<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Publishing DTD v1.3 20210610//EN" "https://jats.nlm.nih.gov/publishing/1.3/JATS-journalpublishing1-3.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:ali="http://www.niso.org/schemas/ali/1.0/" article-type="research-article" dtd-version="1.3" xml:lang="en">
<front>
<journal-meta>
<journal-id journal-id-type="publisher-id">gcj1</journal-id>
<journal-title-group><journal-title>Remote Sensing of Land</journal-title></journal-title-group>
<issn publication-format="electronic">2582-3019</issn>
<publisher><publisher-name>Gatha Cognition</publisher-name></publisher>
</journal-meta>
<article-meta>
<article-id pub-id-type="doi">10.21523/gcj1.17010102</article-id>
<article-id pub-id-type="publisher-id">gca8</article-id>
<article-categories><subj-group subj-group-type="heading"><subject>Original Research Paper</subject></subj-group></article-categories>
<title-group><article-title>Improved Change Detection of Forests Using Landsat TM and ETM plus data</article-title></title-group>
<contrib-group>
<contrib contrib-type="author"><contrib-id contrib-id-type="orcid" authenticated="false">https://orcid.org/0000-0002-8990-9880</contrib-id><name><surname>Sonawane</surname><given-names>Kishor R.</given-names></name><xref ref-type="aff" rid="aff1"/></contrib>
<contrib contrib-type="author"><contrib-id contrib-id-type="orcid" authenticated="false">https://orcid.org/0009-0001-5405-2020</contrib-id><name><surname>Bhagat</surname><given-names>Vijay</given-names></name><xref ref-type="aff" rid="aff1"/></contrib>
</contrib-group>
<aff id="aff1"><institution>Post-graduate Research Centre in Geography, Agasti Arts, Commerce and Dadasaheb Rupwate Science College, Akole-422601, Ahmednagar, Maharashtra (India).</institution></aff>
<pub-date publication-format="electronic" date-type="pub"><day>18</day><month>08</month><year>2017</year></pub-date>
<volume>1</volume>
<issue>1</issue>
<fpage>18</fpage><lpage>40</lpage>
<history><date date-type="received"><day>15</day><month>11</month><year>2016</year></date><date date-type="rev-recd"><day>10</day><month>02</month><year>2017</year></date><date date-type="accepted"><day>02</day><month>02</month><year>2017</year></date></history>
<permissions><copyright-year>2017</copyright-year><license license-type="open-access" xlink:href="https://creativecommons.org/licenses/by-nc-nd/4.0/"><ali:license_ref>https://creativecommons.org/licenses/by-nc-nd/4.0/</ali:license_ref><license-p>This is an open access article distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0).</license-p></license></permissions>
<self-uri xlink:href="https://gathacognition.com/article/gca8/improved-change-detection-of-forests-using-landsat-tm-and-etm-plus-data"/>
<self-uri content-type="pdf" xlink:href="https://gathacognition.com/uploads/admin/articlefiles/GCJ1M71_KishorR1.pdf"/>
<abstract>
<p>Landsat TM and ETM+ datasets are useful for forest change detection (FCD) at good accuracy level. Classified forest maps have been prepared using NDVI calculated from Landsat-5 TM (2009) and Landsat-7 ETM+ (2002) datasets for FCD using post-classification technique. About 58.59% of reviewed area shows positive changes, 33.69% no-changes and 7.72% negative changes with 77.84% accuracy. This accuracy insists limitations of present FCD analysis. Therefore, improved post-classification technique was formulated for precise FCD using field data and statistical techniques. Information about stable land surface (water bodies, rocky lands, deep forests, etc.) was used for normalisation of exaggerated reflectance in vegetation indices i.e. greenness. About 70.08% land estimated using second approach shows stable vegetation, 23.59% positive changes and 6.33% negative changes. Higher accuracy (95.21%) itself shows improvement in FCD technique and efficient applicability for sustainable land management.</p>
</abstract>
<kwd-group kwd-group-type="author"><kwd>Landsat-5 TM</kwd><kwd>Landsat-7 ETM+</kwd><kwd>Normalized Difference Vegetation Index</kwd><kwd>LST</kwd><kwd>Tasseled Cap Coefficient</kwd><kwd>Forest Change Detection</kwd><kwd>Stable Land Objects</kwd></kwd-group>
</article-meta>
</front>
<body>
<sec id="s1"><title>Introduction</title>
<p>A forest is a community of living organisms which interact mutually with the physical environment. Forest covers approximately 30% (<xref ref-type="bibr" rid="R162">Carlowicz, 2012</xref>) to 31% (<xref ref-type="bibr" rid="R185">FAO, 2015</xref>) of the earth surface with complexities and self-regenerating capacities (<xref ref-type="bibr" rid="R155">Bauer et al., 1994</xref>; <xref ref-type="bibr" rid="R243">Virk and King, 2006</xref>; <xref ref-type="bibr" rid="R197">Healey <italic>et al.</italic>, 2008</xref>). It is a source of organic carbon, which helps to maintain planetary climate, freshwater, biodiversity and useful to manage hazards like soil erosion, landslides, floods, etc. They are habitat of wildlife and regulate different cycles including hydrological, nutrient, atmospheric, etc. and conserve soils, water, etc. (<xref ref-type="bibr" rid="R238">Southworth, 2004</xref>).</p>
<p>Land under natural forest is estimated about 33.36 million <ext-link ext-link-type="uri" xlink:href="http://www.forestfacts.org/glossary/jkl/km2.htm">km²</ext-link> (World Resources Institute [WRI]) to 39.88 million <ext-link ext-link-type="uri" xlink:href="http://www.forestfacts.org/glossary/jkl/km2.htm">km²</ext-link> (World Conservation Monitoring Centre [WCMC]) excluding marine forests (<xref ref-type="bibr" rid="R214">Mishra <italic>et al.</italic>, 2003;</xref> <xref ref-type="bibr" rid="R208">Kumsap <italic>et al.</italic>, 2005</xref>; <xref ref-type="bibr" rid="R183">Fettig <italic>et al.</italic>, 2007</xref>; <xref ref-type="bibr" rid="R186">Forkuo and Frimpong, 2012</xref>). However, from last some decades forest destruction and land appropriation increase, globally (<xref ref-type="bibr" rid="R242">Turker and Derenyi, 2000</xref>). Land under forest in India declined to 21.31% of Total Geographical Area (TGA) (ISFR, 2009 and FSI 2015) due to imbalance climatic conditions, soil degradation, deforestation, desertification and water stresses in drought conditions. India endowed with an immense variety of forest resources (<xref ref-type="bibr" rid="R238">Southworth, 2004</xref>). However, adverse changes in ecosystem are taking place with continuous pressures of an exploding population and the subsequent domestic needs including food, fuel, fodder, timber as well as industrial demands (<xref ref-type="bibr" rid="R205">Keenan <italic>et al.</italic>, 1999</xref>). There are significant losses of forests at an alarming rate (<xref ref-type="bibr" rid="R225">Pant <italic>et al.</italic>, 2000</xref>; <xref ref-type="bibr" rid="R196">Hayes and Sader, 2001</xref>). The ‘hotspots’ are identified and declared for tropical biodiversity in Western Ghats. These regions are house of rich biodiversity and globally endemic species (<xref ref-type="bibr" rid="R180">Fang and Xu, 2000</xref>). However, it is widely believed that the natural vegetation in this tropical region is losing the biodiversity at unprecedented rates (<xref ref-type="bibr" rid="R224">Panigrahy <italic>et al.</italic>, 2010</xref>). Around, 275 million rural people (27%) in India are depends on forests for their subsistence and livelihoods (<xref ref-type="bibr" rid="R207">Kim <italic>et al.</italic>, 2011</xref>), earning from trade of fuel wood, fodder, bamboo and minor forest products. It is notable that 17% of rural population in India depends on forests to meet their domestic energy (<xref ref-type="bibr" rid="R177">Drescher and Perera, 2010</xref>).</p>
<p>Governmental and non-governmental agencies are involved in conservation of forests from last some decades (<xref ref-type="bibr" rid="R171">Cohen <italic>et al.</italic>, 1998</xref>). Planning and management for conservation of forest demands reliable information, investigation and analyses (<xref ref-type="bibr" rid="R171">Cohen <italic>et al.,</italic> 1998</xref>; <xref ref-type="bibr" rid="R194">Hansen <italic>et al.</italic>, 2000</xref>, <xref ref-type="bibr" rid="R196">Hayes and Sader, 2001</xref>; <xref ref-type="bibr" rid="R158">Bhagat, 2009</xref>). Many analysts have reported sophisticated techniques like remote sensing (<xref ref-type="bibr" rid="R206">Kennedy <italic>et al.,</italic> 2009</xref>), geographic information system (<xref ref-type="bibr" rid="R197">Healey <italic>et al.,</italic> 2008</xref>), global positioning system along with mathematical and statistical methods (<xref ref-type="bibr" rid="R183">Fettig <italic>et al.,</italic> 2007</xref>) for change detection. Estimations of forests cover using satellite data can give satisfactorily results (<xref ref-type="bibr" rid="R161">Cano <italic>et al.,</italic> 2006</xref>). Furthermore, change detection analysis can provide analytical data about forest degradation and conservation (<xref ref-type="bibr" rid="R227">Pouliot <italic>et al.,</italic> 2002</xref>). However, reliable change detection techniques for forests using remote sensing data remains a challenging task (<xref ref-type="bibr" rid="R174">Coppin <italic>et al.,</italic> 2004</xref>; <xref ref-type="bibr" rid="R200">Im <italic>et al.</italic>,2008</xref>; <xref ref-type="bibr" rid="R232">Schwilch <italic>et al.,</italic> 2011</xref>; <xref ref-type="bibr" rid="R237">Sommer <italic>et al.,</italic> 2011</xref>; <xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). The results are susceptible to data quality, technical efficiency and suitability of selected techniques (<xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). Therefore, modified change detection technique designed for this study based on field checks and statistical analyses to get more precise analysis about forest changes (<xref ref-type="bibr" rid="R238">Southworth, 2004</xref>). The Landsat-5 TM and Landsat-7 ETM+ datasets have been used for detection of changes in forest cover in the study area. Two approaches have been adopted for this analysis: Approach I) post-classification technique and Approach II) improved post-classification technique (<xref ref-type="bibr" rid="R165">Chandio and Matori, 2011</xref>). Changes in forest were detected using post-classification techniques using NDVI (<xref ref-type="bibr" rid="R183">Fettig <italic>et al.,</italic> 2007</xref>) calculated for Landsat-5 TM and Landsat-7 ETM+ datasets, whereas normalized indices and coefficients were used for improved approach of change detection.</p>
<p>Change detection analysis provides a thematic views to understand the natural and artificial behaviour of changes in land (<xref ref-type="bibr" rid="R237">Sommer <italic>et al.,</italic> 2011</xref>) including 1) increase and decrease in area, 2) seasonal changes in forests, snow cover, coastline, ocean water, 3) mapping of floods, landslides, volcanic eruptions, corral rifts, wild animals, birds, 4) changes in near surface atmospheric conditions like temperature, snowfall, rainfall, clouds, fog, storms, and 5) human activities like military actions, observation, planning and management for war areas, coal mining, etc. (<xref ref-type="bibr" rid="R212">Lunetta <italic>et al.,</italic> 2006</xref>). Scientists have used different methods of Digital Change Detection (DCD) including classification of multiband satellite data based on image ratio, tasseled cap coefficient, spectral vegetation indices, principal component analysis, change vector methods, threshold based classification, post-classification comparison, univariate image differencing, simultaneous analysis of multi temporal data, fusion approach, spectral classification, algebraic methods, regressions, etc. (<xref ref-type="bibr" rid="R226">Petit and Lambin, 2001</xref>; <xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). However, all methods are not ideally suitable, reliable and applicable to all surface change conditions (<xref ref-type="bibr" rid="R178">Du <italic>et al.</italic>, 2002</xref>; <xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). Therefore, corrective measure has been suggested to achieve more precise results of forest change detection in this study. Correlation techniques have been used to find the suitable parameter for corrections from different multiband ratios <xref ref-type="bibr" rid="R174">(Coppin and Bauer, 1996)</xref> and spectral indices (<xref ref-type="bibr" rid="R199">Huete <italic>et al.,</italic> 2002</xref>). Here, correlation and regression (<xref ref-type="bibr" rid="R216">Mountrakis <italic>et al.,</italic> 2010</xref>) techniques have been used to determine variables for detection, estimations of exaggerations and corrections using information collected for stable land surface (water bodies, rocky lands, deep forest, etc.). The suggested technique for forest change detection can be useful for land management and especially forests.</p>
</sec>
<sec id="s2"><title>Study Area</title>
<p>The study area (13859 ha) is a mountainous range between Pravara and Mula river basin from Western (Ghatghar) to Eastern borders (Washere) of Akole tahsil in Ahmednagar District (India) (<xref ref-type="fig" rid="F33">Figure 1</xref>). The altitude varies from 560 to 1646 m Mean Sea Level (MSL). Geologically, this area is formed by basaltic rock (<xref ref-type="bibr" rid="R187">Gareeau <italic>et al.</italic>, 2009</xref>) which prevents water to percolate. The depth and water-holding capacity of soils in the region (<xref ref-type="bibr" rid="R249">Zolekar and Bhagat, 2015</xref>) varies according to variations in slopes. The soils are very shallow at hilltop and depth increasing to foothill zones (<xref ref-type="bibr" rid="R208">Kumsap <italic>et al.,</italic> 2005</xref>). Very shallow loamy, shallow clayey soils are found on the moderate (1°- 3°) and stiff (3°- 6°) slopes. Soil moisture impacts on the distribution of vegetation cover (<xref ref-type="bibr" rid="R214">Mishra <italic>et al.,</italic> 2003</xref>). The forest cover varies with height, slopes, soil qualities, rainfall, etc. Foothill zones in Western parts show dense forest than hilltops with thin soils. Annual rainfall varies from 4937 mm at West and 1904 mm at East (<xref ref-type="bibr" rid="R249">Zolekar and Bhagat, 2015</xref>). The mean annual maximum and minimum temperatures are 39.80°C and 8.70°C, respectively. Indigenous people are engaged in agricultural activities (<xref ref-type="bibr" rid="R240">Theiler and Perkins, 2011</xref>) on land reclaimed from forests and dependent on forests for domestic needs.</p>
<fig id="F33"><label>Figure 1</label><caption><title>Study area: location map</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/95b520bfff9705385a9ce1fe187524d0.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
</sec>
<sec id="s3"><title>Dataset And Software</title>
