Horizontal 'scratch' on spectrogram

Hi,

I'm working with one-year geomagnetic data. I also plot spectrogram. I have data from almost 40 geomagnetic stations, but there's this one particular station which the spectrogram has some sort of horizontal 'scratch'. Please see the "spectrogram of G" in the figure.

What might be the source of this scratch in this station?

Thanks!