The accuracy of surface meteorological measurements is vital to derive accurate Global Positioning System (GPS) Precipitable Water Vapour (PWV) data. However, in absence of surface meteorological data, data from Numerical Weather Prediction Models (NWP) are used. The accuracy of these models varies depending on the model, region, season and other climatic conditions. In this study, GPS data for derivation of PWV is collected from 13 geodetic permanent stations for the years 2013 to 2016. Five out of 13 GPS stations are equipped with meteorological sensors. The interpolated European Centre for Medium-Range Weather Forecasts (ECMWF) 5th Re-Analysis (ERA5) dataset at these locations is first validated using the meteorological data from these sensors. The assessment shows that the average root mean square errors (RMSE) of surface pressure and temperature values are about 0.72 hPa and 1.04 K respectively. PWV determined also requires information on the weighted mean temperature (Tm); the Tm at the 13 GPS stations was also evaluated using three methods: the temperature and humidity profiles from ERA5 (ERA5 Tm), as well as the Tm derived from Bevis and Yao Tm - Ts relationships. The Tm derived from ERA5 was used as a reference to assess the errors of the Bevis and Yao Tm. The comparisons show that the average RMSE values of Tm derived from the two methods are 2.05 K and 2.02 K respectively. The GPS PWVs calculated using the ERA5 interpolated data (GPS ERA5 PWV) were compared to GPS PWV calculated using the sensor meteorological data (GPS MET PWV) at the 5 stations. The GPS ERA5 PWV values were used as the reference. The RMSE values of the two PWV datasets are in the range of 0.19 mm to 0.90 mm with overall average value of 0.44 mm. Also the computed ERA5 PWV data are compared to GPS ERA5 PWV at the 13 stations. ERA5 PWV values were used as the reference and the RMS errors of the two PWV values are in the range 1.35 mm to 2.25 mm with the overall average value of 1.66 mm. Regarding the seasonal variations, the stations between latitudes 4 °S and 4 °N and latitudes 12 °S to 4 °S have the highest PWV values in spring and summer while the lowest values occur in summer and winter respectively.@en

The Global Navigation Satellite System (GNSS) can be used to derive accurately the Zenith Tropospheric Delay (ZTD) under all-weather conditions. The derived ZTDs play a vital role in climate studies, weather forecasting and are operationally assimilated into numerical weather prediction models. In this study, variations and statistical analysis of GNSS-derived ZTD over the East African tropical region are analysed. The data is collected from 13 geodetic permanent stations for the period of 4 years from 2013 to 2016. The 13 stations consist of 5 International GNSS Service (IGS) stations plus 8 stations as follows: 4 Africa Array stations and 4 Malawi Rifting stations from Uganda, Kenya, Tanzania and Rwanda. The ZTD time series were processed using goGPS software version 1.0 beta1, a MATLAB based GNSS processing software, originally developed for kinematic applications but recently re-engineered for quasi static applications. The annual variation of the ZTD time series was investigated using Lomb Scargle periodograms. The semi-annual frequency has the dominant power in subregion 1 (latitudes 4 °S and 4 °N) and the annual frequency has the dominant power in subregion 2 (latitudes 12 °S to 4 °S). The highest ZTD estimates occur during the rainy seasons, at all stations, and the lowest estimates occur during the dry seasons. The results also show that the ZTD estimates are largest at stations located at low elevation (regions close to the Indian Ocean). The derived ZTDs are compared to the values derived from the GIPSY-OASIS via Jet Propulsion Laboratory (JPL) online Automatic Precise Positioning Service (APPS) and the Unified Environmental Modelling System (UEMS) numerical weather prediction (NWP) model. The comparison of goGPS and APPS ZTD at the 13 stations shows an overall average bias, Root Mean Square (RMS) and standard deviation (stdev) of −0.9 mm, 3.2 mm and 3.0 mm respectively, with correlation coefficients ranging from 0.974 to 0.999. The comparison of goGPS ZTD against UEMS NWP ZTD at 8 selected stations shows average bias, RMS and stdev of −12.4 mm, 22.0 mm and 17.6 mm respectively, with correlation coefficients ranging from 0.802 to 0.974. The agreement between the GPS ZTD and the NWP ZTD indicates that goGPS ZTD can be assimilated into NWP models in the East African region.@en

Dual-frequency Global Navigation Satellite Systems (GNSSs) enable the estimation of Zenith Tropospheric Delay (ZTD) which can be converted to PrecipitableWater Vapor (PWV). The density of existing GNSS monitoring networks is insufficient to capture small-scale water vapor variations that are especially important for extreme weather forecasting. A densification with geodetic-grade dual-frequency receivers is not economically feasible. Cost-efficient single-frequency receivers offer a possible alternative. This paper studies the feasibility of using low-cost receivers to increase the density of GNSS networks for retrieval of PWV. We processed one year of GNSS data from an IGS station and two co-located single-frequency stations. Additionally, in another experiment, the Radio Frequency (RF) signal from a geodetic-grade dual-frequency antenna was split to a geodetic receiver and two low-cost receivers. To process the single-frequency observations in Precise Point Positioning (PPP) mode, we apply the Satellite-specific Epoch-differenced IonosphericDelay (SEID)model using two different reference network configurations of 50-80 km and 200-300 km mean station distances, respectively. Our research setup can distinguish between the antenna, ionospheric interpolation, and software-related impacts on the quality of PWV retrievals. The study shows that single-frequency GNSS receivers can achieve a quality similar to that of geodetic receivers in terms of RMSE for ZTD estimations. We demonstrate that modeling of the ionosphere and the antenna type are the main sources influencing the ZTD precision.@en

Accurate sensing of water vapor is crucial for the now- and forecasting of rainfall in numerical weather prediction (NWP) models. Especially severe weather events are characterized by their small spatial scale. Improved forecasting of such events is essential for flood-prone areas and urban stormwater systems which rely on precise rainfall forecasts. Current operational meteorological systems undersample the amount of precipitable water vapor (PWV) which has proven to be an important indicator for rainfall forecasts. There are several possibilities to detect water vapor in the atmosphere. Traditional radiosonde launches offer precise data but are limited in their spatial extent and their trajectory is limited by the wind. Satellite-image-based measurements offer a good spatial resolution but are limited in temporal extent. One possibility to overcome these limitations is to expand existing ground-based GNSS networks. European nationwide GNSS monitoring systems are characterized by inter-station distances typically in the order of tens of kilometers. Due to economic reasons, the densification cannot be achieved with expensive geodetic-grade dual-frequency receivers. Instead, low-cost single-frequency sensors can be used to achieve higher receiver network density. In this work we present first results and an analysis of the error budget from the newly installed regional continuous monitoring network in the urban testbed of Rotterdam. In order to process the data in Precise Point Positioning (PPP) mode, the existing dual-frequency receiver network was used to account for the ionospheric error. The newly installed network consisted of four additional low-cost single-frequency GNSS receivers and was characterized by inter-station distances between 4-5 kilometers. We present comparisons with radiosonde, co-aligned dual-frequency PPP results and existing PWV monitoring campaign reference datasets. This experiment aims to address the feasibility of small-scale water vapor monitoring using low-cost devices. It focuses on the improvement possibilities of such a densification for numerical weather modeling and may help to improve extreme rainfall forecasting. @en