基于稀疏型机载激光雷达数据的风景林参数估测

稀疏型机载激光雷达（LiDAR）数据，由于点云密度低，难以对单木尺度的森林参数进行估测，在森林资源监测领域的潜在应用价值没有得到充分发挥。以江苏省南京市紫金山国家森林公园为研究区，以2007年机载激光雷达Optech ALTM LiDAR数据、2007年森林资源规划设计调查数据为主要信息源，在机载LiDAR数据预处理、特征参数提取的基础上，采用多元线性回归、随机森林、支持向量机3种方法，估测小班林分平均树高、平均胸径和单位蓄积量并进行对比分析，同时对森林参数进行空间制图。结果表明:1） 高度百分位数和累计高度百分位数是影响小班平均胸径、平均树高和平均蓄积量的主要特征参数；2）3个遥感估测模型精度对比分析表明，平均胸径、平均树高、单位蓄积量3个参数的估测精度，均是随机森林算法最高，支持向量机次之，多元线性回归最低；3）平均树高、平均胸径、单位蓄积量的空间分布规律一致，人为干扰严重的公园边缘地区和土层瘠薄、立地条件差的山脊较低，山南及山北中部最高。稀疏型机载激光雷达（ALS）数据在林分尺度的森林参数估计中具有较高的精度，可以用于森林资源规划设计调查小班的测树因子估测。本研究应用稀疏型机载激光雷达数据估测风景林森林参数，为稀疏型LiDAR数据在森林资源规划设计调查中的推广应用提供参考。

Estimation of the Parameters of Scenic Forests Using Sparse Airborne LiDAR Data

It is difficult to estimate forest parameters at individual tree scale using sparse airborne laser radar (ALS) data due to its low density of point clouds,so the potential of sparse density ALS data in forest resource monitoring has not been fully exerted.In this study,the airborne Optech ALTM LiDAR data in 2007 and forest resource planning and design survey data in 2007 of Purple Mountain National Forest Park in Nanjing City,Jiangsu Province were used.Based on preprocess of ALS data and extraction of feature parameters,the modeling methods of multiple linear regression (MLR),random forest (RF) and support vector machine (SVM) were used to estimate the average tree height,average DBH and stand volume of subcompartments.Moreover,the spatial distribution map of the three forest parameters were created.The results showed that 1) percentile height and percentile cumulative height were the main variables that affected the average DBH,average tree height and stand volume of subcompartments.2) The random forest had the highest accuracy for the estimations of average DBH,average tree height and stand volume,followed by support vector machine and multiple linear regression.3) Three forest parameters had the same spatial distribution pattern,the marginal area of the park with strong human disturbance and the ridge with poor soil and site conditions had lower average tree height,average DBH and stand volume,the southern and northern central parts of the mountain had higher average tree height,average DBH and stand volume.Sparse airborne laser radar (ALS) data is feasible to estimate forest parameters in forest resource planning and design survey data.In this study,sparse airborne LiDAR data were applied to estimate forest parameters of scenic forests,which provides scientific reference for the popularization and application of sparse LiDAR data in forest resources survey.