昆明市4个主要针叶树种林分断面积生长模型研究

Stand Basal Area Growth Models for 4 Main Coniferous Species in Kunming

采用Richards、Schumacher和Korf模型作为林分断面积生长备选模型,运用麦夸特算法、差分进化算法、遗传算法、模拟退火算法和粒子群算法进行模型参数求解,根据R2和RMSE选择模型拟合结果、算法迭代次数和残差分布,比较各优化算法的效率和参数稳定性。结果表明:华山松、云南松、油杉和柏木地位级表落点检验值分别为97.9%、98.3%、98.1%和98.9%,精度符合要求,能够用于林区林业生产经营活动;优化算法求解模型参数的效率由高到低的顺序为LM>DE>PSO>GA>SA,PSO求解参数的拟合优度较差;针叶树种断面积生长模型更适宜采用Richards模型,Schumacher模型参数拟合结果更稳定。运用优化算法进行林分断面积生长模型参数估计并分析其优劣,对提高模型精度具有重要作用,研究结果为优化算法在生长模型参数估计中的运用提供了依据。

Algorithms of levenberg-marquardt,differential evolution,genetic algorithm,simulated annealing, and particle swarm optimization were applied for parameters solving of stand basal area growth model for Richards, Schumacher and Korf. Determination coefficient and root mean square error were used for selecting suitable parameters. Iterations were used for comparisons of different algorithms efficiency. Residuals were used for parameters stability analysis. Results show that falling point test of site class model are 97. 9%,98. 3%,98. 1%,and 98. 9% for Pinus armadii,Pinus yunnanensis,Keteleeria fortune, Cupressus funebris, respectively. Site class tables could be used in forest management. Parameter solving efficiency from high to low is LM 〉 DE 〉 PSO 〉 GA 〉 SA,Goodness of fitting applying PSO is poor; Richards model is more suitable for the growth model of conifer species,Parameter fitting results of Schumacher model are more stable.