A quantile regression approach to model stand survival in Chinese fir plantations

Abstract

The development of stand survival models can provide an important basis for the sustainable management of forest resources. In a new approach developed in this study, parameters of four survival quantile regression models were predicted from a quantile associated with a current stand density. The curves from these quantile regression models were then used to project future stand density for that stand. A three-fold cross-validation revealed that the quantile regression approach outperformed the least squares method based on three evaluation statistics, especially for longer projection lengths. These results were consistent for all four survival models evaluated. The best survival model is Clutter–Jones model, without constraints, but its ln( N)–ln( Dq) trajectories ( N = stand density and Dq = quadratic mean diameter) from the quantile regression showed the linear self-thinning trend.