Depending upon the distance of the seismic station from the epicentre, the quality of waveform data and the improvisation of the synthetic waveform on it is often a question. If the Pnl waveform segments are showing good fit then S wave components may show poor fit. How this can be improved through proper window and time selections.
Various research articles present a very straight forward approach for the presence of dense seismic stations but how the gCAP inversions can present good inversion results with less density of seismic station network.
For instance
"Synthetics and observed ground velocity were filtered in the same frequency bands (from 0.02 Hz to 0.1 Hz for the surface waves and from 0.05 Hz to 0.3 Hz for the Pnl). These frequency bands were chosen to maximize signal-to-noise ratios of data and to avoid short-wavelength structural heterogeneities. Synthetic seismograms have been computed using an average velocity model derived for the Italian territory as described by Scognamiglio et al. (2009)".
This may not suit my inversions and result in a lower variance reduction.
Your inputs will be much appreciated.