We use wavelets in a Bayesian context to identify changes in the pattern of data collected over time, when missing observations are present. Our work is motivated by the interest in identifying and modeling change points in the measurements of sub-glacial water pressure during melt season along the length of the Bench glacier in Alaska. This modeling will provide insights into the sub-glacial hydrology including the discharge mechanism during the melt season. A Bayesian analysis based on the empirical wavelet coefficients is used to find the change point (Oden and Lynch,1998) in 18 water pressure data sets available.We compare this method with wavelet based method suggested byWang (1995), which examines the empirical wavelet coefficients of the data at the fine scale levels. The above methods have to be adapted for accommodating missing observations, which are present in our data sets.
AGU Fall Meeting Abstracts
- Pub Date:
- December 2007
- 0720 Glaciers;
- 4540 Ice mechanics and air/sea/ice exchange processes (0700;