Global Navigation Satellite System (GNSS) Radio Occultation (RO) is a highly valuable remote sensing technique for probing the Earth’s atmosphere, due to its global coverage, high accuracy, long-term stability, and essentially all-weather capability. In order to ensure the highest quality of essential climate variables (ECVs), derived from GNSS signal tracking by RO satellites in low Earth orbit (LEO), the orbit positions and velocities of the GNSS transmitter and LEO receiver satellites need to be determined with high and proven accuracy and reliability. Wegener Center’s new Reference Occultation Processing System (rOPS) hence aims to integrate uncertainty estimation at all stages of the processing. Here we present a novel setup for precise orbit determination (POD) within the rOPS, which routinely and in parallel performs the LEO POD with the two independent software packages Bernese GNSS software (v5.2) and NAPEOS (v3.3.1), employing two different GNSS orbit data products. This POD setup enables mutual consistency checks of the calculated orbit solutions and is used for position and velocity uncertainty estimation, including estimated systematic and random uncertainties. For LEOs enabling laser tracking we involve position uncertainty estimates from satellite laser ranging. Furthermore, we intercompare the LEO orbit solutions with solutions from other leading orbit processing centers for cross-validation. We carefully analyze multi-month, multi-satellite POD result statistics and find a strong overall consistency of estimates within LEO orbit uncertainty target specifications of 5 cm in position and 0.05 mm/s in velocity for the CHAMP, GRACE-A, and Metop-A/B missions. In 92% of the days investigated over two representative 3-month periods (July to September in 2008 and 2013) these POD uncertainty targets, which enable highly accurate climate-quality RO processing, are satisfied. The moderately higher uncertainty estimates found for the remaining 8% of days (∼5–15 cm) result in increased uncertainties of RO-retrieved ECVs. This allows identification of RO profiles of somewhat reduced quality, a potential benefit for adequate further use in climate monitoring and research.

Earth observation from space provides a highly valuable basis for atmospheric and climate science, in particular also through climate benchmark data from suitable remote sensing techniques. Measurements by global navigation satellite system (GNSS) radio occultation (RO) qualify to produce such benchmark data records as they globally provide accurate and long-term stable datasets for essential climate variables (ECVs) such as temperature. This requires a rigorous processing of the raw RO measurements to ECVs, with narrow uncertainties. In order to fully exploit this potential, Wegener Center's Reference Occultation Processing System (rOPS) Level 1a (L1a) processing subsystem includes uncertainty estimation in both precise orbit determination (POD) and excess-phase profile derivation. Here we introduce the new rOPS L1a excess-phase processing, the first step in the RO profiles retrieval down to atmospheric profiles, which extracts the atmospheric excess phase from raw SI-traceable RO measurements. This excess-phase processing, for itself algorithmically concise, includes integrated quality control and uncertainty estimation, requiring a complex framework of various subsystems that we first introduce before describing the implementation of the core algorithms. The quality control and uncertainty estimation, computed per RO event, are supported by reliable forward-modeled excess-phase profiles based on the POD orbit arcs and collocated short-range forecast profiles of the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA5). The quality control removes or alternatively flags excess-phase profiles of insufficient or degraded quality. The uncertainty estimation accounts both for relevant random- and systematic-uncertainty components, and the resulting (total) uncertainty profiles serve as a starting point for the subsequent uncertainty propagation through the retrieval processing chain down to the atmospheric ECV profiles. We also evaluated the quality and reliability of the resulting excess-phase profiles based on Metop-A/B/C (Meteorological Operational) RO datasets for three 3-month periods in 2008, 2013, and 2020 by way of a sensitivity analysis for three representative atmospheric layers (tropo-, strato-, mesosphere), investigating consistency with ERA5-derived profiles, influences of different orbit and clock inputs, and consistency across the different Metop satellites. These consistencies range from centimeter to submillimeter levels, indicating that the new processing can provide highly accurate and robust excess-phase profiles. Furthermore, cross-evaluation and intercomparison with excess-phase data from the established data providers EUMETSAT (European Organisation for the Exploitation of Meteorological Satellites) and UCAR (University Corporation for Atmospheric Research) revealed subtle discrepancies but overall very close agreement, with larger differences compared to UCAR in the boundary layer. The new rOPS L1a processing can hence be considered capable of producing reliable long-term data records including uncertainty estimation for the benefit of climate applications.