<p>The analyses are based on remotely sensed data (<xref ref-type="bibr" rid="R233">Shalaby <italic>et al.,</italic> 2006</xref>), statistical models (<xref ref-type="bibr" rid="R193">Golmehr, 2008</xref>) and field check information. The satellite data of Landsat-7 ETM+ (06<sup>th</sup> Nov. 2002) and Landsat-5 TM (1<sup>st</sup> Nov. 2009) (accessed on 01 January 2012) has been used for detection of changes in forest cover. Field check data was collected for verification of inferences. The remotely sensed dataset and data obtained in field checks were compiled, merged and loaded in the GIS image processing softwares, ILWIS v3.4 Academic, ERDAS Imagine v9.2 (© Hexagon) and ArcGIS v9.3 (© ESRI Ltd.). Garmin’s Global Positioning System (GPS) was used for ground verification. Correlation analyses was performed using ‘Karl Pearson techniques’ (<xref ref-type="bibr" rid="R167">Chen <italic>et al.,</italic> 2001;</xref> <xref ref-type="bibr" rid="R245">Yang <italic>et al.</italic>, 2014</xref>) available in Statistical Packages for the Social Sciences (SPSS © IBM) (<xref ref-type="bibr" rid="R248">Zhang <italic>et al.,</italic> 2002;</xref> <xref ref-type="bibr" rid="R181">Fastring and Griffith, 2009</xref>).</p>
</sec>
<sec id="s4"><title>Methodology And Approaches</title>
<p>Two approaches were used for detection of changes in forest cover: Approach I - post classification technique and Approach II - improved post-classification technique. Accuracy assessment was performed and results of analyses compared to check the applicability of the approaches. Statistical techniques were useful to inculcate robustness in the analysis for change detection (<xref ref-type="bibr" rid="R240">Theiler and Perkins, 2011</xref>).</p>
<sec id="s4-1"><title>4.1 Co-registration</title>
<p>Matching of multiple images captured at different time is critical task for change detection studies. Mis-registration of images drops accuracy in results obtained in change detection analysis (<xref ref-type="bibr" rid="R160">Burnicki <italic>et al.,</italic> 2010</xref>; <xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>; <xref ref-type="bibr" rid="R223">Pajares <italic>et al.,</italic> 2012</xref>). Authors such as <xref ref-type="bibr" rid="R190">Giri Babu <italic>et al.,</italic> (2014)</xref> and <xref ref-type="bibr" rid="R241">Tsai and Lin (2007)</xref> have suggested automatic tie point registration, pixel to pixel matching, ground control point registration, semi-automotive registration, etc. to obtained reliable results (<xref ref-type="bibr" rid="R170">Clifton 2003; Julien <italic>et al.,</italic> 2011</xref>). Therefore, Landsat-7 ETM+ (t<sub>1</sub>) (2002) and Landsat-5 TM (t<sub>2</sub>) (2009) scenes were co-registered with the help of: 1) ground truth information (<xref ref-type="bibr" rid="R195">Hara <italic>et al.,</italic> 2012</xref>) collected using Garmin GPS device and 2) topo-maps (Survey of India) with pixel to pixel match with 0.00012 Root Mean Square Error (RMSE) in ERDAS Imagine software.</p>
</sec>
<sec id="s4-2"><title>4.2 Image Enhancement</title>
<p>Scientific understanding of behaviour and functioning of vegetation landscape, using remotely sensed data with different spectral, radiometric, temporal and spatial resolutions, have been important task in last some decades (<xref ref-type="bibr" rid="R210">Lippitt <italic>et al.,</italic> 2011</xref>; <xref ref-type="bibr" rid="R232">Schwilch <italic>et al.,</italic> 2011</xref>). Enhancement of images selected for analysis is highly required before further applications like classifications, estimations, etc. (<xref ref-type="bibr" rid="R169">Chowdhury <italic>et al.,</italic> 2005</xref>; <xref ref-type="bibr" rid="R244">Weng <italic>et al.,</italic> 2009</xref>; <xref ref-type="bibr" rid="R179">Ehlers <italic>et al.,</italic> 2010</xref>). Vegetation indices like NDVI, LAI, TVI, etc. (<xref ref-type="bibr" rid="R234">Silleos <italic>et al.,</italic> 2006</xref>) have been calculated using different bands e.g. red, infrared, thermal infrared, middle infrared and widely used for detection of land objects. Furthermore, coefficients like Soil Wetness Index (SWI), Normalized Difference Salinity Index (NDSI), Brightness, etc. have been used to study the distribution of soil moisture, soil salinity, etc. Spectral indices viz. NDVI, LAI and Land Surface Temperature Index (LSTI) and Tasseled Cap Coefficient Transformation have been calculated for enhancement of satellite images in present study.</p>
</sec>
<sec id="s4-3"><title><italic>4.2.1 </italic><italic>Spectral Indices </italic></title>
<p>In present study, Difference Vegetation Index (DVI), Ratio Vegetation Index (RVI), NDVI, Soil Adjusted Vegetation Index (SAVI), Leaf Area Index<bold> (</bold>LAI), Ratio Difference Vegetation Index (RDVI), Modified SAVI (MSAVI), Infrared Percentage Vegetation Index (IPVI) and Modified Simple Ratio (MSR) (<xref ref-type="bibr" rid="R217">Mróz and Sobieraj, 2004</xref>) have been used for change detection analysis. However, NDVI has been widely used for detection of changes in forest cover (<xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). Therefore, NDVI (equation 1) has been calculated using Near Infrared (Band 4) and Red (Band 3) bands of Landsat-7 ETM+ and Landsat-5 TM sensors (<xref ref-type="bibr" rid="R156">Bhagat and Sonawane, 2010</xref>).</p>
<p>\(NDVI = {NIR-RED \over NIR+RED}\) (1)</p>
<p>Calculated NDVI values vary between -1 to +1 depending on relative Digital Number (DN) of Near Infrared and Red bands. Maximum NDVI (<xref ref-type="fig" rid="F34">Figure 2</xref>) indicates dense vegetation and minimum values represent less or absence of vegetation.</p>
<fig id="F34"><label>Figure 2</label><caption><title>Normalized Difference Vegetation Index (NDVI)</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/6e62e57dc30ac3b56143108d369ba968.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>Leaf Area Index (LAI) [m<sup>2</sup>/m<sup>2</sup>] represents the amount of leaf material in an ecosystem and is geometrically defined as the total one-sided area of photosynthetic tissue per unit ground surface area (<xref ref-type="bibr" rid="R159">Breda, 2003</xref>; <xref ref-type="bibr" rid="R203">Jonckheere <italic>et al.,</italic> 2004</xref>; <xref ref-type="bibr" rid="R215">Morisette <italic>et al.,</italic> 2006</xref><bold>)</bold>. It appears as a key variable in many models describing vegetation-atmosphere interactions, particularly with respect to the carbon and water cycles (<xref ref-type="bibr" rid="R191">GCOS, 2010</xref>). LAI was calculated using following equation after <xref ref-type="bibr" rid="R201">Jensen (2002)</xref> (equation 2):</p>
<p>\(LAI = -2.42 +12.18 *(NDVI)\) (2)</p>
<p>Where, -2.42 and 12.18 are the constants (<xref ref-type="bibr" rid="R201">Jensen, 2002</xref>). LAI is widely used to identify and delineate light interception, gross productivity, soil moisture and transpiration from vegetation (Chen <italic>et al.,</italic> 2012). Therefore, LAI were calculated for statistical analysis to test applicability for Forest Change Detection (FCD) in second approach. Furthermore, many scholars have been used Land Surface Temperature Index (LSTI) for vegetation analysis (<xref ref-type="bibr" rid="R247">Yuan and Bauer, 2007</xref>). LSTI is measurements of earth surface temperature including canopy, soils, barren lands, rocks, water bodies, snow, ice, roof of a building using satellite data. The DN recorded for satellite images (<xref ref-type="bibr" rid="R204">Julien <italic>et al.,</italic> 2011</xref>) were converted (equation 3) using space reaching radiance i.e. Top of Atmosphere (ToA) Radiance (equation 4) (<xref ref-type="bibr" rid="R164">Chander and Markham, 2003</xref>).</p>
<p>\(LST(T) = {{K_2} \over in({k_1\over L_\lambda}+1)}\) (3)</p>
<p>Here, \(T\) is an effective at-satellite temperature in k (kelvin), \(L_\lambda\) is a spectral radiance in W/(m<sup>2</sup> in µm), \(K_1\) and \(K_2\) are pre-launch calibration constants for ETM+ and TM sansors.</p>
<p>\(L_\lambda = ({{L_{max}-L_{min}} \over QCAL_{max}-QCAL_{min}})(DN-QCAL_{min})+L_{min}\) (4)</p>
<p>Where, \(L_{max}\) is a maximum spectral radiance (W/m<sup>2</sup> in µm) at <italic>QCAL</italic> equal 0 DN, \(L_{mim}\) is a minimum spectral radiance (W/m<sup>2</sup> in µm) at <italic>QCAL</italic> equal 255 DN and <italic>QCAL</italic> are the quantized calibrated pixel values in DN.</p>
</sec>
<sec id="s4-4"><title><italic>4.2.2</italic> <italic>Tasseled Cap Coefficient Transformations (TCCTs)</italic></title>
<p>TCCTs have potentials to derive forest attributes for different regional applications where, atmospheric noise correction not possible (<xref ref-type="bibr" rid="R234">Silleos <italic>et al.,</italic> 2006</xref>; <xref ref-type="bibr" rid="R188">Ghosh <italic>et al.,</italic> 2010</xref>). These transformations simply reduce the number of radiance noise density and provide high association in single response. Therefore, brightness, greenness, wetness, fourth, fifth and sixth indices were calculated using coefficients estimated by <xref ref-type="bibr" rid="R175">Crist <italic>et al.,</italic> (1986)</xref> for Landsat-5 TM (<xref ref-type="table" rid="T17">Table 1</xref>) and <xref ref-type="bibr" rid="R198">Huang <italic>et al.,</italic> (2002)</xref> (<xref ref-type="table" rid="T18">Table 2</xref>) for Landsat-7 ETM+.</p>
<table-wrap id="T17"><label>Table 1</label><caption><title>Landsat-5 TM: tasselled cap coefficients at satellite reflectance</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Index</bold></td><td><bold>Band1</bold></td><td><bold>Band2</bold></td><td><bold>Band3</bold></td><td><bold>Band4</bold></td><td><bold>Band5</bold></td><td><bold>Band7</bold></td></tr>
<tr><td>Brightness</td><td>0.2909</td><td>0.2493</td><td>0.4806</td><td>0.5568</td><td>0.4438</td><td>0.1706</td></tr>
<tr><td>Greenness</td><td>-0.2728</td><td>-0.2174</td><td>-0.5508</td><td>0.7221</td><td>0.0733</td><td>-0.1648</td></tr>
<tr><td>Wetness</td><td>0.1446</td><td>0.1761</td><td>0.3322</td><td>0.3396</td><td>-0.6210</td><td>-0.4186</td></tr>
<tr><td>Fourth</td><td>0.8461</td><td>-0.0731</td><td>-0.4640</td><td>-0.0032</td><td>-0.0492</td><td>0.0119</td></tr>
<tr><td>Fifth</td><td>0.0549</td><td>-0.0232</td><td>0.0339</td><td>-0.1937</td><td>0.4162</td><td>-0.7823</td></tr>
<tr><td>Sixth</td><td>0.1186</td><td>-0.8069</td><td>0.4094</td><td>0.0571</td><td>-0.0228</td><td>0.0220</td></tr>
<tr><td colspan="7">Source: <xref ref-type="bibr" rid="R175">Crist <italic>et al.,</italic> 1986</xref></td></tr>
</tbody>
</table>
</table-wrap>
<table-wrap id="T18"><label>Table 2</label><caption><title>Landsat-7 ETM+: tasselled cap coefficients at satellite reflectance</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Index</bold></td><td><bold>Band1</bold></td><td><bold>Band2</bold></td><td><bold>Band3</bold></td><td><bold>Band4</bold></td><td><bold>Band5</bold></td><td><bold>Band7</bold></td></tr>
<tr><td>Brightness</td><td>0.3561</td><td>0.3972</td><td>0.3904</td><td>0.6966</td><td>0.2286</td><td>0.1596</td></tr>
<tr><td>Greenness</td><td>-0.3344</td><td>-0.3544</td><td>-0.4556</td><td>0.6966</td><td>-0.0242</td><td>-0.2630</td></tr>
<tr><td>Wetness</td><td>0.2626</td><td>0.2141</td><td>0.0926</td><td>0.0656</td><td>-0.7629</td><td>-0.5388</td></tr>
<tr><td>Fourth</td><td>0.0805</td><td>-0.0498</td><td>0.1950</td><td>-0.1327</td><td>0.5752</td><td>-0.7775</td></tr>
<tr><td>Fifth</td><td>-7252</td><td>-0.0202</td><td>0.6683</td><td>0.0631</td><td>-0.1494</td><td>-0.0274</td></tr>
<tr><td>Sixth</td><td>0.4000</td><td>-0.8172</td><td>0.3832</td><td>0.0602</td><td>-0.1095</td><td>0.0985</td></tr>
<tr><td colspan="7">Source: <xref ref-type="bibr" rid="R198">Huang <italic>et al.,</italic> 2002</xref></td></tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="s4-5"><title>4.3 Approaches of Forest Change Detection</title>
</sec>
<sec id="s4-6"><title><italic>4.3.1 </italic><italic>Approach I: Post-classification technique</italic></title>
<p>Satellite images selected for present study was processed for co-registration, enhancements, classification and finally, classified images were compared for Change Detection (CD) in forest cover. Calculated pixel values of NDVI have been broadly grouped into five classes (<xref ref-type="table" rid="T19">Table 3</xref>) i.e. no-vegetation, low to medium, medium, medium to dense and dense to very dense, using threshold observed in NDVI (2002) and NDVI (2009) images using ‘slicing’ operation in Ilwis.</p>
<table-wrap id="T19"><label>Table 3</label><caption><title>Broad classification of forest density based on NDVI</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Classes</bold></td><td colspan="2"><bold>Index values</bold></td></tr>
<tr><td><bold>ETM+ (2002)</bold></td><td><bold>TM (2009)</bold></td></tr>
<tr><td>No-vegetation</td><td>&lt; -0.16</td><td>&lt; 0.20</td></tr>
<tr><td>Low to medium</td><td>-0.16 to -0.02</td><td>0.20 to 0.23</td></tr>
<tr><td>Medium</td><td>-0.02 to 0.01</td><td>0.23 to 0.36</td></tr>
<tr><td>Medium to dense</td><td>0.01 to 0.16</td><td>0.36 to 0.45</td></tr>
<tr><td>Dense to very dense</td><td>0.16 &lt;</td><td>0.45 &lt;</td></tr>
</tbody>
</table>
</table-wrap>
<p>The maximum value (0.41) in reference image (t<sub>1</sub>) (NDVI 2002) was observed for dense vegetation (<xref ref-type="table" rid="T19">Table 3</xref>) whereas minimum (-0.48) for barren land including water body, rocky and barren land with 0.05 mean and 0.16 standard deviation. Targeted image (t<sub>2</sub>) (NDVI 2009) shows the maximum value (0.71) for dense vegetation and minimum (-0.33) for no-vegetation with 0.31 mean and 0.16 standard deviation. <xref ref-type="bibr" rid="R154">Adegoke and Carleton (2002)</xref> have been used innovative hybrid image classification technique for change detection analysis of forest. Therefore, limit values of NDVI classes are different for images acquired for 2002 and 2009 (<xref ref-type="table" rid="T19">Table 3</xref>) and was decided based on repetitive field checks using GPS, comparison with high-resolution images of Google Earth Pro and FCC (<xref ref-type="fig" rid="F37">Figure 4</xref>). Class ‘no-vegetation’ includes rocky and barren lands distributed at higher levels (&gt; 1100 m) of mountain and water bodies at bottom (<xref ref-type="fig" rid="F36">Figure 3</xref>).</p>
