Estimation of mixing ratio profiles over maritime areas by linear regression methods.

Abstract

An investigation is made into the feasibility of using mixing-ratio predictions employing multiple linear regression techniques applied to saturated mixing ratio profiles at specified levels. Three such sets of coefficients are determined from Ocean Station Vessel sounding data recorded for the months of January (1967-70, inclusive), and tested for effectiveness to determine a "best" prediction scheme. The resultant "best" set of coefficients is then compared to a previously designed regression-prediction scheme due to Weinreb and Crosby (1973) based on a selection of island and coastal data divided equally in time and space. Tests were performed to determine possible sources of error affecting the water vapor specifications made by the multiple regression procedure.