Mapping Savanna Tree Species at Ecosystem Scales Using Support Vector Machine Classification and BRDF Correction on Airborne Hyperspectral and LiDAR Data

Abstract

Abstract:Mapping the spatial distribution of plant species in savannas provides insight into the roles of competition, fire, herbivory, soils and climate in maintaining the biodiversity of these ecosystems. This study focuses on the challenges facing large-scale species mapping using a fusion of Light Detection and Ranging (LiDAR) and hyperspectral imagery. Here we build upon previous work on airborne species detection by using a two-stage support vector machine (SVM) classifier to first predict species from hyperspectral data at the pixel scale. Tree crowns are segmented from the lidar imagery such that crown-level information, such as maximum tree height, can then be combined with the pixel-level species probabilities to predict the species of each tree. An overall prediction accuracy of 76% was achieved for 15 species. We also show that bidirectional reflectance distribution (BRDF) effects caused by anisotropic scattering properties of savanna vegetation can result in flight line artifacts evident in species probability maps, yet these can be largely mitigated by applying a semi-empirical BRDF model to the hyperspectral data. We find that confronting these three challenges-reflectance anisotropy, integration of pixel-and crown-level data, and crown delineation over large areas-enables species mapping at ecosystem scales for monitoring biodiversity and ecosystem function.

Large-scale climatic and geophysical controls on the leaf economics spectrum

“…To improve aerosol corrections in ACORN-5, we iteratively ran the model with different visibilities until the reflectance at 420 nm (which is relatively constant for vegetated pixels) was 1%. Reflectance imagery was corrected for cross-track brightness gradients using a bidirectional reflectance distribution function modeling approach described by Colgan et al (66). The imaging spectrometer data were then orthorectified to the LiDAR digital canopy model.…”

Remote sensing for biodiversity monitoring: a review of methods for biodiversity indicator extraction and assessment of progress towards international targets

“…Similar to vegetation and wetland mapping, hyperspectral data perform high in species mapping and plant invasion applications during the last years (He et al 2011), especially in areas of high habitat and species diversity (Nagendra et al 2013), and are widely employed (Clark and Roberts 2012;Artigas and Pechmann 2010;Féret and Asner 2013;Colgan et al 2012;Hestir et al 2008;Baldeck et al 2014;Ghiyamat et al 2013;Pengra et al 2007;Miao et al 2006;Somers and Asner 2012). Thenkabail et al (2004) demonstrated the use of hyperspectral data by simulating the bands of Hyperion with a hand-held spectroradiometer to discriminate vegetation and agricultural crops.…”

“…finer-than the monitored object, to provide an effective tradeoff between within-object and between-object variance (Nagendra 2001). Some of the best performing studies in alien and invasive species detection are based on fine resolution data, either aerial (Dorigo et al 2012;Shouse et al 2013;Artigas and Pechmann 2010;Hantson et al 2012;Clark and Roberts 2012;Colgan et al 2012) or satellite (Laba et al 2008;Walsh et al 2008;Immitzer et al 2012). Dorigo et al (2012) extracted a bi-temporal band ratio (BTBR) and a number of Haralick texture features from bi-seasonal digital orthophotos and successfully detected Fallopia japonica, one of the world's worst invasive alien species, with up to 90.3% PA and 98.1% UA.…”

Mesoscale assessment of changes in tropical tree species richness across a bioclimatic gradient in Panama using airborne imaging spectroscopy

“…almost constant for vegetation) was 1%. The reflectance data were further corrected for cross-track brightness gradients using a bidirectional reflectance distribution function model (Colgan, Baldeck, Féret, & Asner, 2012). Full details on the preprocessing of the VSWIR data can be found in Asner et al (2014) and Colgan et al (2012).…”