<fig id="F36"><label>Figure 3</label><caption><title>Distribution of forest depicted based on NDVI</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/20bcc20b3063fe61286de501baee6cc2.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<fig id="F37"><label>Figure 4</label><caption><title>False Color Composite (FCC) images</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/5ad2b4496260f4b04d4dbcad028d6ec2.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>Measurements of differences between detected land classes in base image (t<sub>1</sub>) and target image (t<sub>2</sub>) based on radiance difference called change detection (<xref ref-type="bibr" rid="R244">Weng <italic>et al.,</italic> 2009</xref>), for instance forest. The radiance difference mainly occurred due to real change in surface features, deviation in atmospheric conditions, deviation in sensor calibration, minimum to maximum error in Route Mean Square (RMS) at the time of georeferencing, deviation in illumination and difference in soil moisture conditions, etc. Here, classified images t<sub>1</sub> and t<sub>2</sub> combined using ‘cross’ operator (<xref ref-type="fig" rid="F38">Figure 5</xref>) in Ilwis which produce 25 classes (5 classes t<sub>1</sub> X 5 classes t<sub>2</sub>). Similar classes have been merged into three classes viz. no-change, positive change and negative change in forest cover (<xref ref-type="table" rid="T20">Table 4</xref>) using tool ‘merging the classes’ in Ilwis. Further, class wise changes (<xref ref-type="table" rid="T27">Table 11</xref>) in forest cover also detected successfully (<xref ref-type="fig" rid="F47">Figure 14</xref>) using same technique (<xref ref-type="table" rid="T20">Table 4</xref>). However, overall accuracy of the classes detected to show changes in forest cover was estimated about 78%. Users’ accuracy for class ‘positive change’ was about 58% and producer’s accuracy 70%. <xref ref-type="bibr" rid="R231">Rozenstein and Karnieli (2011)</xref> have insisted vigorous and creative efforts to establish new algorithms for change detection. Therefore, statistical methods and corrective measures were adopted to achieve improvements in post-classification technique for CD in forest cover.</p>
<fig id="F38"><label>Figure 5</label><caption><title>Approach-I: schematic preparation</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/778d61b873adbb9cd2f17fd645b58af4.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<table-wrap id="T20"><label>Table 4</label><caption><title>Class merging scheme for overall FCD</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Vegetation classes: ETM+ (2002)</bold></td><td colspan="5" align="center"><bold>Vegetation classes: TM (2009)</bold></td></tr>
<tr><td><bold>No-vegetation </bold></td><td><bold>Low to medium </bold></td><td><bold>Medium </bold></td><td><bold>Medium to dense</bold></td><td><bold>Dense to very dense</bold></td></tr>
<tr><td><bold>No-vegetation </bold></td><td>No-change</td><td>Positive<break/>change</td><td>Positive<break/>change</td><td>Positive<break/>change</td><td>Positive<break/>change</td></tr>
<tr><td><bold>Low to medium </bold></td><td>Negative<break/>change</td><td>No-change</td><td>Positive<break/>change</td><td>Positive<break/>change</td><td>Positive<break/>change</td></tr>
<tr><td><bold>Medium </bold></td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>No-change</td><td>Positive<break/>change</td><td>Positive<break/>change</td></tr>
<tr><td><bold>Medium to dense</bold></td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>No-change</td><td>Positive<break/>change</td></tr>
<tr><td><bold>Dense to very dense</bold></td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>Negative<break/>change</td><td>No-change</td></tr>
</tbody>
</table>
</table-wrap>
<table-wrap id="T21"><label>Table 5</label><caption><title>Class merging scheme for class wise FCD</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Vegetation classes:</bold><break/><bold>ETM+ (2002)</bold></td><td colspan="5" align="center"><bold>Vegetation classes: TM (2009)</bold></td></tr>
<tr><td><bold>No- vegetation </bold></td><td><bold>Low to medium </bold></td><td><bold>Medium </bold></td><td><bold>Medium to dense</bold></td><td><bold>Dense to very dense</bold></td></tr>
<tr><td><bold>No-vegetation </bold></td><td>NVNoC</td><td>NVPL</td><td>NVPM</td><td>NVPD</td><td>NVPVD</td></tr>
<tr><td><bold>Low to medium </bold></td><td>LNNV</td><td>LNoC</td><td>LPM</td><td>LPD</td><td>LPVD</td></tr>
<tr><td><bold>Medium </bold></td><td>MNNV</td><td>MNL</td><td>MNoC</td><td>MPD</td><td>MPVD</td></tr>
<tr><td><bold>Medium to dense</bold></td><td>DNNV</td><td>DNL</td><td>DNM</td><td>DNoC</td><td>DPVD</td></tr>
<tr><td><bold>Dense to very dense</bold></td><td>VDNNV</td><td>VDNL</td><td>VDNM</td><td>VDND</td><td>VDNoC</td></tr>
<tr><td colspan="6">* Refer section- Abbreviations or <xref ref-type="table" rid="T27">Table 11</xref> and <xref ref-type="table" rid="T29">Table 13</xref> for explanation.</td></tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="s4-7"><title><italic>4.3.2 </italic><italic>Approach II: Improved post-classification technique</italic></title>
<p>Many scholars have been successfully used combinations of different approaches for CD of land objects (<xref ref-type="bibr" rid="R161">Cano <italic>et al.,</italic> 2006</xref>; <xref ref-type="bibr" rid="R228">Pu <italic>et al.,</italic> 2008</xref>; <xref ref-type="bibr" rid="R244">Weng <italic>et al.,</italic> 2009</xref>). The algorithm of FCD has been modified to improve (<xref ref-type="fig" rid="F39">Figure 6</xref>) the results and processed through five steps: 1) sampling (<xref ref-type="bibr" rid="R166">Chapelle <italic>et al.,</italic> 2002</xref>; <xref ref-type="bibr" rid="R211">Lu <italic>et al.,</italic> 2011</xref>), 2) statistical analysis (<xref ref-type="bibr" rid="R236">Singh, 1989</xref>; <xref ref-type="bibr" rid="R163">Dhakal <italic>et al.,</italic> 2002</xref>), 3) normalization of indices, 4) change detection and 5) accuracy assessment. Statistical analysis and information about stable land surface including deep forests, rocky lands, barren lands and water bodies have used for normalization of indices calculated for target image (t<sub>2</sub>) (<xref ref-type="bibr" rid="R182">Felkar <italic>et al.,</italic> 1981</xref>; <xref ref-type="bibr" rid="R236">Singh, 1996</xref>; <xref ref-type="bibr" rid="R242">Turker and Derenyi, 2000</xref>).</p>
<fig id="F50"><label>Figure 6</label><caption><title>Approach-II: schematic preparation</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/4a4c448aa76103b422ec1bb1b31e6a0e.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>a. Sampling</p>
<p>Many researchers have been successfully used single band, ratio indices and coefficients for detection and delineation (Chen <italic>et al.,</italic> 2012) of objects like forests, water bodies, barren lands, rocky lands, snow covers, marsh lands, desert lands, etc. The variations in land phenomenon regulate surface radiance according to their reflectivity (<xref ref-type="bibr" rid="R163">Carlson <italic>et al.,</italic> 1990</xref>; Chen <italic>et al.,</italic> 2012). Surface reflectance values are sensitive to variations in vegetation cover, soil, background temperature, thermal inertia, and topography (<xref ref-type="bibr" rid="R157">Bhagat 2012</xref>). <xref ref-type="bibr" rid="R192">Goetz (1997)</xref> has used inverse relationship between SVI and Ts based on inversion soil-vegetation-atmosphere-transfer (SVAT) model for estimating soil surface moisture using NDVI-Ts triangle space. This approach is defined as the ‘Triangle Method’ includes vegetation indices, soil moisture indices and thermal indices (<xref ref-type="bibr" rid="R244">Weng <italic>et al.,</italic> 2009</xref>; <xref ref-type="bibr" rid="R211">Lu <italic>et al.,</italic> 2011</xref>). Objects detected using this approach are stood more precise than single bands. Surface temperature can be estimated using thermal band and vegetation can be detected with the help of near infrared band (<xref ref-type="bibr" rid="R221">Nemani and Running, 1989</xref>; <xref ref-type="bibr" rid="R163">Carlosan <italic>et al.,</italic> 1990</xref>; <xref ref-type="bibr" rid="R221">Nemani <italic>et al.,</italic> 1993</xref>; <xref ref-type="bibr" rid="R239">Tanriverdi, 2010</xref>). Thus, all surface elements have equal importance in estimations of objects (<xref ref-type="bibr" rid="R222">Owen <italic>et al.,</italic> 1998</xref>).</p>
<p>Water body absorbs maximum amount of radiation and emits very little energy (<xref ref-type="bibr" rid="R221">Nemani <italic>et al.,</italic> 1993</xref>). Forest cover reflects maximum amount of Green and Near Infrared (NIR) bands and a good absorber of Blue and Red band. Chemical component existing in tree leaves known as ‘chlorophyll’ effectively absorbs radiation from the Blue and Red band (<xref ref-type="bibr" rid="R202">Jiany <italic>et al.,</italic> 2008</xref>). The inner structure of these leaves response as good emission of NIR wavelength (<xref ref-type="bibr" rid="R230">Roberts <italic>et al.,</italic> 2011</xref>). On the other hand, visible and NIR radiation absorb more in the form of longer wavelength from water body and less from shorter wavelength (<xref ref-type="bibr" rid="R211">Lu <italic>et al.,</italic> 2011</xref>). Thus, water appears blue-green or blue due to maximum reflectance of shorter wavelength and viewed dark because reflectance of red and near infrared depending on the suspended sediments amount in water <xref ref-type="bibr" rid="R184">(Fletcher and Everitt, 2007)</xref>. Rocky lands are purely dry and rarely covered by thin grass and algae in wet seasons. However, rocky lands absorb very low amount of visible as well as NIR waves and reflect maximum as compare to other classes (<xref ref-type="bibr" rid="R213">Majed <italic>et al.,</italic> 2011</xref>). Water bodies, deep forests and rocky lands are stable objects and have extreme reflectance (minimum and maximum) in the region.</p>
<p>The variations in reflectance recorded is satellite image have influence on accuracy of forest CD using post-classification (<xref ref-type="bibr" rid="R228">Pu <italic>et al.,</italic> 2008</xref>). Therefore, corrective approach was adopted for detection of FCD (discussed in next section). Ground truth information (170 samples) was collected for these stable land objects which were appeared and identified on satellite images e.g. Landsat-7 ETM+ (2002) and Landsat-5 TM (2009). DN values of these land objects were used further statistical analysis and normalisation of indices calculated using satellite data (t<sub>2</sub>) (<xref ref-type="bibr" rid="R157">Bhagat, 2012</xref>). About 31.18% samples were collected from forest cover, 27.65% from rocky land, 29.41% from water body and 11.76% from barren land for further process (<xref ref-type="fig" rid="F40">Figure 7</xref>).</p>
<fig id="F40"><label>Figure 7</label><caption><title>Ground Reference Triangle</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/bfb23b84498b22a1eba1e294647229bc.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>b. Statistical analysis</p>
<p>NDVI has been widely used for vegetation analysis, however, greenness (equation 5 and 6) (<xref ref-type="bibr" rid="R175">Crist <italic>et al.,</italic> 1986</xref>; <xref ref-type="bibr" rid="R198">Huang <italic>et al.,</italic> 2002</xref>) have more potential of vegetation analysis. Greenness values estimated for both images (t<sub>1</sub> and t<sub>2</sub>) have positively correlated with NDVI. Therefore, greenness estimated for ETM+ (t<sub>1</sub>) and TM (t<sub>2</sub>) data was used as dependent variable for estimations corrected greenness values for t2 (<xref ref-type="fig" rid="F41">Figure 8</xref>).</p>
<fig id="F41"><label>Figure 8</label><caption><title>Greenness Index</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/7fe126cea35d81aaed12a26d5567df09.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>Greenness (ETM+) = ((Band1 * (-0.3344)) + (Band2 * (-0.3544)) + (Band3 * (-0.4556)) + (Band4 * (0.6966)) + (Band5 *(-0.0242)) + (Band7* (-0.2630)) (5)</p>
<p>Greenness (TM) = ((Band1*(-0.2728)) + (Band2*(-0.2174))+ (Band3 * (-0.5508)) + (Band4 * (0.7221)) + (Band5 *(0.0733)) + (Band7 * (-0.1648)) (6)</p>
<p>NDVI (2002) values show positive correlation with greenness estimated for t<sub>2</sub> (2009) (r = 0.976) and NDVI (2009) with greenness for t<sub>2</sub> (2009) (r = 0.975) (<xref ref-type="table" rid="T22">Table 6</xref>). Scatterplot (<xref ref-type="fig" rid="F42">Figure 9</xref>) for NDVI (2002) and Greenness (2009) show clustered distribution of vegetation. Therefore, calculated greenness values used for estimations of dependent variable in regression analysis. Maximum greenness (-0.07) calculated for t<sub>1</sub> (2002) and 65.06 for t<sub>2</sub> (2009) whereas minimum (-110.16) for no-vegetation i.e. water, rocky land, shadow, etc. in t<sub>1</sub> (2002) and -36.57 in t<sub>2</sub> (2009). These are stable land surface in the study area. Theoretically, values estimated for the stable lands might be same which is not possible in certain conditions at the time of image capturing (<xref ref-type="bibr" rid="R218">Munyati, 2004</xref>; <xref ref-type="bibr" rid="R246">Yarbrough <italic>et al.</italic>, 2005</xref>; <xref ref-type="bibr" rid="R189">Gill <italic>et al.</italic>, 2012</xref>). Therefore, accuracy of FCD in the region shows less. Further, these estimated values have been normalized using values estimated for stable pixels in this approach. Differences between estimated greenness values for stable pixels of t<sub>1</sub> and t<sub>2</sub> have been calculated and correlation with estimated indices SWI (t<sub>1</sub>), brightness (t<sub>1</sub>), fifth (t<sub>1</sub>) and DN values of band 7 (t<sub>1</sub>), band 7 (t<sub>2</sub>), band 5 (t<sub>1</sub>) have been estimated to find independent variables for estimation of normalized dependent variable for t<sub>2</sub> e.g. greenness.</p>
<fig id="F42"><label>Figure 9</label><caption><title>Scatterplot: NDVI (2002) and Greenness (2009)</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/8ce951e49d435ed442096c3e4feae594.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<table-wrap id="T22"><label>Table 6</label><caption><title>Correlations between NDVI and Greenness</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td colspan="2"></td><td><bold>Greenness (2002)</bold></td><td><bold>Greenness (2009)</bold></td></tr>
<tr><td rowspan="3"><bold>NDVI (2002)</bold></td><td>Pearson correlation</td><td>0.684<sup>**</sup></td><td><bold>0.976<sup>**</sup></bold></td></tr>
<tr><td>Sig. (2-tailed)</td><td>.000</td><td>.000</td></tr>
<tr><td>N</td><td>170</td><td>170</td></tr>
<tr><td rowspan="3"><bold>NDVI (2009)</bold></td><td>Pearson correlation</td><td>0.571<sup>**</sup></td><td><bold>0.975<sup>**</sup></bold></td></tr>
<tr><td>Sig. (2-tailed)</td><td>.000</td><td>.000</td></tr>
<tr><td>N</td><td>170</td><td>170</td></tr>
<tr><td colspan="4"><bold>** Significance at the 0.01 level (2-tailed).</bold></td></tr>
</tbody>
</table>
</table-wrap>
<p>Significant correlation of calculated difference between greenness estimated for stable pixels of t<sub>1</sub> and t<sub>2</sub> was estimated with band 7 (t<sub>1</sub>), SWI (t<sub>1</sub>), band 5 (t<sub>1</sub>) band 7 (t<sub>2</sub>), brightness (t<sub>1</sub>), difference in SWI, difference in fifth (<xref ref-type="table" rid="T23">Table 7</xref>). However, the strongest correlation was estimated for band 7 of t<sub>1</sub> and therefore this band was used as independent variable for estimations of pixel values of correction in t<sub>2</sub> using regression techniques. Samples (170) collected from stable land surface i.e. forest land, rocky land, water, barren land, etc. (<xref ref-type="fig" rid="F41">Figure 7</xref>) was used for this analysis.</p>
