Abstract: Objective Selected Larix gmelinii which is the dominant and constructive species of typical boreal forest in northern Great Hinggan Mountains as a research object. Analyzing sap flow in response to multiple environmental factors and building sap flow model through various tree differentiation classes. Methods Using Granier's thermal dissipation probe method and gradient meteorological observation system of eddy covariance tower to continuous monitor the change of sap flow and environment factors. Results The results showed that:1) During the observation period, dominant trees have strong transpiration capacity. The average sap flow density of dominant trees are 1.9 times and 2.5 times comparing to intermediate trees and suppressed trees, respectively. In general, trees with the higher differentiation class have longer duration of daily sap flow, also the peak value of sap flow density appears earlier and have higher peak value. 2) eight environmental factors (precipitation, net radiation, air temperature, air relative humidity, wind speed, soil temperature, soil water content and vapor pressure deficit) could be divided into evaporative demand index (EDI), soil hydrothermal index and precipitation index by principal components analysis. The evaporative demand index (significant correlated with net radiation, air temperature, air relative humidity and vapor pressure deficit) which can explain 45% information of environmental dataset is the key factor influence the sap flow in research area. Soil hydrothermal index and precipitation index can explain 20% and 13%, respectively. 3) Sap flow density presents asymmetrical response to environmental factors through various differentiation classes, which shows a clockwise delayed time lag with EDI, anticlockwise delayed time lag with net radiation and clockwise with vapor pressure deficit, while the time lag with EDI is smallest. Various differentiation classes have the same level of time lag with EDI and vapor pressure deficit, while the time lag of the net radiations smaller in the dominant than that in the intermediate and suppressed. 4) The sap flow density and EDI present a sigmoid function model through various differentiation classes, when the sap flow reaches an asymptote where higher evaporative demand could not cause sap flow to increase further. In this model, the transition slope of the intermediate (0.457) and suppressed (0.458) are greater than the dominant (0.443), which means the dominant is less sensitive to environmental factors. This model can explain on average 90% of the simulation precision in sap flow density in the dominant, intermediate and suppressed trees, respectively. Considering time lag effect of EDI or inserting soil hydrothermal index and precipitation index cannot promote the precision of sap flow model. Conclusions The sap flow of Larix gmelinii has a strong response to multiple environment factors, and these relationships have differences among different tree differentiation classes. By using the sigmoid function model and multiple environment factors, the sap flow of Larix gmelinii with various differentiation classes could be estimated effectively.