Microseism source direction from noise cross-correlation

Inhomogeneous noise sources surrounding stations produce asymmetric amplitudes in cross-correlation functions that yield preferential source directions. Here we show that preprocessing biases the dominant source direction estimate towards the source producing long-duration signals by down-weighting high-amplitude signals. Tests with both synthetic data and observations show that conventional preprocessing, where only earthquakes and local transients (e.g. trawling, fish impacts) are removed, is more sensitive to coherent energy, while one-bit preprocessing and running-absolute-mean preprocessing are more influenced by signal duration. Comparisons between different preprocessing methods are made on data from the Cascadia Initiative ocean bottom seismometer array, where we find that the total energy arriving from pelagic and coastal areas is similar. Moreover, pelagic-generated signals tend to be weaker but have longer duration, in contrast to coastal-generated signals that tend to be stronger but have shorter duration.