<table-wrap id="T23"><label>Table 7</label><caption><title>Correlation of difference greenness with selected criterion</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td colspan="2"></td><td><bold>Band7 of t<sub>1</sub></bold></td><td><bold>SWI</bold><break/><bold>of t<sub>1</sub></bold></td><td><bold>Band7</bold><break/><bold>of t<sub>2</sub></bold></td><td><bold>Band5</bold><break/><bold>of t<sub>1</sub></bold></td><td><bold>Brightness of t<sub>1</sub></bold></td><td><bold>Difference </bold><break/><bold>SWI</bold></td><td><bold>Difference </bold><break/><bold>Fifth</bold></td></tr>
<tr><td rowspan="3">Difference greenness</td><td>Pearson Correlation</td><td>0.983<sup>**</sup></td><td>-0.964<sup>**</sup></td><td>0.960<sup>**</sup></td><td>0.960<sup>**</sup></td><td>0.949<sup>**</sup></td><td>0.934<sup>**</sup></td><td>-0.933<sup>**</sup></td></tr>
<tr><td>Sig.<break/>(2-tailed)</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td><td>0.000</td></tr>
<tr><td>N</td><td>170</td><td>170</td><td>170</td><td>170</td><td>170</td><td>170</td><td>170</td></tr>
<tr><td colspan="9">** Significance at the 0.01 level (2-tailed).</td></tr>
</tbody>
</table>
</table-wrap>
<p>Regression (<xref ref-type="bibr" rid="R178">equation 7</xref>) gives mathematical tools (<xref ref-type="bibr" rid="R209">Lawrence and Wrlght, 2001</xref>; <xref ref-type="bibr" rid="R172">Cohen <italic>et al.</italic>, 2003</xref>) to estimate the fit between two different multi-spectral images acquired at different time for same area. Pixel values close to the regression line indicate the suitability of the model (<xref ref-type="table" rid="T24">Table 8</xref>).</p>
<p>\(Y=a+b(x)\) (7)</p>
<p>Where, \(Y\) is an estimated variable (dependent), \(a\) and \(b\) are constant ( \(a\) is intercept and \(b\) is rate of change in slope) and \(x\) is independent variable.</p>
<table-wrap id="T24"><label>Table 8</label><caption><title>Regression analysis of difference greenness with selected criterion</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td align="center"><bold>\(Y\)</bold></td><td align="center">\(x\)</td><td><bold>Correlation (r)</bold></td><td align="center">\(a\)</td><td align="center">\(b\)</td><td align="center"><bold>R<sup>2</sup></bold></td><td><bold>Trend</bold></td></tr>
<tr><td rowspan="7">Difference greenness</td><td>Band7 of t<sub>1</sub></td><td>0.983</td><td>21.664</td><td>0.602</td><td>0.967</td><td>Positive</td></tr>
<tr><td>SWI of t<sub>1</sub></td><td>-0.964</td><td>33.511</td><td>-0.432</td><td>0.930</td><td>Negative</td></tr>
<tr><td>Band7 of t<sub>2</sub></td><td>0.96</td><td>24.208</td><td>0.891</td><td>0.922</td><td>Positive</td></tr>
<tr><td>Band5 of t<sub>1</sub></td><td>0.96</td><td>20.554</td><td>0.399</td><td>0.921</td><td>Positive</td></tr>
<tr><td>Brightness of t<sub>1</sub></td><td>0.949</td><td>-2.307</td><td>0.370</td><td>0.901</td><td>Positive</td></tr>
<tr><td>Difference SWI</td><td>0.934</td><td>30.447</td><td>0.769</td><td>0.871</td><td>Positive</td></tr>
<tr><td>Difference Fifth</td><td>-0.933</td><td>90.231</td><td>-2.110</td><td>0.871</td><td>Negative</td></tr>
</tbody>
</table>
</table-wrap>
<p>Scholars have been reported the highest change detection accuracy by using regression approach. <xref ref-type="bibr" rid="R202">Jiany <italic>et al.,</italic> (2008)</xref> reported three steps of radiometric normalization using pixel values for stable land surface: (1) delineation of stable land based on near-infrared (NIR) bands; (2) DN for stable areas using regression models, and (3) the regression coefficients used for normalizing DN of the target image. Here, linear regression model depicts (<xref ref-type="fig" rid="F43">Figure 10</xref>) the fit (R<sup>2</sup>= 0.967) of differences in Greenness of t<sub>1</sub> and t<sub>2</sub> with band 7 of t<sub>1</sub> with high correlation (0.983) and clustered distribution. The values closer to zero on x-axis are for water bodies, next cluster for forests and at far from the zero for barren and rocky lands. Band 5 and band 7 of ETM+ and TM images show higher sensitivity to soils and vegetation.</p>
<fig id="F43"><label>Figure 10</label><caption><title>Regression analysis</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/4234bd1e4b5af60c9453acba6a829fef.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>c. Estimation of exaggeration and normalization of images</p>
<p>The maximum value (0.41) in reference image (t<sub>1</sub>) (NDVI 2002) was observed for dense vegetation (<xref ref-type="table" rid="T19">Table 3</xref>) whereas minimum (-0.48) for barren land and water body with 0.05 mean and 0.16 standard deviation. Maximum NDVI value (0.71) in target image, (t<sub>2</sub>) (2009) was observed for dense vegetation and minimum (-0.33) for no-vegetation with 0.31 mean and 0.16 standard deviation. Therefore, it is assumed that the values recorded in target image are exaggerated from ground reality (<xref ref-type="fig" rid="F41">Figure 8</xref>). Regression model (equation 8) was formed to estimate the values of exaggeration in greenness (Ge) estimated for target image (t<sub>2</sub>) for normalizing (t<sub>2</sub>).</p>
<p>\(Ge=a+b(B7t_1)\) (8)</p>
<p>Where, \(a\) is an intercept 21.664 and \(b\) is rate of change in slope 0.602 and \(B7 t_1\) is band 7 of reference image (t<sub>1</sub>). Finally, raster image for greenness estimated for target image (t<sub>2</sub>) was normalised \(Nt_2\) (equation 9) for final FCD.</p>
<p>\(Nt_2=Greenness2009-Ge\) (9)</p>
<p>The histograms for estimated images viz. corrected greenness 2009 using band 7 (t<sub>1</sub>), SWI (t<sub>1</sub>), band 7 (t<sub>2</sub>) and band 5 (t<sub>1</sub>), etc. were prepared in SPSS and compared with histograms prepared for greenness for t<sub>1</sub> and t<sub>2</sub> (<xref ref-type="fig" rid="F44">Figure 11</xref>). Histogram values distributed from -110.1568 to -0.0782 for greenness of ETM+ (2002) and from -36.5670 to 65.0585 of TM (2009). However, distributed values on histogram prepared for corrected greenness of t<sub>2</sub> (2009) vary from -108.6296 to 17.5085. Similar results observed for normalised raster images of band 7 (2002), SWI (2002), band 7 (2009) and band 5 (2002).</p>
<fig id="F44"><label>Figure 11</label><caption><title>Histogram distributions</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/77e4018b5ac1f156252b02c79a78e615.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
</sec>
<sec id="s4-8"><title><italic>4.3.3 </italic><italic>Change detection</italic></title>
<p>Finally, raster image of greenness estimated for t<sub>1</sub> (ETM+ 2002) and normalised raster image of greenness estimated (<xref ref-type="fig" rid="F45">Figure 12</xref>) for t<sub>2</sub> (TM 2009) were processed for FCD in this study. The maximum value was -00.08 estimated for greenness of t<sub>1</sub> (2002) and 17.51 for normalized greenness (2009) whereas minimum value was -110.16 for t<sub>1</sub> (2002) and -108.63 for normalized image t<sub>2</sub> for ‘no-vegetation’ i.e. water, rocky land, shadow, etc. (<xref ref-type="fig" rid="F46">Figure 13</xref>). Reference (t<sub>1</sub>) and normalised (Nt<sub>2</sub>) images were classified into five classes (<xref ref-type="table" rid="T25">Table 9</xref>) e.g. no-vegetation, low to medium, medium, medium to dense and dense to very dense, respectively.</p>
<table-wrap id="T25"><label>Table 9</label><caption><title>Broad classification of forest density based greenness index</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Vegetation classes</bold></td><td><bold>Index value</bold></td></tr>
<tr><td>No-vegetation</td><td>&lt; -64</td></tr>
<tr><td>Low to medium</td><td>-64 to -48</td></tr>
<tr><td>Medium</td><td>-48 to -32</td></tr>
<tr><td>Medium to dense</td><td>-32 to -16</td></tr>
<tr><td>Dense to very dense</td><td>-16 &lt;</td></tr>
</tbody>
</table>
</table-wrap>
<fig id="F45"><label>Figure 12</label><caption><title>Corrected Greenness index</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/ee5a9f98bfa590980ffcdefadeadcba8.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<p>Classified raster images (<xref ref-type="fig" rid="F46">Figure 13</xref>) for 2002 and 2009 were combined using cross operation in Ilwis for forest CD. 25 classes were created in this combination and similar classes merged to get meaningful categories (<xref ref-type="table" rid="T20">Table 4</xref>). Overall changes were estimated as ‘negative changes’, ‘positive changes’ and ‘no-changes’. Further, class wise changes also estimated to know the transformations from one class to another (<xref ref-type="table" rid="T21">Table 5</xref>) as ‘positive change, negative change and no-change.</p>
<fig id="F46"><label>Figure 13</label><caption><title>Distribution of forests depicted based on greenness index</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/b806d8a10a127e4da47a1a86e8427434.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
</sec>
</sec>
<sec id="s5"><title>Results</title>
<p>Changes in forest cover in the study area (13858.83 ha) was estimated using two approaches: Approach I) post-classification technique and Approach II) improved post-classification technique. NDVI based classified maps prepared for t<sub>1</sub> and t<sub>2</sub> were used for CD (<xref ref-type="fig" rid="F47">Figure 14</xref>) in first approach and statistical analyses were performed for improvement in accuracy of post-classification CD in second approach (<xref ref-type="fig" rid="F48">Figure 15</xref>). Primary field check data and pixel information of stable land units i.e. water bodies, deep forest and rocky lands were used in statistical analysis for improvement of the technique.</p>
<sec id="s5-1"><title>5.1 Approach I: Post-classification Technique</title>
<p>The change in forest area within period of images acquired has been detected and successfully demarcated (<xref ref-type="fig" rid="F47">Figure 14</xref>). Approximately, 58.59% of reviewed area shows forests with increasing trends, 33.69% area shows no-changes and 7.72% area losing the forest. Forest cover increases for the class ‘medium’ vegetation about 22.41% and class ‘dense to very dense’ 16.84% (<xref ref-type="table" rid="T26">Table 10</xref>). Slightly negative change (-0.12%) observed for class ‘no-vegetation’ and the class ‘medium to dense’ (-2.81%) (<xref ref-type="fig" rid="F47">Figure 14</xref>). Positive changes in forest cover was estimated for 8119.98 ha, no-change for 4669.29 ha and negative change for 1069.56 ha. However, these estimated values show generalized picture of changes within the classes therefore, dynamics of class wise changes also analyzed.</p>
<fig id="F47"><label>Figure 14</label><caption><title>Approach-I: distribution of changes in forest cover</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/e590c0000d1c7cf761b980f7493cc113.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<table-wrap id="T26"><label>Table 10</label><caption><title>Approach I- class wise changes in area under forest</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Vegetation classes</bold></td><td colspan="2"><bold>Area (2002)</bold></td><td colspan="2"><bold>Area (2009)</bold></td><td rowspan="2"><bold>Change </bold><break/><bold>%</bold></td></tr>
<tr><td><bold>ha</bold></td><td><bold>%</bold></td><td><bold>ha</bold></td><td><bold>%</bold></td></tr>
<tr><td>No-vegetation</td><td>1713.51</td><td>12.36</td><td>1695.96</td><td>12.24</td><td>-0.12</td></tr>
<tr><td>Low to medium</td><td>5840.46</td><td>42.14</td><td>806.76</td><td>5.82</td><td>-36.32</td></tr>
<tr><td>Medium</td><td>968.4</td><td>6.99</td><td>4075.11</td><td>29.4</td><td>22.41</td></tr>
<tr><td>Medium to dense</td><td>3706.65</td><td>26.75</td><td>3317.76</td><td>23.94</td><td>-2.81</td></tr>
<tr><td>Dense to very dense</td><td>1629.81</td><td>11.76</td><td>3963.24</td><td>28.6</td><td>16.84</td></tr>
<tr><td><bold>Total Area</bold></td><td><bold>13858.8</bold></td><td><bold>100</bold></td><td><bold>13858.8</bold></td><td><bold>100</bold></td><td></td></tr>
</tbody>
</table>
</table-wrap>
<p>Sparse vegetation has been changed to medium vegetation (22.68%) and dense vegetation to very dense vegetation (15.02%) (<xref ref-type="table" rid="T27">Table 11</xref>) with no-change on 10.89 % area. However, vegetation decreases with high rate from ‘low vegetation’ to ‘no-vegetation’ for 4.30% area. The accuracy of assessment has been estimated about 77.84%, discussed in the next section (<xref ref-type="table" rid="T30">Table 14</xref>).</p>
<table-wrap id="T27"><label>Table 11</label><caption><title>Approach I – change class wise distribution of forest</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Class Code</bold></td><td><bold>Change classes of forest </bold></td><td><bold>Area in ha</bold></td><td><bold>Area in %</bold></td></tr>
<tr><td>NVNoC</td><td>No-vegetation to no-change</td><td>1083.96</td><td>7.8214</td></tr>
<tr><td>NVPL</td><td>No-vegetation positively changed to low vegetation</td><td>259.92</td><td>1.8755</td></tr>
<tr><td>NVPM</td><td>No-vegetation positively changed to medium vegetation</td><td>347.31</td><td>2.5061</td></tr>
<tr><td>NVPD</td><td>No-vegetation positively changed to dense vegetation</td><td>21.24</td><td>0.1533</td></tr>
<tr><td>NVPVD</td><td>No-vegetation positive changed to very dense vegetation</td><td>1.08</td><td>0.0078</td></tr>
<tr><td>LNNV</td><td>Low vegetation negatively changed to no-vegetation</td><td>597.24</td><td>4.3095</td></tr>
<tr><td>LNoC</td><td>Low vegetation with no-change</td><td>531.09</td><td>3.8321</td></tr>
<tr><td>LPM</td><td>Low vegetation positively changed to medium vegetation</td><td>3143.34</td><td>22.6811</td></tr>
<tr><td>LPD</td><td>Low vegetation positively changed to dense vegetation</td><td>1393.38</td><td>10.0541</td></tr>
<tr><td>LPVD</td><td>Low vegetation positive changed to very dense vegetation</td><td>175.41</td><td>1.2657</td></tr>
<tr><td>MNNV</td><td>Medium vegetation negative changed to no-vegetation</td><td>9.54</td><td>0.0688</td></tr>
<tr><td>MNL</td><td>Medium vegetation negatively changed to low vegetation</td><td>8.19</td><td>0.0591</td></tr>
<tr><td>MNoC</td><td>Medium vegetation with no-change</td><td>253.98</td><td>1.8326</td></tr>
<tr><td>MPD</td><td>Medium vegetation positively changed to dense vegetation</td><td>500.94</td><td>3.6146</td></tr>
<tr><td>MPVD</td><td>Medium vegetation positively changed to very dense vegetation</td><td>195.75</td><td>1.4125</td></tr>
<tr><td>DNNV</td><td>Dense vegetation negatively changed to no-vegetation</td><td>5.22</td><td>0.0377</td></tr>
<tr><td>DNL</td><td>Dense vegetation negatively changed to low vegetation</td><td>7.47</td><td>0.0539</td></tr>
<tr><td>DNM</td><td>Dense vegetation negatively changed to medium vegetation</td><td>321.48</td><td>2.3197</td></tr>
<tr><td>DNoC</td><td>Dense vegetation with no-change</td><td>1290.87</td><td>9.3144</td></tr>
<tr><td>DPVD</td><td>Dense vegetation positively changed to very dense vegetation</td><td>2081.61</td><td>15.0201</td></tr>
