core, 1TB HD space for model data
– Less than 1GB / core memory for 1.2 Million locations for all perils, 10 samples
– Speed also depends on what level of detailed output is selected
– Software Requirement: Open Source R and packages, Open Source C++
Machine CPU Disk Memory TC wind Flood Storm
Surge
All Perils
Laptop 4
cores
500 MB/s
SSD drive
8 GB 25 min 28 min 7 min 72 min
High End
Desktop
6
cores
500 MB/s
SSD drive
32 GB 12 min 13 min 3 min 40 min
Server 25
cores
> 2 GB/s
PCIe drive
256 GB 189 sec 207 sec 41 sec 603 sec
10. Peril – Peril Correlation: Harvey
KatRisk released modeled footprint during event, and updated throughout event
Loss estimates based on KatRisk footprints
– 8.8 million point IED in Texas
– $40 - $50 Billion GU Texas Inland Flood Loss
– Large demand surge (1.4?) + wind and storm surge (<$2 Billion) + other areas ~ $80 Billion
11. Overview of US Economic Insurable Losses
USA AAL All Perils (TC, IF, SS) Combined = $39 Billion +- $6 Billion
Model run with economic exposure
– About $80 Trillion insurable
– Three LOBs
– Average Vulnerability
– Ran every 10th location
– 100 Samples (5 Million year EP)
– Model IF/SS results sensitive to
assumptions of BFE
– SpatialKat run-time GU/GR ~ 80 min
on 25 cores Xeon E5-2690
$80 Billion
16 Year RP
12. AAL and EP all Perils (Flood, Storm Surge, Wind)
Combined
Wind
Inland Flood
Storm Surge
Combined
Wind
Inland Flood
Storm Surge
OEP AEP
$80 Billion
16 Yr RP
77 Yr RP
$80 Billion
Combined OEP and AEP curves for all perils and combined
– AAL TC Wind = $12 Billion - $15 Billion
– AAL Inland Flood = $16 Billion - $22 Billion
– AAL Storm Surge = $4 Billion - $7 Billion
Wind drives tail risk
Flood AEP highest in low return periods
OEP
13. EP all Perils (Flood, Storm Surge, Wind)
Zoom: all perils combined
Combined
Wind
Inland Flood
Storm Surge
Combined
Wind
Inland Flood
Storm Surge
OEP AEP
$80 Billion
16 Yr RP
77 Yr RP
14. Deeper look into TC vs. non-TC losses
Combined
Wind
Inland Flood
Storm Surge
Combined
Wind
Inland Flood
Storm Surge
OEP AEP
$80 Billion
420 Yr RP
25 Yr RP
Just TC only, wind, inland flood, storm surge
How special was Harvey for just TC flood? Answer: very
15. Deeper look into TC vs. non-TC losses
Zoom: Just TC only, wind, inland flood, storm surge
Combined
Wind
Inland Flood
Storm Surge
Combined
Wind
Inland Flood
Storm Surge
OEP AEP
420 Yr RP
$80 Billion
16. AAL TC Flood = $3.5 Billion to $5 Billion (18% to 25%)
– Contribution of Atlantic is about 17% to 23.5%, Pacific the rest
Overview of Losses – TC contribution to Flood
17. AAL TC Flood = $3.5 Billion to $5 Billion (18% to 25%)
– Contribution of Pacific is about 1% to 1.5%, Atlantic the rest
Overview of Losses – TC contribution to Flood
18. Effects of ENSO on Precipitation in Oct – March
https://www.climate.gov/news-features/featured-images/how-el-ni%C3%B1o-and-la-ni
%C3%B1a-affect-winter-jet-stream-and-us-climate
19. Flood AAL differrence El Nino
Wet
Dry Strongest Effect on
Precipitation is during Oct-Mar
1. Filter out non Oct-Mar
Events (IF Only)
2. Compute State AAL
3. Filter Out Oct-Mar and
Strong + ENSO Years
4. Compute % Difference
20. Flood AAL Difference La Nina
Wet
Dry Strongest Effect on
Precipitation is during Oct-Mar
1. Filter out non Oct-Mar
Events (IF Only)
2. Compute State AAL
3. Filter Out Oct-Mar and
Strong + ENSO Years
4. Compute % Difference
21. Coastal AAL Difference El Nino (TC Wind)
High
Low Effects Hurricane
Generation (top 20% of
ENSO index)
1. Compute State AAL
2. Compute % Difference
22. AAL Difference El Nino (TC Surge)
High
Low Effects Hurricane
Generation
1. Compute State AAL (top
20% of ENSO index)
2. Compute % Difference
23. AAL Difference Flood Positive AMO
Wet
Dry
Strongest Effect on
Precipitation is during Oct-
Mar
1. Compute State AAL (top
20% of AMO index)
2. Compute % Difference
24. AAL Difference Flood Negative AMO
Wet
Dry
Strongest Effect on
Precipitation is during Oct-
Mar
1. Compute State AAL
(bottom 20% of AMO
index)
2. Compute % Difference
25. USA AAL by Atlantic SST and ENSO
Hurricane losses dependency on Atlantic SST Anomaly and ENSO
AAL by Atlantic SST
AAL by ENSO
Introduction of SST leads to clustering
for TCs that cause losses in the USA
# Atlantic TCs with SST
Dispersion = 1.15
# Atlantic TCs Poisson
26. KatRisk has 22 failure modes by default
which define the probability of defense
failure vs return period up to 1 in 1000
year probability
>1k years and above undefended
User can add in any new curves
By default, all locations have a defense
mode of 4 for pluvial and 5 for fluvial
For the US, by default IF:
1, fluvial defended at mode 20
2, fluvial defended at mode 21
3, fluvial defended at mode 22
User-modifiable by imported location
or area
All Failure Modes Default Failure Modes plus a
low and a high
KatRisk SpatialKat Defense Failure Modes
28. SpatialKat also accepts building footprints instead of
latitude/longitude point locations
For the example of the hospital in the image to the right,
using street level geocoding, it would get a location of the
red-star which has near zero loss
KatRisk allows building discretization wherein value is
distributed over all the grid-cells over a footprint (blue dots)
Allows for a coherent view of risk through location aware
sampling
SpatialKat Building Footprint Capability
Data from BuildingFootPrintUSA
Study to be published around March 2018
29. Catastrophe models need coherent measures of risk
Many current catastrophe models do not support coherent measures
of risk
– Can you diversify your portfolio when hypothetically insuring the same building
twice and then take half the risk?
