L1 regularization for detecting offsets and trend change points in GNSS time series

Abstract

Offset and trend change point detection are major problems for GNSS time series preprocessing. Without accurate detection of change points and offsets, signals estimated from GNSS time series are prone to be biased. To solve this problem, we introduced an extensive L1 regularization model, which can estimate piecewise trends, level shifts and seasonal signals simultaneously from raw GNSS time series. It thus can be used to detect trend change points and discontinuities successfully in GNSS time series. Furthermore, a new Python tool has been incorporated into our previous TSAnalyzer software to realize the benefits our L1 regularization model and some examples are listed to show its usage.