This record has been edited as far as possible, missing data will be added when the version of record is issued.
A global uncertainty model for GNSS integrated water vapor derived from ERA5, GPT3 and co-located meteorological sensors
- Journal Article
The uncertainty assessment of Integrated Water Vapor (IWV) estimates remains a tough task in routine Global Navigation Satellite System (GNSS) IWV retrievals as the uncertainties of related station pressure (P_s) and weighted mean temperature (T_m) are usually difficult to be evaluated properly. Hence, we made the first attempt to establish a global model for GNSS IWV uncertainty (σ_IWV) by evaluating the uncertainties of the P_s and T_m derived from the state-of-the-art ERA5 reanalysis, the latest GPT3 model, and from meteorological sensors. Millions of worldwide in situ pressure and radiosonde observations were taken as references. The evaluation shows that the median Root Mean Square (RMS) errors of ERA5-derived P_s and T_m are 0.6 hPa and 1.0 K, respectively. Assuming that all the uncertainty contributors are smaller than their individual third quartiles, the σ_IWV of post analysis with ERA5 is not larger than 0.8 kg m-2, which meets the requirement for climate research (1.0 kg m-2). With the same assumption on the uncertainty contributors, the σ_IWV of near-real-time analysis with and without co-located meteorological sensors are lower than 1.1 and 3.2 kg m-2, respectively. Both are within the accuracy requirement for nowcasting (5 kg m-2). IWV uncertainty budget analysis shows that the uncertainty of GNSS zenith total delay generally contributes the most. However, the uncertainty of GPT3-derived P_s can also be the dominant contributor. In short, ERA5 is outstanding enough for most GNSS IWV retrieved with post analysis, whereas a co-located meteorological sensor should be installed for (near) real-time analysis. Show more
Journal / seriesRemote Sensing of Environment
Organisational unit03824 - Rothacher, Markus / Rothacher, Markus
MoreShow all metadata