Abstract:We used 12 height curve models using height-diameter data from felled trees for Larix gmelinii Rupr. from three ecoregions in Daxing’an Mountains Region to evaluate the coefficient determination, bias, and root mean square error for model accuracy and precision. The functions for Korf, Tang, Hossfeld were better for height curve from Region 1 and Region 3. Richard, Weibull, Mitscherlich were better for height curve from Region 2. We used full model and reduced model test method for regional comparison. The height curves were significantly different among different ecoregions. Height curve in Region 2 had significant difference with Regions 1 and 3, and height curve in Region 1 had small difference with Region 3. Incorrectly applying height curve in different ecoregions would result in larger prediction error.