Field characterization of olive (Olea europaea L.) tree crown architecture using terrestrial laser scanning data

Abstract

Since the introduction of Terrestrial Laser Scanning (TLS) instruments, there now exists a means of rapidly digitizing intricate structural details of vegetation canopies using Light Detection and Ranging (LiDAR) technology. In this investigation, Intelligent Laser Ranging and Imaging System (ILRIS-3D) data was acquired of individual tree crowns at olive (Olea europaea L.) plantations in Córdoba, Spain. In addition to conventional tripod-mounted ILRIS-3D scans, the unit was mounted on a platform (12 m above ground) to provide nadir (top–down) observations of the olive crowns. 24 structurally variable olive trees were selected for in-depth analysis. From the observed 3D laser pulse returns, quantitative retrievals of tree crown structure and foliage assemblage were obtained. Robust methodologies were developed to characterize diagnostic architectural parameters, such as tree height (r² = 0.97, rmse = 0.21 m), crown width (r² = 0.97, rmse = 0.13 m), crown height (r² = 0.86, rmse = 0.14 m), crown volume (r² = 0.99, rmse = 2.6 m³), and Plant Area Index (PAI) (r² = 0.76, rmse = 0.26 m²/m²). With the development of such LiDAR-based methodologies to describe vegetation architecture, the forestry, agriculture, and remote sensing communities are now faced with the possibility of replacing current labour-intensive inventory practices with, modern TLS systems. This research demonstrates that TLS systems can potentially be the new observational tool and benchmark for precise characterization of vegetation architecture for improved agricultural monitoring and management.

Introduction

Current methods for remote detection of plant physiology are inhibited by limitations in the explicit information about vegetation structure. The spatial architecture of plant material, within natural and plantation-like environments, plays a pivotal role in controlling functional activities like photosynthesis and evapotranspiration. As such, recent advancements have addressed this challenge using Light Detection And Ranging (LiDAR), an active remote sensing technology. Traditional remote sensing approaches “indirectly” determine plant architecture and physiology using data from passive optical imaging sensors. These methods rely on variability in vegetation spectral responses from the visible and near-infrared spectral regions. Widely accepted algorithms such as the Normalized Difference Vegetation Index (NDVI) have been empirically correlated to structural parameters such as canopy-level Leaf Area Index (LAI). Unlike passive optical imaging sensors, which are only capable of providing detailed measurements of vertically integrated horizontal distributions in vegetation canopies, LiDAR systems can yield highly specific information in both the horizontal and vertical (depth) dimensions (Lim et al., 2003). LiDAR-based instruments from airborne and ground-level platforms provide a “direct” means of measuring crown-level architecture, previously unattainable using passive remote sensing imagery. LiDAR units employ the Time-Of-Flight (TOF) principle or phase-based differences to measure the distances of objects based on the time interval between laser pulse exitance and return, upon backscattering from an object. The acquired LiDAR point cloud of returns yield a 3D digital representation of the vegetation environment in which each point is characterized by an XYZ coordinate. The challenge within the remote sensing community is to now develop robust methodologies that utilize such highly specific 3D point cloud data to directly retrieve canopy structural attributes thereby addressing the current scientific need for precise in situ measures of vegetation biophysical parameters (Maas et al., 2008). Applications of LiDAR systems from airborne platforms have characterized tree crown structure from both discrete return recordings (Coops et al., 2004, Donoghue et al., 2007, Jang et al., 2008, Lefsky et al., 1999, Thomas et al., 2006) and waveform recordings (Harding et al., 2001, Lefsky et al., 2002, Means et al., 1999, Patenaude et al., 2004). The backscattered laser pulses from Airborne Laser Scanning (ALS) systems have been used to extract various dimensional parameters, such as tree height (Andersen et al., 2006, Morsdorf et al., 2004, Hopkinson, 2007, Yu et al., 2004), crown dimension (Means et al., 2000, Popescu and Zhao, 2008), and crown volume (Hinsley et al., 2002, Riaño et al., 2004). Retrievals from discrete airborne LiDAR systems have the advantage of capturing information over a large area, but are constrained by their laser pulse return density (pts/m²). Multi-echo reading capabilities of ALS platforms produce somewhere between 3 and 20 backscattered pulses/m². Often this level of detail is insufficient to provide a detailed profile, especially in the vertical axis, of the tree crown that is required for precision 3D radiative transfer modeling. The ability to acquire laser pulse echoes from the bottom part of vegetation canopies is confounded by the ALS system properties (i.e. laser footprint size, recording frequency, etc.) as well as the organization of the crown elements themselves (i.e. closed vs. open canopies). As such, it is logical to introduce the use of LiDAR systems at ground level for much higher resolution laser pulse densities and thereby enabling detailed specification of canopy organization and individual tree crown characterization.

Ground-level, Terrestrial Laser Scanning (TLS), conventionally used for mining, urban planning and surveying applications, are now used to rapidly measure intricate structural details of vegetation canopies (Côte et al., 2009, Jupp et al., 2009, Lichti et al., 2002, Omasa et al., 2007, Rosell et al., 2009, Strahler et al., 2008). The unique perspective of such portable TLS systems allows the characterization of the vertical distribution of vegetation structure (Radtke and Bolstad, 2001), potentially replacing current labour-intensive, and manual field inventory practices. By mounting a laser ranging system on a pan/tilt platform Lovell et al. (2003) cross-validated laser-derived Leaf Area Index (LAI) estimates to those obtained by hemispherical photography to within 8% for mixed evergreen canopies in Australia. More recently, TLS data has been used to estimate other photosynthetically significant parameters, such as plant area densities (Takeda et al., 2008, Hosoi and Omasa, 2009), and the ratios of woody to total plant areas (Clawges et al., 2007). Additional research that focused on the measurement and segmentation of tree stem diameters and branching structures has also been conducted (Henning and Radtke, 2006, Hopkinson et al., 2004, Thies et al., 2004). Although the delineation of stem diameters is tree specific, the previous retrievals of LAI are spatially integrated for the entire canopy. Appropriate ground LiDAR data acquisitions also offer the potential of calculating LAI at the individual crown level provided that 3D point cloud data can be acquired for isolated crowns. Accordingly, this research is focused on estimating critical biophysical parameters that describe tree crown dimensional properties, as well as the foliar assemblage characteristics, using TLS data for individual tree crowns within an organized plantation.