Radio occultation is a well‐established remote sensing method that provides reliable estimates of atmospheric profiles of diverse variables, including temperature and pressure. However, as with all indirect methods, radio occultation has some inherent systematic and random error effects, which lead to observational uncertainties. While propagation of uncertainties along the processing chain for individual radio occultation profiles was described in recent studies, this uncertainty information has not yet been carried forward to climatological fields. We close this gap and present an uncertainty propagation procedure that provides uncertainty estimates for aggregated means for climate applications. Estimated random uncertainties, basic and apparent systematic uncertainties and sampling uncertainties (due to the discrete sampling by profiles) are propagated through the aggregation process, resulting in uncertainty estimates for gridded fields. We demonstrate the new procedure for two test months and representative variables, inspecting monthly mean profiles for refractivity, dry temperature and physical temperature measurements. Results show that estimated random uncertainties and residual sampling uncertainties (after sampling bias correction) have similar magnitudes, both decreasing with increasing spatial aggregation sizes and corresponding increasing number of aggregated observations. At small aggregation they are the main contributors to uncertainty in refractivity, and important contributors to uncertainty of temperature. Systematic uncertainty, whose magnitude is independent of the number of profiles, is for refractivity the main source of uncertainty for larger aggregation sizes, and for pressure and dry temperature at all commonly used aggregation sizes. All uncertainty components exhibit pronounced spatial variation over the globe, with polar regions showing the greatest uncertainty.

Earth observation from space provides a highly valuable basis for atmospheric and climate science, in particular also through climate benchmark data from suitable remote sensing techniques. Measurements by Global Navigation Satellite System (GNSS) radio occultation (RO) qualify to produce such benchmark data records as they globally provide accurate and long-term stable datasets for essential climate variables (ECVs) such as temperature. This requires a rigorous processing from the raw RO measurements to ECVs, with narrow uncertainties. In order to fully exploit this potential, Wegener Center’s Reference Occultation Processing System (rOPS) Level 1a (L1a) processing subsystem includes uncertainty estimation in both precise orbit determination (POD) and excess phase profile derivation. Here we introduce the new rOPS L1a excess phase processing, the first step in the RO profiles retrieval down to atmospheric profiles, which extracts the atmospheric excess phase from raw SI-traceable RO measurements. This excess phase processing, for itself algorithmically concise, includes integrated quality control and uncertainty estimation, which requires a complex framework of various subsystems that we first introduce before describing the implementation of the core algorithms. The quality control and uncertainty estimation, computed per RO event, are supported by reliable forward-modeled excess phase profiles based on the POD orbit arcs and collocated short-range forecast profiles of the European Reanalysis ERA5. The quality control removes or alternatively flags excess phase profiles of insufficient or degraded quality. The uncertainty estimation accounts both for relevant random and systematic uncertainty components and the resulting (total) uncertainty profiles serve as starting point for the subsequent uncertainty propagation through the retrieval processing chain down to the atmospheric ECV profiles. We also evaluated the quality and reliability of the resulting excess phase profiles based on Metop-A/B/C RO datasets for three 3-month periods in 2008, 2013, and 2020 by way of a sensitivity analysis for three representative atmospheric layers (tropo-, strato-, mesosphere), investigating consistency with ERA5-derived profiles, influences of different orbit and clock inputs and consistency across the different Metop satellites. These consistencies range from centimeter to submillimeter levels, indicating that the new processing can provide highly accurate and robust excess phase profiles. Furthermore, cross-validation and inter-comparison with excess phase data from the established data providers EUMETSAT and UCAR revealed subtle discrepancies but overall very close agreement, with larger differences against UCAR in the boundary layer. The new rOPS L1a processing can hence be considered capable to produce reliable long-term data records including uncertainty estimation for the benefit of climate applications.

GNSS radio occultation (RO) provides high-quality atmospheric profiles of variables such as temperature and pressure. Recent efforts have succeeded in propagating the related systematic and random error effects from the raw measurements to the resulting profiles, attaching a measure of observational uncertainty to each one. In this work we build upon these profile-level uncertainty estimates and propagate them to aggregated mean fields for climate applications. In this context, sampling uncertainties also need to be considered. This approach is applied to the GNSS RO time series of refractivity, dry temperature, and physical temperature. The results show that random and residual sampling uncertainties decrease with increasing aggregation size and are comparable in magnitude. They dominate refractivity uncertainty at small aggregation scales and contribute substantially to temperature uncertainty. Systematic uncertainty is the main source of uncertainty for refractivity at larger aggregation scales, as well as for pressure and dry temperature at commonly used aggregation sizes. Uncertainties exhibit strong spatial variability, with the largest values occurring in polar regions. There are also substantial, mission-dependent variations within the time series.