“…However, RO data possess retrieval errors (occurring in the course of imperfect atmospheric data processing) as well as measurement errors (those in the primary observable). A number of studies analyzed the error sources and propagation of the errors through the retrieval process (e.g., Kursinski et al, 1997;Feng and Herman, 1999;Syndergaard, 1999;Rieder and Kirchengast, 2001;Hajj et al, 2002;Kuo et al, 2004;. Some of the error sources are particularly hard to tackle including ionospheric residuals (e.g, Syndergaard, 2000), the influence of a priori error on the statistical optimization of bending angle (e.g., Wee and Kuo, 2014), horizontal inhomogeneity in the atmosphere (e.g., Poli and Joiner, 2004), abnormal propagations of radio waves (e.g., Sokolovskiy, 2003), strong multipath effects in the lower troposphere (e.g., Gorbunov et al, 2006), and the ambiguity in separating moisture and dry air density from the bending angle or the refractivity (e.g., Kursinski et al, 1997).…”