1: \begin{abstract}

2: Convergence of the Bayes posterior measure is considered in canonical statistical settings where observations sit on a geometrical object such as a compact manifold, or more generally on a compact metric space verifying some conditions. 

3: A natural geometric prior based on randomly rescaled solutions of the heat equation is considered. Upper and lower bound posterior contraction rates are derived.%, which under some conditions are shown to be  minimax adaptive.

4: \keywords{Bayesian nonparametrics \and Gaussian process priors \and Heat kernel}

5: % \PACS{PACS code1 \and PACS code2 \and more}

6: \subclass{MSC 62G05 \and MSC 62G20}

7: \end{abstract}

8: