PERFORMANCE IN LONG-RANGE WEATHER FORECASTING

View More View Less
  • 1 Clark University, Worcester, Mass
© Get Permissions
Full access

Abstract

SYNOPSIS

Long-range forecasts are so much desired that any number of unqualified persons issue them without regard for criteria of performance. Among what might be called “fake” forecasts are the almanac, astrological, pseudosolar, and “mathematical” sorts, many of them calamity howls. The “prediction” of climatic normals, forecasts from phenomena on certain dates and from the behavior or aspects of certain animals or plants also belong in this category.

To be of value a forecast must be specific, limited as to place and time, and it must have a probability of more than chance verification. Furthermore, the economic consequences of failures, both in the long run and in a small sequence of years, must be reckoned. A forecast that will not hit the mark four times out of five, or at least once out of every three in succession, can not be of much value, though some claim that a forecast verified only three times out of five would be useful. The uncertainties of meteorological relationships on which any long-range forecasts can now be based are generally too great to permit reputable meteorologists to forecast on expectations of less than 75 or 80 per cent verification. A critical study of the methods now used in the attempts at scientific long-range weather forecasting and an evaluation of their relative merits for different parts of the world is much needed.

Abstract

SYNOPSIS

Long-range forecasts are so much desired that any number of unqualified persons issue them without regard for criteria of performance. Among what might be called “fake” forecasts are the almanac, astrological, pseudosolar, and “mathematical” sorts, many of them calamity howls. The “prediction” of climatic normals, forecasts from phenomena on certain dates and from the behavior or aspects of certain animals or plants also belong in this category.

To be of value a forecast must be specific, limited as to place and time, and it must have a probability of more than chance verification. Furthermore, the economic consequences of failures, both in the long run and in a small sequence of years, must be reckoned. A forecast that will not hit the mark four times out of five, or at least once out of every three in succession, can not be of much value, though some claim that a forecast verified only three times out of five would be useful. The uncertainties of meteorological relationships on which any long-range forecasts can now be based are generally too great to permit reputable meteorologists to forecast on expectations of less than 75 or 80 per cent verification. A critical study of the methods now used in the attempts at scientific long-range weather forecasting and an evaluation of their relative merits for different parts of the world is much needed.

Save