Some Notes on Forecasting of Winds Aloft by Statistical Methods

Abstract

Winds aloft statistics from Eniwetok were used to derive a linear regression forecast technique that verified satisfactorily in later application. Other studies showed that these principles were applicable throughout the tropical Pacific Ocean area. Time-lag correlation coefficients were found to decay exponentially with increased time except in the first 6 hr, when decay was more rapid. This has been interpreted as the effect of measurement error and small-scale turbulence. A method was derived for removal of these random errors and, on application, time-lag correlation coefficients for each altitude were then found to decay at constant exponential rates, to at least 48 hr. This indicates that no improvement in regression forecasting can be achieved by use of multiple-term regressions from sequential data. Also, regression forecast verifications can be significantly improved by 1) removing the instrument error by hardware techniques and 2) smoothing out small-scale turbulence by multiple observation. This general technique is also shown to be useful in forecasting for other latitudes but comparison with other forecast methods for middle and high latitudes has not been made.