<tr><td>VDNNV</td><td>Very dense vegetation negatively changed to no-vegetation</td><td>0.02</td><td>0.0002</td></tr>
<tr><td>VDNL</td><td>Very dense vegetation negatively changed to low vegetation</td><td>0.07</td><td>0.0004</td></tr>
<tr><td>VDNM</td><td>Very dense vegetation negatively changed to medium vegetation</td><td>9</td><td>0.0649</td></tr>
<tr><td>VDND</td><td>Very dense vegetation negatively changed to dense vegetation</td><td>111.33</td><td>0.8033</td></tr>
<tr><td>VDNoC</td><td>Very dense vegetation with no-change</td><td>1509.39</td><td>10.8912</td></tr>
<tr><td colspan="2"><bold>Total area</bold></td><td><bold>13858.83</bold></td><td><bold>100</bold></td></tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="s5-2"><title>5.2 Approach II: Improved Post-classification Technique</title>
<p>Statistical techniques i.e. correlation analysis and regression model have been used to improve the performance of FCD (<xref ref-type="fig" rid="F39">Figure 6</xref>). Area under forest has been detected and estimated using TCCT greenness index. However, estimated greenness has been corrected to reduce exaggeration in reflectance recorded in DN using regression model (<xref ref-type="fig" rid="F48">Figure 15</xref>). About 13858.83 ha (23.59% TGA) shows positive changes, 70.8% no-change and only 6.33% area shows negative changes. Accuracy assessment shows preciseness 95.21% (<xref ref-type="table" rid="T31">Table 15</xref>) of the results compared to post-classification based FCD. ‘No-changes’ in vegetation observed for the class, ‘low to medium’ vegetation (<xref ref-type="table" rid="T28">Table 12</xref>) and positive changes for ‘medium to dense’ and ‘dense to very dense’ vegetation. This improved technique of FCD helps to reduce the exaggeration in estimations. Class wise changes also estimated to understand the dimensions of changes in forest cover (<xref ref-type="table" rid="T29">Table 13</xref>).</p>
<fig id="F48"><label>Figure 15</label><caption><title>Approach-II: distribution of changes in forest cover</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/c46aea2e02990ef2da55484367653cb8.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
<table-wrap id="T28"><label>Table 12</label><caption><title>Approach II - changes in forest cover</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Vegetation classes</bold></td><td colspan="2"><bold>Area (2002)</bold></td><td colspan="2"><bold>Area (2009)</bold></td><td><bold>Change</bold></td></tr>
<tr><td></td><td><bold>ha</bold></td><td><bold>%</bold></td><td><bold>ha</bold></td><td><bold>%</bold></td><td><bold>%</bold></td></tr>
<tr><td>No-vegetation</td><td>3310.2</td><td>23.89</td><td>2975.22</td><td>21.47</td><td>-2.42</td></tr>
<tr><td>Low to medium</td><td>4228.47</td><td>30.51</td><td>3505.68</td><td>25.3</td><td>-5.21</td></tr>
<tr><td>Medium</td><td>4115.97</td><td>29.7</td><td>4251.24</td><td>30.68</td><td>0.98</td></tr>
<tr><td>Medium to dense</td><td>2083.95</td><td>15.04</td><td>2908.71</td><td>20.99</td><td>5.95</td></tr>
<tr><td>Dense to very dense</td><td>120.24</td><td>0.87</td><td>217.98</td><td>1.57</td><td>0.7</td></tr>
<tr><td><bold>Total Area</bold></td><td><bold>13858.83</bold></td><td><bold>100</bold></td><td><bold>13858.83</bold></td><td><bold>100</bold></td><td></td></tr>
</tbody>
</table>
</table-wrap>
<table-wrap id="T29"><label>Table 13</label><caption><title>Approach II – change class wise distribution of forest</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td><bold>Class code</bold></td><td><bold>Change classes of forest</bold></td><td><bold>Area in ha</bold></td><td><bold>Area in %</bold></td></tr>
<tr><td>NVNoC</td><td>No-vegetation to no-change</td><td>1931.31</td><td>13.94</td></tr>
<tr><td>NVPL</td><td>No-vegetation positively changed to low vegetation</td><td>1176.48</td><td>8.49</td></tr>
<tr><td>NVPM</td><td>No-vegetation positively changed to medium vegetation</td><td>196.02</td><td>1.41</td></tr>
<tr><td>NVPD</td><td>No-vegetation positively changed to dense vegetation</td><td>6.3</td><td>0.05</td></tr>
<tr><td>NVPVD</td><td>No-vegetation positive changed to very dense vegetation</td><td>0.09</td><td>0.01</td></tr>
<tr><td>LNNV</td><td>Low vegetation negative change to no-vegetation</td><td>338.67</td><td>2.44</td></tr>
<tr><td>LNoC</td><td>Low vegetation negatively changed to no-vegetation</td><td>1838.99</td><td>13.27</td></tr>
<tr><td>LPM</td><td>Low vegetation with no-change</td><td>1919.72</td><td>13.85</td></tr>
<tr><td>LPD</td><td>Low vegetation positively changed to medium vegetation</td><td>168.75</td><td>1.22</td></tr>
<tr><td>LPVD</td><td>Low vegetation positively changed to dense vegetation</td><td>2.34</td><td>0.02</td></tr>
<tr><td>MNNV</td><td>Medium vegetation negative changed to no-vegetation</td><td>8.73</td><td>0.06</td></tr>
<tr><td>MNL</td><td>Medium vegetation negatively changed to low vegetation</td><td>317.43</td><td>2.29</td></tr>
<tr><td>MNoC</td><td>Medium vegetation with no-change</td><td>2262.79</td><td>16.33</td></tr>
<tr><td>MPD</td><td>Medium vegetation positively changed to dense vegetation</td><td>1511.28</td><td>10.9</td></tr>
<tr><td>MPVD</td><td>Medium vegetation positively changed to very dense vegetation</td><td>27.73</td><td>0.2</td></tr>
<tr><td>DNNV</td><td>Dense vegetation negatively changed to no-vegetation</td><td>0.34</td><td>0.01</td></tr>
<tr><td>DNL</td><td>Dense vegetation negatively changed to low vegetation</td><td>16.54</td><td>0.12</td></tr>
<tr><td>DNM</td><td>Dense vegetation negatively changed to medium vegetation</td><td>401.13</td><td>2.89</td></tr>
<tr><td>DNoC</td><td>Dense vegetation with no-change</td><td>1466.17</td><td>10.58</td></tr>
<tr><td>DPVD</td><td>Dense vegetation positively changed to very dense vegetation</td><td>98.73</td><td>0.71</td></tr>
<tr><td>VDNNV</td><td>Very dense vegetation negatively changed to no vegetation</td><td>0.17</td><td>0.01</td></tr>
<tr><td>VDNL</td><td>Very dense vegetation negatively changed to low vegetation</td><td>6.94</td><td>0.05</td></tr>
<tr><td>VDNM</td><td>Very dense vegetation negatively changed to medium vegetation</td><td>73.92</td><td>0.53</td></tr>
<tr><td>VDND</td><td>Very dense vegetation negatively changed to dense vegetation</td><td>31</td><td>0.22</td></tr>
<tr><td>VDNoC</td><td>Very dense vegetation with no-change</td><td>57.26</td><td>0.41</td></tr>
<tr><td colspan="2"><bold>Total area</bold></td><td><bold>13858.83</bold></td><td><bold>100</bold></td></tr>
</tbody>
</table>
</table-wrap>
</sec>
</sec>
<sec id="s6"><title>Accuracy Assessment</title>
<p>Accuracy of detected changes in forest cover was estimated as user’s accuracy, producer’s accuracy and overall accuracy. <xref ref-type="bibr" rid="R229">Rahman and Saha (2008)</xref> have suggested that the sample size should take minimum 30 for each class to estimate accuracy at 90%. Therefore, 170 samples well distributed within the classes were collected using GPS points and Google Earth Pro high resolution images.</p>
<table-wrap id="T30"><label>Table 14</label><caption><title>Approach-I: error matrix</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Classified data</bold></td><td colspan="6"><bold>Reference data</bold></td></tr>
<tr><td><bold>Positive change</bold></td><td><bold>No-change</bold></td><td><bold>Negative change</bold></td><td><bold>Water</bold></td><td><bold>Row total</bold></td><td><bold>User’s </bold><break/><bold>accuracy in (%)</bold></td></tr>
<tr><td><bold>Positive change</bold></td><td><bold>19</bold></td><td>6</td><td>4</td><td>4</td><td>33</td><td><bold>58</bold></td></tr>
<tr><td><bold>No-change</bold></td><td>5</td><td><bold>31</bold></td><td>5</td><td>3</td><td>44</td><td><bold>70</bold></td></tr>
<tr><td><bold>Negative change</bold></td><td>2</td><td>3</td><td><bold>32</bold></td><td>3</td><td>40</td><td><bold>80</bold></td></tr>
<tr><td><bold>Water</bold></td><td>1</td><td>0</td><td>1</td><td><bold>48</bold></td><td>50</td><td><bold>96</bold></td></tr>
<tr><td><bold>Column total</bold></td><td>27</td><td>40</td><td>42</td><td>58</td><td><bold>167</bold></td><td></td></tr>
<tr><td><bold>Producer&apos;s </bold><break/><bold>accuracy (%)</bold></td><td><bold>70</bold></td><td><bold>78</bold></td><td><bold>76</bold></td><td><bold>83</bold></td><td></td><td></td></tr>
<tr><td><bold>Over accuracy </bold></td><td><bold>77.84</bold></td><td></td><td></td><td></td><td></td><td></td></tr>
</tbody>
</table>
</table-wrap>
<p>Overall accuracy of the FCD using approach I: post-classification technique was estimated about 77.84% (<xref ref-type="table" rid="T30">Table 14</xref>). User’s accuracy was estimated about 58% for positively changed forest cover, 70% for no-changed or stable and 80% for negatively changed forest and 96% for water bodies. Producer’s accuracy was estimated about 70% for positive changes, 78% for no-changes, 76% for negative changes and 83% for water bodies. Maximum accuracy was estimated for water bodies and negative changes in vegetation whereas positively changed vegetated area and stable land units show very less accuracy. Accuracy assessment suggests need of improvement in FCD techniques. Therefore, algorithm for improved FCD has been formulated in this study. Approach II has improved accuracy of FCD to 95.21% (<xref ref-type="table" rid="T31">Table 15</xref>) with user’s accuracy about 97% for positive changes, 95% for no-change, 93% for negative changes and 96% for water bodies. Producer’s accuracy was estimated about 91% for positive changes, 98% for no-change, 95% for negative changes and 96% water bodies. Thus, improved post-classification technique improves the accuracy of FCD.</p>
<table-wrap id="T31"><label>Table 15</label><caption><title>Approach-II: error matrix</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Classified data </bold></td><td colspan="6"><bold>Reference data</bold></td></tr>
<tr><td><bold>Positive change</bold></td><td><bold>No-</bold><break/><bold> change </bold></td><td><bold>Negative change</bold></td><td><bold>Water</bold></td><td><bold>Row total</bold></td><td><bold>User’s</bold><break/><bold>accuracy (%)</bold></td></tr>
<tr><td><bold>Positive change</bold></td><td><bold>32</bold></td><td>0</td><td>0</td><td>1</td><td>33</td><td><bold>97</bold></td></tr>
<tr><td><bold>No-change </bold></td><td>1</td><td><bold>42</bold></td><td>1</td><td>0</td><td>44</td><td><bold>95</bold></td></tr>
<tr><td><bold>Negative change </bold></td><td>1</td><td>1</td><td><bold>37</bold></td><td>1</td><td>40</td><td><bold>93</bold></td></tr>
<tr><td><bold>Water</bold></td><td>1</td><td>0</td><td>1</td><td><bold>48</bold></td><td>50</td><td><bold>96</bold></td></tr>
<tr><td><bold>Column total</bold></td><td>35</td><td>43</td><td>39</td><td>50</td><td><bold>167</bold></td><td></td></tr>
<tr><td><bold>Producer&apos;s</bold><break/><bold>accuracy in (%)</bold></td><td><bold>91</bold></td><td><bold>98</bold></td><td><bold>95</bold></td><td><bold>96</bold></td><td></td><td></td></tr>
<tr><td><bold>Over accuracy</bold></td><td><bold>95.21%</bold></td><td></td><td></td><td></td><td></td><td></td></tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="s7"><title>Conclusion</title>
<p>About 58.59% of reviewed area estimated using first approach of CD shows (<xref ref-type="fig" rid="F48">Figure 15</xref>) positive changes, 33.69% no-change and 7.72% negative changes in forest cover in the area. However, improved post-classification analysis estimates (<xref ref-type="table" rid="T32">Table 16</xref>) about 23.59% area with positive changes, 70.08% no-change and 6.33% negative changes. Thus, normalised raster greenness image using statistical analysis removes illusional exaggeration of reflectance recorded in satellite data and useful to estimate more precisely FCD.</p>
<table-wrap id="T32"><label>Table 16</label><caption><title>Comparison of changes in forest cover estimated using two approaches</title></caption><table frame="hsides" rules="groups">
<tbody>
<tr><td rowspan="2"><bold>Change classes</bold></td><td colspan="2"><bold>Approach I</bold></td><td colspan="2"><bold>Approach II</bold></td></tr>
<tr><td><bold>Area (hectare)</bold></td><td><bold>Area (%)</bold></td><td><bold>Area (hectare)</bold></td><td><bold>Area (%)</bold></td></tr>
<tr><td>Positive change</td><td>8119.98</td><td>58.59%</td><td>3269.25</td><td>23.59%</td></tr>
<tr><td>No-change</td><td>4669.29</td><td>33.69%</td><td>9712.89</td><td>70.08%</td></tr>
<tr><td>Negative change</td><td>1069.56</td><td>7.72%</td><td>876.69</td><td>6.33%</td></tr>
<tr><td><bold>Total area</bold></td><td><bold>13858.83</bold></td><td><bold>100%</bold></td><td><bold>13858.83</bold></td><td><bold>100%</bold></td></tr>
</tbody>
</table>
</table-wrap>
<p>Authors are aware about limitations of finer temporal resolution (7 years) and medium spatial resolution (30 x 30m) of Landsat ETM+ and Landsat TM images. The methodology and techniques formulated in this study can be useful for researchers, scholars, environmental planners and managements for change detection of forest for sustainable land management and development.</p>
</sec>
</body>
<back>
<ack>
<title>Acknowledgements</title>
<p>Authors gratefully acknowledge the anonymous reviewers for constructive comments and suggestions for improvement in the draft.</p>
</ack>
<glossary>
<title>Abbreviations</title>
