New Zealand Journal of Forestry (2015) 60(1): 38–43
©New Zealand Institute of Forestry

Professional Paper
Tree counts from airborne LiDAR

David Pont ¹, Mark Kimberley ², Rod Brownlie ³, Justin Morgenroth ⁴ and Michael S. Watt <sup>5

1</sup> Scientist, Scion, Rotorua
² Statistician, Scion, Rotorua
³ Technical Officer, Scion, Rotorua
⁴ Senior Lecturer, School of Forestry, University of Canterbury, Christchurch
⁵ Senior Scientist, Scion, Rotorua

The purpose of this report is to inform the New Zealand forest industry of test results for a new tree counting system, discuss the potential for inventory applications, and identify future developments. The new methodology utilises calibration counts, made either on the ground or on an image, to obtain accurate estimates of total tree counts for forest stands. The method also provides a measure of error, which has been used to evaluate test results. Accuracy of tree counts was evaluated on two image types for a number of stands of varying stockings and ages from two contrasting sites. The effect of two different image-processing operators and two different tree detection algorithms were also evaluated. Results showed that overall the tree count error with LiDAR images was 6% with the ground calibration method and 11% with the image calibration method. The increased error of image calibration method is due to operator subjectivity in image interpretation and LiDAR images were found to give better accuracy than orthophotos. There was no effect of stand age, stocking, image-processing operator or tree detection algorithm on accuracy, indicating the tree counting system is robust. The reduction in operator input of the new system has increased automation and accuracy. Future work will focus on improving tree counting accuracy and developing inventory applications.
(no keywords)