$\delta^{18}O$$\mathrm{\delta^{18}O}$ (i. e. the ratio of $^{18}O/^{16}O$$\mathrm{^{18}O/^{16}O}$ compared to a standard) is massively used today in paleoclimatology as a global temperature proxy.
But how did it become a proxy for temperature? When and by whom was the correlation observed for the first time?
Was the model explaining why it followed global temperature found straight away or was this proxy purely empirical for a long time?