In last week’s post we used the seismic data obtained from Apollo Mission 12 to look at Data QC and how PSD PDF plots can be informative in this regard. We described the situation where an instance of the meta-data was inaccurate – missing information in the response file – concluding that the PSD results at and around 2.2 seconds were invalid. One might be tempted, then, to discard the entire data-set, ignoring everything it is trying to tell us. But this would be a mistake.
While we know that we can’t use the data at 2.2 seconds, the PSD results at longer periods are not invalid; and this is true even though the response is nominal and not strictly rendered according to the calibration pulses contained in the data itself. So for this post, instead of looking at the data from a QC perspective, I’m going to have a look at the PSD’s at periods longer than 2.2 seconds, grouped by week, to render a Seasonal Spectrogram/PDF.
In the plot below, all 16,675 PSD’s from 1970 have been grouped and averaged by week-of-year (not necessarily starting 1-Jan), with the chart following listing the exact dates of the New and Full Moon phases. The color bar on the right side of the spectrogram defines the dB noise levels associated by color, the Y-Axis representing Period (in seconds, starting at 2.3 of them), and the X-Axis representing the month of the year:
With blue representing minimum noise levels, and red representing maximum noise levels, when we compare the New and Full Moon phase dates with the Seasonal Spectrogram, something very quickly and obviously pops out. Namely, the more sunlight shining on the moon during daylight hours, the higher the overall noise; while conversely, the less sunlight hitting the moon’s surface, the lower the noise levels; where this difference can be up to 20 dB or more!
As per this NASA Science News Release, there are four types of moonquakes we humans know of: 1) deep, 2) vibrations from meteorite strikes, 3) thermal, and 4) shallow. Here, then, we have a visual representation of thermal moonquakes resulting from the massive temperature change between night and day (-153°C to 107°C).
Two things to take away: 1) Even though one aspect of a data-set is demonstrated to be invalid, this does not necessarily mean the entire data-set is worthless; and 2) Our collection of PSD’s, when grouped in different ways can tell us something totally unique about varying noise levels depending on the grouping scheme we select.
Next week: Diurnal Spectrograms and what this PSD grouping scheme has to tell us about this data-set. Hint: Yet something else completely different.
– Richard Boaz