Evaluation of GNSS Radio Occultation Observations in Tropical Cyclone Environments

Tropical cyclones (TCs) can cause billions of dollars in damage annually. Long-range tropical cyclone intensity forecasts remain one of the biggest challenges in numerical weather prediction models and forecast errors have been attributed to a number of observational deficits. Remote sensing methods offer additional information about TCs, particularly in areas where in-situ observations are sparse. Vertical profiling of TC thermodynamics from conventional passive microwave and infrared sensors has historically been limited due to coarse vertical resolution and signal degradation from clouds and precipitation. High vertical resolution Global Navigation Satellite System (GNSS) radio occultation (RO) soundings are insensitive to clouds and precipitation and provides a unique opportunity to study TC thermodynamic vertical structure. However, the quality of the RO observations in the TC environment has not been critically examined.

In this study, GNSS RO profiles from COSMIC-1 (2006-2019) and COSMIC-2 (2020-2021) are analyzed in conjunction with colocated dropsondes (336 for COSMIC-1 and 529 for COSMIC-2) in the TC environment. Overall median refractivity difference between GNSS RO (COSMIC-1 & COSMIC-2) and corresponding colocated dropsondes is approximately 0.1% between 5 and 14 km, indicating that GNSS RO is highly capable of observing the TC middle- and upper-troposphere. Near-surface (2 km and below) overall fractional refractivity differences are -1.6% for COSMIC-1 colocations and -1.2% for COSMIC-2 colocations. Refractivity bias is mostly unaffected by colocation criteria and the effects of atmospheric ducting are examined. A negative correlation between RO refractivity biases and moisture content is found in the lower troposphere. Thorough understanding of GNSS RO observation quality in the TC environment will allow for the regular use of GNSS RO in TCs to fill in observational gaps and ultimately improve TC simulation and prediction through better model parameterizations and data assimilation.