Pravesh Debba

Improving Discrimination of Savanna Tree Species Through a Multiple-Endmember Spectral Angle Mapper Approach: Canopy-Level Analysis

Differences in within-species phenology and structure are controlled by genetic variation, as well as topography, edaphic properties, and climatic variables across the landscape, and present important challenges to species differentiation with remote sensing. The objectives of this paper are as follows: 1) to evaluate the classification performance of a multiple-endmember spectral angle mapper (SAM) classification approach in discriminating ten common African savanna tree species and 2) to compare the results with the traditional SAM classifier based on a single endmember per species. The canopy spectral reflectance of the tree species ( Acacia nigrescens, Combretum apiculatum , Combretum imberbe, Dichrostachys cinerea, Euclea natalensis, Gymnosporia buxifolia, Lonchocarpus capassa, Pterocarpus rotundifolius, Sclerocarya birrea, and Terminalia sericea) was extracted from airborne hyperspectral imagery that was acquired using the Carnegie Airborne Observatory system over Kruger National Park, South Africa, in May 2008. This study highlights three important phenomena: 1) Intraspecies spectral variability affected the discrimination of savanna tree species with the SAM classifier; 2) the effect of intraspecies spectral variability was minimized by adopting the multiple-endmember approach, e.g., the multiple-endmember approach produced a higher overall accuracy (mean of 54.5% for 20 bootstrapped replicates) when compared to the traditional SAM (mean overall accuracy = 20.5%); and 3) targeted band selection improved the classification of savanna tree species (the mean overall percent accuracy is 57% for 20 bootstrapped replicates). Higher overall classification accuracies were observed for evergreen trees than for deciduous trees.

Abundance Estimation of Spectrally Similar Minerals by Using Derivative Spectra in Simulated Annealing

This paper presents a method for estimating the partial abundance of spectrally similar minerals in complex mixtures. The method requires formulation of a linear function of individual spectra of individual minerals. The first and second derivatives of each of the different sets of mixed spectra and the individual spectra are determined. The error is minimized by means of simulated annealing. Experiments were made on several different mixtures of selected endmember, which could plausibly occur in real situations. The variance of the differences between the first derivatives of the observed spectrum and the first derivatives of the endmember spectra gives the most precise estimates for the partial abundance of each endmember. We conclude that the use of first-order derivatives provides a valuable contribution to unmixing procedures provided that the signal-to-noise ratio is at least 50 : 1

Potential utility of the spectral red-edge region of SumbandilaSat imagery for assessing indigenous forest structure and health

Indigenous forest degradation is regarded as one of the most important environmental issues facing Sub-Saharan Africa and South Africa in particular. We tested the utility of the unique band settings of the recently launched South African satellite, SumbandilaSat in characterising forest fragmentation in a fragile rural landscape in Dukuduku, northern KwaZulu-Natal. The AISA Eagle hyperspectral image was resampled to the band settings of SumbandilaSat and SPOT 5 (green, red and near infrared bands only) for comparison purposes. Variogram analysis and the red edge shift were used to quantify forest heterogeneity and stress levels, respectively. Results showed that the range values from variograms can quantify differences in spatial heterogeneity across landscapes. The study has also shown that the unique band settings of SumbandilaSat provide additional information for quantifying stress in vegetation as compared to SPOT image data. This is critical in light of the fact that stress levels in vegetation have previously been quantified using hyperspectral sensors, which are more expensive and do not cover large areas as compared to SumbandilaSat satellite. The study moves remote sensing a step closer to operational monitoring of indigenous forests.

Multiple endmember spectral-angle-mapper (sam) analysis improves discrimination of savanna tree species

Differences in within-species phenology and structure driven by factors including topography, edaphic properties, and climatic variables across the landscape present important challenges to species differentiation with remote sensing. The objective of this paper was to evaluate the classification performance of a multipleendmember spectral angle mapper (SAM) classification approach in discriminating seven common African savanna tree species and to compare the results with the traditional SAM classifier based on a single endmember per species or class. The leaf spectral reflectances of seven common tree species in the Kruger National Park, South Africa, Combretum apiculatum, Combretum hereroense, Combretum zeyheri, Gymnosporia buxifolia, Gymnosporia senegalensis, Lonchocarpus capassa and Terminalia sericea were used in this study. Discriminating species using all training spectra for each species as reference endmembers (i.e. the multiple endmember approach or more conventionally termed Knearest neighbour classifier) yielded a higher classification accuracy of 60% compared to the conventional SAM classifier based on the mean of the training spectra for each species (overall accuracy = 44%). Further analysis using endmembers selected after cluster analysis of all the spectra for each species yielded the highest classification accuracy for the species (overall accuracy = 74%). This study underscores two important phenomena; (i) within-species spectral variability affects the discrimination of savanna tree species with the SAM classifier and (ii) the effect of within-species spectral variability can be minimised by adopting a multiple endmember approach with the SAM classifier. This study further highlights the importance of the quality of the reference endmember or spectral library.

