Stratified Light Detection and Ranging Double-Sample Forest Inventory

Abstract

Abstract An industrial application of a light detection and ranging (LiDAR) individual-tree, stratified double-sample forest inventory of approximately 18,000 ha of southeastern pine plantations was accomplished with an 9:1 ratio of 0.02-ha phase 1 LiDAR and phase 2 ground plots in ages 6 to 28 years. Phase 2 ground inventory data of tree dbh and sample tree heights for 2 trees per plot were used to obtain dbh-height relationships and volumes of standing trees. Phase 1 LiDAR data with 1.9 points per m2 were used to obtain ground–LiDAR height relationships for phase 2 matching LiDAR trees and phase 1 estimates of basal area and volume. A conventional ground inventory of 971 ground plots by private contractors applying standard company field specifications resulted in an overall sampling error of ±2.7% (α = 0.05) for a single-phase volume estimate and ±2.2% for the double-sample volume estimate. Sampling error was defined as one-half the 1-α confidence interval expressed as a percentage of the mean. Reducing the phase 2 ground sample to 15 plots per age class stratum achieved sampling errors of approximately ±15% for half the strata, with a combined error of ±3.9%. Adjusting the LiDAR-ground height bias of approximately 1.8 m resulted in more realistic volume estimates compared with the industry's continuing forest inventory volumes. The double-sample volume estimates were obtained at a cost of approximately $3.88/ha of timberland inventoried as compared with $1.67/ha for the conventional inventory.