Monitoring tropospheric water vapor variations with ppp during severe weather

The Global Navigation Satellite Systems (GNSS) have been used for atmosphere monitoring since the early 1990's owing to its accurate, continuous, and high coverage in all-weather operable data. The approach has been known as "GNSS Meteorology" and it is based on the correlation between a GNSS signal delay in the receiver - satellite signal path and humidity (i.e., water vapor content) in the atmosphere. Atmospheric water vapor is an important greenhouse gas and a significant element for the energy exchange in the atmosphere and thus it must be monitored with a high accuracy. However, its spatiotemporal variability brings some limitations in terms of a correct interpretation of weather events. It has been revealed in many studies that zenith wet delays (ZWDs), estimated from a GNSS processing, are associated with the water vapor content, namely precipitable water vapor (PWV), a crucial parameter that has been extensively used in climate studies. Until recently, most of the studies that have been performed on the GNSS Meteorology are based on GNSS solutions using network GNSS observations within a dense regional or global Continuously Operating Reference System (CORS) networks. The advantage of a network solution is the mitigation of potential error sources by differencing observations formed between stations tracking same satellites. However, a computational burden due to a high number of observations and the necessity of using a reference site to form a baseline between sites are the weak points of the solution. In contrast, Precise Point Positioning (PPP) stands out as an alternative GNSS measurement technique, which uses undifferenced code and phase observations of a dual frequency single receiver and highly accurate precise orbits and clocks distributed by the International GNSS Service (IGS). In order to achieve a similar accuracy to a network based GNSS solution, all error sources in PPP have to be modeled or eliminated from observations either by applying correct models or by using linear combinations of observations. Therefore, from one point of view, PPP is one of the best quality control tool for assessing the reliability of each parameter defined in observation equations. Meteorological data are crucial part of the GNSS based atmosphere monitoring. The data may be obtained from in-situ measurements when available or may be derived from external sources such as analytical prediction models and numerical weather models (NWM). Analytical models, such as the Global Pressure and Temperature (GPT and GPT2) and the University of New Brunswick (UNB), are widely used models because of their ability to predict meteorological data at any time and for any site location. However, they are produced based on annual and semiannual data averages and amplitudes of meteorological data that makes these analytical models insufficient for observing sudden air changes. Contrary, NWM like the European Centre for Medium-Range Weather Forecasts (ECMWF) and the Vienna Mapping Function grids (VMFG) that provides the ECMWF data four times within a day, are based on real weather analyses using frequently updated meteorological data, and represent spatio-temporal atmospheric variations better than the analytical models. But, the VMFG offers the ECMWF data with a latency of several hours that prevents users from performing real-time GNSS analyses. Similarly, several modern mapping functions have been developed as a result of analyzing of several years' meteorological data from different stations and elevation angles. Those functions have increased the performance of GNSS-based atmosphere monitoring by means of being susceptible to low elevation angle observations and by using advanced virtual models, i.e., analytical or numerical, that statistically strong in order to represent the dry and wet part of the troposphere. Global Mapping Function (GMF), Niell Mapping Function (NMF) and Vienna Mapping Function (VMF1) are few examples of these commonly used mapping functions that were all developed to be used collaboratively with the GPT, UNB and VMFG (i.e., ECMWF data), respectively. Using different analytical and NWMs and their associated mapping functions to model the troposphere affects estimates derived from GNSS observations and thus weather predictions utilized based on these estimates vary from each other. Furthermore, the performance of these mentioned models and corresponding mapping function for each one of them have never been tested during extreme weather conditions before. In this dissertation, modern troposphere modeling approaches derived from both analytical prediction models and NWMs have been evaluated via GNSS PPP sessions during severe weather conditions. The sessions were processed in both post-process (PP) and real-time (RT) mode with dual frequency ionosphere-free observations. ZWDs, which are estimated through the Least Squares Parameter Adjustment (LSA), converted to the PWV estimates at each site in order to observe variations in the water vapor and to monitor the potential precipitation fronts. The Central European Flooding in summer 2013 was one of the extreme periods when some of the European regions experienced a monthly average rainfall within a day. The water vapor variations in May and June of 2013 were exceptionally high and thus this period was selected to reveal the relationship between the GNSS derived troposphere parameters and severe weather events. In order to accomplish the task, eleven IGS sites that are located within the affected regions were selected and daily observations of these sites were processed through static PPP sessions. IGS Final orbits and clocks for satellites were introduced to the analyses to enhance the accuracy of the zenith wet delay estimates and, therefore, weather forecasts. Besides IGS Final ephemerides, other published IGS ephemerides products such as IGS Rapid (IGR) and IGS Ultra-Rapid (URP) products were also processed in the PPP sessions to determine if any significant changes are occurring in the estimates due to employ different IGS product sets varying in accuracy and latency. The ZWD estimates derived from the PPP sessions that only differ in applied troposphere model and mapping function (i.e., GPT / GMF, UNB / Niell and ECMWF / VMF1) were compared to each other and the ZWDs derived from the network GNSS solutions. The results were also validated by the ZWD estimates derived from IGS Final troposphere products (i.e., IGS Zenith Path Delays (ZPD)) and by the radiosonde derived ZWDs that calculated from co-located (or close) sites. The ZWD estimates of each session were derived simultaneously using both LSA and Kalman filter (KF) algorithms and no significant differences arising from using different processing strategy was observed. Based on the results of the PP data analyses, the PWV parameters, which are calculated from the ZWD estimates via a conversion factor, offer more consistent results with the IGS derived PWVs if an analytical function of the GPT/GMF is employed in the PPP solution. The analyses result also confirmed that the PPP derived PWV values, which are processed with the ECMWF / VMF1, better agreed with the radiosonde derived PWV values while the biggest discrepancies were observed for the PPP sessions applied the UNB / Niell approach. In overall, the performance of the PPP sessions using the ECMWF and VMF1 exceeds the other solutions in terms of atmosphere monitoring during severe weather events. In this thesis, the implementation procedure of real-time orbit and clock correction streams, which are disseminated by the IGS Real-Time Service (IGS RTS), has been investigated through real-time (RT) PPP analyses. Long convergence times due to un-calibrated receiver and satellite based phase biases were also assessed in order to accelerate real-time solutions. It was noticed that using real-time derived biases along with real-time orbits and clocks help to reduce long convergence periods to several minutes. In addition, regardless of the applied troposphere model, the accuracy level of PWV estimates derived from the RT PPP sessions, were found to be consistent with the results derived from the PP PPP analyses and the radiosonde readings. The reliability of the RT PPP in atmosphere monitoring was also validated by showing the strong correlation between the PWV variations and the daily accumulated precipitation rates recorded at the observed sites. As observed in the PP analyses, the best agreement with the radiosonde and precipitation data was obtained for the RT sessions using the ECMWF / VMF1. This study also highlighted the use of the Moderate-Resolution Imaging Spectroradiometer (MODIS) to verify both the RT and the PP PPP PWV estimates. It is observed that the PP PPP estimates showed more consistent results with the MODIS data that the real-time solutions. The thesis provides a comprehensive guide for GNSS based troposphere monitoring, in particular with PPP solutions and during extreme weather conditions. The impact of using different troposphere models, IGS product sets and processing strategies on the estimates were investigated for both PP and RT GNSS solutions.