GPS signal derived precipitable water vapor and its applications in rainfall prediction

• Main Author: Manandhar, Shilpa
• Other Authors: Lee Yee Hui
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: https://hdl.handle.net/10356/106016; http://hdl.handle.net/10220/48092
• Institution: Nanyang Technological University

The atmospheric water vapor is generally expressed in terms of Precipitable water vapor (PWV). It is an important indicator of water vapor climatology and variability in the lower troposphere and related climate processes. Global Positioning System (GPS) satellite signals are used to derive PWV values with high spatio-temporal resolution which is available under all-weather conditions. The PWV values are derived using the delay information from GPS signals. The signal delays in the troposphere are divided into two categories; Zenith Hydrostatic Delay (ZHD) and Zenith Wet Delay (ZWD). The ZWD delay is caused by the water vapor of the atmosphere and PWV values are derived using these ZWD values. The accuracy of the GPS-PWV values depends on how accurately the ZWD values are derived. A comprehensive analysis is done to understand the effects of different mapping functions and different ZHD models on the retrieved PWV values. The results show that there is a very small effect of the mapping functions on the PWV values. Use of different ZHD models have greater effect on the PWV values. When converting the ZWD values to PWV, it involves the use of a dimensionless constant called PI. This PI value is found to rely on a water-vapor weighted mean temperature (Tm) value which varies according to different locations and seasons. It is therefore, both site and time specific. Analysis of the PI value and its effect on the retrieved PWV from the data obtained from different regions show that although the PI value is time and site specific, the change in the median value of PI for different years is minimal and is dependent only on factors like the latitude coordinates of the particular site and the day-of-the year. Therefore, using the data obtained from 174 different sites, a latitude-coordinate and day-of-year based PI value model is proposed for the retrieval of PWV. The proposed model has been successfully validated using data from different databases. The GPS-PWV values are useful in applications like rainfall prediction, cloud detection, etc. With the rapid deployment of the GPS CORS (Continuously Operating Reference Stations), many researchers are studying the PWV values and its usefulness in the prediction of a rainfall event. This thesis firstly reviews the rainfall prediction algorithms from the temperate and sub-tropical regions. Secondly, We propose a prediction algorithm suitable for the tropical region. It was found that the absolute value of PWV plays an important role in long-term rainfall prediction, unlike for the temperate and the sub-tropical regions. In addition to the long-term rain prediction, a simple algorithm is proposed to perform the rainfall nowcasting in the tropical region. The algorithm applies GPS-derived PWV values and its second derivatives. The proposed algorithm incorporates the seasonal dependency of PWV values for the prediction of a rain event. The occurrence of rainfall is complicated and dependent on a myriad of atmospheric parameters. Therefore, this research is extended into the systematic study of various parameters that affect the precipitation in the atmosphere. Different ground based weather parameters are identified as probable features for rainfall prediction and a detailed feature correlation study is presented. It has been shown that only a few parameters; PWV, solar radiation, seasonal and diurnal factors stand out for the rainfall prediction. In summary, in this thesis, we have studied the GPS signal delays and related processes in detail. We have done a comprehensive study on different hydrostatic delay models and their effect on the PWV values. We have proposed a simplified PI model that is globally applicable. We have proposed rainfall prediction algorithms suitable for the tropical region. We have implemented a data-driven approach that uses other meteorological parameters to improve the rainfall prediction.