Estimating the available soil water reservoir is needed for developing a strategy for optimum management of rainfed crops in marginal dry regions. Previous research on measurement of the lower limit water content of soil water availability (required for calculation of the total soil water reservoir) has shown unexplained discrepancies between field estimates of the water content at which plants are dead or dormant due to soil water deficit and laboratory estimates of the water content at a water potential of -1.5 kJ kg(-1). Whether such discrepancies are due to sample size, handling, or the choice of -1.5 kJ kg(-1) as the lower limit is not clear, This study was designed to determine what the limit is under controlled conditions, and to test the appropriateness of the -1.5 kJ kg(-1) value. We used in situ soil psychrometers to measure soil mater potential (psi) as water was withheld from cotton (Gossypium hirsutum L.) plants in a rain-sheltered lysimeter monolith and from sorghum [Sorghum bicolor (L.) Moench] in glasshouse pots. The loner limit psi was -2.2 kJ kg(-1) in the lysimeter experiment and <-2.5 kJ kg(-1) in the glasshouse, With the lysimeter, the difference between the soil water content corresponding to -2.2 W and -1.5 W kg(-1) was <0.01 m(3) m(-3). Water potential at the greatest measurement depth (2.1 m) decreased rapidly from near 0 to -1.8 kJ kg(-1) over a period of 30 d, while root length density increased from nearly 0 to 2400 m m(-3). At an advanced stage of soil water deficit, there was limited upward movement of water from depths >2.1 m. No psi oscillations, corresponding to so-called hydraulic lift or nocturnal water transport between soil layers, were found at any depth in the lysimeter (0.5-2.1 m). Our results suggest that for Rater balance calculations requiring absolute accuracy for the water balance of the soil reservoir, the choice of the -1.5 kJ kg(-1) soil water potential is appropriate and corresponds closely to the held lower limit of soil water availability.