Using the felled tree data, subregions-based height-diameter models were developed for Dahurian larch (Larix gmelinii Rupr) in Daxing'an Mountains in China Thirty frequently used nonlinear growth functions were fitted to individual tree height-diameter data. A total of 2411 for Dahurian larch trees were collected and measured. Parameter estimates and evaluation statistics were calculated using SAS PROC NLIN procedure for overall model, three regional models, and ten subregional models. The non-linear extra sum of squares method was used to assimilate the difference of the height-diameter relationships among different subregions. The results indicate that the Chapman-Richards growth function show the best for fitting tree height-diameter relationships at regional and subregional levels with a root mean square error RMSE = 2.193 and coefficient determination R-2 = 0.8221. It was revealed that, Dahurian larch height-diameter relationships showed the several differences among these regions and subregions (P < 0.0001) when overall, regional and subregional model was applied to predict tree height separately in each region and subregion. Among the 45 pairs of subregions, only four subregion pairs (e.g., XL versus SBZ, XL versus HJY, SBZ versus HJY, and XL versus JGDQ) showed the nonsignificant difference F-value (P > 0.05) and the data from those subregions were combined and applied to others subregions; all comparisons show the difference. The tree height-diameter relationships of Dahurian larch are significantly different among the geographic regions in Daxing'an Mountains, depending on local climatic, soil, and ecological conditions.