MIDAS robust trend estimator for accurate GPS station velocities without step detection

Area of Science:

• Geosciences
• Geodesy
• Data Analysis

Background:

• Automatic estimation of GPS velocities is crucial for handling large datasets.
• Existing methods often overlook data problems like outliers and seasonality.
• Robust trend estimation is needed for accurate velocity determination.

Purpose of the Study:

• To develop an automatic and accurate trend estimator resistant to common data problems.
• To introduce the Median Interannual Difference Adjusted for Skewness (MIDAS) method.
• To assess MIDAS's performance against other automatic estimators.

Main Methods:

• Developed MIDAS, a variant of the Theil-Sen median trend estimator.
• MIDAS uses data pairs separated by 1 year to mitigate seasonality and step discontinuities.
• Outliers are removed and the median is recomputed to reduce bias.

Main Results:

• MIDAS achieved ±0.23 mm/yr RMS accuracy in horizontal velocity for North American plate GPS data.
• In blind tests with synthetic data, MIDAS showed ±0.33 mm/yr horizontal and ±1.1 mm/yr vertical RMS accuracy.
• MIDAS demonstrated superior performance compared to 20 other automatic estimators.

Conclusions:

• MIDAS provides a robust and accurate method for estimating velocities from GPS time series.
• The method effectively addresses common data issues, enhancing reliability.
• MIDAS has broad potential for application across geosciences.