Global land surface models use spatially distributed soil information for the parameterization of soil hydraulic properties (SHP). Parameters of measured SHP are correlated with easy-to-measure soil properties to construct general pedotransfer functions (PTFs) used to predict SHP from spatial soil information. Global PTFs are based on a limited number of samples yielding highly variable and poorly constrained SHP. The study implements a physical constraint, soil-specific capillary length, to reduce unphysical combinations of SHP. The procedure fits concurrently soil water retention and capillary length using the same parameters. Results suggest that meeting the capillary length constraint has minor effects on the goodness of fit to soil water retention data. Constrained SHP were applied to represent 4 years of lysimeter fluxes yielding evapotranspiration values in close agreement with measurements relative to slight overestimation by unconstrained SHP. The procedure was applied for testing constraint SHP at a regional scale in New Zealand using the surface evaporation capacitance model and Noah-MP for detailed simulations of land surface processes. The use of constrained SHP in both models yields higher surface runoff in agreement with observations (unconstrained SHP severely underestimated runoff generation). The concept of constrained SHP could be extended to include other physical constraints to improve PTFs, for example, by consideration of vegetation cover and soil structure effects on infiltration.