Ordinary Average Guy - EIA's Latest Approach to Weather Modeling Promises Forecast Improvements

The Energy Information Administration (EIA) recently changed the weather forecast methodology for one of its most important energy models — the Short-Term Energy Outlook (STEO) — and while we talk about the effects of weather on energy markets fairly often (571 times in the past 12 years, or about once a week, but who’s counting?), we rarely take a step back and explain how those weather forecasts are used. In today’s RBN blog, we look at different approaches to weather forecasting, the recent change made by the EIA, and how the new approach might affect our understanding of EIA forecasts.

Weather has a significant impact on the demand for — and sometimes, the supply of — a number of hydrocarbon commodities, especially natural gas and propane, but also heating oil and other fuels. The benefits of accurate weather forecasts are obvious: For example, a month-ahead or quarter-ahead prediction of a frigid January in Massachusetts or a lingering July heat wave in Texas would be of major value to a wide range of market participants, including electric-grid planners, power generators, and natural gas marketers and pipelines, not to mention propane and heating-oil dealers in New England. In addition to helping markets anticipate fuel demand, forecasts of extreme weather also can signal prospective challenges with fuel supply, such as well freeze-offs that temporarily slash production of natural gas (see Terminal Frost) or blizzards that make some commercial/residential fuel deliveries virtually impossible.

For these reasons — and more — the vast majority of energy supply/demand models factor in predictions about the weather. Of course, there are many approaches to weather forecasting as well as many “time horizons” for the forecasts being made. We’ll get to the various forecast approaches (and the EIA’s recent change) in a moment — first we’ll look at the time-horizon angle. Most government weather services and private weather-forecast vendors break up their forecasts into these five categories:

• NowCast: Less than 24 hours
• Short-Range: 1 to 5 days
• Medium-Range: 6 to 16 days
• Extended-Range: 17 to 45 days
• Long-Range (aka Seasonal): 46 days to 1 year

But these definitions are clearly not what the EIA has in mind when it talks about its “Short Term” Energy Outlook, which extends out a couple of years. So to hopefully keep the confusion level to a minimum, we are not going to use either the weather service or the EIA terminology. Instead, given that our focus is on how the EIA incorporates weather forecasting in its widely read STEO and Winter Fuels Outlook (~15 months forward), we’re going to zero in on what we think is best called the medium-term planning horizon — in effect, a combination of the “Weather Service” Extended-Range and Long-Range/Seasonal time periods, or 17 to 365 days. As you would expect, forecasts using this substantial time horizon are of much more value to market participants looking weeks or months ahead than the large subset of energy traders and others focused on the here-and-now.