1. SEVERE CONVECTION & CLIMATE: AN INSURANCE PERSPECTIVE Severe Convection & Climate Workshop Columbia University 9 - 10 March 2016 Photo: FEMA

2. Today’s Panel Kelly Hereid, Chubb Tempest Re Rick Thomas, Willis Towers Watson Kevin Van Leer, RMS Eric Robinson, AIR Worldwide Tom Larson, Corelogic Steve Drews, Aon Benfield Impact Forecasting Moderator: Mark Bove, Munich Reinsurance America

3. Insured Property Losses in the United States Due to Convective Storms, 1980-2015 Source: MRNatCatSERVICE

4. Socioeconomic Factors Influencing Convective Storm Losses Current Atlanta Metro Population: 5.6 million The built-up areas of Atlanta and Barcelona are represented at the same scale 4 U.S. Population Shifts / Sprawl Increase in amount & values of real & personal property Increase in vulnerability of personal property (electronics) Poor construction quality / building codes / code enforcement

5. Insurance & Convective Storms Some Industry Perspectives  Convective storm loss trends has eroded profitability of personal lines insurance across eastern 2/3rds of the United States. Insurers have responded by contractually limiting their liabilities from these events.  Convective storm losses often perceived as a frequency issue, not a severity issue. Tropical Cyclone & Earthquake have much higher single-occurrence loss potentials.  Some large national insurers don’t buy catastrophe reinsurance to cover convective storm losses. Their reinsurance programs, if any, are designed to protect against tropical cyclone / earthquake loss potentials.  Convective storm claims & insured loss totals at the individual company level are, to some extent, perceived as a matter of “luck” – losses are critically dependent on exact location of tornado / hail / wind swaths.  Socioeconomic influences on convective storm loss trend likely dominates any potential anthropogenic signal.  Convective storm models are considered by many in industry as “second tier” due to difficulties in modeling thunderstorm perils.  Many companies do not license or use model operationally; prefer to use pure experience rating of risks. 5

7. Catastrophe Modeling of Convective Storms Key Issues & Gaps  What is the frequency - severity relationship at a given location for each convective storm peril?  What constitutes an convective storm “event” in the model?  How does one address high frequency / low severity convective storm activity?  Event Set Size vs. Model Grid Resolution  How does model handle the small spatial characteristics of individual tornado / hail / SLW swaths?  What statistical distribution best describes the frequency of severe convective storm events?  Can convective storm outbreaks be clustered in time and/or space?  Is there clustering of event severity?  Is the “20/80 Rule” of insurance captured in model results?  Should climate variability / AGW impacts be considered in light of data & modeling limitations? 7

8. © Copyright 2016 Munich Reinsurance America, Inc. All rights reserved. "Munich Re" and the Munich Re logo are internationally protected registered trademarks. The material in this presentation is provided for your information only, and is not permitted to be further distributed without the express written permission of Munich Reinsurance America, Inc. or Munich Re. This material is not intended to be legal, underwriting, financial, or any other type of professional advice. Examples given are for illustrative purposes only. Each reader should consult an attorney and other appropriate advisors to determine the applicability of any particular contract language to the reader's specific circumstances. APPENDIX

9. Thunderstorm Peril Modeling Historical Data Problems: Underreporting Discontinuity due to introduction of NEXRAD Radar and better observing practices. Source: NOAA / SPC 9 Increase primarily consists of additional EF0 & EF1 tornadoes being observed.

10. Thunderstorm Peril Modeling Historical Data Problems: Underreporting Source: NOAA / SPC Increasing trend primarily due to NEXRAD and better observing practices. 10

11. Thunderstorm Peril Modeling Historical Data Problems: Urban Bias, Time of Day, and Borders Source: NOAA / SPC 11

12. Thunderstorm Peril Modeling: Subjective Nature of Defining an “Event” Are these reports from one event? Or three? What is appropriate gap (geographic and time) between events? Source: NOAA/SPC

13. Thunderstorm Peril Modeling: Ensuring Adequate Event Set Coverage Grid size: 1 km x 1 km Original Shapefile Data: NOAA / NWS / Norman, OK WFO Information of event intensity kept for each grid box, but otherwise the actual swaths generated are not explicitly used in the actual model.

14. Annual Convective Storm Outbreak Severity: Maximum Loss Event vs. Average Event