Date of Award

2018

Document Type

Open Access Master's Thesis

Degree Name

Master of Science in Forest Ecology and Management (MS)

Administrative Home Department

School of Forest Resources and Environmental Science

Advisor 1

Robert Froese

Committee Member 1

Curtis Edson

Committee Member 2

Michael Hyslop

Abstract

This study implemented LiDAR (Light Detection and Ranging) remote sensing technology and applied ITD (Individual Tree Detection) methods as an approach to estimate some essential tree variables, such as DBH (Diameter at Breast Height), height, volume, and biomass for Ford Forest Research Center in Upper Peninsula, Michigan. There were 34 deciduous (1 bigtooth aspen, 9 red oaks, 20 sugar maples, 2 white birches, and 2 yellow birches) and 17 coniferous (2 eastern hemlocks, 11 red pines, and 4 white pines) subject tree species. There were two different available LiDAR datasets from the same area that were collected in 2011 and 2017. Height measurements were done at 96% and 97% accuracy for hardwood and softwood tree species, respectively. Several other tree variables derived from LiDAR point cloud were used to estimate DBH by using regression analysis for both 2017 and 2011 datasets. Estimation equations were tested on the other dataset. The best-fitted formula was 2017’s, with 0.55 adjusted R² and less than 0.0001 p-values on 2017 LiDAR data while 0.42 adjusted R² and less than 0.0001 p-values on 2011’s dataset. Some additional analysis that includes calculating PRMSE (Predicted Root Mean Square Error), BIAS (Mean Error), and MAD (Mean Absolute Difference) have been applied. The equation that was generated by using data from 2017 has -0.57 BIAS for Hardwood and 1.13 BIAS for softwood. That result indicates that the equation has -0.57 centimeters (cm) estimation error for hardwood and 1.13 cm for softwood on DBH estimations.

Share

COinS