Extracting long-term trends and high frequency pressure fluctuations from seafloor pressure records using Empirical Mode Decomposition (EMD)

Empirical Mode Decomposition (EMD) is a well-established tool for exploratory data analysis of non-stationary and nonlinear data. It decomposes the data set into a series of Intrinsic Mode Functions (IMF) which separate the analyzed data set into a set of functions with different characteristics, representing the underlying physical processes. Pressure sensors at the seafloor record a time-varying pressure signal which is a composition of periodic and aperiodic signal components such as (1) atmospheric and oceanic tides as well as solid earth tides, (2) water level changes due to dynamic effects of the ocean (3) internal density fluctuations in the water column caused by salinity and/or temperature variations, (4) atmospheric disturbances, (5) vertical motions of the ocean floor (subsidence/uplift), (6) settling of the instrument, (7) sensor drift, and (8) sensor noise. Our goal is to examine the signal components like long-period pressure changes associated with uplift or subsidence and on the other hand short-period fluctuations associated with earthquakes or tremor-like signals. The amplitude of these pressure signals are on the order of only 1 kPa and smaller and difficult to extract from a signal which is dominated by tidal and strong oceanographic components. A number of pressure time series from different sources (DART, CORK, OBP, PIES) is analyzed and results are presented. The EMD is very powerful in extracting the long-period changes but also allows to isolating short term pressure fluctuations related to earthquakes and periods with increased background pressure noise. The major difficulty is, however, still to associate the observed pressure changes with tectonic or magmatic events because a number of effects -in the subsurface as well as in the water column - may be the cause.