Modeling and Forecasting the GPS Zenith Troposphere Delay in West Antarctica Based on Different Blind Source...

Area of Science:

• Geodesy and Geophysics
• Atmospheric Science and Meteorology
• Data Science and Machine Learning

Background:

• Tropospheric delay is a significant error source in Global Positioning System (GPS) measurements.
• Water vapor derived from tropospheric delay is crucial for meteorological research, including climate analysis and weather forecasting.
• Existing zenith tropospheric delay (ZTD) models are primarily used for positioning corrections, with limited application in water vapor estimation, particularly in Antarctica.

Purpose of the Study:

• To establish high-accuracy hourly ZTD models for GPS meteorology in West Antarctica.
• To evaluate the effectiveness of blind source separation algorithms (PCA, ICA) combined with neural networks (BP, LSTM) for ZTD modeling and water vapor estimation.
• To assess the short-term forecasting capabilities of the developed ZTD models.

Main Methods:

• Utilized GPS-ZTD data from 52 GPS stations in West Antarctica.
• Applied two blind source separation algorithms: Principal Component Analysis (PCA) and Independent Component Analysis (ICA).
• Developed models using a Back-Propagation (BP) neural network and a Long Short-Term Memory (LSTM) network.

Main Results:

• Models achieved mean accuracy better than 10 mm, with Independent Component Analysis (ICA) slightly outperforming Principal Component Analysis (PCA).
• Mean Root Mean Square (RMS) errors were 9.3 mm for PCA and 8.9 mm for ICA, with correlation coefficients exceeding 90% against GPS-ZTD.
• Six-hour forecasts showed the best performance (mean correlation 90.6%, mean RMS 7.2 mm), while 24-hour forecasts were significantly less accurate (correlation 63.2%). Forecast accuracy was lower in coastal areas than inland.

Conclusions:

• The combination of ICA and deep learning (LSTM) effectively models ZTD and restores original signals for GPS meteorology.
• Short-term ZTD forecasting shows promise for meteorological applications in Antarctica.
• Further technological development is needed to improve long-term forecasting accuracy and address regional variations.