Evapotranspiration (ET) was estimated from a planted coniferous forest in southwestern Japan by applying
three methods: the eddy covariance method; the measurement of rainfall (P) and runoff (Q) in a small
watershed; and a combination of rainfall interception loss (IC), upper canopy transpiration based on a
sap-flux density measurement in Japanese cedar (Cryptomeria Japonica D. Don) stands (EUC), and modeled
sub-canopy ET (ESC). After inverse multiplication of the energy imbalance ratio, ET by the eddy covariance
method (ETEC) was 839.9 mm in 2007 and 811.8 mm in 2008. The yearly values of P–Q were partially
affected by P in the previous autumn. After continuous data collection for more than 5 years, P–Q became
stable. The 9-year (2000–2008) average P–Q, which was considered most reliable in this study, was
897.5 mm y1. The cumulative ETEC during the daylight hours from the right stream bank, covered mainly
with large Japanese cedars, was 894.1 mm from April 2007 to March 2008. The value was almost the
same as that calculated as the components sum (ETCOMP = IC + EUC + ESC: 911.4 mm), and the comparison
suggested that the annual totals of ETEC with an energy imbalance correction provide a reliable estimate
of ET in a forest stand on a complex topography. Spatial variation in the watershed was likely caused by
differences in soil water retention at each slope position. The slight difference in annual ETEC in 2007
compared with 2008 was attributed to differences in the radiative energy input. In the monthly–weekly
analysis, ETCOMP was frequently higher than ETEC after heavy rainfall, while ETEC was higher under dry
conditions and during active ET. Even under dry canopy conditions, daily ETEC was often higher than
EUC + ESC. The results suggested a time-lag in evaporation from the ecosystem and/or under-estimated
ETEC after rainfall.