International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research Publications. http://www.ripublication.com Spectral Based Vegetation discrimination and Forest Health Assessment Using Hyperion (EO-1) in Yelagiri Hills,

Kumaresan P R Research Scholar, Centre for Remote Sensing, Bharathidasan University, India.

ORCID ID: 0000-0001-9138-1993

Abstract management is needed in today scenario to ensure sustainable utilization (Singh 2014).The Eastern Ghats forest ecosystem Remotely sensed information became milestone in the field of includes a rich assemblage of floral, faunal wealth including forestry and natural resource monitoring. Species level forest many endangered, endemic and medicinal species. Among mapping is an active research topic that aims to provide the them Yelagiri is one of the plant biodiversity rich hub. The systematic and updatable information necessary for hills are endowed with rich biodiversity of species like understanding and monitoring the rapid changing in the forest sandalwood, Ethno medicinal plants and etc. Plant species of environment. Hyperspectral remote sensing data Hyperion diverse families like Lilaceae, Solanaceae, Euphorbiaceae (EO‐1) was used for analysis. Image was pre-processed followed by Lamiaceae, apocyanaceae were found (Tarique followed by Minimum Noise Fraction (MNF) and Pixel Purity and Infam 2013). Index (PPI). The output of MNF and PPI has been analyzed by visualizing it in n-dimensional visualizer for endmember The remote sensing with the collaboration of geo informatics extraction and collection. Spectral discrimination of surface technology is a reliable method and a practical means of features like Vegetation, Grass, Barren land, Water and Man- estimating various biophysical and biochemical vegetation made structure was done. Further 6 different types of variables. Hyperspectral remote sensing data sets will provide Vegetation species among them were identified. Advanced narrow band combination with contiguous spectral signature per-pixel supervised classifications such as Spectral Angle based on molecular absorption which enables us to Mapper (SAM) has been done with Image collected discriminate species on the basis of their spectral response and endmember spectra as an input. Finally analyzed vegetation absorption pattern (Carter et al. 2009). The objective of this growth using the normalized difference vegetation index study is to investigate the potential of satellite hyper spectral (NDVI) method and 0.75 is the highest value shown in this imagery in forest vegetation mapping, spectral discrimination region. Relative forest heath has been found and classified and Heath assessment. into 9 classes in which higher value implies good health.

Keywords: Yelagiri Hills, Endmember extraction, Spectral STUDY AREA discrimination, Forest Health Assessment. Yelagiri hills is situated in the Jolarpet Panchayat Union of

Thirupattur taluk, District. The study area is INTRODUCTION represented in Fig. 1. The Maximum temperature of Yelagiri hills during summer (April) is of 27oC and the minimum Forests are the most diverse ecosystems contain millions of temperature in winter (December) is 11oC (Chandrasekaran et different species of plants, fungi, micro-organisms and al. 2014). It has comparative dry climate with low humidity of animals. Humans depend on forests for their survival and 45-50 % and considered to be Aw according to the Köppen- basic needs. Forests also offer watershed protection, prevent Geiger climate classification. The mean annual rainfall for soil erosion and mitigate climate change. Yet, despite of Yelagiri hills is 1026.16mm. Soil is derived from feldspar and dependence on forests, they are threatened by rapid hornblende. It has been observed that mineral resources such deforestation, fragmentation, degradation, hunting and the as Sulphides, Quartz, Hayte, Apatite and Vermiculite arrival of invasive species from other habitats. (Nagaraju et al 2014). The Average elevation of the area The accurate, regular and periodic monitoring of forest ranges from 300 to 1200 meter from Mean Sea Level (MSL) changes and its dynamics is essential to understanding the loss with its highest point culminating at swamimalai about 1339 of biodiversity and reduction of carbon sequestration capacity meters. Primarily resided by malayali tribal population. that results from deforestation. Species specific planning and

13826 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com

Figure 1. Study area

13827 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com

Figure 2 (a) False color composite and (b) True color composite 3D hyper cube

DATA ACQUISITION The study used data EO-1 Hyperion data of 4th March, 2004 (https://earthexplorer.usgs.gov/). EO-1 Hyperion is the satellite-borne hyper spectral sensor to orbit the Earth, capable of recording spectral information superior to previous satellite sensors. It has a spatial resolution of 30 meters and records radiance in 242 bands spanning from the blue (visible) at 450 nm to the middle infrared at 2500 nm. Each band has a width of approximately 10 nm. Out of it 198 bands are calibrated and rests are unused due to their low signal to noise ratio (Beck 2003). 198 channels provides detailed spectral mapping with high radiometric accuracy. Thus, there are 50 VNIR and 148 SWIR bands when the data is processed. True color composite (TCC) and False color composite (FCC) of the study area subset from full image was generated using RGB band 50, 23, and 16 having a spectral wavelength 854.18, 579.45, and 505.28 respectively for FCC and bands 22, 13 and 4 having spectral wavelength 640.5, 548.92, 47.34 Figure 3 Methodology respectively for TCC Fig. 2.

