Michael: initially did the analysis with 4 stations and got 50x suppression at 0.2 Hz, with some residual structure left over. Now trying to repeat it with the larger array: more channels and larger area. For coherence, using all stations including surface and far-away ones.
Improved to 100x subtraction at 0.2 Hz, no left-over structure in the residual. Using only sub-surface stations for the filter, adding the surface stations did not make much difference. Some excess noise was included at low frequencies - not sure why, in principle the filter should not get worse if you include more data (but in practice, numerical issues could cause a difference).
Michael: will also try to make the filter based on the surface stations only, for comparison.
At 1 Hz still have 5-10x subtraction, and higher frequencies less suppression (2-3x), affected by local disturbances. Might be able to improve at high frequencies. Have 1.3 Hz and its harmonics which are not plane waves.
Jan: looked at the k-f maps, produced using the phases between stations and fitting a 3D plane wave to them, averaged over some time. Trying to understand if there is body-wave component in the microseism, which might explain why the initial Homestake data subtraction was not fully successful.
Victor: a number of papers now confirm that body waves contribute to the microseism, typically small contribution and pointed to the storm.
Jan: now we fully subtract the microseismic peak, so wanted to look at the k-f maps. But turns out to be hard, the maps seem useless. The wavelength at 0.2 Hz is very large compared to the array size. Possibly the reason may be that there are likely two main contributors, one from Greenland and one from the Pacific/Alaska, which cannot be resolved with a simple k-f map.
Vuk: should try to look at the radiometer algorithm, may be able to separate two sources.
Jan: For the Wiener filter paper this is it. We would like to determine if there is a sign of body waves in the microseism, would demonstrate that the body waves could be overcome by a 3D array, which would be a good result.
Victor: could try to look at different times of the year, they should give different data and different source.
Jan: maybe try to use spherical wave models to deal with the local sources at 1.3 Hz. Not soon, but would like to do.