Statistical Derivation of Wind Speeds From CYGNSS Data
Maria Paola Clarizia, Christopher S. Ruf
Abstract
In this article, a statistical methodology to estimate wind speed from CYGNSS observables is proposed and implemented. The approach uses the cumulative distribution function (cdf) of the observable and of the ground-truth reference winds. It depends only on the statistical distributions of the CYGNSS data and the wind speed, and therefore, is simpler to implement than alternative approaches requiring coincident matchups between the data and the ground truth. This cdf matching method produces retrieved winds with a probability density function that is very close to that of the ground-truth winds. When compared to the current CYGNSS baseline winds for fully developed seas, the cdf matching winds show better behavior and agreement with reference wind speeds over the low to medium wind speed range, which constitutes the majority of the wind population that drives the statistics used by the algorithm. The performance is robust with respect to measurement geometry and transmitter and receiver hardware parameters, with the exception of a dependence of the error on the GPS satellite identifier (ID), probably due to uncorrected variations in GPS equivalent isotropically radiated power (EIRP). Validation using modeled winds and winds measured by other satellites reveals that CYGNSS winds behave in a very similar manner as the winds modeled by the Global Data Assimilation System (GDAS).