Accurate prediction of stem diameter is an important prerequisite of forest management. In this study, an appropriate stem taper function was developed for upper stem diameter estimation of white birch (Betula platyphylla Sukaczev) in ten sub-regions of the Daxing’an Mountains, northeast China. Three commonly used taper functions were assessed using a diameter and height dataset comprising 1344 trees. A first-order continuous-time error structure accounted for the inherent autocorrelation. The segmented model of Max and Burkhart (For Sci 22:283–289, 1976. https://doi.org/10.1093/forestscience/22.3.283) and the variable exponent taper function of Kozak (For Chron 80:507–515, 2004. https://doi.org/10.5558/tfc80507-4) described the data accurately. Owing to its lower multicollinearity, the Max and Burkhart (1976) model is recommended for diameter estimation at specific heights along the stem for the ten sub-regions. After comparison, the Max and Burkhart (1976) model was refitted using nonlinear mixed-effects techniques. Mixed-effects models would be used only when additional upper stem diameter measurements are available for calibration. Differences in region-specific taper functions were indicated by the method of the non-linear extra sum of squares. Therefore, the particular taper function should be adjusted accordingly for each sub-region in the Daxing’an Mountains.