<p><bold>CD</bold>: Change Detection; <bold>DCD</bold>: Digital Change Detection; DN: Digital Numbers, <bold>DN</bold>L: Dense vegetation negatively changed to low vegetation; <bold>DNM</bold>: Dense vegetation negatively changed to medium vegetation; <bold>DNNV</bold>: Dense vegetation negatively changed to no-vegetation; <bold>DNoC</bold>: Dense vegetation with no-change; <bold>DPVD</bold>: Dense vegetation positively changed to very dense vegetation; <bold>DVI</bold>: Difference Vegetation Index; <bold>ERDAS</bold>: Earth Resource Data Analysis System: <bold>ETM+</bold>: Enhanced Thematic Mapper Plus; <bold>FAO</bold>: Food and Agriculture Organization; <bold>FCC</bold>: False Colour Composite; <bold>FCD</bold>: Forest Change Detection; <bold>FSI</bold>: Forest Survey of India; <bold>GIS</bold>: Geographical Information System; GPS: Global Positioning System; <bold>ILWIS</bold>: Integrated Land Water Information System; <bold>IPVI</bold>: Infrared Percentage Vegetation Index; <bold>ISFR</bold>: India State of Forest Report; <bold>LAI</bold>: Leaf Area Indices; <bold>LNNV</bold>: Low vegetation negatively changed to no-vegetation; <bold>LNoC</bold>: Low vegetation with no-change; <bold>LPD</bold>: Low vegetation positively changed to dense vegetation; <bold>LPM</bold>: Low vegetation positively changed to medium vegetation; <bold>LPVD</bold>: Low vegetation positive changed to very dense vegetation; <bold>LSTI</bold>: Land Surface Temperature Index; <bold>MNL</bold>: Medium vegetation negatively changed to low vegetation; <bold>MNNV</bold>: Medium vegetation negative changed to no-vegetation; <bold>MNoC</bold>: Medium vegetation with no-change; <bold>MPD</bold>: Medium vegetation positively changed to dense vegetation; <bold>MPVD</bold>: Medium vegetation positively changed to very dense vegetation; <bold>MSAVI</bold>: Modified <bold>SAVI</bold>; <bold>MSL</bold>: Mean Sea Level; <bold>MSR</bold>: Modified Simple Ratio; <bold>NASA</bold>: National Aeronautics and Space Administration; <bold>NDSI</bold>: Normalised Difference Salinity Index; <bold>NDVI</bold>: Normalized Difference Vegetation Index; <bold>NIR</bold>: Near Infrared; <bold>NVNoC</bold>; No-vegetation to no-change; <bold>NVPD</bold>: No-vegetation positively changed to dense vegetation; <bold>NVPL</bold>: No-vegetation positively changed to low vegetation; <bold>NVPM</bold>: No-vegetation positively changed to medium vegetation; <bold>NVPVD</bold>: No-vegetation positive changed to very dense vegetation; <bold>RDVI</bold>: Ratio Difference Vegetation Index; <bold>RMS</bold>: Route Mean Square; <bold>RVI</bold>: Ratio Vegetation Index; <bold>SAVI</bold>: Soil Adjusted Vegetation Index; <bold>SOI</bold>: Survey of India; SPSS: Statistical Packages for the Social Sciences; <bold>SVAT</bold>: Soil Vegetation Atmosphere Transfer; <bold>SVI</bold>: Soil Vegetation Index; <bold>SWI</bold>: Soil Wetness Index; <bold>TCCT</bold>: Tasseled Cap Coefficient Transformation; <bold>TGA</bold>: Total Geographical Area; <bold>ToA</bold>: Top of Atmosphere; <bold>TM</bold>: Thematic Mapper; <bold>TV</bold>I: Transformed Vegetation Index; <bold>VDND</bold>: Very dense vegetation negatively changed to dense vegetation; <bold>VDNL</bold>: Very dense vegetation negatively changed to low vegetation; <bold>VDNM</bold>: Very dense vegetation negatively changed to medium vegetation; <bold>VDNNV</bold>: Very dense vegetation negatively changed to no vegetation; <bold>VDNoC</bold>: Very dense vegetation with no-change; <bold>WCMC</bold>: World Conservation Monitoring Centre; <bold>WRI</bold>: World Resources Institute.</p>
</glossary>
<fn-group>
<fn fn-type="coi-statement">
<p>The authors declare no conflict of interest.</p>
</fn>
</fn-group>
<ref-list>
<title>References</title>
<ref id="R154"><mixed-citation>Adegoke J. O. and Carletion A. M., 2002. Relations between soil moisture and satellite vegetation indices in the U.S. Corn Belt. <italic>Journal of Hydrometeorology</italic>, 3, 395-405. <ext-link ext-link-type="uri" xlink:href="http://journals.ametsoc.org/doi/abs/10.1175/1525-7541(2002)003%3C0395:RBSMAS%3E2.0.CO;2">http://journals.ametsoc.org/doi/abs/10.1175/1525-7541(2002)003%3C0395:RBSMAS%3E2.0.CO;2</ext-link></mixed-citation></ref>
<ref id="R155"><mixed-citation>Bauer M. E., Burk T. E., Ek A. R., Coppin P. R., Lime S. D., Walsh T. A., Walters D. K., B. William and Heinzen, D. F. 1994. Inventory of Minnesota forest resources with Landsat TM data. <italic>Photogramm. Eng. Remote Sensing</italic>, 60, 287-298. <ext-link ext-link-type="uri" xlink:href="https://www.asprs.org/wp-content/uploads/pers/1994journal/mar/1994_mar_287-298.pdf">https://www.asprs.org/wp-content/uploads/pers/1994journal/mar/1994_mar_287-298.pdf</ext-link></mixed-citation></ref>
<ref id="R156"><mixed-citation>Bhagat V. S. and Sonawane K. R., 2010. Use of LANDSAT ETM+ Data for delineation of water bodies in hilly zones. <italic>Journal of Hydroinformatics</italic>, 13, 661-671. <pub-id pub-id-type="doi">10.2166/hydro.2010.018</pub-id> <ext-link ext-link-type="uri" xlink:href="http://jh.iwaponline.com/content/ppiwajhydro/13/4/661.full.pdf">http://jh.iwaponline.com/content/ppiwajhydro/13/4/661.full.pdf</ext-link></mixed-citation></ref>
<ref id="R157"><mixed-citation>Bhagat V. S., 2012. Use of Remote Sensing Techniques for Robust Digital Change Detection of Land: A Review. <italic>Recent Patents on Space Technology</italic>, 2 (2), 123-144. <pub-id pub-id-type="doi">10.2174/1877611611202020123</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.eurekaselect.com/102894">http://www.eurekaselect.com/102894</ext-link></mixed-citation></ref>
<ref id="R158"><mixed-citation>Bhagat V.S., 2009. Use of LANDSAT ETM+ data for detection of potential areas for afforestation. <italic>International Journal of Remote Sensing, 30, 2607-2617</italic>. <pub-id pub-id-type="doi">10.1080/01431160802552793</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431160802552793">http://www.tandfonline.com/doi/abs/10.1080/01431160802552793</ext-link></mixed-citation></ref>
<ref id="R159"><mixed-citation>Breda N.J., 2003. Ground-based measurements of leaf area index: a review of methods, instruments and current controversies. <italic>Journal of Experimental Botany</italic>, 54, 2403-2417. <pub-id pub-id-type="doi">10.1093/jxb/erg263</pub-id> <ext-link ext-link-type="uri" xlink:href="https://academic.oup.com/jxb/article/54/392/2403/621920/Ground-based-measurements-of-leaf-area-index-a">https://academic.oup.com/jxb/article/54/392/2403/621920/Ground-based-measurements-of-leaf-area-index-a</ext-link></mixed-citation></ref>
<ref id="R160"><mixed-citation>Burnicki A.C., Brown D. G. and Goovaerts P., 2010. Propagating error in land-cover-change analyses: impact of temporal dependence under increased thematic complexity. <italic>International Journal of Geographical Information Science</italic>, 24, 1043-1060. <pub-id pub-id-type="doi">10.1080/13658810903279008</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/13658810903279008">http://www.tandfonline.com/doi/abs/10.1080/13658810903279008</ext-link></mixed-citation></ref>
<ref id="R161"><mixed-citation>Cano F., Cerrillo R. M. N., Ferrer A. G., De La Orden M.S., 2006. Detection of forest decline using IKONOS sensor for Cork Oak woods in South Spain. <italic>Geocarto International</italic>, 21, 13-18. <pub-id pub-id-type="doi">10.1080/10106040608542388</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040608542388">http://www.tandfonline.com/doi/abs/10.1080/10106040608542388</ext-link></mixed-citation></ref>
<ref id="R162"><mixed-citation>Carlowicz M., 2012. Seeing Forests for the Trees and the Carbon: Mapping the World’s Forests in Three Dimensions. <ext-link ext-link-type="uri" xlink:href="http://earthobservatory.nasa.gov/Features/ForestCarbon/">http://earthobservatory.nasa.gov/Features/ForestCarbon/</ext-link></mixed-citation></ref>
<ref id="R163"><mixed-citation>Dhakal A. S., Amada T., Aniya M. and Sharm R. R., 2002. Detection of areas associated with flood and erosion caused by a heavy rainfall using multitemporal Landsat TM data, <italic>Photogram. Eng. Remote Sens</italic>., 68, 233-239.</mixed-citation></ref>
<ref id="R164"><mixed-citation>Chander G. and Markham B., 2003. Revised Landsat-5 TM radiometric calibration procedures and postcalibration dynamic ranges. <italic>IEEE Transactions on Geoscience and Remote Sensing</italic>, 41 (11), 2674−2677. <pub-id pub-id-type="doi">10.1109/TGRS.2003.818464</pub-id> <ext-link ext-link-type="uri" xlink:href="http://ieeexplore.ieee.org/abstract/document/1245255/">http://ieeexplore.ieee.org/abstract/document/1245255/</ext-link></mixed-citation></ref>
<ref id="R165"><mixed-citation>Chandio I.A., Matori A.N., Lawal D.U., Sabri S., 2011. GIS-based land suitability analysis using AHP for public parks planning in Larkana City. <italic>Mod. Appl. Sci</italic>. 5 (4), 177–189. <ext-link ext-link-type="uri" xlink:href="http://search.proquest.com/openview/454ff7e782e0abcd037128adfdbc5f2f/1.pdf?pq-origsite=gscholar&amp;cbl=307059">http://search.proquest.com/openview/454ff7e782e0abcd037128adfdbc5f2f/1.pdf?pq-origsite=gscholar&amp;cbl=307059</ext-link></mixed-citation></ref>
<ref id="R166"><mixed-citation>Chapelle O., Vapnik V., Bousquet O. and Mukherjee S., 2002. Choosing multiple parameters for support vector machines. <italic>Machine Learning</italic>, 46, 131–159. <pub-id pub-id-type="doi">10.1023/a:1012450327387</pub-id> <ext-link ext-link-type="uri" xlink:href="http://link.springer.com/article/10.1023/a:1012450327387">http://link.springer.com/article/10.1023/a:1012450327387</ext-link></mixed-citation></ref>
<ref id="R167"><mixed-citation>Chan J. C. W., Chan K. P. and Yeh A. G. O., 2001. Detecting the nature of change in an urban environment: A comparison of machine learning algorithms. <italic>Photogrammetric Engineering &amp; Remote Sensing</italic>, 67 (2), 213-225. <ext-link ext-link-type="uri" xlink:href="http://info.asprs.org/publications/pers/2001journal/february/2001_feb_213-225.pdf">http://info.asprs.org/publications/pers/2001journal/february/2001_feb_213-225.pdf</ext-link></mixed-citation></ref>
<ref id="R168"><mixed-citation>Chen G., Hay G. J., Carvalho Luis M. T. and Wulder M. A., 2012. Object-based change detection. <italic>Int. J. Remote Sens.</italic>, 33, 4434-4457. <pub-id pub-id-type="doi">10.1080/01431161.2011.648285</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431161.2011.648285">http://www.tandfonline.com/doi/abs/10.1080/01431161.2011.648285</ext-link></mixed-citation></ref>
<ref id="R169"><mixed-citation>Chowdhury A. B., Paul C. J., Martyn G. F. and Geoff K. F. 2005. The role of cuticular waxes and surface roughness in determining the insecticidal efficacy of deltamethrin and dimethoate applied as emulsifiable concentrates to leaf surfaces. <italic>International Journal of Pest Management</italic>, 51, 253-263. <pub-id pub-id-type="doi">10.1080/09670870500404674</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/09670870500404674">http://www.tandfonline.com/doi/abs/10.1080/09670870500404674</ext-link></mixed-citation></ref>
<ref id="R170"><mixed-citation>Clifton C., 2003. Change detection in overhead imagery using Neural Networks. <italic>Applied Intelligence</italic>, 18, 215-234. <pub-id pub-id-type="doi">10.1023/A:1021942526896</pub-id> <ext-link ext-link-type="uri" xlink:href="http://link.springer.com/article/10.1023/A:1021942526896">http://link.springer.com/article/10.1023/A:1021942526896</ext-link></mixed-citation></ref>
<ref id="R171"><mixed-citation>Cohen W. B., Fiorella M., Gray J., Helmer E. and Anderson K., 1998. An efficient and accurate method for mapping forest clearcuts in the Pacific Northwest using Landsat imagery. <italic>Photogram.</italic> <italic>Eng. Remote Sens.</italic>, 64, 293-300. <ext-link ext-link-type="uri" xlink:href="http://myweb.facstaff.wwu.edu/wallin/envr442/pdf_files/cohen_etal_1998_apr_293-300.pdf">http://myweb.facstaff.wwu.edu/wallin/envr442/pdf_files/cohen_etal_1998_apr_293-300.pdf</ext-link></mixed-citation></ref>
<ref id="R172"><mixed-citation>Cohen W. B., Maierspergerb T. K., Gowerc S. T. and Turner D. P., 2003. An improved strategy for regression of biophysical variables and Landsat ETM+ data. <italic>Remote Sens. Environ</italic>., 84, 561-571. <pub-id pub-id-type="doi">10.1016/S0034-4257(02)00173-6</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425702001736">http://www.sciencedirect.com/science/article/pii/S0034425702001736</ext-link></mixed-citation></ref>
<ref id="R173"><mixed-citation>Coppin P. and Bauer M., 1996. Digital change detection in forest ecosystems with remote sensing imagery. <italic>Remote Sensing Reviews</italic>, 13, 207- 234. <pub-id pub-id-type="doi">10.1080/02757259609532305</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/02757259609532305">http://www.tandfonline.com/doi/abs/10.1080/02757259609532305</ext-link></mixed-citation></ref>
<ref id="R174"><mixed-citation>Coppin P., Jonckheere I., Nackaerts K., Muys B. and Lambin E., 2004. Digital change detection methods in ecosystem monitoring: a review. <italic>Int. J. Remote Sens.</italic>, 25, 1565-1596. <pub-id pub-id-type="doi">10.1080/0143116031000101675</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/0143116031000101675">http://www.tandfonline.com/doi/abs/10.1080/0143116031000101675</ext-link></mixed-citation></ref>
<ref id="R175"><mixed-citation>Crist E. P. and Kauth R. J., 1986. The tasseled cap de-mystified, <italic>Photogram.</italic> <italic>Eng. Remote Sen</italic>., 52, 81-86. <ext-link ext-link-type="uri" xlink:href="https://www.asprs.org/wp-content/uploads/pers/1986journal/jan/1986_jan_81-86.pdf">https://www.asprs.org/wp-content/uploads/pers/1986journal/jan/1986_jan_81-86.pdf</ext-link></mixed-citation></ref>
<ref id="R176"><mixed-citation>Dhakal A. S., Amada T., Aniya M. and Sharm R. R., 2002. Detection of areas associated with flood and erosion caused by a heavy rainfall using multitemporal Landsat TM data, <italic>Photogram. Eng. Remote Sens</italic>., 68, 233-239.</mixed-citation></ref>
<ref id="R177"><mixed-citation>Drescher M. and Perera A. H., 2010. Comparing two sets of forest cover change knowledge used in forest landscape management planning. <italic>Journal of Environmental Planning and Management</italic>, 53 (5), 591-613. <pub-id pub-id-type="doi">10.1080/09640561003727110</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/09640561003727110">http://www.tandfonline.com/doi/abs/10.1080/09640561003727110</ext-link></mixed-citation></ref>
<ref id="R178"><mixed-citation>Du Y., Teillet P. M. and Cihlar J., 2002. Radiometric normalization of multitemporal high-resolution satellite images with quality control for land cover change detection. <italic>Remote Sensing Environment</italic>, 82, 123-134. <pub-id pub-id-type="doi">10.1016/S0034-4257(02)00029-9</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425702000299">http://www.sciencedirect.com/science/article/pii/S0034425702000299</ext-link></mixed-citation></ref>
<ref id="R179"><mixed-citation>Ehlers M., Klonus S., Astrand P. J. and Rosso P., 2010. Multi- sensor image fusion for pansharpening in remote sensing. <italic>International Journal of Image and Data Fusion</italic>, 1, 25-45. <pub-id pub-id-type="doi">10.1080/19479830903561985</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/19479830903561985">http://www.tandfonline.com/doi/abs/10.1080/19479830903561985</ext-link></mixed-citation></ref>
<ref id="R180"><mixed-citation>Fang H. and Xu J., 2000. Land cover and vegetation change in the Yellow river delta nature reserve analysed with Landsat Thematic Mapper data. <italic>Geocarto International</italic>, 15, 43-50. <pub-id pub-id-type="doi">10.1080/10106040008542171</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040008542171">http://www.tandfonline.com/doi/abs/10.1080/10106040008542171</ext-link></mixed-citation></ref>
<ref id="R181"><mixed-citation>Fastring D. R. and Griffith J. A., 2009. Malaria incidence in Nairobi, Kenya and decadal trends in NDVI and climatic variables. <italic>Geocarto International</italic>, 24 (3), 207-221 <pub-id pub-id-type="doi">10.1080/10106040802491835</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040802491835">http://www.tandfonline.com/doi/abs/10.1080/10106040802491835</ext-link></mixed-citation></ref>
<ref id="R182"><mixed-citation>Felkar P., Cannell G. H. and Clark P. R., 1981. Variation in growth among 13 Prosopis (mesquite) species. <italic>Experimental Agriculture</italic>, 28, 209-218. <pub-id pub-id-type="doi">10.1017/S0014479700011479</pub-id> <ext-link ext-link-type="uri" xlink:href="https://www.cambridge.org/core/journals/experimental-agriculture/article/div-classtitlevariation-in-growth-among-13-prosopis-mesquite-speciesdiv/73E765DDF1AF175C3C1399D6EFEE7398">https://www.cambridge.org/core/journals/experimental-agriculture/article/div-classtitlevariation-in-growth-among-13-prosopis-mesquite-speciesdiv/73E765DDF1AF175C3C1399D6EFEE7398</ext-link></mixed-citation></ref>
