Forecasting Weather Risk: The State of the Art Moves Forward

There may be no more common external risk facing SC managers today than weather risk. Up and down the value chain, producers, processers and retailers are all impacted in one way or another by climate and climate volatility. Yet only a small number of firms make a serious and consistent effort to model weather risk and the exposure it brings to their operations.

A recently report, published by the Emerging Risks Team at Lloyd’s (Forecasting risk: The value of long-range forecasting for the insurance industry) highlights the improving quality of weather modeling and forecasting and makes the case for re-examining their use in risk quantification. (1)

As the report notes:

Climate models are now yielding skillful regional forecasts beyond seasonal timescales. Skillful climate model predictions have been demonstrated well beyond the seasonal timescale for the first time by the MET Office Hadley Centre DePreSys system – which won the Lloyd’s Science of Risk Prize in 2010….there is significant potential skill at forecasting Atlantic hurricane season activity rates at timescales out to seven years. This is beyond what can be achieved using, for example, the average of the last five years as a forecast (P 15).

The report provides a good summary of some of the challenges in this field, noting the complexity of the models and the large data sets involved:

Current cutting edge catastrophe modeling methods use large hazard event sets describing, for example, the equivalent of 10,000 years of hurricanes. These are generated by random (stochastic) re-sampling of the parameters that describe the observed hazards from the last ~50 years, such as maximum wind speed in a storm. These are now increasingly enhanced by simulations from high-resolution numerical weather models. Such hazard event sets can be re-sampled to reflect forecasts derived from statistical or expert elicitation exercises (P 9).

The report’s recommendation is that underwriters increase their use of these increasingly useful models, and the same advice holds true for SC risk managers. The authors highlight that today the World Meteorologal Organization has eleven centers producing advanced weather models “The Climate Prediction Center (NOAA), Met Office and the European Centre for Medium Range Weather Forecasts (ECMWF) operate three of the most cutting-edge systems,” note the authors, adding that “Europe, among others, has historically funded and led several large international collaborative research programmes” (P 13). This data, which is being accesses and analyzed by the risk industry can also be used by SC risk managers to understand their own exposure to weather risk in areas such as sourcing decision, transport lane design, and manufacturing site selection. The lay impression is often that weather is “unpredictable” but as the Lloyd’s report notes, this is hardly the case.

It behooves any SC risk manager whose operation is seriously dependent on weather to review the Lloyd’s report and perhaps consider increasing their understanding of the latest quantitative techniques in use and how these may soon filter into pricing of weather related risk products brokers will bring to the table in the near future.



Carlos Alvarenga

Founder and CEO at KatalystNet and Adjunct Professor in the Logistics, Business and Public Policy Department at the University of Maryland’s Robert E. Smith School of Business.

One thought on “Forecasting Weather Risk: The State of the Art Moves Forward

  1. Studying the impacts of climate change and the associated increase of extreme weather events has been a hobby of mine since about 2007. So this blog and the Lloyds report were a delight to read. The primary emphasis, at least here in the US, should have shifted from, “is it really happening” to “how do we integrate our knowledge about what is happening to useful planning?” years ago.
    As the authors of this report said in the ‘Confidence’ section under ‘Requirements of a Forecast’, “a model is only useful if
    it helps with business-relevant decisions – and skill measures should be designed to test this.”


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s