Abstract Knowledge of soil water availability for plant growth is vital for the development of plant growth simulation models. Data on soil water availability are often not available because field and laboratory measurements of soil water content are time‐consuming and tedious. The objective of this study was to develop alternative procedures to predict water content at ‐10 kPa (UL10), ‐33 kPa (U133), ‐1500 kPa (LL), and the potential available water capacity (AWC) from easily and routinely available soil properties. Multiple regression equations for soil orders of Soil Taxonomy were developed using a database containing information of about 12,000 pedons of the continental U.S., Hawaii, Puerto Rico, and some foreign countries. Regression equations with bulk density, sand, silt, clay, and organic carbon contents accounted for up to 83% of the variation in UL10 for all orders except Ultisols. For Ultisols, sand content accounted for up to 90% of the variation in UL10. Equations with clay and organic carbon contents accounted for up to 75% of the variability in UL33 for all except Aridisols, Oxisols, Vertisols, and Spodosols. For these four orders, equations with bulk density, clay, silt, and sand contents accounted for up to 81% of the variation in UL33. LL was linearly related to clay content. Clay content accounted for up to 91% of the variation in LL for all but Oxisols and Vertisols. More accurate predictions of AWC resulted when AWC was computed from UL10 and LL water content data. Equations with bulk density alone or bulk density plus silt and/or sand contents accounted for up to 83% of the variation in AWC for all except Entisols, Inceptisols, and Spodosols.