"Inaudible" is fine, of course, and knowing or looking up a threshold for that is no problem. Having a good feel for how much more noticeable +3dB are (at various absolute levels) may already be less clear.
Quite clear if one sticks to the fundamentals: an increase in 10 phone (above some 40 phone) is equivalent to a perceived doubling of the sound volume. Once again, the Fletcher-Munson curves help us to find the correlation of SPL and the loudness measured in phon. There’s usually no scaling, i.e. a 10 dB increase in the phon measure will correspond to 10 dB more SPL, but there is a frequency dependent shift in sensitivity.
But what I was mainly thinking about are deviations from white or pink noise. We frequently read descriptions of (and complaints about) noise that is perceived as "whiny", "high pitched", "with a clicking noise" etc. To recognize the signatures of such noise components in the spectra is where I think a lot of experience is needed. And to assess from looking at the spectrum how obnoxious the noise will sound is probably impossible, since it is a rather subjective perception anyway.
I don’t even believe that it’s that subjective. Yes, white (and even more so pink) noise should be quite unobtrusive – and it shows, since it’s used for medical purposes in noise masker systems for tinnitus treatment.
What is more annoying are dominant frequencies, and these in turn all the more so as closer they are to the maximum sensitivity of the human ear, i.e. 3 kHz.
Even more annoying are low frequencies (which would be inaudible themselves) with very high harmonic content that
is audible (“ticking sounds”).
Or any form of frequency modulated dominant tones (“whining noise”). Here it might be a little subjective indeed, which one are perceived as more annoying.
And for all types of unwanted sounds and noises it is clear that they are all the more annoying the louder they are.