Inversion, error analysis, and validation of GPS/MET occultation data

The global positioning system meteorology (GPS/MET) experiment was the ®rst practical demonstration of global navigation satellite system (GNSS)-based active limb sounding employing the radio occultation technique. This method measures, as principal observable and with millimetric accuracy, the excess phase path (relative to propagation in vacuum) of GNSS-transmitted radio waves caused by refraction during passage through the Earth's neutral atmosphere and ionosphere in limb geometry. It shows great potential utility for weather and climate system studies in providing an unique combination of global coverage, high vertical resolution and accuracy, long-term stability, and all-weather capability. We ®rst describe our GPS/MET data processing scheme from excess phases via bending angles to the neutral atmospheric parameters refractivity, density, pressure and temperature. Special emphasis is given to ionospheric correction methodology and the inversion of bending angles to refractivities, where we introduce a matrix inversion technique (instead of the usual integral inversion). The matrix technique is shown to lead to identical results as integral inversion but is more directly extendable to inversion by optimal estimation. The quality of GPS/ MET-derived pro®les is analyzed with an error estimation analysis employing a Monte Carlo technique. We consider statistical errors together with systematic errors due to upper-boundary initialization of the retrieval by a priori bending angles. Perfect initialization and properly smoothed statistical errors allow for better than 1 K temperature retrieval accuracy up to the stratopause. No initialization and statistical errors yield better than 1 K accuracy up to 30 km but less than 3 K accuracy above 40 km. Given imperfect initialization, biases >2 K propagate down to below 30 km height in unfavorable realistic cases. Furthermore, results of a statistical validation of GPS/MET pro®les through comparison with atmospheric analyses of the European Centre for Medium-range Weather Forecasts (ECMWF) are presented. The comparisons indicate the high utility of the occultation data in that very good agreement of upper troposphere/lower stratosphere temperature (better than 1.5 K rms, <0.5 K bias) is found for a region (Europe+USA) where the ECMWF analyses are known to be good, but poorer agreement for a region (Southern Paci®c) where the analyses are known to be degraded.