Modelling the effect of terrain variability in even-aged Eucalyptus species using LiDAR-derived DTM variables

Abstract

Accurate multi-source forest inventory attributes are necessary for estimating productivity and timber stock in commercial forest plantations. This study aims to uncover the effects of terrain variation on the growth of even aged Eucalyptus forest species using Light Detection and Ranging (LiDAR) topographical variables. Using 32 generated variables at 5 different spatial resolutions (1m, 3m, 5m, 7m, 9m), the random forest (RF) regression successfully revealed variations for structural attributes such as volume (Vol/ha), dominant tree height (HtD), mean tree height (Htm), and diameter breast heights (DBH). Results indicate that smaller spatial resolutions performed better for younger stands while larger resolutions produced the best results for mature stands. Using the multi-resolution approach results improved with variable selection. Incoming solar radiation and slope variables were among the most important terrain variables for modelling forest structural variability. The findings from this study demonstrates the value of stratifying forest productivity across the commercial forest landscapes of South Africa.