AIR Worldwide Issues Public Comment on Treasury Department Notice

WASHINGTON, Nov. 19 -- Roger Grenier, senior vice president for global resilience practice leader at AIR Worldwide, Boston, Massachusetts, has issued a public comment on the Department of the Treasury notice entitled "Federal Insurance Office Request for Information on the Insurance Sector and Climate-Related Financial Risks". The comment was written on Nov. 15, 2021, and posted on Nov. 16, 2021:

* * *

This letter is submitted in response to the Request for Information on the Insurance Sector and Climate-Related Financial Risks issued on August 31, 2021. AIR Worldwide appreciates the opportunity to offer our perspective as a leading provider of analytics for managing extreme event risk. Over 400 organizations rely on AIR's models and software, and our tools have become the standard through which insurers, reinsurers and other market participants assess and manage catastrophe risk - from natural and manmade catastrophes to cyber-attacks, pandemics, and climate change. AIR supports clients in the filing of regulatory requirements, answering questions on climate change risk analysis and scenario tests such as the Bank of England 2021 Climate Biennial Exploratory Scenario and the 2021 Bermuda Monetary Authority Climate Change Exposure Assessment, and responding to risk assessment questions from rating agencies.

Our comments below address questions related to the use of extreme event models in insurance, data requirements associated with the models, consideration of climate risk, and our observations about climate-related disclosures. Our primary focus in this response is on the physical risks from climate change, based on our experience building models, our research on climate change and our decades of helping our clients prepare for and recover from extreme events.

Use of Extreme Event Models

As noted above, extreme event (or catastrophe) models have become the standard method for measuring, managing and transferring risk in the insurance market. These models simulate a variety of natural hazards, including climate-related perils such as hurricane, flood, wildfire, and severe thunderstorm. Models of peril event activity are then superimposed onto the built environment to generate probabilistic estimates of damage and loss.

The models are rooted in physical-statistical approaches that combine weather and physical variables with scientific expertise and simulate many possible scenarios that could be expected in a given year. The "catalog" of simulated peril events is conditioned on the near-present climate, i.e., it considers the most recent data available and includes adjustments as necessary to account for climate change which has already occurred. This near-present view is appropriate for managing today's risk. However, although the existing models are informed by historical data, they are by no means constrained by only events which have occurred. Catalogs include many plausible events which could occur not only today, but also in a warmer climate. Hence the extreme event modeling framework is an important tool in scenario modeling.

How these models use data varies, but in all cases, long term, high resolution data are key to building robust views of a peril and estimating losses to an asset or investment. Detailed, high quality historical data inform models at all stages, from developing underlying distributions, improving understanding about interactions between perils and the built environment to model validation. Many sources of data used across the industry to model views of risk come from large scale government projects. Weather data from the National Weather Service, streamflow information from the US Geological Survey, satellite data from NASA, and historical fire records from the US Forest Service are just a few examples of the many datasets being utilized. The combination of these data along with the modeling framework allows the insurance sector to translate these variables into a view of risk. Ensuring adequate funding to maintain and enhance the data available across government is critical to this effort.

Just as critical is developing new datasets that consider how climate change will impact extreme events in the future. As mentioned above, high resolution, detailed data are being employed in understanding the risk for the current climate. To move into understanding the effects of climate change thoroughly, developing similar variables in climatic models will be key going forward. Given the longstanding history of the insurance industry evaluating risk using these models, any approach which looks to answer future questions should take advantage of the data, methods and expertise insurers have already developed. Extreme event models excel at capturing hazard estimates with sufficient geographical resolution--and loss estimates for specific asset types--to inform the types of business decisions insurance companies are trying to make.

It's worth noting that since 2012, AIR has published a report on Global Modeled Catastrophe Losses which includes global loss metrics based on AIR's latest suite of models. This year's report/1 concludes that while the actual average loss over the past decade is approximately USD 75 billion, our modeled estimates suggest that global industry modeled average annual loss is far higher - USD 106 billion. Recent insurance industry loss experience has prompted discussion around extreme event models, and in particular whether models are underestimating risk. The results from our models suggest that the industry should expect that significant losses can and will occur under the climatic conditions of today.

Approaches for Assessing Climate Change Impacts

Extending the extreme event modeling framework for future climate risk requires new data and methods - including climate models. Climate projections created by the Intergovernmental Panel on Climate Change (IPCC) and others rely on mathematical models of the atmosphere, ocean and land systems often referred to as General Circulation Models (GCMs). While the latest generation of climate models is increasing in sophistication, most models are at coarse geographic resolution (order of 100km or more) and do not directly include the detailed physics that drive the extreme events of interest to insurers. Methods aimed towards "downscaling" the results are being continuously refined, but these can introduce biases that can limit the usefulness of location-specific projections necessary for insurance risk.

