All reviewers agreed that the paper presents an original idea for dealing with spatially aggregated data by constructing a suitable likelihood used together with a GP framework. Experimentally the method seems to be effective. For the final version please address all reviewers' comments.