Item

Abstract

Zenith tropospheric delay (ZTD) is critical information for global navigation satellite systems (GNSS) meteorology. ZTD is estimated from GNSS stations, radiosonde stations, meteorological stations, meteorological reanalysis data, global pressure and temperature 3 model (GPT3), etc. Extreme weather events can detect from ZTD values. In this situation, the importance of obtaining reliable ZTD data increases. With the development of machine learning, its usability has also been raised in ZTD estimation. In this study, we have developed three ZTD models based on the support vector machine model (SVM) using radiosonde ZTD data, the International GNSS Service (IGS)-ZTD products, ZTD estimated by European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5) data throughout July 2017. The reason for choosing the year 2017 is the extreme weather events as severe precipitation and hail events that took place in Istanbul (IGS-ISTA) on July 18-27, 2017. In order to investigate which ZTD model gives better results, Root Mean Square Errors (RMSE), and bias were adopted for our comparisons in extreme weather events and normal weather conditions.