Looking at some data over the past month or so, and the higher microseism definitely complicates looking at local sensors for earthquakes. It looks like the fast PEAKMON channel kind of scales with the microseism. The first attach plot shows ~2 weeks of trends for the ITMY .03-.1hz z grnd blrms, the .1-3hz z grnd blrms and peakmon on the top plot. When the .1-.3hz background goes up the peakmon channel generally follows. Not suprising since peakmon is partially a bandpass that overlaps with the microseism some. This complicates using peakmon as a trigger for earthquake transitions. Looking at this data, 700 nm/s looks like a good peakmon threshold for the STAT_ENVs local earthquake threshold. This generally avoids falsely triggering on high microseism, but is low enough to trigger on earthquakes we care about.
The bottom plot shows the ratio of peakmon over the useism blrms. This ratio is a little more consistent outside of earthquake times, staying below about .5, but frequently goes above 1 when there is actually an earthquake, even for relatively small earthquakes. However, there are couple instances where this ratio briefly touches 1 that are not clearly earthquakes, generally only during higher microseism periods. I've been talking to TJ about adding a microseism state to STAT_ENV, maybe part of that state could include an different earthquake threshold for higher microseism times, using either peakmon alone, or the peakmon/useism threshold. The difficulty for that is useism state for that guardian should probably be monitoring hours long averages, and have some kind of sleep during earthquakes.
I did this looking at minute trends, but it would be good to examince full data for some of the corner cases here, to make sure this holds together using actual 16hz data.
