Computational fluid dynamics has been used for consequence modelling in place of traditional methodologies, to obtain more accurate results and better understanding of complex problems. With today's computational technology, it is not difficult to run CFD simulations for several selected scenarios. However, its use to produce a full quantitative risk assessment, involving thousands of scenarios, is impractical given the extensive computational resources required. In this paper, a modified QRA methodology proposed recently, with the integration of sampling algorithms, surrogate models and error measures, is extended to a higher dimensional hydrocarbon pool fire problem. The characteristics, advantages and challenges of a local linear model and a global non-linear model are identified and addressed. The addition of time as an input variable to surrogate models provides an extension of the methodology to safety studies where the resolution of time-varying output is important. We demonstrate the effectiveness of the modified QRA methodology through a case study of pool fire in a hydrocarbon storage facility. (C) 2018 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.