The Potential of Precipitable Water Vapour Measurements from Global Navigation Satellite Systems in Luxembourg

The primary objectives of this research were to investigate the potential of precipitable water vapour (PWV) estimates derived from Global Navigation Satellite Systems (GNSS) measurements, firstly, for short-term weather forecasting based on numerical weather prediction (NWP) in Luxembourg and its surroundings and, secondly, for monitoring climate on regional and global scales.
The suitability of real-time (RT) zenith total delay (ZTD) estimates obtained from three different precise point positioning (PPP) software packages was assessed by comparing them with the state-of-the-art product from the International GNSS Service (namely the IGS Final troposphere product) as well as collocated radiosonde (RS) observations. It was found that the RT-PPP ZTD estimates from two of the three software packages meet the threshold requirements for NWP nowcasting. The biases between the RT-PPP ZTD and the reference ZTD were found to be stable over time for all the RT-PPP ZTD solutions. A millimetre-level impact on the RT-PPP ZTD estimates was also observed when integer ambiguities were resolved.
The impact of assimilating GNSS-derived near real-time (NRT) ZTD in the Applications of Research to Operations at Mesoscale (AROME) NWP model using a three-dimensional, variational (3D-VAR) assimilation scheme on the quality of weather forecasts for Luxembourg was studied. It was found that the assimilation of GNSS-derived ZTD systematically improves the atmospheric humidity short-range forecasts in comparison to other water vapour observing systems (radio soundings, satellite radiances, surface networks). Examination of several case studies revealed the ability of the ZTD observations to modify the intensity and location of predicted precipitation in accordance with previous studies. The addition of ZTD from the dense GNSS network in Wallonie (Belgium) was also found to be beneficial by improving the prediction of rainfall patterns in Luxembourg.
The 2D maps of IWV obtained from the hourly NRT system were compared with cloud distribution and precipitation maps from satellite and weather radar data, respectively, and a good agreement in the location of the front system was found. A rise in IWV was recorded during a precipitation event in Luxembourg and it was shown that by observing the IWV change over the ground-based GNSS stations in Luxembourg in NRT, it is possible to determine the speed and direction of the passing fronts and hence storms can also be tracked.
A 5-year long global reprocessed GNSS data set containing over 400 ground-based GNSS stations and based on the double differencing strategy has been used to validate the ZTD estimates obtained from the climate reanalysis model of the European Centre for Medium-range Weather Forecasts (ECMWF) namely the ECMWF ReAnalysis-Interim (ERA-Interim) in different climate zones. It was found that the correlation coefficient between the GNSS-derived ZTD observations and the ZTD modeled by ERA-Interim ranges from 0.87 to 1.00. Higher correlation coefficients were found for the stations belonging to the climate zones with lower amount of water vapour. Furthermore, it was found that the mean, SDev and RMS of the differences depends on periodicity in the residuals, altitude of the stations in a particular zone as well as the topographic variation in the zone.
Monthly and seasonal means of GNSS-derived ZTD (ZTDgnss) were computed using a global ZTDgnss dataset consisting of 19-years of data and over 400 stations to study the climate variability in different climate zones. In terms of seasonal means, it was found that the climate zones in the northern hemisphere have ZTD maxima in Boreal Summer (June-July-August) whereas those in the southern hemisphere have ZTD maxima in Austral Summer (December-January-February). Monthly and seasonal variability in ZTDgnss was also studied for the locations of 6 ground-based GNSS (SPSLux) stations in Luxembourg. It was found that all the 6 SPSLux stations experience the same monthly and seasonal variability of ZTDgnss. In terms of monthly variation, it was found that the maxima in ZTDgnss occurs around the month of July for all the 6 SPSLux stations whereas in terms of seasonal variation, the location of maxima was found to be in Summer (June-July-August).
The suitability of the ZTD derived using precise point positioning (PPP) strategy for climate monitoring applications was studied through its comparison with the ZTD estimates derived using double differenced positioning (DDP) using a global network of 114 stations and duration of 1 year. The mean differences between the two were found to be ranging from -3.35 to 2.37 mm over different climate zones. Furthermore, correlation coefficients ranging from 0.90 and 1.00 were found between the ZTD obtained using the two processing strategies. It was found that use of higher elevation cut-off angles and tropospheric mapping functions based on NWP improves the agreement between the PPP and DDP solutions.