Pre-train time series models

Is it possible to pre-train a general model, then re-train for a more specific subset using warm start? I am trying to train a model for forecasting production for our various locations. We have a newer location that has over a year of data, but the earlier data is very sparse and likely not representative of the future seasonalities. We expect the new location to be similar, but do expect some slight differences:



I understand it is not recommended to use warm start if there are large changes to the data, but in this case I do expect the changepoints to be very similar so I am more concerned with learning seasonalities. My gut is telling my to use MCMC in this as well.

Thanks!