Necessary pre-processing techniques applied to radiance in METHODOLOGY order to obtain reflectance: to fix bad pixels using the bad Pre-processing, Atmospheric correction and Collection of pixel list process and further re-calibration is done to data by Endmembers tracking down the gain/off values using the ASCII file; the removal of the outliers was done using the median and mean The pre-processing of hyper spectral data is needed in order to absolute deviation (MAD) as statistics for the decisions and extract reliable and efficient information. Pre-processing is fixing out of range data. Atmospheric correction is done by crucial process because satellite based platform with modest using The Fast Line-of-sight Atmospheric Analysis of Hyper surface signal levels and a column of atmosphere diminishes cubes (FLAASH) tool by selecting the input radiance image the signal. Using ENVI 5.3 software, pre- processing, removal and to set the radiance scale factors. Then the calibrated hyper of errors, spectral classification by image collected spectral raw radiance data are scaled into 2-byte integers. To endmembers as an input, analysis of vegetation spectra and convert the 2-byte integer data into floating point radiance forest health assessment from Hyperion data. Tentative values input image is divided by the scale factors in units of methodology adopted for the current study is shown in Fig. 3. µW/cm2/nm/sr. For this current study, the gain values are 1000 as the data are radiance units (W/m2/lm/sr) multiplied by 100 Fig. 4.

13828 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com

Since we know the zones representing the different spectral composition from the United States Geological Survey (USGS), Johns Hopkins University (JHU) and Jet Propulsion Laboratory (JPL) spectral library as model spectra, the curve pattern and the absorption zone were also considered Fig. 5. This helped in identifying the different type of spectra in the study area. In the next step, data were classified using SAM. This method calculates the similarity of the two spectra by the angle between them, which is treated as the vectors in a space with dimensionally equal to its number of bands (Campbell 2009). The endmembers collected and derived from the image are used as an input for SAM classification. The angle between the endmember spectrum vector and each pixel, vector in n-dimensional euclidean space has been compared. In simple

concept, the smaller the angle, the closer and equivalent Figure 4 (a) False color composite shows spectra of location matches to the reference spectrum. SAM classification works point (b) Spectral Profile without FLAASH (c) Spectral on the reflectance data that are pre-processed and Profile with FLAASH atmospherically corrected. Hyperspectral imaging has a huge quantity of data with narrow contiguous bands. However, it is not required to use all these bands for identifying and differentiating the spectra Normalized difference vegetation index (NDVI) and Forest with in them because of its error in some bands (Folkman health assessment 2001). For the spectra based classification it is necessary to To determine the density of live green vegetation on a patch collect the endmembers from image. To attain this basic of land the NDVI is calculated from reflectance bands by procedure followed is using Minimum Noise Fraction (MNF) these individual measurements (Rouse et al. 1973; Souza et al. and Pixel Purity Index (PPI) techniques. To reducing the 2010) as follows. spectral dimensionality and lessen the noise in the data MNF transformation is applied. Later PPI technique has been NDVI = (R864 - R671)/ (R864 + R671) carried upon selected MNF bands (first few bands) to find the Highest absorption and reflectance regions of chlorophyll most spectrally pure pixels that can be used as an input make it robust over a wide range of conditions. The value of endmember and is a method to reduce spatial dimension. this index ranges from -1 to 1. Finally, results achieved from the above processes helped in the n-Dimensional visualization, including auto-clustering, The Forest health assessment (FHA) is done through Forest selecting, retrieving and collecting of distinct endmember. health vegetation analysis. Which could be attained by four simple procedure (1) Broadband and narrowband greenness, (2) Leaf pigments, to show the concentration of carotenoids Identification of different spectra and Classification and anthocyanin pigments for stress levels, (3) Canopy water content, to show the concentration of water and finally (4) Initially endmembers are saved as spectral library and this Light use efficiency, to show forest growth rate (Ceccato et al. step is required to identify and estimate the different types of 2001; Datt 1999; Merzlyak et al. 1999; Riley 1989; Sims and surface features present in the study area. The spectra profile Gamon 2002; White et al. 2007). It indicates unhealthy and of each sample area was collected. Basically five spectral dying vegetation in forest region. classes as Water body, Vegetation, Grass, Barren land, and Man- Made structure. RESULTS AND DISCUSSION Spectral discriminant analysis The yelagiri is one of the global biodiversity hotspots. Conservation and sustainable management of biodiversity in the yelagiri hills is a priority agenda. Collected different surface features among the region: Vegetation, Grass, Barren land, Water and Man-made structure shown in Fig 6 (Clark 1993). Further 6 different types of vegetation species are identified among them shown in Fig 7 (Thenkabail et al 2013; Somers 2013). SAM effective classification technique carried out to classify the Image based on Endmembers collected.19 Figure 5 Model spectra based on USGS, JPL and JHU types of endmembers were identified and classification done spectral library to entire subset image using SAM classification technique

