Loss of containment often results in flashing releases of hazardous chemicals into the atmosphere. Rainout of these chemicals reduces airborne concentrations, but can also lead to extended cloud duration because of re-evaporation of the rained-out liquid. Therefore, for hazard assessment one must use models which accurately predict both the amount of rainout and its rate of re-evaporation. However, the findings of a literature survey reveal weaknesses in the state-of-the-art for modelling the sub-processes of droplet atomisation, atmospheric expansion, two-phase dispersion, rainout, pool formation and re-evaporation. A recent joint industry project has implemented recommendations from this survey, deriving from scaled water experiments droplet size correlations for conditions ranging from negative to high superheat. This experimental programme is reported in more detail in a separate companion paper. As a whole these correlations describe a tri-linear function of droplet size (expressed as Sauter mean diameter) as a function of superheat. This function describes the regimes of non-flashing, the transition between non-flashing and flashing, and fully flashing. The new correlations have been compared with previous correlations recommended by the Dutch Yellow Book and CCPS Books. The correlations are validated against published experiments including the STEP experiments (flashing propane jets), experiments by the Von Karman Institute (flashing R134-A jets), and water and butane experiments carried out by Ecole des Mines and INERIS. The rainout calculations by the dispersion model have been validated against a subset of the CCPS experiments (flashing jets of water, CFC-11, chlorine, cyclohexane, monomethylamine). (C) 2006 Elsevier B.V. All rights reserved.