Spectral band discrimination for species observed from hyperspectral remote sensing

In vegetation spectroscopy, compositional information of leaves contained at band level or across the electromagnetic spectrum (EMS) and parts thereof, plays a huge rule in the analysis of spectra and their relations to the reflectance patterns across the spectrum. Spectral matching is often achieved by means of matching algorithms such as the Spectral Angle Mapper (SAM), Spectral information divergence (SID) and mixed measures of SAM and SID using either the tangent or the sine trigonometric functions, SID(TAN) or SID(SIN). The performance of these measures in distinguishing between objects of interest, such as species, is often compared using the relative spectral discriminatory probability (RSDPB). In this study, these measures are used to assess whether various sets of bands including the full spectrum, the visible (VIS), the near infrared (NIR), the shortwave infra-red (SWIR) region, as well as sets of bands identified by the stepwise discriminant analysis (SDA), can be used to discriminate the different species. This is done to identify the important regions of the EMS to distinguish seven common savannah tree species observed in the Kruger National Park, South Africas largest game reserve. The magnitude of variation of the species in any part of the spectrum can be linked to the importance of that spectral region in distinguishing the species. In addition, classification accuracy of these sets of bands was assessed and the SDA bands often gave better classification accuracy compared to using all bands, bands in the NIR, and SWIR parts of the EMS.

Spectral variability within species and its effects on Savanna tree species discrimination

Differences in within-species phenology and structure driven by factors including topography, edaphic properties, and climatic variables present important challenges for species differentiation with remote sensing in the Kruger National Park, South Africa. The objective of this study was to examine probable factors including intraspecies spectral variability and the spectral sample size that could affect remote sensing of Savanna tree species across a land-use gradient in the Kruger National park. Eighteen species were examined: Acacia gerradii, Acacia nigrescens, Combretum apiculatum, Combretum collinum, Combretum hereroense, Combretum imberbe, Combretum zeyheri, Dichrostachys cinerea, Euclea sp (E. divinurum and E. natalensis, Gymnosporia sp (G. buxifolia and G. senegalensis), Lonchocarpus capassa, Peltoforum africanum, Piliostigma thonningii, Pterocarpus rotundifolia, Sclerocarya birrea, Strychnos sp (S. madagascariensis, S. usambarensis), Terminalia sericea and Ziziphus mucronata. Discriminating species using the K-nearest neighbour (K = 1) classifier with spectral angle mapper (SAM) yielded a higher classification accuracy (48% overall accuracy) compared to 16% for the classification involving the mean spectra for each species as the training spectral set. Within-species spectral variability and the training sample size were identified as important factors affecting classification accuracy of the tree species. We recommend a non-parametric classifier such as K-nearest neighbour classifier for classifying and mapping tree species in a highly complex environment such as the savanna system of the Kruger National Park.

Field Sampling from a Segmented Image

This paper presents a statistical method for deriving the optimal prospective field sampling scheme on a remote sensing image to represent different categories in the field. The iterated conditional modes algorithm (ICM) is used for segmentation followed by simulated annealing within each category. Derived field sampling points are more intense in heterogenous segments. This method is applied to airborne hyperspectral data from an agricultural field. The optimized sampling scheme shows superiority to simple random sampling and rectangular grid sampling in estimating common vegetation indices and is thus more representative of the whole study area.

Are Earth Sciences lagging behind in data integration methodologies

This article reflects discussions German and South African Earth scientists, statisticians and risk analysts had on occasion of two bilateral workshops on Data Integration Technologies for Earth System Modelling and Resource Management. The workshops were held in October 2012 at Leipzig, Germany, and April 2013 at Pretoria, South Africa, and were attended by about 70 researchers, practitioners and data managers of both countries. Both events were arranged as part of the South African-German Year of Science 2012/2013. The South African National Research Foundation (NRF, UID 81579) has supported the two workshops as part of the South African–German Year of Science activities 2012/2013 established by the German Federal Ministry of Education and Research and the South African Department of Science and Technology.

Optimal individual supervised hyperspectral band selection distinguishing savannah trees at leaf level

This paper uses simulated annealing and focus on the spectral angle mapper (SAM), to demonstrate how the separability of two mean spectra from different species can be increased by choosing the bands that maximize the metric. It is known that classification performance is enhanced when the differences in mean spectra for each endmember species are maximized. Comparison was made using the selected bands derived from the proposed method, to all bands in the electromagnetic spectrum (EMS), only the bands in the visible, near infrared and short wave infrared regions of the EMS and selected bands using stepwise discriminant analysis. The bands from the proposed method often indicates a better choice of band selection as viewed by the summary statistics for (a) the SAM measurements, (b) the correlations between bands and (c) the spectral information divergence (SID), for each pair of species; and the classification accuracy of SAM and SID.

Within- and between-class variability of spectrally similar tree species

In this paper, a comparison is made through evaluating the within-and between-class species variability for the original, the first derivative and second derivative spectra. For each, the experiment was conducted (i) over the entire electromagnetic spectrum (EMS), (ii) the visible (VIS) region, (iii) the near infrared (NIR) region, (iv) the short wave infrared (SWIR) region, (v) using band selection, for example, best 10, 20, 30 and 65 bands selected, through linear step-wise discriminant analysis (vi) using sequential selection of bands, for example, every 5th, 9th, 15th, 19th or 25th band selected and (vii) spectral degradation of the spectral bands by averaging the reflectance values for every 5th, 9th, 15th, 19th or 25th band. We concluded that for this data set, there are important bands from the original spectra, the first and second derivative spectra and from various regions of the EMS (VIS, NIR, SWIR) that is important for species separability. Furthermore, there did not seem to be any decrease in species separability, for this data set, by degrading the spectral bands through averaging the reflectance. This implies that hyperspectral (high spectral) measurements did not prove useful in species separability compared to lower spectral resolution data.