Evaluation of surface wind and flux analysis techniques using conventional data in marine cyclones

Abstract

Abstract Data from seven storms from the Storm Transfer and Response Experiment in November 1980 have been used to evaluate the relative accuracy of surface wind and flux fields based on two analysis procedures. Two essentially independent techniques were used; objective analysis which is based on composite data taken at several synoptic intervals to enhance the number of observations and processing carefully hand-analyzed (subjective) surface pressure analyses into wind and flux fields using a planetary boundary layer (PBL) similarity model. Scale and accuracy limits imposed by instrument accuracy, sampling error, and gridding and analysis procedures are evaluated for each of these techniques by comparison with independent data and with each other. Wind field differences between the objective composite analysis and the PBL model predictions are found to be comparable to the measurement-related uncertainty in the observations. Unresolved variability in the 10–100 km scale of the dynamic and thermodynamic variables produces the main source of error in both the objective and model wind fields. Additional wind field differences are contributed by PBL and gradient wind assumptions used in the PBL model. Wind differences between either of the two analyses and individual observations are about ±3 m s−1 and ±30° in the mean, and can be greater than ±5 m s−1 and ±50° for small regions. Comparable differences are found between the two wind field analyses. The wind and thermodynamic field differences combine to produce substantial differences in the derived fields. Mean differences of ±19W m−2 and ±41 W m−2 for the fluxes of sensible and latent heat, respectively, represent differences of about 50% of the mean fluxes, with local differences as much as double or triple the magnitude of these means. Fronts are equally well represented by the two analyses, but values of divergence and curl of surface stress may differ by a factor of 2 or more in regions of fronts. These local differences in the derived fields result primarily from the large wind field differences in these inadequately resolved regions.