I have been told that the signal in the graph below was obtained by passing a seismogram through a high-pass filter to obtain the high frequency-small amplitude waves common in tremors far from the epicenter of an earthquake.
The high-pass filter used to unmask the tremor was selected to avoid the 'attenuation' of the tremor signal. What exactly is meant by 'attenuation' in this context?
Attenuation in this context refers to a reduction in amplitude of some signal component as it passes through a system (in your case, the system being the highpass filter. Linear filters are designed to pass certain frequency components while stopping (or attenuating) others. As the name implies, a highpass filter is designed to pass high-frequency content while attenuating low-frequency content.
As you noted, the tremor features that you're interested in are high in frequency and small in amplitude. Thus, they may be masked by low-frequency components in the original seismogram. A highpass filter can be used to accentuate the high-frequency tremor waves by attenuating any low-frequency waves that might also be present.
As a complement to Jason's fine answer I would like to add this: Normally, higher attenuation rates exist in higher frequencies and it is likely that higher frequency signals become buried/masked in the noise (again due to higher attenuation and subsequently lower Signal to Noise ratio). In your application, as you mentioned, lower frequencies also exist and can distort signals even more, making signal interpretaion challenging. In this case, a high pass filter can be used to get rid of those lower frequencies and help identifying signals of interest (tremor) at higher frequencies. Please also note that attenuation rate can change in different elastic media at different frequencies.