<ref id="R183"><mixed-citation>Fettig C. J., Klepzig K. D., Billings R. F., Munson A. S., Nowak J. T., Nebeker T. E., and Negro J. F., 2007. The effectiveness of vegetation management practices for prevention and control of bark beetle infestations in coniferous forests of the western and southern United States, <italic>Forest Ecology and Management</italic>, 238, 24-53. <pub-id pub-id-type="doi">10.1016/j.foreco.2006.10.011</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0378112706010310">http://www.sciencedirect.com/science/article/pii/S0378112706010310</ext-link></mixed-citation></ref>
<ref id="R184"><mixed-citation>Fletcher R. S. and Everitt J. H., 2007. A six-camera digital video imaging system sensitive to visible, red edge, near-infrared, and mid-infrared wavelengths. <italic>Geocarto International</italic>, 22 (2), 75-86. <pub-id pub-id-type="doi">10.1080/10106040701201608</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040701201608">http://www.tandfonline.com/doi/abs/10.1080/10106040701201608</ext-link></mixed-citation></ref>
<ref id="R185"><mixed-citation>Food and Agriculture Organisation (FAO), 2015. Global Forest Resources Assessment 2015. Available online at <ext-link ext-link-type="uri" xlink:href="http://www.fao.org/3/a-i4793e.pdf">http://www.fao.org/3/a-i4793e.pdf</ext-link></mixed-citation></ref>
<ref id="R186"><mixed-citation>Forkuo E. K. and Frimpong A., 2012. Analysis of Forest Cover Change Detection. <italic>International Journal of Remote Sensing Applications</italic>, 2 (4), 82-92.</mixed-citation></ref>
<ref id="R187"><mixed-citation>Gareeau R. M., Gutierrez M., Ham T., Nemethy J. and Hackett J., 2009. Geospatial image change detecting system and associated method. US 7528938 B2. <ext-link ext-link-type="uri" xlink:href="https://docs.google.com/viewer?url=patentimages.storage.googleapis.com/pdfs/US7528938.pdf">https://docs.google.com/viewer?url=patentimages.storage.googleapis.com/pdfs/US7528938.pdf</ext-link></mixed-citation></ref>
<ref id="R188"><mixed-citation>Ghosh A., Mishra N. S. and Ghosh S., 2011. Fuzzy clustering algorithms for unsupervised change detection in remote sensing images. <italic>Information Sci.</italic>, 181 (4), 699–715 <pub-id pub-id-type="doi">10.1016/j.ins.2010.10.016</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0020025510005153">http://www.sciencedirect.com/science/article/pii/S0020025510005153</ext-link></mixed-citation></ref>
<ref id="R189"><mixed-citation>Gill T.K., Danaher T., Gillingham S.S. and Mitchell R.M., 2012. Comparing bright- target surface spectral-reflectance estimates obtained from IRS P6 LISS III to those obtained from Landsat 5 TM and Landsat 7 ETM+. <italic>Remote Sensing Letters</italic>, 3 (2), 121-130. <pub-id pub-id-type="doi">10.1080/01431161.2010.543180</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431161.2010.543180">http://www.tandfonline.com/doi/abs/10.1080/01431161.2010.543180</ext-link></mixed-citation></ref>
<ref id="R190"><mixed-citation>Giri Babu D., Raja Shekhar S. S., Chandrasekar K., Sesha Sai M. V. R., Diwakar P. G. and Dadhwal V.K., 2014. Automatic image registration using free and open source software. <italic>The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences</italic>, XL-8, 1121-1128. <pub-id pub-id-type="doi">10.5194/isprsarchives-XL-8-1121-2014</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XL-8/1121/2014/isprsarchives-XL-8-1121-2014.pdf">http://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XL-8/1121/2014/isprsarchives-XL-8-1121-2014.pdf</ext-link></mixed-citation></ref>
<ref id="R191"><mixed-citation>Globle Climate Observaion System (GCOS), 2010. Implementation plan for the Global Observing System for Climate in support of the UNFCCC. GCOS-138, WMO-TD/No. 1523, 1-136. <ext-link ext-link-type="uri" xlink:href="http://www.wmo.int/pages/prog/gcos/documents/GCOSIP-10_DRAFTv1.0_131109.pdf">http://www.wmo.int/pages/prog/gcos/documents/GCOSIP-10_DRAFTv1.0_131109.pdf</ext-link></mixed-citation></ref>
<ref id="R192"><mixed-citation>Goetz S. J., 1997. Multi-sensor analysis of NDVI, surface temperature and biophysical variables at a mixed grassland site. <italic>Int. J. Remote Sens</italic>., 18, 71-94. <pub-id pub-id-type="doi">10.1080/014311697219286</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/014311697219286">http://www.tandfonline.com/doi/abs/10.1080/014311697219286</ext-link></mixed-citation></ref>
<ref id="R193"><mixed-citation>Golmehr E., 2008. A remote sensing evaluation for agronomic land use mapping in Tehran Province, <italic>Iran. J. Appl. Sci. Environ.</italic> Manage., 12 (2), 43-46. <pub-id pub-id-type="doi">10.4314/jasem.v12i2.55527</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.ajol.info/index.php/jasem/article/view/55527">http://www.ajol.info/index.php/jasem/article/view/55527</ext-link></mixed-citation></ref>
<ref id="R194"><mixed-citation>Hansen H. O., 2000. An AIDS model with reproduction with an application based on data from Uganda. <italic>Mathematical Population Studies</italic>, 8, 175-203. <pub-id pub-id-type="doi">10.1080/08898480009525480</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/08898480009525480">http://www.tandfonline.com/doi/abs/10.1080/08898480009525480</ext-link></mixed-citation></ref>
<ref id="R195"><mixed-citation>Hara G. C., Cheriyadat A., Seo S. and Vijayaraj V., 2012. Change Analyst, US 8160296 B2. <ext-link ext-link-type="uri" xlink:href="http://patentimages.storage.googleapis.com/pdfs/US8160296.pdf">http://patentimages.storage.googleapis.com/pdfs/US8160296.pdf</ext-link></mixed-citation></ref>
<ref id="R196"><mixed-citation>Hayes D. and Sader S. A., 2001. Change detection techniques for monitoring forest clearing and regrowth in a tropical moist forest. <italic>Photogrammetric Engineering and Remote Sensing</italic>, 67, 1067-1075. <ext-link ext-link-type="uri" xlink:href="https://www.asprs.org/wp-content/uploads/pers/2001journal/september/2001_sept_1067-1075.pdf">https://www.asprs.org/wp-content/uploads/pers/2001journal/september/2001_sept_1067-1075.pdf</ext-link></mixed-citation></ref>
<ref id="R197"><mixed-citation>Healey S. P., Cohen W. B., Zhiqiang Y. and Krankina O. N., 2008. Comparison of tasseled cap-based Landsat data structures for use in forest disturbance detection. <italic>Remote Sens. Environ.</italic>, 97, 301-310. <pub-id pub-id-type="doi">10.1016/j.rse.2005.05.009</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425705001586">http://www.sciencedirect.com/science/article/pii/S0034425705001586</ext-link></mixed-citation></ref>
<ref id="R198"><mixed-citation>Huang C., Wylie B., Yang L., Homar C. and Zylstra G., 2002. Derivation of a tasselled cap transformation based on Landsat-7 at-satellite reflectance. <italic>International Journal of Remote Sensing</italic>, 23, 1741-1748. <pub-id pub-id-type="doi">10.1080/01431160110106113</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431160110106113?journalCode=tres20">http://www.tandfonline.com/doi/abs/10.1080/01431160110106113?journalCode=tres20</ext-link></mixed-citation></ref>
<ref id="R199"><mixed-citation>Huete A., Didan K., Miura T., Rodriguez E. P., Gao X. and Ferreira L. G., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. <italic>Remote Sens. Environ.</italic>, 83, 195-213. <pub-id pub-id-type="doi">10.1016/S0034-4257(02)00096-2</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425702000962">http://www.sciencedirect.com/science/article/pii/S0034425702000962</ext-link></mixed-citation></ref>
<ref id="R200"><mixed-citation>Im J., Jensen J. R. and Tullis J. A., 2008. Object‐based change detection using correlation image analysis and image segmentation. <italic>International Journal of Remote Sensing</italic>, 29 (2), 399-423, <pub-id pub-id-type="doi">10.1080/01431160601075582</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431160601075582">http://www.tandfonline.com/doi/abs/10.1080/01431160601075582</ext-link></mixed-citation></ref>
<ref id="R201"><mixed-citation>Jensen R. R., 2002. Spatial and temporal leaf area index dynamics in a North Central Florida, USA Preserve. <italic>Geocarto International</italic>, 17 (4), 47-54. <pub-id pub-id-type="doi">10.1080/10106040208542253</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040208542253">http://www.tandfonline.com/doi/abs/10.1080/10106040208542253</ext-link></mixed-citation></ref>
<ref id="R202"><mixed-citation>Jiany G., Haiganga S., Guoruia M. and Qiming Z., 2008. A review of multi-temporal remote sensing data change algorithms, The International Archives of the Photogrammetry. <italic>Remote Sens. Spatial Infor. Sci.,</italic> 37 (B7), 757-762. <ext-link ext-link-type="uri" xlink:href="http://www.isprs.org/proceedings/XXXVII/congress/7_pdf/5_WG-VII-5/05.pdf">http://www.isprs.org/proceedings/XXXVII/congress/7_pdf/5_WG-VII-5/05.pdf</ext-link></mixed-citation></ref>
<ref id="R203"><mixed-citation>Jonckheere I., Fleck S., Nackaerts K., Muysa B., Coppin P., Weiss M. and Baret F., 2004. Review of methods for in situ leaf area index determination Part I. Theories, sensors and hemispherical photography. <italic>Agricultural and Forest Meteorology</italic>, 121, 19-35. <pub-id pub-id-type="doi">10.1016/j.agrformet.2003.08.027</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0168192303001643">http://www.sciencedirect.com/science/article/pii/S0168192303001643</ext-link></mixed-citation></ref>
<ref id="R204"><mixed-citation>Julien Y., Sobrino J. A., Mattar C., Ruescas B., Jimenez-mun J. A., Sorai G., Hidalgo V., Atitar M., Franch B. and Cuenca J., 2011. Temporal analysis of normalized difference vegetation index (NDVI) and land surface temperature (LST) parameters to detect changes in the Iberian land cover between 1981 and 2001. <italic>International Journal of Remote Sensing</italic>, 32, 2057-2068. <pub-id pub-id-type="doi">10.1080/01431161003762363</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431161003762363">http://www.tandfonline.com/doi/abs/10.1080/01431161003762363</ext-link></mixed-citation></ref>
<ref id="R205"><mixed-citation>Keenan R. J., Lamb D., Parrotta J. and Kikkawa J., 1999. Ecosystem management in tropical timber plantations. <italic>Journal of Sustainable Forestry</italic>, 9 (1-2), 117-134. <pub-id pub-id-type="doi">10.1300/J091v09n01_10</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1300/J091v09n01_10">http://www.tandfonline.com/doi/abs/10.1300/J091v09n01_10</ext-link></mixed-citation></ref>
<ref id="R206"><mixed-citation>Kennedy R. E., Townsend P. A., Gross J. E., Cohen W. B., Bolstad P., Wang Y. Q. and Adams P., 2009. Remote sensing change detection tools for natural resource managers: Understanding concepts and tradeoffs in the design of landscape monitoring projects. <italic>Remote Sens. Environ</italic>., 113, 1382-1396. <pub-id pub-id-type="doi">10.1016/j.rse.2008.07.018</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425709000601">http://www.sciencedirect.com/science/article/pii/S0034425709000601</ext-link></mixed-citation></ref>
<ref id="R207"><mixed-citation>Kim A. M., Lui D. and Gong P., 2004. Change detection from SOPT Panchromatic imagery at the urban-rural fringe of Ho Chi Minh City, Vietnam. <italic>Geographic Information Sci.</italic>, 10, 42-48. <pub-id pub-id-type="doi">10.1080/10824000409480653</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10824000409480653">http://www.tandfonline.com/doi/abs/10.1080/10824000409480653</ext-link></mixed-citation></ref>
<ref id="R208"><mixed-citation>Kumsap C., Borne F. and Moss D., 2005. The technique of distance decayed visibility for forest landscape visualization. <italic>International Journal of Geographical Information Science</italic>, 19 (6), 723-744. <pub-id pub-id-type="doi">10.1080/13658810500104880</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/13658810500104880">http://www.tandfonline.com/doi/abs/10.1080/13658810500104880</ext-link></mixed-citation></ref>
<ref id="R209"><mixed-citation>Lawrence R. and Wrlght A., 2001. Rule-based classification systems using classification and regression tree (CART) Analysis. <italic>Photogram. Eng. Remote Sens.</italic>, 67, 1137-1142. <ext-link ext-link-type="uri" xlink:href="https://pdfs.semanticscholar.org/0a3c/ec5d1c1c1ba8ee2ce44dcae1b2bcc93c5519.pdf">https://pdfs.semanticscholar.org/0a3c/ec5d1c1c1ba8ee2ce44dcae1b2bcc93c5519.pdf</ext-link></mixed-citation></ref>
<ref id="R210"><mixed-citation>Lippitt C. D., Coulter L. L., Freeman M., Lamantia J., Pang W. and Stow D. A., 2011. The effect of input data transformations on object based image analysis. <italic>Remote Sensing Letters</italic>, 3, 21-29. <pub-id pub-id-type="doi">10.1080/01431161.2010.531059</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431161.2010.531059">http://www.tandfonline.com/doi/abs/10.1080/01431161.2010.531059</ext-link></mixed-citation></ref>
<ref id="R211"><mixed-citation>Lu D., Moran E. and Hetrick S., 2011. Detection of impervious surface change with multitemporal Landsat images in an urban-rural frontier. <italic>ISPRS J. Photogrammetry Remote Sens.</italic>, 66, 298-306. <pub-id pub-id-type="doi">10.1016/j.isprsjprs.2010.10.010</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0924271610001176">http://www.sciencedirect.com/science/article/pii/S0924271610001176</ext-link></mixed-citation></ref>
<ref id="R212"><mixed-citation>Lunetta R. S., Knight J. F., Ediriwickrema J., Lyon J. G. and Worthy L. D., 2006. Land-cover change detection using multi-temporal MODIS NDVI data. <italic>Remote Sens. Environ.</italic>, 105, 142- 154. <pub-id pub-id-type="doi">10.1016/j.rse.2006.06.018</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425706002549">http://www.sciencedirect.com/science/article/pii/S0034425706002549</ext-link></mixed-citation></ref>
<ref id="R213"><mixed-citation>Majed M. A. Z., Abdullah T. and Abdulrahman A. A., 2011. Soil erosion control and moisture conservation of arid lands with stone cover. <italic>Arid Land Research and Management</italic>, 25 (3), 294-307. <pub-id pub-id-type="doi">10.1080/15324982.2011.565859</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/15324982.2011.565859">http://www.tandfonline.com/doi/abs/10.1080/15324982.2011.565859</ext-link></mixed-citation></ref>
<ref id="R214"><mixed-citation>Mishra A., Sharma S. D. and Gupta M. K., 2003. Soil rehabilitation through afforestation: Evaluation of the performance of Prosopis juliflora, Dalbergia sissoo, and Eucalyptus tereticornis plantations in a sodic environment, <italic>Arid Land Research and Management</italic>, 17, 257–269. <pub-id pub-id-type="doi">10.1080/15324980301597</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/15324980301597">http://www.tandfonline.com/doi/abs/10.1080/15324980301597</ext-link></mixed-citation></ref>
