1. Weighted mean temperature models derived from the regression artificial neural network for estimating GNSS PWV in Thailand region
- Author
-
Charoenphon, C., Trakolkul, C., and Satirapod, C.
- Abstract
Estimation of Precipitable Water Vapor (PWV) using GNSS data has been widely recognized as an alternative tool for a meteorological application. All-weather PWV tracking is possible twenty-four hours a day, seven days a week using GNSS data. For highly accurate GNSS-PWV determination, the weighted mean temperature (Tm) is a critical variable. Unfortunately, the typical method of getting Tm involves expensive metrological tools like radiosondes, which are not widely available at every GNSS station in Thailand. A linear relationship between Tm and Surface Temperature (Ts) is another alternative. Locally, seasonal variations, time-dependent factors, spatial coordinates can affect the Ts-based linear model. Linear empirical models have a finite degree of precision. Because they cannot simulate temporal and spatial variations, traditional techniques are unable to depict more complicated temporal changes. In this study, we compare the performance of the countrywide Tm model derived from the simple linear model and the regression Artificial Neural Network (ANN) algorithm to estimate GNSS-PWVs. The Tm values for the investigation are calculated using ERA5 Reanalysis data. Our primary investigation shows that the Root Mean Square error (RMSE) of the different Tm values calculated from the countrywide model and ANN model compared to the ERA5-Tm values as reference is 1.7 and 1.1 K, respectively. In terms of RMSE, we can conclude that the ANN-model performed 35% better than the countrywide Ts-based linear model. However, there is no statistically significant improvements in GNSS-PWVs estimation when the Ts-based linear model and the ANN-Tm model were both applied., The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)
- Published
- 2023
- Full Text
- View/download PDF