Effect of size and number of calibration plots on the estimation of stem diameter distributions using airborne laser scanning

摘要

Stem diameter distribution is a crucial forest inventory variable in operational forest management. Compared to ground based forest mensuration (e.g., diameter at breast height (DBH)), airborne laser scanning (ALS) offers a cost effective alternative for modelling forest inventory variables. The objective of this study is to determine the impact of the size and number of sample plots on modelling diameter distributions in an unevenaged tolerant hardwood forest using discrete return ALS data. With the size of the sample plots ranging from 0.04 to 0.25 ha, DBH distributions were divided into six structural classes, estimated by two categories of non-parametric methods: k-nearest neighbor (k-NN) imputation and the random forest (RF). Sensitivity analysis demonstrated that the size of sample plots has a stronger impact on model performance than the number of plots. In addition, RF was found to be the most accurate model, regardless of the size and number of sample plots.