Creating Added-Value Data Sets From Weather Model Forecasts

August 18, 2014 by Dr. Todd Crawford

In the first part of this two-part series of articles, we examined the various weather models that drive most of the forecasts used by consumers and businesses to make weather-sensitive decisions. Here, we will discuss how these individual model forecasts can be intelligently combined to create an optimal “super-forecast” that is better than any of the individual forecasts and can be used to drive business decisions, such as tropical cyclone track forecasts or severe thunderstorm forecasts. Executive SummaryWSI’s Dr. Todd Crawford explains how P/C insurers, and vendors who work with them, can intelligently combine individual weather model forecasts to create optimal “super-forecasts” to evaluate weather risks like severe thunderstorms and tropical cyclone tracks.

Executive Summary

WSI's Dr. Todd Crawford explains how P/C insurers, and vendors who work with them, can intelligently combine individual weather model forecasts to create optimal "super-forecasts" to evaluate weather risks like severe thunderstorms and tropical cyclone tracks.

Forecast Ensembles

While weather forecast models have improved, and will continue to improve, there are still numerous sources of errors that are impossible to completely eliminate. One of these types of errors is related to what are called the “initial conditions” of each model run.

In order to run a weather forecast model, the best possible estimate of the current state of the atmosphere must first be gathered and input into the model. Since we do not have observations of all the important weather variables at all levels of the atmosphere around the globe, our best estimate of these initial conditions will always be fraught with error.