DiscoTimeS: New approach to detect change points in GNSS, satellite altimetry, tide gauge and other geophysical time series
Precise knowledge of coastal vertical land motion (VLM) is essential for assessing the impact of sea level rise on coastal regions and its consequences for local populations. It links relative sea level, measured by tide gauges with a fixed connection to land, with absolute sea level, measured by satellite altimetry. VLM can be determined pointwise using Global Navigation Satellite Systems (GNSS) or from the combination of tide gauge measurements with absolute sea level changes measured by satellite altimetry.
Among the largest sources of uncertainty in the determination of VLM are discontinuities and trend changes in the observed time series. Discontinuities are often caused by instrumental changes, especially in GNSS. Trend changes often have seismic causes, but can also be due to long-term changes in surface loading or to local effects. Although these issues have been addressed extensively for GNSS data analysis, there is limited knowledge of how to directly detect and mitigate such events when determining VLM from the difference of altimetry and tide gauge (SATTG).
The novel Bayesian approach DiscoTimeS automatically and simultaneously detects discontinuities and trend changes, for the first time not only for GNSS time series but also for VLM derived from SATTG. We show that accounting for time-varying VLM significantly increases the agreement of SATTG with GNSS measurements (by 0.36 mm/year on average for 339 globally distributed station pairs). Bayesian change point detection is applied to 606 SATTG and 381 GNSS time series. One of the main results of this work is that the determination of time-varying VLM now makes it possible to avoid extrapolation errors of coastal VLM and sea level change projections. Details of the study are presented in the article Bayesian modeling of piecewise trends and discontinuities to improve the estimation of coastal vertical land motion (Journal of Geodesy, 2022, DOI: 10.1007/s00190-022-01645-6, [PDF]).
Arcisstraße 21
80333 München