The objective of this study is to evaluate the performance of the complementary relationship-based evapotranspiration models, namely, advection-aridity (AA) model of Brutsaert and Stricker and the CRAE model of Morton for estimating actual evapotranspiration. Both models were applied to the Bokhacheon middle-upper watershed, and their estimates were evaluated against the water balance estimate. The calculation was made on a daily basis and comparison was made on monthly and annual bases. For comparison, the water balance estimates were not obtained from the observed precipitation and streamflow data but were based on the simulated data by using integrated watershed model, SWAT-K which is the revised version of SWAT. The reason not to directly use the observed data for water balance estimate is that the credible record period is not sufficient and the streamflow has been altered due to water use and release. Overall, the results showed that both AA model and CRAE model with their original parameters overestimate annual and monthly evapotranspiration, and the large difference between the complementary relationship-based approach and the water balance approach occurs especially for the dry season from Nov. to Mar. It was found out that the parameters, particularly for the advection related parameter, must be recalibrated to accurately produce monthly and annual regional evapotranspiration for this study area.