Spectral discrimination of bulloak (Allocasuarina luehmannii) and associated woodland for habitat mapping of the endangered bulloak jewel butterfly (Hypochrysops piceata) in southern Queensland

Abstract

Abstract. The bulloak jewel butterfly (Hypochrysops piceata) is an endangered species due to a highly restricted distribution and complex life history, yet little is known of the availability of suitable habitat for future conservation. The aim of this study was to examine the potential of hyperspectral reflectance data for the discrimination of woodland species in support of bulloak jewel butterfly's habitat mapping. Sites from known butterfly sightings in Leyburn, Southern Queensland, Australia, were examined using hyperspectral scanning and vegetation species discrimination. Reflectance data of eight woodland vegetation species (Allocasuarina luehmannii, Eucalyptus crebra, Eucalyptus populnea, Callitris glauca, Corymbia maculata, Angophora leicarpa, Acacia sparsiflora, and Jacksonia scoparia) were collected at the leaf and canopy levels using a full-range (350 to 2500 nm) hand-held nonimaging spectroradiometer. Partial least square (PLS) regression was used to interpret the bulloak tree spectra against other vegetation species. The PLS results indicated high-prediction accuracies ranging from 78% to 95% and 52% to 5% for canopy and leaf levels, respectively. The highest spectral separability was observed at the near-infrared bands (approximately at 700 to 1355 nm), followed by selected ranges in the short-wave infrared band where separability peaked at 1670 and 2210 nm. The results confirmed the feasible use of hyperspectral sensing for discriminating vegetation species and its potential use for habitat mapping of the endangered bulloak jewel butterfly.

Accuracy of carrot yield forecasting using proximal hyperspectral and satellite multispectral data

“…The ASCII data format was imported into R software (R Core Team 2014) and converted into hyperspectral object (R class) with the hyper-Spect R-package to facilitate hyperspectral data manipulation and analysis (Beleites 2014). Pre-processing of spectral data included exploratory and cleaning processes to identify and remove outlier reflectance data as they tend to increase the error in statistical models (Wold et al 2001;Zainol Abdullah et al 2014). Other pre-processing techniques such as smoothing or normalization of reflectance data were omitted as these techniques can change the spectral characteristics of the data leading to inaccurate results (Suarez et al 2017;Vaiphasa 2006).…”