Is there access to large, representative datasets of the data sets needed to make hazard predictions?