METHOD FOR DEGREE/DAYS FORECASTING PROVES ITSELF

Lyle E. Brosche
Melbourne, Fla.

The long-range meteorological/engineering method for predicting heating degree/days can now be evaluated on the basis of nearly 40 years of experience. A comparison of predicted vs. actual degree/days over this period shows that it is a highly accurate method.

Long range oil-industry planners are well aware of the importance of heating and cooling degree day forecasts, but many are not aware that several different methods can be employed to give them forecasts or which one, or combination, their forecasts are coming from.

Before going into a review of the primary techniques long range forecasters use and before focusing on the meteorological/engineering method, it should be noted that there is a technical distinction between the words "forecasts" and "outlooks." The American Meteorological Society specifies that "outlook" be used in general practice. However, in this article the words are used interchangeably.

LONG RANGE

Oil and gas industry planners are generally concerned with long range outlooks. These cover the average or mean temperature expectancy for periods of a month ahead, seasonal, semi-annual, or on an annual basis. However, long range outlooks usually are not practically divided into daily predictions, nor on a monthly average basis beyond a month ahead.

Almost all long-range forecasters prepare their forecasting material using more than one technique. However, in most every case one primary method prevails or dominates. Of the many varieties of long-range methods, most would fall into one or a combination of the following six categories:

Synoptic
Statistical
Solar
Persistence
Physical-statistical
Meteorological/engineering.

Following are some comments on each:

The synoptic method of long range forecasting is an extension of the short range method of surface and upper air weather map extrapolation. The forecast is contingent on a prediction of large scale pressure or circulation patterns. Synoptic weather maps are often averaged over periods corresponding to the period of the forecast. Analogue weather charts are also often used, but while surface charts are comparable, the upper air patterns (which control storm movement) seldom are analogous in all respects to surface conditions at a given time.

The meteorological statistical approach is basically the use of historical weather data. Comprehensive statistical techniques depend essentially on the statistical determination of just how the variations of past weather patterns are related to the weather that follows. Such determination must include not just surface conditions but those at high limits of the atmosphere. The problems are extremely complex, and conclusions are greatly empirical.

The solar approach involves a study of the variability of emissions from the sun, primarily "sunspots." This is somewhat an extension of the statistical approach inasmuch as it mainly deals with the general 11-year cycle, or peaking of sunspots. Complete evaluation, however, of the influence of solar emissions on the earth's atmosphere is presently inconclusive, as is the regular occurrence and intensity of sunspot cycles.

The past history of such cycles remains in question as to the reality of effects on weather.

Persistence in long range weather forecasting is merely, in it's simplest form, the continuation of an effect with the cause likely unknown, and which may or may not be continuing. In applying the concept of persistence, one looks to the very recent past for guidance and hopes that atmospheric behavior continues similarly into the period of his forecast. This practice is constantly exposed to jeopardy in that the causative factor, or factors, may be altered so as to a produce radically different temperature trend. This may occur at any time, without advance warning.

The persistence forecaster then has only one alternative: updating the outlook to agree with the new temperature trend, with the likelihood of coming too late for the energy planner to make adjustments.

The physical-statistical method is one which considers causes of weather changes to be terrestrial, but not atmospherical. Such factors as changes in ocean temperature and the occurrence of ash ejection out of volcanoes into the atmosphere have been given the most attention.

All of the preceding types of long range forecasting methods relate to efforts in predicting temperature trends over broad areas- not city by city.

The meteorological/engineering method for long range prediction was developed by the author after intensive research. Basically the technique involves a unique arrangement and processing of meteorological data.

The method removes much of the confusion in handling large numbers of variables in the atmosphere that make it seem chaotic. Actually when data are used effectively in conjunction with the engineering method, the broad circulation appears well regulated and orderly. The method fundamentally constructs, systematically, an arrangement of the temperature element so as to provide an important "guide" utilized by the meteorologist to geometrically project future seasonal trends city by city, or more specifically stated by official U.S. National Weather Service locations.

This, combined with a correlation process related to the atmosphere, produces "pulse" designs acting as harbingers of future trends that are useful in confirming the original determination.

RELIABILITY

In order to evaluate and express an opinion as to the reliability of long range outlooks, it is necessary to define the terms in which the reliability is measured. Without a professional meteorologically prepared outlook, let us assume an energy company would use the climatological normals as a basis in long range planning for the ensuing heating season. Because actual temperature seldom falls sufficiently close to normal (in fact, normals are calculated from a great many extremes), the energy company must apply an additional percentage of heating degree/days as an insurance factor against the possibility of experiencing a colder than normal heating season. This insurance factor must necessarily be quite high, and costly, to protect against severe winters that have actually occurred during the periods in which the normal has been calculated.

Therefore, reliability and usefulness in an outlook prepared by a professional meteorologist would be indicated by:

The skill in correctly predicting temperature trends that average warmer than normal or colder than normal.
An accurate warmer than normal prediction would be extremely valuable in reducing, or eliminating, the costly insurance factor.
The skill in predicting quantity of heating degree/days expected for the heating season.
This would further define winter's intensity and aid the planner in refining heating requirements.
The ability to "pinpoint" the predictions on a city-by-city basis, rather than by regions.
Forecasts over large geographical areas become too generalized, especially when planning for heavily populated cities.

Fig. 1 shows the long range forecast performance at Chicago for the October through March periods, using the meteorological/engineering methods.

The outlook and verification data are for O'Hare Airport, the site used by the National Weather Service as the observation point. Outlooks were issued as early as July, and not later than September preceding the heating season. It is significant to note:

Since 1950-51 the outlooks have never been more than 6% off in total variation from actual.
On the average they have varied only about 3.5% from actual.
Extreme weather such as occurred during very warm seasons (1953-54 and 1975-76) and very cold seasons (1976-77 and 1981-82) represent a large difference in heating degree/days, and were followed closely by the outlook.
The fallacy of utilizing the climatological normal" as a planning instrument is highlighted in that the actual seldom falls close to normal, but conversely normal frequently misses the target by a very costly number of heating degree/days.

Table 1 illustrates the accuracy of the engineering method for heating degree days outlooks on a national basis.

This outlook, which was prepared for Chevron U.S.A., shows a colder than normal condition east of the Rocky Mountains; a warmer than normal condition from the Rockies westward to the northwest.

Where the actual varied significantly from normal, practically all of the author's city outlooks were more accurate than using normal.