GPS signal for weather parameter

Abstract

In recent years, meteorologists have been looking at utilizing the Global Positioning System (GPS) technology for weather prediction. GPS is utilized not only for location tracking but also to collect weather parameter datasets such as Atmospheric Gradient (including convergence and magnitude), Precipitable Water Vapor (PWV) and Precipitation represented in Binary. In addition, Artificial Intelligence techniques have been used as tools using methods such as ResNet and UNet models for weather accuracy prediction. Three methods of analysis have been used in this project, namely, numerical datasets on individual data points, dataset averaging of dimensions 6x6 and 8x8 and image processing. The first method analysis shows that using all four parameters will achieve better accuracy compared to dropping any parameters. Additionally, UNet and ResNet were also being compared, but UNet is better in achieving accuracy. The second analysis is using the averaging of 6x6 and 8x8 and it shows that dimensions of 8x8 perform better with the test size of 0.3. The major parameter that affects the accuracy is the PWV. Finally, image processing was also done by converting all the numerical data into image format, but the result was not showing good result. Overall, it can be seen that averaging the spatial dimension of the numerical datasets would show better performance and results.