[Objective] Evapotranspiration (ET) including evaporation (E) and transpiration (T), was the main way that ecosystem precipitation (P) returned to the atmosphere. Under the background of climate change, understanding the characteristics and distribution of ET in a cold-temperate Larix gmelinii natural forest at permafrost region of northern part of Daxing'an Mountains, northeastern China could help to further comprehend the response modes of the boreal forest ecosystem to the climate change. [Method]Model simulation and field experiments were applied for estimating and measuring ecosystem scale evaporation (E), transpiration (T), and evapotranspiration (ET) in a boreal larch forest. E includes forest floor evaporation (E_f) and canopy interception (E_c) which only occurred during rainfall events. Total forest transpiration (T_(tot)) in this study is the sum of dominant tree transpiration (T_d), intermediate tree transpiration (T_i) and suppressed tree transpiration (T_s). Furthermore, we analyzed the variations and allocation proportions of ET and components under conditions of non-rainy or rainy. Then we discussed the responses of E_f, T_(tot) and ET to the net radiation (R_n) and the vapor pressure deficit (VPD) under different water input conditions. [Result] Diurnal variations of ET and its components all performed as single peak patterns in both non-rainy and rainy days, and peak value in each curve at non-rainy day grouping was higher than in rainy day grouping. E_f, T_d, T_i, T_s and ET in non-rainy days were 10.3, 25.6, 15.2, 10.8 and 66.3 mm, respectively. Meanwhile, E_f, E_c, T_d, T_i, T_s and ET in rainy days were 2.2, 24.3, 11.2, 5.1, 3.8 and 47.8 mm, respectively.In non-rainy days, E_f/ET was 15.5%, and T_(tot)/ET was 78.0%, while T_d/ET, T_i/ET and T_s/ET contributed 38.7%, 23.0% and 16.4%, respectively.In rainy days, E_f/ET could go low to 4.6%, E_c/ET reached 50.9%, and T_(tot)/ET decreased to 42.2%, in which T_d/ET, T_i/ET and T_s/ET were 23.5%, 10.6% and 8.0%, respectively.These results represented that ET was dominated by T (specifically for T_d) in non-rainy days. However, E (specifically for E_c) contributed the highest proportion of ET in rainy days. 94.7% of P returned to the atmosphere through ET during whole observation days, there into, T and E accounted for nearly 57% and 38%, respectively. Overall, without regard to rainfall events, ET had better correlation with net radiation at 23 m height (R_n) than with VPD and T_(tot) owned similar correlations with both R_n and VPD, but E_f expressed opposite result with ET. This result demonstrated that R_n was obviously the main driver for ecosystem's energy cycle and material exchange,T_(tot) restrained by R_n and VPD at the same time, and E_f was more susceptible to VPD compared to R_n. [Conclusion] The transpiration capacity of dominant trees of Larix gemelinii species was stronger than intermediate and suppressed trees. So that, forest transpiration upscale from the individual tree could be overestimated if this method only considers in situ measured T_d (even including T_i with larger DBH) and ignores T_s. The actual error depends on the degree of forest differentiation and moisture income conditions. The meteorological condition of non-rainfall days is more propitious to the water-vapor exchange at vegetation-atmosphere interface, and the occurrence of rainfall will affect the distribution pattern of ecosystem ET.