One of the biggest challenges in predicting ecohydrologic fluxes is scaling from easily measured variables to more difficult, often emergent patterns and processes. This is especially true in spatially heterogeneous systems such as black spruce (Picea mariana)-dominated boreal forests containing excessive and low soil moisture conditions. Traditional hypotheses suggest that transpiration is controlled by hydraulic responses to vapor pressure deficit (D) and soil moisture; however, these may potentially be misinformed because of omission of soil drainage gradients. Thus, we predict that 1) spatial heterogeneity in tree transpiration along a soil drainage gradient is positively correlated with D, 2) sap flux (JS) and leaf-level transpiration (EL) are higher and whole-tree transpiration (EC) lower in the poorly drained than well-drained stands, and 3) spatial heterogeneity of EC is regulated primarily by tree-related covariates such as sapwood and leaf area and secondarily by environmental covariates including peat and moss depth. With the use of 122 black spruce sap flux measurements, the range of autocorrelation (inverse of spatial variation) decreased from 20 m at low D (<0.7 kPa) to 2 m at midday D values (>0.9 kPa). JS and EL were significantly greater and EC less in poorly drained than well-drained stands; controlled primarily by tree-related covariates (sapwood and leaf area) representing long-term growth conditions and secondarily by soil moisture and spatially sampled D reflecting shorter-term environmental variation. Quantification of spatial heterogeneity informs predictive models of the distance at which homogeneity can no longer be assumed and will improve mechanistic predictions of transpiration at multiple spatial scales. Copyright © 2012 John Wiley & Sons, Ltd.