Evaluating the Robustness of Plot Databases in Species-Specific Light Detection and Ranging-Based Forest Inventory

Abstract

In recent years, airborne laser scanning (also known as light detection and ranging [LiDAR]), in combination with digital aerial photography has been used to estimate plot-level forest characteristics of new sites. Forest characteristics are defined both as parameters derived without regard to species, total stand parameters, and species-specific stand parameters. The use of LiDAR has produced promising results, but its costs have been high, because the numbers of sample plots needed for model development and calibration are relatively high. Recently, the use of databases of sample plots from other formerly measured sites in the estimation of new site total stand parameters has been tested. Only a small number of sample plots were needed for acceptable results. In this study, the use of databases is extended to species-specific forest stand parameter estimation with LiDAR histograms and digital aerial photography. The data processing includes LiDAR histogram calibration and statistically tuned plot selection from the databases. The samples of the calibration set and databases are weighted to avoid bias in the estimates. Data from seven different sites are used for cross-validation of the given method. The estimates of species-specific parameters are quite accurate, although their accuracies fall short of those attained for total forest parameters. The use of plot databases reduces the variance in estimation error.