Since the GCMs, by themselves, are generally unsuitable to quantify the location and peril-specific risks of concern to insurers, hybrid methods are often employed to develop forward-projected views of risk in extreme event models. Such methods are often informed by the scientific literature and other expert guidance developed from studies of GCM output. One such approach is to use GCMs to adjust current climate frequencies and intensities of events to develop new catalogs of events reflecting future climate scenarios, an approach which has been utilized by AIR in recent projects./2

A key question in developing climate scenarios is whether the impact of climate change is well understood for the peril in question. Detecting and attributing climate change impacts on various weather phenomena is a relatively new branch of climate science. Attribution confidence depends on many factors, including: the degree of alignment across climate models; whether there is a detectable trend in the historical data that agrees with model results; and how well we can physically connect and understand the modeled or observed effect of climate.

Temperature phenomena are most confidently assessed because of the direct physical connection between increasing carbon dioxide (and other greenhouse gases) and a warming atmosphere. There is also a high degree of confidence in the impact of sea level rise, which originates from a combination of melting glaciers and ice sheets and thermal expansion of warming seawater. Sea level rise contributes to chronic ("sunny-day") and acute flooding through coastal storm surges. Precipitation, drought and wildfires are increasingly accepted to be heavily influenced by climate change.

There is less confidence that climate change is impacting other perils, including some of the most costly (tropical cyclones) and most frequent (severe convective storms and floods) and thus of most interest to insurers. The reasons for this low confidence include:

* The relative infrequent occurrence

* A historical record with changes in observational uncertainty over time

* The inherently nonlinear physics driving these events

* The small physical scale relative to the geographic resolution of many climate models

The main point is it is important to distinguish where the impacts of climate change are most certain from those where scientists have lower confidence. Since many of the lower confidence perils are of greatest concern to insurers, there are limits to how confidently we can assess climate impacts.

In addition, while climate change is happening, natural variability may suggest a stronger trend than what actually exists. Scientists have identified natural cycles that influence hurricane activity, and the signal from climate variability may exceed any signal from climate change, at least in the short term. Thus, we should not be surprised to see periods of increased activity followed by less active periods - the most recent example being very active 2004 and 2005 hurricane seasons, with 8 hurricanes of major strength (category 3+) making landfall in the US followed by a nearly 13-year "drought" of major hurricane landfalls until Hurricane Harvey in 2017. Such periods of activity may be influenced by climate change, but other factors may play a role.

Any risk assessment framework should consider that while climate change will influence extreme events, there may be considerable uncertainty and significant annual variability. Furthermore, the evolution of physical risk is only one aspect of climate change relevant for insurers - as communities adapt, markets adjust, and populations migrate the risk landscape will be in continual flux. These socioeconomic factors are not considered on today's risk models but will likely have material impacts.

Climate-Aware Risk Assessment

Our position as a data and analytics provider gives us a unique perspective on climate risk disclosures. In addition to participating in working groups and discussions with regulators in Europe and elsewhere, we also develop guidance and engage with our insurance company clients as they respond to various requests. Our main observation is that given the uncertainty in climate projections and the evolving models and tools, the emphasis should be less on quantitative data collection and reporting procedures and more on promoting climate-aware risk management frameworks. Given the uncertainty in forward projections described above, strictly quantitative measures may not provide useful information. Most insurers are well positioned with the tools, data and expertise to manage climate risk within existing risk management and examination processes, and insurers have the knowledge to assess the materiality of risk in their portfolios. The ultimate goal for insurers should be to develop more informed risk management and governance processes, better data and tools to inform those process, and more climate-aware business practices.

Thank you for the opportunity to provide our perspective. We hope that these comments are useful as you complete your work and develop your recommendations. Please feel free to contact me with any questions.

Thank You

Roger Grenier, Ph.D. Senior Vice President, Global Resilience Practice Leader

AIR Worldwide

* * *

Footnotes:

1/ 2021 Global Modeled Catastrophe Losses Report, available from the AIR website

2/ See AIR White Paper: Quantifying the Impact from Climate Change on U.S. Hurricane Risk available from the AIR website

* * *

The notice can be viewed at: https://www.regulations.gov/document/TREAS-DO-2021-0014-0001

TARGETED NEWS SERVICE (founded 2004) features non-partisan 'edited journalism' news briefs and information for news organizations, public policy groups and individuals; as well as 'gathered' public policy information, including news releases, reports, speeches. For more information contact MYRON STRUCK, editor, [email protected], Springfield, Virginia; 703/304-1897; https://targetednews.com