The aim of this study is to evaluate rainfall partitioning at the forest canopy and reveal the physical process of canopy interception loss. Observations were conducted for 19 months in neighboring stands of Chamaecyparis obtusa Sieb. et Zucc. (Hinoki) and Cryptomeria japonica D. Don (Sugi). Cumulative amounts for the period showed that portions of throughfall (TF), stemflow (SF) and interception (IC) to rainfall (RF) for Hinoki were 65.3%, 9.1%, and 25.5%, respectively. Corresponding values for Sugi were 67.9%, 6.6%, and 25.5%. The smaller TF and larger SF in Hinoki than those in Sugi were induced by greater mean funneling ratio of a tree and greater tree density in Hinoki. Similar IC/RF would result from similar leaf area index. In analyses for rainfall events, rainfall period (RP) was defined as the period excluding short no-rainfall periods within an event, and rainfall intensity (RFI) was as RF/RP. In events with canopy saturation (RF >= 10 mm), IC/RF was insensitive to RP and RFI. This was related to an increasing rate of IC with RFI Evaporation for IC estimated by the model, based on the Penman-Monteith equation, was approximately 40% of cumulative IC observed. Underestimation was great in events with long RP, but not with large RFI. We suggest that large amount of IC occurred during rainfall, which is induced by splash droplets transport (SDT) by canopy ventilation. (C) 2013 Elsevier B.V. All rights reserved.