Abstract. Global Navigation Satellite System (GNSS) radio occultation (RO) data enable the retrieval of near-vertical profiles of atmospheric parameters like bending angle, refractivity, pressure, and temperature. The retrieval step from bending angle to refractivity, however, involves an Abel integral with an upper limit of infinity. RO data are practically limited to altitudes below about 80 km and the observed bending angle profiles show decreasing signal-to-noise ratio with increasing altitude. Some kind of high-altitude background data are therefore needed in order to perform this retrieval step (this approach is known as high-altitude initialization). Any bias in the background data will affect all RO data products beyond bending angle. A reduction of the influence of the background is therefore desirable -in particular for climate applications.Recently a new approach for the production of GNSS radio occultation climatologies has been proposed. The idea is to perform the averaging of individual profiles in bending angle space and then propagate the mean bending angle profiles through the Abel transform. Climatological products of refractivity, density, pressure, and temperature are directly retrieved from the mean bending angles.The averaging of a large number of profiles suppresses noise in the data, enabling observed bending angle data to be used up to 80 km without the need for a priori information. Some background information for the Abel integral is still necessary above 80 km.This work is a follow-up study, having the focus on the comparison of the average profile inversion climatologies (API) from the two processing centers WEGC and DMI, which study monthly COSMIC (Constellation Observing System for Meteorology, Ionosphere, and Climate) data from January to March 2011. The impact of different backgrounds above 80 km is tested, and different implementations of the Abel integral are investigated. Results are compared for the climatological products with ECMWF analyses, MIPAS, and SABER data.It is shown that different implementations of the Abel integral have little impact on the API climatologies. On the other hand, different extrapolations of the bending angle profile above 80 km play a key role in the resulting monthly mean refractivities above 35 km in altitude. Below that respective altitude the API climatologies show a good agreement between the two processing centers WEGC and DMI. Due to the downward propagation within the retrieval, effects of the high-altitude initialization lead to differences in dry-temperature climatologies down to 20 km in altitude.When applying an exponential extrapolation to the bending angles above 80 km at both centers, the dry-temperature climatologies agree among WEGC, DMI, ECMWF analysis, and MIPAS up to 35 km in altitude within ± 0.5 K and up to 40 km in altitude within ±1 K. We conclude that the API retrieval is a valid approach up to the lower stratosphere. It is a computationally efficient alternative method for producing dry atmospheric RO climatologies.