>Past precipitation can be used to reconstruct past palaeoclimatic temperatures. δD and δ18O is related to surface temperature at middle and high latitudes. The relationship is consistent and linear over Antarctica[9]...

I think this is also flawed, because this computation requires the ratio of the isotopes in the atmosphere at the period of interest, which we don't have. And the people making this measurement just seem to assume it would certain value derived from the present, which is also unreliable...

So yea, all things considered, pretty unreliable..