采用野外调查易获取的林分特征因子、立地因子和气候因子构建杉木林分蓄积量非线性混合效应模型,为无人机激光雷达技术监测森林蓄积量提供技术支撑。基于湖南省和广西壮族自治区2021—2023年3个年度林草生态综合监测杉木实地调查的580个样地数据,建立含林分特征因子的林分蓄积量基础模型,并综合考虑立地因子和气候因子,确定固定效应变量和随机效应变量后构建非线性混合效应模型,运用十折交叉验证法进行检验。结果显示:以林分平均优势高、郁闭度为代表的林分特征因子,海拔为代表的立地因子,年平均气温为代表的气候因子与林分公顷蓄积量相关性高,利用林分平均优势高、郁闭度、龄组构建了林分蓄积量基础模型;加入海拔立地因子作为固定效应变量,引入样地所在年平均气温差异作为随机效应,构建杉木林分蓄积量非线性混合效应模型,其拟合精度较基础模型有所提高,决定系数(R²)从0.616提高到0.654,估计值的标准差(E_SE)、平均预估误差(E_MP)和平均百分标准误差(E_MPS)值均有所降低。对林分平均优势高、郁闭度、立地和气候因子与林分蓄积量之间的关系进行了有效探究,构建的杉木林分蓄积量模型具备一定的实用性,建模采用的变量均能通过无人机激光雷达点云数据或坐标信息获取,为利用激光雷达技术获取区域森林蓄积量提供了参考途径。

The nonlinear mixed-effect model of Cunninghamia lanceolata forest volume was developed using easily accessible stand characteristic factors,site,and climate factors obtained from field investigation.This model provides technical support for forest volume monitoring by UAV LiDAR technology.Based on mensuration data from 580 sample plots of Cunninghamia lanceolata forest and comprehensive ecological monitoring of grassland in Hunan Province and Guangxi Zhuang Autonomous Region between 2021 and 2023,a basic model for stand volume,including stand characteristic factors,was constructed.Subsequently,a nonlinear mixed-effects model was developed by incorporating site and climate factors,and determining both fixed and random effect variables.The models were evaluated using 10-fold cross-validation.The results revealed that stand characteristic factors(such as average dominant height and canopy density),site factors(elevation),and climate factors(annual mean temperature)were strongly correlated with the hectare volume of stand.The basic stand volume model,based on average dominant height,canopy density and age group,was constructed.Compared to the basic model,the nonlinear mixed-effect model,which included altitude as a fixed effect and the annual mean temperature difference among plots as a random effect,showed improved fitting accuracy.The determination coefficient(R²)increased from 0.616 to 0.654,while the standard error of the estimate(E_SE),mean prediction error(E_MP),and mean percent standard errors(E_MPS)were reduced.The relationship between stand volume and factors such as average dominant height,canopy density,site and climate factors are effectively explored.The resulting stand volume model demonstrates practical applicability,and all the variables used in modeling could be derived from UAV LiDAR point cloud data or coordinate information,providing a valuable reference for regional forest volume monitoring using LiDAR technology.