THE APPLICATION OF A TECHNIQUE FOR VECTOR CORRELATION TO PROBLEMS IN METEOROLOGY AND OCEANOGRAPHY

In a recent study, Crosby et al. proposed a definition for vector corr elation that has not been commonly used in meteorology or oceanography . This definition has both a firm theoretical basis and a rather compl ete set of desirable statistical properties. In this study, the author s apply the definition to practical problems arising in meteorology an d oceanography. In the first of two case studies, vector correlations were calculated between subsurface currents for five locations along t he southeastern shore of Lake Erie. Vector correlations for one sample size were calculated for all current meter combinations, first includ ing the seiche frequency and then with the seiche frequency removed. R emoval of the seiche frequency, which was easily detected in the curre nt spectra, had only a small effect on the vector correlations. Under reasonable assumptions, the vector correlations were in most cases sta tistically significant and revealed considerable fine structure in the vector correlation sequences. In some cases, major variations in vect or correlation coincided with changes in surface wind. The vector corr elations for the various current meter combinations decreased rapidly with increasing spatial separation. For one current meter combination, canonical correlations were also calculated; the first canonical corr elation tended to retain the underlying trend, whereas the second cano nical correlation retained the peaks in the vector correlations. In th e second case study, vector correlations were calculated between marin e surface winds derived from the National Meteorological Center's Glob al Data Assimilation System and observed winds acquired from the netwo rk of National Data Buoy Center buoys that are located off the contine ntal United States and in the Gulf of Alaska. Results of this comparis on indicated that 1) there was a significant decrease in correlation b etween the predicted and observed winds with increasing forecast inter val out to 72 h, 2) the technique provides a sensitive indicator for d etecting bad buoy reports, and 3)there was no obvious seasonal cycle i n the monthly vector correlations for the period of observation.