The management of water resources in compliance with regulatory requirements increasingly uses national environmental impact assessments incorporating models of soil solute transport. Most such models use either continuous functions designed to solve the Richards equation, or capacity-type functions based on water contents between specified soil water pressure heads. The work described here has used three separate sets of measured water-release data representative of the whole of the UK to develop a set of pedotransfer functions (PTFs) for predicting volumetric water contents at seven pressure heads on the water release curve ranging from 0 to −1500 kPa. A theoretically-based structural stratification of the data was proposed and tested. Within each grouping, multiple regression analysis was used to derive equations for predicting the water contents at each pressure head, based on a selected set of predictors. The resulting PTFs gave a significantly better prediction than PTFs derived from the unstratified data and also than the widely used HYPRES PTFs. For all non-sandy horizons not subject to regular cultivation, additional PTFs incorporating a climatic variable, the average annual potential soil moisture deficit (PSMD) gave a further improvement in prediction in drier areas of the UK with PSMD greater than 130 mm. Mean errors associated with using the PTFs to parameterize continuous function models and a capacity model for calculating water availability for cereals (APcereals) are ±16.4 and 16.9%, respectively.