Allan variance applied to time series baseline results for GPS-based deformation monitoring applications

Deformation monitoring systems seek to detect constant movements with time, varying movements with time or a sudden movement. A slow movement at the millimetre level will not be detected by comparing consecutive session solutions due to inaccuracies between sessions. However, comparisons over several sessions, or days, or weeks, could highlight slow consistent changes. The Allan variance is introduced to compare a new measurement with previous baseline data from varying epochs. A low-cost (singlefrequency), GPS-based volcano deformation monitoring system was developed, tested and deployed on the Mt Papandayan volcano, West Java, Indonesia. 21 days of data was collected in July 2001. Data analysis of this equatorial network was compared with a 21-day time series from the mid-latitude SAGE-NZ volcano deformation monitoring network. Errors due to differential tropospheric signal delay between GPS receivers with a large change in height, ionospheric effects, multipath, ambiguity resolution and monumentation etc are discussed and mitigated to reduce the time series repeatability. The Allan variance is applied to these time series’ to identify a temporal upper limit accuracy of the respective monitoring systems and help distinguish between a systematic error and a genuine geophysical signal for a new measurement.