Remote sensing tools represent an attractive proposition for measuring wind speeds over the oceans because in principle they also offer a mechanism for determining the spatial variability of flow. Here we present the continuation of research focussed on the uncertainties and biases currently present in these data and quantify the number of independent observations (scenes) required to characterize various parameters of the probability distribution of wind speeds. We derive theoretical and empirical estimates of the critical number of observations (wind speeds derived from remotely sensed scenes) required to obtain probability distribution parameters with an uncertainty of ±10% and a confidence level of 90% under the assumption of independent samples, and find approximately 250 independent observations are required to fit the Weibull distribution parameters. We present an evaluation of Weibull fitting methods and determine the fitting method based on the first and third moments exhibits the 'best' performance for pure Weibull distributions. We further examine the generalizability of parameter uncertainty bounds presented in Barthelmie and Pryor (2003) for distribution parameter estimates from sparse data sets and find them to be robust, and hence generally applicable to remotely sensed wind speed data series.
展开▼