Detection of abrupt baseline length changes using cumulative sums

Abstract

Dynamic processes are usually monitored by collecting a time series of observations, which is then analysed in order to detect any motion or non-standard behaviour. Geodetic examples include the monitoring of dams, bridges, high-rise buildings, landslides, volcanoes and tectonic motion. The cumulative sum (CUSUM) test is recognised as a popular means to detect changes in the mean and/or the standard deviation of a time series and has been applied to various monitoring tasks. This paper briefly describes the CUSUM technique and how it can be utilised for the detection of small baseline length changes by differencing two perpendicular baselines sharing a common site. A simulation is carried out in order to investigate the expected behaviour of the resulting CUSUM charts for a variety of typical deformation monitoring scenarios. This simulation shows that using first differences (between successive epochs) as input, rather than the original baseline lengths, produces clear peaks or jumps in the differenced CUSUM time series when a sudden change in baseline length occurs. These findings are validated by analysing several GPS baseline pairs of a network deployed to monitor the propagation of an active ice shelf rift on the Amery Ice Shelf, East Antarctica.