经验模型和过程模型对油松林枯损预测的比较

Comparison of Empirical and Process-based Methods on Mortality Predictions for Pinus tabulaeformis Stands

研究采用西北农林科技大学开发的可变密度全林模型QUASSI 1.0,基于赫尔辛基大学开发的CROBAS碳平衡模型框架,融合二类调查数据、文献数据以及解析木数据,进行参数校正及优化,本地化了CROBASPT(油松)枯损模块的参数。并根据立地条件和初始密度选取了9个代表性油松林分,以20年为预测期,比较分析了经验模型和过程模型在不同密度、不同地位级条件下对枯损预测的差异,探索有效的林分枯损预测方法。同时采用平均误差、平均绝对误差和平均相对误差分析了过程模型CROBAS-PT与经验模型QUASSI 1.0预测值的偏差。结果表明:对于不同初始密度的油松林,CROBAS-PT和QUASSI 1.0在预测期内均呈现初始密度越大,林分年枯损率越大的规律;对于不同地位级的油松林,无论CROBAS-PT还是QUASSI 1.0在预测期内,林分枯损受立地条件的影响均不敏感。CROBAS-PT枯损预测的机理过程分析说明,油松树冠投影在20~40 a预测期内呈现先增加后降低的趋势。立地质量越好的林分,树冠投影越大,枯损率越大。误差检验分析显示,过程模型CROBASPT枯损预测结果符合统计检验要求,尽管CROBAS-PT对于枯损的预测相比经验模型QUASSI 1.0存在一定程度的低估。在缺乏连续观测样地数据,无法保证经验模型的建模数据需求时,采用过程模型方法预测林分枯损不失为一种有效补充。

Based on carbon balance framework of CROBAS developed by University of Helsinki and variable density empirical model QUASSI 1.0 developed by Northwest Agriculture and Forestry University, and multi-source inventory data, the parameters of CROBAS-PT （Pinus tabulaeformis） were calibrated. The effects of stand site and density on mortality predic- tions were analyzed, based on a 20-year simulation of 9 plots to compare the empirical model and process based model on mortality predictions. Mean error, mean absolute error and mean relative error were calculated to compare the results from CROBAS-PT and QUASSI 1.0. The mortality predictions are sensitive to different initial density, hut insensitive to site. The stands of good site perform high crown projection area and mortality. Although the CROBAS-PT model predicts lower level of mortality than QUASSI 1.0 model, the process-based CROBAS-PT model can be used for mortality predictions, as its results match the statistical test. Since processed model explains the theory of forest mortality, it can be an effective method to predict the mortality of P. tabulaeformis in the region where the data are lacking.