Developing objective methods for evaluating the effects of silvicultural treatments on carbon sequestration and stock in plantation forests

我们开发的方法来获得参数估计林分生产力从地面激光雷达获得的三维数据。为了从LiDar数据估计树的尺寸，我们应用了神经网络算法。我们能够估计胸径误差< 2m，树高误差< 50cm。该算法平均区分了林分中95%的茎。这意味着地面激光雷达的误差水平与人类工作人员的误差水平相似。因此，激光雷达可以作为一种客观的森林测定方法，而不会引入人为误差。

We developed methods to obtain parpameters for estimating stand productivity from 3D data obtained by ground-based LiDAR. To estimate tree dimensions from LiDar data, we applied the neural network algorithm. We were able to estimate DBH with < 2m error and tree height with < 50cm error. The algorithm distinguished 95% of the stems in the stand, on average. This meant that error levels for ground-based LiDAR were similar to that of human workers. Thus, LiDAR could serve as a objective method for forest mensuration without introducing human error.