13829 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com shown in Fig. 8. Entire image is well classified which is clearly able to identify from the classified output image. SAM algorithm had showing immense potential of species level vegetation mapping (Vyas et al 2011; Singh et al 2014)

Figure 6 Basic types of Surface features Spectral profile Figure 8 Spectral Angle Mapper based classification (a) with man-made structure and (b) Without man-made structure. Vegetation analysis: NDVI and Forest Health Assessment Forest vigor and health can be evaluated by using satellite‐ derived vegetation indices such as NDVI that can monitor large remote areas with an effective database (Souza et al 2010) higher index Value shows good growth and lower shows minimal Fig. 9. Highest value shown in this region is 7.58. Such subtle spectral information is able to describe the slight variation of forest thus it is beneficial to establish high quality prediction model (Pu et al 2008). FHA technique exhibits spatial map of the overall health and vigor of a forest. It is good at detecting pest and blight conditions in a forest as well as assessing areas of timber harvest. A forest under high stress conditions shows signs of dry or dying plant material, very dense or very sparse canopy, and inefficient light use. Classified into 9 classes in which 9 impels high health and 1 impels poor health (Sims and Gamon Figure 7 Different types of Vegetation spectra in Yelagiri 2002; Datt 1999). Results indicate most of the region is in hills moderate health comes under class 5 and some region (northern central) as good health and comes under class 9 Fig

13830 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com

10. Figure 10 Relative Forest Heath Assessment map of Yelagiri region CONCLUSION The study was carried out to know the ability of Hyperion data sets in vegetation mapping, spectral discrimination, spectral based image classification and heath assessment in forest region promising tool for future forest health monitoring. The purpose of the study was fulfilled to maximum extent and it is stated that such study can be carried out to any region, if any reference spectrum such as collected using field base instruments it becomes more valuable for validation and exact mapping. Findings of present study are encouraging for the discrimination of forest vegetation types and its health assessment using space borne hyper spectral data.

REFERENCES [1] Beck, R. (2003). EO-1 user guide, version 2.3. Satellite Systems Branch, USGS Earth Resources Observation Systems Data Center (EDC). [2] Carter, G. A., Lucas, K. L., Blossom, G. A., Lassitter, C. L., Holiday, D. M., Mooneyhan, D. S., ... & Griffith, J. A. (2009). Remote sensing and mapping of tamarisk along the Colorado river, USA: a comparative use of summer-acquired Hyperion, Thematic Mapper and QuickBird data. Remote Sensing, 1(3), 318-329. [3] Ceccato, P., Flasse, S., Tarantola, S., Jacquemoud, S., & Grégoire, J. M. (2001). Detecting vegetation Figure 9 Normalized Difference Vegetation Index (NDVI) leaf water content using reflectance in the optical Map domain. Remote sensing of environment, 77(1), 22- 33. [4] Chandrasekaran, A., Ravisankar, R., Senthilkumar, G., Thillaivelavan, K., Dhinakaran, B., Vijayagopal & Venkatraman, B. (2014). Spatial distribution and lifetime cancer risk due to gamma radioactivity in Yelagiri Hills, Tamilnadu, India. Egyptian Journal of basic and applied sciences, 1(1), 38-48. [5] Clark, R. N., Swayze, G. A., Gallagher, A. J., King, T. V., & Calvin, W. M. (1993). The US Geological Survey, Digital Spectral Library: Version 1 (0.2 to 3.0 um) (No. 93-592). Geological Survey (US). [6] Datt, B. (1999). A new reflectance index for remote sensing of chlorophyll content in higher plants: tests using Eucalyptus leaves. Journal of Plant Physiology, 154(1), 30-36. [7] Folkman, M. A., Pearlman, J., Liao, L. B., & Jarecke, P. J. (2001, February). EO-1/Hyperion hyperspectral imager design, development, characterization, and calibration. In Hyperspectral Remote Sensing of the Land and Atmosphere (Vol. 4151, pp. 40-52). International Society for Optics and Photonics.

