Assessing uncertainties by evaluation against independent observational datasets