Tropical Forests of Réunion Island Classified from Airborne Full-Waveform LiDAR Measurements

Xiaoxia Shang,Patrick Chazette,Olivier Flores,Eric Hamonou,Julien Totems,Elsa Dieudonné,Dominique Strasberg,Valentin Duflot,Pierre Tulet,Jacques Fournel

doi:10.3390/rs8010043

Abstract

From an unprecedented experiment using airborne measurements performed over the rich forests of Réunion Island, this paper aims to present a methodology for the classification of diverse tropical forest biomes as retrieved from vertical profiles measured using a full-waveform LiDAR. This objective is met through the retrieval of both the canopy height and the Leaf Area Index (LAI), obtained as an integral of the foliage profile. The campaign involved sites ranging from coastal to rain forest, including tropical montane cloud forest, as found on the Bélouve plateau. The mean values of estimated LAI retrieved from the apparent foliage profile are between ~5 and 8 m2/m2, and the mean canopy height values are ~15 m for both tropical montane cloud and rain forests. Good agreement is found between LiDAR- and MODIS-derived LAI for moderate LAI (~5 m2/m2), but the LAI retrieved from LiDAR is larger than MODIS on thick rain forest sites (~8 against ~6 m2/m2 from MODIS). Regarding the characterization of tropical forest biomes, we show that the rain and montane tropical forests can be well distinguished from planted forests by the use of the parameters directly retrieved from LiDAR measurements.

Highlights

Tropical forest areas are difficult to monitor/to classify using either remote sensing or in situ approaches, because of their tremendous heterogeneity and complex structure
Each parameter estimated from the airborne LiDAR measurements over the tropical forests of Réunion Island will be analyzed and discussed
Airborne LiDAR measurements conducted in May 2014 over several tropical forest sites of Island allow one to clearly identify the different types of coverage thanks to key parameters

Summary

Introduction

Tropical forest areas are difficult to monitor/to classify using either remote sensing or in situ approaches, because of their tremendous heterogeneity and complex structure. Forest horizontal patterns are accessible using passive multispectral sensors [1,2] and hyper-spectral sensors [3,4,5], but these sensors are not adequate to penetrate beyond the upper canopy layer [6]. Active remote sensing instruments, including LiDAR and radar, have more of a chance to peer through the forest canopy down to the ground level [7]. Radar yields volumetric scattering information in addition to surface scattering observations, but the retrieval of the vegetation vertical structure is not direct, unlike with LiDAR.

Objectives

Methods

Results

Conclusion