Converting GNSS Zenith Wet Delays to Precipitable Water: Investigating the Mean and Variability of the Water-Vapor-Weighted Mean Temperature (Tm) using Reanalysis Data

The high spatial and temporal variability of atmospheric water vapor makes it one of the more difficult parameters to monitor in the atmosphere. However, the use of Global Navigation Satellite System (GNSS) tropospheric-delay data can help to improve efforts in monitoring atmospheric water vapor, as the number and density of continuously operating GNSS networks around the globe increases with each passing year. The conversion of GNSS signals to precipitable water depends on the water-vapor-weighted mean column temperature T_m, which can be estimated or calculated through various means. Previous studies of T_m largely focus on the validation of empirical T_m models (i.e., GPT2w, GTm-III) against radiosonde-derived T_m or the GGOS Atmosphere model, but there has been little investigation into the large-scale spatial pattern of T_m and its variability. In this study, we calculate the global-gridded T_m using ERA-Interim reanalysis data, and compare our results with T_m from the GGOS Atmosphere model. We find that the spatial pattern, and diurnal and seasonal variability of our derived T_m are comparable to that of the GGOS T_m; however, the GGOS T_m has higher interannual variability than our T_m over mountainous regions. Comparison of our results against the GPT2w model indicates that the GPT2w model underestimates seasonal T_m variability by over 5% at high latitudes. We note that using a linear model to estimate T_m based on surface temperature T_s tends to overestimate the diurnal variability of T_m over land and underestimate it over the ocean. We find that diurnal T_m variability is affected by large-scale dynamics, such as the jet stream, which is known to modulate diurnal weather. More importantly, our results indicate that the variability of T_m is smaller than 10% within the tropics (<30º), and therefore T_m can be approximately treated as a constant value within this region. This shows that in the tropics the GNSS wet delays can be taken as direct proxies for precipitable water if the research interest lies in variability rather than absolute values. As future work, we hope to expand this to ERA5, which provides higher resolution of data in addition to improved representation of water vapor.