One common way of doing it in a LC resonant circuit is to pick the C part to have a negative tempco val equal to the positive tempco val of the L to try and cancel total temp drift.
I am talking about LC VCO, but c & L temp variation coefficient is really depend on the each process, right? also, the value of c & L will affect other performance (PN & tuning range) , right?
for wide temperature range, as you have, you must basically try to use low TC devices. For inductance, only air coil is appropriate, no ferrite. Then a residual TC from VCO circuit could be hopefully compensated in part. But I doubt, if compensation will work for wide range. Most non-zero TC capacitors are specified for -20..+85 at maximum.