[8] Merzlyak, M. N., Gitelson, A. A., Chivkunova, O. B., & Rakitin, V. Y. (1999). Non‐destructive optical

13831 International Journal of Applied Engineering Research ISSN 0973-4562 Volume 13, Number 18 (2018) pp. 13826-13832 © Research India Publications. http://www.ripublication.com

detection of pigment changes during leaf senescence Survey of Medicinal Plants in Yelagiri Hills of Tamil and fruit ripening. Physiologia plantarum, 106(1), Nadu. Research Journal of Pharmacy and 135-141. Technology, 6(6), 652-654. [9] Nagaraju1T, Parthiban S,Hari Nivas A, Siva [21] Thenkabail, P. S., Mariotto, I., Gumma, M. K., Sivasubramanian T and Mariappan, V. N. (2014) A Middleton, E. M., Landis, D. R., & Huemmrich, K. Study on the Spectral Annual Variability of Yealgiri F. (2013). Selection of hyperspectral narrowbands hills in International Journal of ChemTech Research (HNBs) and composition of hyperspectral twoband 6(4)2223-2331. vegetation indices (HVIs) for biophysical characterization and discrimination of crop types [10] Pantaleoni, E., Wynne, R. H., Galbraith, J. M., & using field reflectance and Hyperion/EO-1 data. Campbell, J. B. (2009). Mapping wetlands using IEEE Journal of Selected Topics in Applied Earth ASTER data: a comparison between classification Observations and Remote Sensing, 6(2), 427-439. trees and logistic regression. International Journal of Remote Sensing, 30(13), 3423-3440. [22] Vyas, D., Krishnayya, N. S. R., Manjunath, K. R., Ray, S. S., & Panigrahy, S. (2011). Evaluation of [11] Pu, R., Gong, P., & Yu, Q. (2008). Comparative classifiers for processing Hyperion (EO-1) data of analysis of EO-1 ALI and Hyperion, and Landsat tropical vegetation. International Journal of Applied ETM+ data for mapping forest crown closure and Earth Observation and Geoinformation, 13(2), 228- leaf area index. Sensors, 8(6), 3744-3766. 235. [12] Pu, R., Gong, P., & Yu, Q. (2008). Comparative [23] White, J. C., Coops, N. C., Hilker, T., Wulder, M. A., analysis of EO-1 ALI and Hyperion, and Landsat & Carroll, A. L. (2007). Detecting mountain pine ETM+ data for mapping forest crown closure and beetle red attack damage with EO‐1 Hyperion leaf area index. Sensors, 8(6), 3744-3766. moisture indices. International Journal of Remote [13] Riley, J. R. (1989). Remote sensing in entomology. Sensing, 28(10), 2111-2121. Annual Review of Entomology, 34(1), 247-271. [14] Rouse Jr, J., Haas, R. H., Schell, J. A., & Deering, D. W. (1974). Monitoring vegetation systems in the Great Plains with ERTS. [15] Sims, D. A., & Gamon, J. A. (2002). Relationships between leaf pigment content and spectral reflectance across a wide range of species, leaf structures and developmental stages. Remote sensing of environment, 81(2-3), 337-354. [16] Singh, D., Singh, S., Lekshmi, V. R., Dutta, S., Ali, M. S., & Ahmed, A. (2014). Species type and forest health assessment via hyperspectral remote sensing in the part of Himalayan range, India. Dhaka University Journal of Biological Sciences, 23(2), 135-146. [17] Somers, B., & Asner, G. P. (2013). Invasive species mapping in Hawaiian rainforests using multi- temporal Hyperion spaceborne imaging spectroscopy. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(2), 351-359. [18] Souza, A. A., Galvao, L. S., & Santos, J. R. (2010). Relationships between Hyperion-derived vegetation indices, biophysical parameters, and elevation data in a Brazilian savannah environment. Remote Sensing Letters, 1(1), 55-64. [19] Souza, A. A., Galvao, L. S., & Santos, J. R. (2010). Relationships between Hyperion-derived vegetation indices, biophysical parameters, and elevation data in a Brazilian savannah environment. Remote Sensing Letters, 1(1), 55-64. [20] Tariq, N. P. M., & Ifham, S. (2013). Ethnobotanical

13832