Estimating the refractivity bias of FORMOSAT-7/COSMIC-2 Global Navigation Satellite System (GNSS) radio occultation in the deep troposphere

FORMOSAT-7/COSMIC-2 radio occultation (RO) measurements show promise for observing the deep troposphere and for providing critical information on the Earth's planetary boundary layer (PBL). However, refractivity retrieved in the low troposphere can have severe biases under certain thermodynamic conditions. This research examines the characteristics of the deep tropospheric biases and presents methods for estimating the region-dependent refractivity bias using statistical regression models. The results show that the biases have characteristics that vary over land and oceans. With substantial correlation between local spectral width (LSW) and bias, the LSW-based bias estimation model can explain the general pattern of the refractivity bias but with deficiencies in measuring the bias in the ducting regions and in certain areas over land. The estimation model involving the relationship with temperature and specific humidity (TQ) can capture the large biases associated with ducting. Finally, a minimum variance estimation that combines the LSW and TQ provides the most accurate estimation of the refractivity bias.