Transpiration and canopy conductance at two slope positions in a Japanese cedar forest watershed

Plant–soil system patterns and processes along a slope are among the greatest causes of uncertainty in estimating watershed-scale transpiration (E). Tree-to-tree and radial variations in xylem sap flux density (Fd), in addition to tree biometrics, were measured over a 2-year period (2005–2006) in two slope stand positions. The areas of interest consisted of an upper slope plot (UP) and a lower slope plot (LP) in a Japanese cedar (Cryptomeria japonica D. Don) forest watershed and the environmental controls of stand E for each plot were compared. Canopy stand E (EC) and canopy stomatal conductance (GC) in the UP were less than those in the LP during the growing season, while those in the UP were greater than those in the LP over winter. In addition, mean stand Fd (JS) in the UP was greater than that in the LP over winter, but JS values were similar in the UP and LP except in the winter, which allows us to extrapolate watershed-scale E based on JS estimated from Fd measurements of a partial stand in the watershed. However, this relationship contains a bias and differed between 2005 and 2006. Although there were significant differences in soil moisture conditions between the UP and LP in both years, a systematic relationship between the similarity in JS and soil moisture conditions was not found. The bias was due to a tendency for JS in the LP to be greater than that in the UP in 2006. This tendency was amplified because JS in the LP was greater than that in the UP around an atmospheric humidity deficit (D) of 1–1.5 kPa and frequencies of this D range were higher in 2006 than in 2005. The greater JS in the LP at D ∼ 1–1.5 kPa could be explained by the difference in the response of GC to D between the UP and LP. Our results suggest this to be the cause of the similarity in JS values for the UP and LP and for the occasional abortion of its similarity. However, even when the bias or the occasional deviation is disregarded, the error in estimating stand E from a partial stand is so small that it is comparable to an Fd measurement error. For example, the error when using only the LP was 6.6% for stand E.