● Coherent risk measure ρ on measurable function Z
● Positive homogeneity: if α ≥ 0, then ρ(αZ) = αρ(Z)
– Can your risk measure go up when you diversify?
● Sub-additivity: ρ(Z1 + Z2) ≤ ρ(Z1) + ρ(Z2)
– Can EP losses go down anywhere on the EP curve when adding e.g. a
building?
30. SpatialKat Financial Model
Explicit inuring order
between perils
Choose wind or flood first
Choose how wind and
flood losses should be
executed within a contract
Choose how surge and
inland flood should be
executed within a contract
Please visit our booth for a live demo
Financial Model
Limits, Deductibles
and Blankets
Location
Coverage
Site
Account
Portfolio
Facultative
Reinsurance
Special Conditions
Comprehensive client survey to
ensure contracts execute as they
do in reality
31. Climate Sensitivity: Short Story about Sea Level
Representative
Concentration
Pathways
LIG
127k
11k to today rise in sea level
Lohmann, AWI
32. Sea Level Rise Puzzle
During Last Inter-Glacial (LIG) exposed fossil reef indicate 5m - 9m higher sea
level (Dutton & Lambeck, 2012, Dutton et al., 2015)
LIG with Sea Surface Temperature Southern Hemisphere + 1 - 3o
C warmer (Capron
et. al. 2014)
Lohmann, AWI
33. Melting West Antarctic Ice Sheet from below
Last Interglacial: Climate Models and paleo-climate data are consistent
Antarctic Ice Sheet: Marine ice sheet instability -> Sea level rise
Threshold ~2°C based on paleo-climate and climate model studies
Lohmann, AWI
34. KatRisk Surge Climate Change Study
Compares USA surge losses with current conditions, conditions around 1900,
and a uniform 30 cm sea level rise (SLR)
– Current speed of SLR is about 2.8 to 3.6 mm/year (currently accelerating), and was about
1.8 mm/year in the 20th century
Use high resolution exposure of $6.88 trillion along the coasts results
summarized on 200m gridded resolution
Buildings, contents, time element and appurtenant structures modeled
Ground-up AAL increased from $5 billion to $6.9 billion, implying an increase of
about $60 million per centimeter SLR, or currently about $20 million per year
(although the increase is not linear), equal to 0.4% of the AAL – but also with
potential to accelerate.
Ground-up AAL increased from $4 billion to $5 billion based on a 20cm sea level
rise between 1900 and today. Simulations for 1900 assume the same sea
defenses, bathymetry, and tropical cyclone frequency and severity as today.
Number are slightly different compared to before – ran different BFE assumptions for this
40. Increase in GU loss AAL [$Bn] by State
Increase in GU loss AAL
between 1900 and today,
as well as today to uniform
30cm SLR scenario.
Exposures are from current
residential, commercial and
industrial estimates
Risk increase is measured
as increase in AAL by state
41. KatRisk Cat Response
For the last three years KatRisk has released wind and flood
footprints of major events within days of an event
Inform KatRisk models with observations (Data Assimilation)
Cat Response slides are on http://www.katrisk.com/recent-events
Compare flood footprint with FEMA
and point observations (Pensacola 2014)
43. Summary
Available Products KatRisk SpatialKat (laptop to server to cloud)
– US probabilistic TC (wind + storm surge + inland flood), non-TC inland flood
– Canada probabilistic inland flood (coupled to US)
Available Products KatRisk SoloKat (laptop to server to cloud)
– Worldwide flood maps (US and UK 10m - Europe, Canada, Australia 30m)
– Worldwide location loss analytics
KatRisk APIs and Third Party data integrators: SpatialKey, MapRisk, ...
Please visit KatRisk Booth for a demo