Forest productivity varies with soil moisture more than temperature in a small montane watershed

•Temperature inversions were observed more often than not in a 27 km watershed across 460 m elevation.•The cold-air pool had a constant upper boundary throughout the year at ∼1370 m a.s.l.•Modelled elevational trends of tree height generally matched lidar-based observations.•T and VPD were not the key drivers of elevational variation in forest productivity.•Soil moisture was likely the key driver of elevational variation of forest productivity. Mountainous terrain creates variability in microclimate, including nocturnal cold air drainage and resultant temperature inversions. Driven by the elevational temperature gradient, vapor pressure deficit (VPD) also varies with elevation. Soil depth and moisture availability often increase from ridgetop to valley bottom. These variations complicate predictions of forest productivity and other biological responses. We analyzed spatiotemporal air temperature (T) and VPD variations in a forested, 27-km2 catchment that varied from 1000 to 1650 m in elevation. Temperature inversions occurred on 76% of mornings in the growing season. The inversion had a clear upper boundary at midslope (∼1370 m a.s.l.). Vapor pressure was relatively constant across elevations, therefore VPD was mainly controlled by T in the watershed. We assessed the impact of microclimate and soil moisture on tree height, forest productivity, and carbon stable isotopes (δ13C) using a physiological forest growth model (3-PG). Simulated productivity and tree height were tested against observations derived from lidar data. The effects on photosynthetic gas-exchange of dramatic elevational variations in T and VPD largely cancelled as higher temperature (increasing productivity) accompanies higher VPD (reducing productivity). Although it was not measured, the simulations suggested that realistic elevational variations in soil moisture predicted the observed decline in productivity with elevation. Therefore, in this watershed, the model parameterization should have emphasized soil moisture rather than precise descriptions of temperature inversions.