A stand-level model derived from National Forest Inventory data to predict periodic annual volume increment of forests in Italy

A model was developed for predicting the periodic annual volume increment (PAI) of forests using variables commonly recorded through field surveys or the remote sensing. The model was developed using the Italian National Forest Inventory (INFC2005) data, publicly available at www.inventarioforestale.org. Data from 5707 plots were split into two groups. The first was used for fitting the model; the second was used for cross validation. Model reliability for applications at the local, in the Alpine and Mediterranean regions, and at the country level was tested. A sensitivity analysis was carried out to investigate the effects of entering inaccurate values of the number of trees per hectare, one of the predictors of the final model, that may occur in case of biased estimates from the remote sensing. During model calibration, the highest proportion of increment variation was captured using forest category (FC) as dummy variable and, in this respect, this study supports the classification of forests on ecological basis as a stratification criterion in environmental sampling. The model explained 72% of PAI and it predicted annual increment at plot level with no statistical difference to the observed value in any FC, at the country level.本研究开发了一款用于预测森林定期年蓄积生长量（Periodic Annual Volume Increment, PAI）的模型，模型输入变量多通过野外调查或遥感手段采集。该模型基于意大利国家森林资源清查（Italian National Forest Inventory, INFC2005）数据开发，相关数据公开于www.inventarioforestale.org平台。研究共纳入5707块样地数据，并将其划分为两组：一组用于模型拟合，另一组用于交叉验证。随后，本研究检验了该模型在区域（阿尔卑斯山区、地中海区域）及国家尺度的应用可靠性。为探究遥感估算出现偏倚时，作为模型核心预测因子之一的每公顷株数输入不准确数值所带来的影响，研究开展了敏感性分析。在模型校准阶段，以森林类别（Forest Category, FC）作为虚拟变量可最大程度捕捉生长量的变异；据此，本研究支持将基于生态学的森林分类作为环境抽样的分层标准。该模型可解释72%的PAI变异，且在国家尺度及各森林类别下，样地水平的年蓄积生长量预测值与观测值均无统计学差异。