This is a figure including all the cases of broad band stellar typing using PS1, SDSS, and Gaia. The sample consists of 22 stars with no photometry problems in all the catalogs. The stars meet the crude color selection for F-stars defined in PIPE2D-383.

There are five cases that use a different photometry set: PS1 grizy, PS1 grizy + SDSS u-band, SDSS ugriz, SDSS griz (without u), Gaia GBpRp.
The robust (sigma clipping) RMSs of fractional residuals of each case are about 3-4%. There is no big difference among the cases. If we add SDSS-u to PS1, we can get a bit better result. 
The largest RMS at PS1 g-r~0.33 is due to a double peak of its probability distribution in log g. This could be improved by a spectral typing.
This is a figure including all the cases of broad band stellar typing using PS1, SDSS, and Gaia. The sample consists of 22 stars with no photometry problems in all the catalogs. The stars meet the crude color selection for F-stars defined in
PIPE2D-383.There are five cases that use a different photometry set: PS1 grizy, PS1 grizy + SDSS u-band, SDSS ugriz, SDSS griz (without u), Gaia GBpRp.
The robust (sigma clipping) RMSs of fractional residuals of each case are about 3-4%. There is no big difference among the cases. If we add SDSS-u to PS1, we can get a bit better result.
The largest RMS at PS1 g-r~0.33 is due to a double peak of its probability distribution in log g. This could be improved by a spectral typing.