Abstract
The line-intercept sampling (LIS) method has found important applications in such areas as forest and wildlife, ecological and biological sciences, and crop and agriculture fields. LIS is a sampling technique to make observations along line transects in order to make inferences of area properties. The placement of transects can be chosen in many different manners, i.e., randomly or systematically. The motivation of this study is to use LIS to infer regional information of forestry biophysical parameters based on the linear transects measurements of a profiling LiDAR system. However, there is no optimum method to properly derive a reasonable measure to the uncertainty of LIS estimates. As such, the study first developed a theoretical framework to describe the LIS estimation in two settings, one with fixed landscape configuration, and another with random configuration. The subsequent simulation of transect observation is realized for two categorical maps: the artificial one simulated by SIMMAP, and the real one classified from Landsat ETM+ multispectral imagery. The simulated samples were used to test four estimators. The methodology employed in this study provides a good starting point for practically implementing the quantification of variance estimates with LIS.
