APPLICATION OF THE METHOD OF CANONICAL CORRELATIONS IN METEOROLOGY,
Abstract
Application of canonic correlations to problems in meteorology and specifically to forecasting characteristics of meteorological fields is discussed. In meteorological terms, the problem of canonic correlation can be formulated as follows: find a linear combination of predictors and of predicants that will have an optimally close statistical relationship among each other. Algorithms and covariational (or correlational) matrices are constructed. To improve the accuracy of the input matrices (or assumption matrices), two alternate methods are suggested: (1) covariation (or correction) functions are plotted as a preliminary measure, i.e., an autocovariation function for two variables and a cross-covariation function to link these variables; (2) random errors in measurement of the variables are calculated. A specimen calculation of a long-range forecast of the relationships between the average monthly surface water temperature in the North Atlantic in November and the average monthly air temperature over the European part of the SSSR in December is included.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 03, 1970
- Accession Number
- AD0713774
Entities
People
- L. S. Gandin
Organizations
- National Air and Space Intelligence Center