This is an ad-hoc residual, observations minus predicted earth tide minus air pressure effect.

Ad-hoc because the time series involved here are all recorded by the gravimeter's acquisition system.

Since the plot of the latest 31 days is updated on the current day, only a reduced range of data types are available for reduction.

In particular, we don't have the Atmacs pressure effects, so we work with the local air pressure.

The currently installed prediction table is from a least-squares analysis of 2015-02-05.

A preliminary solution, least-squares fit of a luni-solar tide potential to the observed data, simultaneously with air pressure.

A simple regression model, giving a coefficient in units of nm/s

Generated either in the deep ocean by short-wavelength standing waves or on the continental shelves, sometimes as far aways as the Gulf of Biscay.

- Black curve shows gravity residual.
- Gray curve shows the last ten minutes.
- Yellow curve is low-pass filtered gravity residual, band corner at 31 mHz, suppressing the microseismic noise due to ocean waves
- Dark-blue curve shows air pressure
- Light-blue curve shows band-passed filtered air pressure, with a 40 dB gain factor applied. Big variance is indicative of strong local winds.

- Black: Gravity residual with drift
removed and low-pass filtered, 1 smp/min. Low-pass filter is
not narrow enough to suppress seismic waves from big
earthquakes, primarily late-arriving surface waves ("coda")
and early free-oscillations. The drift signal is adjusted on a
weekly schedule (see
the diagram, scroll down to bottom).

- All other series at 1 smp//h
- Red: Predicted earth tide, typically
2000 nm/s
^{2}peak-to-peak. - Dark-blue: air pressure
- Light-blue: Onsala "bubble mareograph" (previously Ringhals tide gauge), sign inverted (so the signal tends to follow air pressure)
- Yellow: Proxy for bottom pressure at Onsala (previously Ringhals), indicative of mass imbalance in the local region

Using the Burg algorithm, a long prediction-error filter is calculated (400 coeffcients) from the 1 smp/s residual. Eleven slices of data each 12 minutes long, shifted through one hour by 6 minutes' steps.

The curves are colour-coded with yellowish colours for the oldest slice to blue for the most recent one. Notice the colour bar along the top of the frame. A baseline (light-blue) has been determined from a few quiet days, retrieving the mean-decibel level (i.e. the geometric mean).

The spectra (and the baseline) are affected by the anti-aliasing filter of the gravimeter's analog-digital converter. The baseline is used in the spectrograms (follow the link "spectrograms vs. time") as their 0-dB level.

The oceanic microseisms show up primarily in the band between 30 and 300 mHz.

Without much careful study, alleged source region are shown by colour code, seen as vertical pajamas stripes.

.bye