CLIMATIC NORMALS AS PREDICTORS. PART 2. EXTENSION
Abstract
Temperature and precipitation for a single month or year, picked at random, at each of seven United States stations are found to be estimated as closely, on the average, by the mean of the preceding 10 to 40 years as by a 30- year 'normal.' The median value for the preceding 15 or so years may be an even slightly better estimator. Graphs show the mean square and mean absolute differences between k-year means and the next observation for k = 1 (1) 50, and the mean absolute differences between k-year medians and the next observation for k = 1 (2) 49, for Dodge City, Vicksburg, Memphis, Cairo, Madison, Pittsburgh, and Lynchburg. Comparison of these graphs with corresponding graphs based on random normal numbers, biased in various ways, suggests that many climatic records contain progressive changes in mean or in variance, or both. The number of antecedent years (k*) for which the mean or median is closest to the next year's observation varies erratically from month to month, but tends to be the same at nearby stations for a given month.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1968
- Accession Number
- AD0686163
Entities
People
- Arnold Court