Handling coherence measures of displacement field time series: Application to Greenland ice sheet glaciers

Displacement Field Time Series (DFTS) are often derived from Satellite Image Time Series to study dynamic systems such as glaciers. This analysis can be performed with pattern-based data mining techniques that search DFTS for all possible displacement evolutions. Nevertheless, existing pattern oriented methods do not take into account the coherence measures coming along with such DFTS data. This paper introduces an approach for handling the coherence when extracting displacement evolutions. In addition to defining a coherence notion for these evolutions, we show that focusing on the coherent patterns allows to prune the search space. Reported experiments exhibit consistent displacement evolutions of Greenland ice sheet glaciers.