Spatial variation in microclimate caused by air temperature inversions plays an important role in determining the timing and rate of many physical and biophysical processes. Such phenomena are of particular interest in mountainous regions where complex physiographic terrain can greatly complicate these processes. Recent work has demonstrated that, in some plants, stomata do not close completely at night, resulting in nocturnal transpiration. The following work was undertaken to develop a better understanding of nocturnal cold air drainage and its subsequent impact on the reliability of predawn leaf water potential (Psi(pd)) as a surrogate for soil water potential (Psi(s)). Eight temperature data loggers were installed on a transect spanning a vertical distance of 155 m along a north facing slope in the Mica Creek Experimental Watershed (MCEW) in northern Idaho during July and August 2004. Results indicated strong nocturnal temperature inversions occurring from the low- to upper-mid-slope, typically spanning the lower 88 m of the vertical distance. Based on mean temperatures for both months, inversions resulted in lapse rates of 29.0, 27.0 and 25.0 degrees C km(-1) at 0000, 0400 and 2000 h, respectively. At this scale (i.e., < 1 km), the observed lapse rates resulted in highly variable nighttime vapor pressure deficits (D) over the length of the slope, with variable impacts on modeled disequilibrium between soil and leaf water potential. As a result of cold air drainage, modeled Psi(pd) became consistently more negative (up to -0.3 MPa) at higher elevations during the night based on mean temperatures. Nocturnal inversions on the lower- and mid-slopes resulted in leaf water potentials that were at least 30 and 50% more negative over the lower 88 m of the inversion layer, based on mean and maximum temperatures, respectively. However, on a cloudy night, with low D, the maximum decrease in Psi(pd) was -0.04 MPa. Our results indicate that, given persistent cold air drainage and nighttime stomatal opening, serious errors will result if Psi(s) is estimated from Psi(pd).