Abstract: The natural regeneration grade is an important foundation for formulating forest management measures. Traditional studies have only considered the regeneration quantities predictive model of the total stand or dominant tree species, but the consistency among the prediction results of different tree species and the total regeneration quantities of stand is not solved. That is, the regeneration prediction results at the stand level are not equal to the sum of the predicted results of all tree species. To address this, on the basis of the traditional counting model, we attempted to construct a compatibility predictive model for regeneration quantities of different tree species within the stand, which would provide a theoretical basis for the rational management and decision-making of natural forest. Based on the survey data from 96 standard plots of Cuigang Forest Farm, Xinlin Forest Farm, and Zhuangzhi Forest Farm in Daxing’an Mountains, we selected 30 basic indices from five aspects of site factor, soil factor, stand factor, tree diversity and stand spatial structure, and used Poisson model and negative binomial model as the basic models to construct the regeneration prediction models of Larix gmelinii, Betula platyphylla and other tree species. By comparing the accuracy and fitting effect of the two traditional counting models, we selected the optimal model and used the seemingly unrelated regressions to further construct the compatibility predictive model for regeneration quantities of different tree species. Poisson model was the best one for the regeneration of L. gmelinii, B. platyphylla, and other tree species. The test index RMSE of the compatibility predictive model for regeneration quantities of L. gmelinii, B. platyphylla, other tree species and total stand regeneration quantities were 388, 413, 504, and 871 trees·hm^-2, respectively. The adjusted R² was 0.389, 0.421, 0.488, and 0.407, respectively. The most influential variables for regeneration quantities of L. gmelinii, B. platyphylla and other tree species were Pielou evenness index of DBH (25.2%), herbal coverage (34.6%) and organic matter in B layer (23.2%). In this study, the compatibility predictive model system for regeneration quantities satisfied the additive logic among L. gmelinii, B. platyphylla, other tree species, and total stands, and provided a basis for accurately estimating natural regeneration.

Key words: regeneration, quantity prediction model, compatibility model, seemingly unrelated regression