Assessing remote sensing estimations for burn area and tree mortality

Abstract

Remote sensing tools will increase the ability of land managers to visually sample large areas more feasibly. This increase in applications of remote sensing such as UAV aerial LiDAR may require an assessment of algorithm accuracy while utilizing LiDAR data versus ground collected data to ensure these applications are appropriate. One such application included within this

study is the detection of trees utilizing the LidR package which allows a cost- effective and quick survey estimating trees contained, and providing their

estimated heights. The aim of this paper is to compare these detection results to a traditional ground tree stocking survey, exploring the viability of applying tree detection algorithms on post-burn forestry blocks to assess the surviving trees allowing an indication of future stocking allowing the forest manager to create a more accurate re-planting schedule. The results derived from this assessment deviated significantly from ground surveys with the aerial analysis providing an estimate of 2.20 WSP/ha and the ground survey estimating 70.18 WSP/ha (Well spaced stems per hectare) within block 525_19C. Although stocking results were inconclusive the analysis resulted in several useful outputs such as a combination of orthomosaic imagery alongside the tree detection points. These outputs resulted in an effective visual aid allowing a more detailed visualization of the spatial extent severe burns included within the forested blocks.