Analysis of the effect of leaf-on and leaf-off forest canopy conditions on LiDAR derived estimations of forest structural diversity

Abstract

UK legislation aims to conserve and enhance biological diversity within the UK and so accurate measurements of forest biodiversity are important to assess efficacy of management activities in this context. Forest structural diversity metrics can be used as indicators of biodiversity and airborne LiDAR data provide a means of producing these metrics. Forest structure metrics derived from LiDAR can be significantly affected by the canopy conditions the datasets are collected under. Existing studies have combined and compared leaf-on and leaf-off LiDAR datasets in existing analyses, however the majority of these utilise field sites where climate, species and terrain are very different to those found in the UK. Additionally, studies comparing leaf-on and leaf-off LiDAR over forested areas assess the changes in pulse penetration through the canopy and how this effects forest structure metrics and not the effect on modelled forest structure diversity. The novel aim of this research is to assess and compare the accuracy of forest structural diversity modelled from two LiDAR surveys collected under leaf-on and leaf-off conditions, and do so in a UK forest environment.
A robust methodology for correcting the absolute and relative accuracy between datasets was adopted and comparative analysis of ground detection and return height metrics (maximum, mean and percentiles of return height) and return height diversity (L-CV, CV, kurtosis, standard deviation, skewness and variance) was undertaken. Regression models describing the field tree size diversity measurements were constructed using diversity metrics from each LiDAR dataset in isolation and, where appropriate, a mixture of the two.
Both surveys were consistently effected by growth and differing survey parameters making the isolation and assessment of the effects of seasonal change difficult. Despite this, models created using diversity variables from both LiDAR datasets were generally very similar. Both leaf-on and leaf-off LiDAR dataset models described 65% of the variance in tree height diversity (R² 0.65, RMSE 0.05, p <0.0001), however models utilising leaf-off LiDAR diversity variables described DBH diversity, crown length diversity and crown width diversity more successfully than leaf-on (leaf-on models resulted in R² values of 0.68, 0.41 and 0.19 respectively and leaf-off models 0.71, 0.62 and 0.26 respectively). When diversity variables calculated from both LiDAR datasets were combined into one model to describe tree height diversity and DBH diversity their efficacy was increased (R² of 0.77 for tree height diversity and 0.72 for DBH diversity). The results suggest strongly that tree height diversity models derived from airborne LiDAR collected (and where appropriate combined) under any seasonal conditions can be used to differentiate between single and multiple storey UK forest structure with confidence. However, leaf-off LiDAR acquisitions can generate models with the ability to better explain the diversity of crown shapes in a forest stand than leaf-on, with no improvement in model performance when the two are combined.