Presumably your aim is to filter out the pump noise, in which case the three things we need to know are: the frequency at which the pump operates, the INTERNAL diameter and the length of the air line.
The Air line and the surge tank form an RC filter with the flow resistance of the air line as the R (calculated according to Poiseuille's Law) and the volume of the tank as the C. The resulting time constant (RC) should be well below the period of the pump (1/2piF where F is the pump frequency). 1 octave will give you 3dB attenuation, 1 decade will give you 10 dB so if, for instance, you wish to reduce the surge to say 1% of its initial value you need two decades (eg RC = 100 x pump period)
If there's a pre regulator on the line things become very complex as this is hard to model. Probably the ebst approach is to insert a small surge tank after the pre reg and model it as a zero output impedance from the reg - this will slightly oversize the final tank but that's probably not a worry.