Modeling Seasonal Variations in Vertical GPS Coordinate Time Series Using Independent Component Analysis and Varying...

Area of Science:

• Geodesy
• Geophysics
• Signal Processing

Background:

• Seasonal variations are inherent in Global Positioning System (GPS) coordinate time series, impacting geodetic studies.
• Accurate modeling and correction of these signals are crucial for precise analysis of GPS observations.

Discussion:

• A novel spatiotemporal model combining independent component analysis (ICA) and varying coefficient regression is proposed.
• ICA effectively separates common seasonal signals, while varying coefficient regression accounts for their temporal evolution.
• This approach was validated on 262 International GPS Service (IGS) sites globally.

Key Insights:

• The varying coefficient method provides more reliable fitting of seasonal variations compared to traditional least squares regression.
• The proposed model accurately captures common seasonal variations in vertical GPS coordinate time series.
• An average root mean square (RMS) reduction of 41.6% was achieved post-correction, demonstrating significant noise reduction.

Outlook:

• This enhanced modeling technique can improve the accuracy of various geodetic applications relying on GPS data.
• Further research could explore the application of this model to other types of geodetic time series or geophysical phenomena.
• The method offers a robust framework for analyzing time-varying signals in complex observational datasets.