Setting Upper Bounds for Bias in Radio Occultation Climate Data Records

Radio occultation (RO) using Global Navigation Satellite Systems (GNSS) has the potential to provide measurements of atmospheric refractivity that meet stringent accuracy requirements for climate studies. Fundamental observables are atmospheric phase delay and the resulting bending angle, measured with systems that are tied to atomic clock standards, providing long term stability and absence of drift. Atmospheric refractivity (proportional to density) is a retrieved quantity. Detailed analyses of the physical GNSS-RO retrieval have yielded significant insights into potential sources of retrieval bias. Some biases vary with geophysical conditions, such as the state of the ionosphere or vertical structure of the stratosphere. Setting upper bounds on retrieval bias must be robust to varying geophysical conditions. This is best achieved by monitoring signatures in individual occultations to characterize how local conditions may be affecting the retrieval bias. Examples will be presented of residual signatures due to large and small-scale ionospheric structure. We will describe data analysis methods to quantify potential biases caused by these conditions. Another source of retrieval error in the lower stratosphere is an incorrect estimate of the profile upper boundary, which is needed to initialize the Abel inversion method. We will apply a recent result demonstrating that averaging RO bending angles prior to the retrieval step reduces retrieval noise for climatological averages. Therefore, retrieving refractivity from climate average bending angle profiles provides a robust means of estimating and reducing upper altitude Abel inversion errors. Case studies of our approach to creating the climate data set will be presented using data from the COSMIC/FORMOSAT-3 satellites.