A Method to Deconvolve Errors in GPS RO-Derived Water Vapor Histograms

Abstract

ater vapor is an important constituent in the earth?s atmosphere that must be understood to predict weather and climate. Water vapor is challenging to measure, and observations of water vapor must be as independent of weather and climate models as possible in order to assess and improve those models. When combined with independent atmospheric temperature estimates, GPS radio occultation (RO) refractivity profiles yield unique, all-weather, high-vertical-resolution, globally distributed profiles of tropospheric water vapor whose vertical extent is maximum at low latitudes. A method for retrieving water vapor, known as the direct method, combines GPS RO-derived refractivity with temperatures from global weather analyses. Unlike variational methods, the direct method does not use the water vapor estimates from analyses because of their potential systematic errors. While utilization of the direct method has been limited because it produces negative values, these unphysical negative values reveal unique information about errors. A deconvolution technique is presented here that uses the negative values to estimate and remove random errors in histograms of water vapor derived via the direct method. Results indicate direct method specific humidity errors at low latitudes range from 0.4 g kg?1 at 725 hPa to 0.14 g kg?1 near 350 hPa, which can be removed via deconvolution. The deconvolved histograms extend to about the 240-K level in the troposphere, corresponding to 10 km at low latitudes. Unlike the limited information of low-order statistical moments like means and variances, histograms capture the full range of variability, which opens a new window into water vapor behavior and increases understanding of processes at work in the hydrological cycle.