Jim, seems to me it depends on just what you mean by "reasonably accurate." For a first pass, I'd use the standard one with a series inductance, capacitance and resistance and a shunt capacitance. You can measure the crystal parameters to get a good estimate of the values of Cs, Ls, Rs and Cp. Crystal manufacturers will commonly spec a typical and/or maximum Rs and a typical and/or minimum Q and a holder capacitance which is essentially the Cp. Rs and Q and frequency can easily be used to find Cs and Ls.
If you want to improve on that, you can add more parts to account for the inevitable minor parasitic resonances near the fundamental--I think the usual way is additional series LCRs in parallel with the main one. And you can add parts to account for the mechanical overtones at nominally 3x, 5x, ... the fundamental.
If you don't drive the crystal too hard, such a linear model should be OK, but real crystals, just like real resistors, capacitors and inductors, are not perfectly linear, and of course they also have a small temperature variation which depends on the cut angle.
Cheers, Tom