<ref id="R215"><mixed-citation>Morisette J. T., Baret F., Privette J. L., Myneni R. B., Nickeson J. E., Garrigues S., Shabanov N. V., Weiss M., Fernandes R. A., Leblanc S. G., Kalacska M., Sánchez-Azofeifa G. A., Chubey M., Rivard B., Stenberg P., Rautiainen M., Voipio P., Manninen T., Pilant A. N., Lewis T. E., Iiames J. S., Colombo R., Meroni M., Busetto L., Cohen W.B., Turner D. P., Warner E. D., Petersen G. W., Seufert G., and Cook R., 2006. Validation of global moderate-resolution LAI products: A framework proposed within the CEOS land product validation subgroup. <italic>IEEE Transactions on Geoscience and Remote Sensing</italic>, 44, 1804-1817. <pub-id pub-id-type="doi">10.1109/TGRS.2006.872529</pub-id> <ext-link ext-link-type="uri" xlink:href="http://ieeexplore.ieee.org/document/1645281/">http://ieeexplore.ieee.org/document/1645281/</ext-link></mixed-citation></ref>
<ref id="R216"><mixed-citation>Mountrakis G., Im J. and Ogole C., 2011. Support vector machines in remote sensing: A review. <italic>ISPRS Journal of Photogrammetry Remote Sensing</italic>, 66(3), 247–259. <pub-id pub-id-type="doi">10.1016/j.isprsjprs.2010.11.001</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0924271610001140">http://www.sciencedirect.com/science/article/pii/S0924271610001140</ext-link></mixed-citation></ref>
<ref id="R217"><mixed-citation>Mróz M. and Sobieraj A., 2004. Comparison of several vegetation indices calculated on the basis of a seasonal SPOT XS time series, and their stability for land cover and agricultural crop identification. <italic>Technical Sciences</italic>, 7, 40-66. <ext-link ext-link-type="uri" xlink:href="http://www.uwm.edu.pl/wnt/technicalsc/ts7_2004/4_7_2004.pdf">http://www.uwm.edu.pl/wnt/technicalsc/ts7_2004/4_7_2004.pdf</ext-link></mixed-citation></ref>
<ref id="R218"><mixed-citation>Munyati C., 2004. Use of principal component analysis (PCA) of remote sensing images in wetland change detection on the Kafue Flats, Zambia. <italic>Geocarto Int.</italic>, 19, 11-22. <pub-id pub-id-type="doi">10.1080/10106040408542313</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040408542313">http://www.tandfonline.com/doi/abs/10.1080/10106040408542313</ext-link></mixed-citation></ref>
<ref id="R219"><mixed-citation>NASA. <italic>Landsat 7 Science Data Users Handbook</italic>. <ext-link ext-link-type="uri" xlink:href="https://landsat.gsfc.nasa.gov/landsat-7-science-data-users-handbook/">https://landsat.gsfc.nasa.gov/landsat-7-science-data-users-handbook/</ext-link></mixed-citation></ref>
<ref id="R220"><mixed-citation>Nemani R. R. and Running S. W., 1989. Estimation of regional surface resistance to evapotranspiration from NDVI and Thermal-IR AVHRR data. <italic>J. Appl. Meteorol</italic>, 28, 276-284. <ext-link ext-link-type="uri" xlink:href="http://journals.ametsoc.org/doi/abs/10.1175/1520-0450(1989)028%3C0276%3AEORSRT%3E2.0.CO%3B2">http://journals.ametsoc.org/doi/abs/10.1175/1520-0450(1989)028%3C0276%3AEORSRT%3E2.0.CO%3B2</ext-link></mixed-citation></ref>
<ref id="R221"><mixed-citation>Nemani R., Pierce L., Running S. W. and Goward S., 1993. Developing satellite derived estimates of surface moisture status. <italic>J. Appl. Meteorol</italic>, 32, 548-557. <ext-link ext-link-type="uri" xlink:href="http://journals.ametsoc.org/doi/abs/10.1175/1520-0450(1993)032%3C0548%3ADSDEOS%3E2.0.CO%3B2">http://journals.ametsoc.org/doi/abs/10.1175/1520-0450(1993)032%3C0548%3ADSDEOS%3E2.0.CO%3B2</ext-link></mixed-citation></ref>
<ref id="R222"><mixed-citation>Owen T. W., Carlson T. N. and Gillies R. R., 1998. An assessment of satellite remotely-sensed land cover parameters in quantitatively describing the climatic effect of urbanization. <italic>International Journal of Remote Sensing</italic>, 19, 1663 -1681. <pub-id pub-id-type="doi">10.1080/014311698215171</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/014311698215171">http://www.tandfonline.com/doi/abs/10.1080/014311698215171</ext-link></mixed-citation></ref>
<ref id="R223"><mixed-citation>Pajares G., Alonso C., Cruz J. M. and Moreno V., 2002. Shade identification in urban areas through the bayesian classifier, <italic>Proc. European Symposium on Intelligent Technologies</italic>, 476-481.</mixed-citation></ref>
<ref id="R224"><mixed-citation>Panigrahy S., Anitha D., Kimothi M. M. and Singh S. P., 2010. Timberline change detection using topographic map and satellite imagery. Tropical Ecology, 51, 87-91. <ext-link ext-link-type="uri" xlink:href="http://www.tropecol.com/pdf/open/PDF_51_1/Jour.10.pdf">http://www.tropecol.com/pdf/open/PDF_51_1/Jour.10.pdf</ext-link></mixed-citation></ref>
<ref id="R225"><mixed-citation>Pant T., Singh D. and Srivastava T., 2010. The potential application of fractal approach for surface roughness retrieval: A study for simulated surfaces. <italic>Geomatics Natural Hazards and Risk,</italic> 1, 243-257. <pub-id pub-id-type="doi">10.1080/19475705.2010.494835</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/19475705.2010.494835">http://www.tandfonline.com/doi/abs/10.1080/19475705.2010.494835</ext-link></mixed-citation></ref>
<ref id="R226"><mixed-citation>Petit C. C. and Lambin E. F., 2001. Integration of multi-source remote sensing data for land cover change detection. <italic>International Journal of Geographical Information Science</italic>, 15, 785-803. <pub-id pub-id-type="doi">10.1080/13658810110074483</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/13658810110074483">http://www.tandfonline.com/doi/abs/10.1080/13658810110074483</ext-link></mixed-citation></ref>
<ref id="R227"><mixed-citation>Pouliot D. A., King D. J, Bell F. W. and Pitt D. G., 2002. Automated tree crown detection and delineation in high-resolution digital camera imagery of coniferous forest regeneration. <italic>Remote Sens. Environ</italic>., 82, 322-334. <pub-id pub-id-type="doi">10.1016/S0034-4257(02)00050-0</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425702000500">http://www.sciencedirect.com/science/article/pii/S0034425702000500</ext-link></mixed-citation></ref>
<ref id="R228"><mixed-citation>Pu R., Gong P., Tian Y., Miao X., CarruthersI R. I. and Anderson G. L., 2008. Invasive species change detection using artificial neural networks and CASI hyperspectral imagery. Environ Monitoring Assessment, 140, 5-32. <pub-id pub-id-type="doi">10.1007/s10661-007-9843-7</pub-id> <ext-link ext-link-type="uri" xlink:href="http://link.springer.com/article/10.1007%2Fs10661-007-9843-7?LI=true">http://link.springer.com/article/10.1007%2Fs10661-007-9843-7?LI=true</ext-link></mixed-citation></ref>
<ref id="R229"><mixed-citation>Rahman R. and Saha S. K., 2008. Multi-resolution segmentation for object-based classification and accuracy assessment of land Use /land cover classification using remotely sensed data. <italic>Journal of the Indian Society of Remote Sensing</italic>, 36, 189-201. <pub-id pub-id-type="doi">10.1007/s12524-008-0020-4</pub-id> <ext-link ext-link-type="uri" xlink:href="http://link.springer.com/article/10.1007%2Fs12524-008-0020-4?LI=true">http://link.springer.com/article/10.1007%2Fs12524-008-0020-4?LI=true</ext-link></mixed-citation></ref>
<ref id="R230"><mixed-citation>Roberts J. W., Aardt J. A. N. V. and Ahmed F. B., 2011. Image fusion for enhanced forest structural assessment. <italic>International Journal of Remote Sensing</italic>, 32 (1), 243-266. <pub-id pub-id-type="doi">10.1080/01431160903463684</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431160903463684">http://www.tandfonline.com/doi/abs/10.1080/01431160903463684</ext-link></mixed-citation></ref>
<ref id="R231"><mixed-citation>Rozenstein O. and Karnieli A., 2011. Comparison of methods for landuse classification incorporating remote sensing and GIS inputs. <italic>Applied Geography</italic>, 31, 533-544. <pub-id pub-id-type="doi">10.1016/j.apgeog.2010.11.006</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0143622810001517">http://www.sciencedirect.com/science/article/pii/S0143622810001517</ext-link></mixed-citation></ref>
<ref id="R232"><mixed-citation>Schwilch G., Bestelmeyer B., Bunning S., Critchley W., Herrich J., Kellner K., Liniger H. P., Nachtergaele F., Ritsema C. J., Schuster B., Tabo R., Lynden G. V. and Winslow M., 2011. Experiences in monitoring and assessment of sustainable land management<italic>. Land Degrad. Develop.</italic>, 22, 214-225. <pub-id pub-id-type="doi">10.1002/ldr.1040</pub-id> <ext-link ext-link-type="uri" xlink:href="http://onlinelibrary.wiley.com/doi/10.1002/ldr.1040/full">http://onlinelibrary.wiley.com/doi/10.1002/ldr.1040/full</ext-link></mixed-citation></ref>
<ref id="R233"><mixed-citation>Shalaby A., Ouma Y. O. and Tateishi R., 2006. Land suitability assessment for perennial crops using remote sensing and geographic information systems: a case study in North-Western Egypt. <italic>Arch.</italic> <italic>Agron. Soil Sci</italic>., 52 (3), 243–261. <pub-id pub-id-type="doi">10.1080/03650340600627167</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/03650340600627167">http://www.tandfonline.com/doi/abs/10.1080/03650340600627167</ext-link></mixed-citation></ref>
<ref id="R234"><mixed-citation>Silleos N. G., Alexandridis T. K., Gitas I. Z. and Perakis K., 2006. Vegetation indices: advances made in biomass estimation and vegetation monitoring in the last 30 years<italic>. Geocarto International</italic>, 21 (4), 21-28. <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040608542399">http://www.tandfonline.com/doi/abs/10.1080/10106040608542399</ext-link></mixed-citation></ref>
<ref id="R235"><mixed-citation>Singh A., 1989. Digital change detection techniques using remotelysensed data. <italic>Internation Journal of Remote Sensing</italic>, 10, 989-1003. <pub-id pub-id-type="doi">10.1080/01431168908903939</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431168908903939">http://www.tandfonline.com/doi/abs/10.1080/01431168908903939</ext-link></mixed-citation></ref>
<ref id="R236"><mixed-citation>Singh G., 1996. Effect of site preparation techniques on Prosopis juliflora in an alkali soil. <italic>Forest Ecology and Management</italic>, 80, 267-278. <pub-id pub-id-type="doi">10.1016/0378-1127(95)03543-5</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/0378112795035435">http://www.sciencedirect.com/science/article/pii/0378112795035435</ext-link></mixed-citation></ref>
<ref id="R237"><mixed-citation>Sommer S., Zucca C., Grainger A., Cherlet M., Zougmore R., Sokona Y., Hill J., Della R., Peruta J. R. and Wang G., 2011. Application of indicator systems for monitoring and assessment of desertification from national to global scales, <italic>Land Degradation Development</italic>, 22, 184-197. <pub-id pub-id-type="doi">10.1002/ldr.1084</pub-id> <ext-link ext-link-type="uri" xlink:href="http://onlinelibrary.wiley.com/doi/10.1002/ldr.1084/abstract">http://onlinelibrary.wiley.com/doi/10.1002/ldr.1084/abstract</ext-link></mixed-citation></ref>
<ref id="R238"><mixed-citation>Southworth J., Munroe D. and Nagendra H., 2004. Land cover change and landscape fragmentation-comparing the utility of continuous and discrete analyses for a Western Honduras region. <italic>Agriculture Ecosystems and Environment</italic>, 101, 185-205. <pub-id pub-id-type="doi">10.1016/j.agee.2003.09.011</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0167880903003232">http://www.sciencedirect.com/science/article/pii/S0167880903003232</ext-link></mixed-citation></ref>
<ref id="R239"><mixed-citation>Tanriverdi C., 2010. Improved agricultural management using remote sensing to estimate water stress indices. <italic>Application in Remote Sensing Journal</italic>, 1, 19-24. <ext-link ext-link-type="uri" xlink:href="http://www.academyjournal.net/asj/index.php/ARS/article/download/31/112">http://www.academyjournal.net/asj/index.php/ARS/article/download/31/112</ext-link></mixed-citation></ref>
<ref id="R240"><mixed-citation>Theiler J. P. and Perkins J. S., 2011. Anomalous change detection in imagery, US 7953280 B2. <ext-link ext-link-type="uri" xlink:href="https://patentimages.storage.googleapis.com/pdfs/US7953280.pdf">https://patentimages.storage.googleapis.com/pdfs/US7953280.pdf</ext-link></mixed-citation></ref>
<ref id="R241"><mixed-citation>Tsai F. and Lin H. C., 2007. Polygon‐based texture mapping for cyber city 3D building models. <italic>International Journal of Geographical Information Science</italic>, 21 (9), 965-981. <pub-id pub-id-type="doi">10.1080/13658810601034929</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/13658810601034929?journalCode=tgis20">http://www.tandfonline.com/doi/abs/10.1080/13658810601034929?journalCode=tgis20</ext-link></mixed-citation></ref>
<ref id="R242"><mixed-citation>Turker M. and Derenyi E., 2000. GIS assisted change detection using remote sensing. <italic>Geocarto Int.</italic>, 15 (1), 51-56. <pub-id pub-id-type="doi">10.1080/10106040008542139</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/10106040008542139">http://www.tandfonline.com/doi/abs/10.1080/10106040008542139</ext-link></mixed-citation></ref>
<ref id="R243"><mixed-citation>Virk R. and King D., 2006. Comparison of techniques for forest change mapping using Landsat data in Karnataka, <italic>India. Geocarto Int</italic>., 21, 49-57. <ext-link ext-link-type="uri" xlink:href="http://www.glel.carleton.ca/PDF/geoPub/gc06.pdf">http://www.glel.carleton.ca/PDF/geoPub/gc06.pdf</ext-link></mixed-citation></ref>
<ref id="R244"><mixed-citation>Weng Q., 2009. Thermal infrared remote sensing for urban climate and environmental studies: Methods, applications and trends. <italic>ISPRS J. Photogrammetry Remote Sens.</italic>, 64, 335-344. <pub-id pub-id-type="doi">10.1016/j.isprsjprs.2009.03.007</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S092427160900046X">http://www.sciencedirect.com/science/article/pii/S092427160900046X</ext-link></mixed-citation></ref>
<ref id="R245"><mixed-citation>Yang G. G., Qu S. L., Jian H. F., Bao Y. H., Hong B. F., and Ze J. T., 2014. Identification of heavy metal sources in the reclaimed farmland soils of the pearl river estuary in China using a multivariate geostatistical approach. <italic>Ecotoxicology and Environmental Safety, </italic>105, 7-12<italic>.</italic> <pub-id pub-id-type="doi">10.1016/j.ecoenv.2014.04.003</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0147651314001432?np=y&amp;npKey=f0d702ab84986eadd25a6f7b5939047b37c17858edc6c2caa1599cffc30b3225">http://www.sciencedirect.com/science/article/pii/S0147651314001432?np=y&amp;npKey=f0d702ab84986eadd25a6f7b5939047b37c17858edc6c2caa1599cffc30b3225</ext-link></mixed-citation></ref>
<ref id="R246"><mixed-citation>Yarbrough L. D., Easson G. and Kuszmaul J. S., 2005. Using at-sensor radiance and reflectance tasseled cap transforms applied to change detection for the ASTER sensor. <italic>Analysis of Multi-Temporal Remote Sensing Images</italic>, 16, 41-145. <pub-id pub-id-type="doi">10.1109/AMTRSI.2005.1469857</pub-id> <ext-link ext-link-type="uri" xlink:href="http://ieeexplore.ieee.org/document/1469857/">http://ieeexplore.ieee.org/document/1469857/</ext-link></mixed-citation></ref>
<ref id="R247"><mixed-citation>Yuan F. and Bauer M. E., 2007. Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery. <italic>Remote Sensing of Environment</italic>, 106, 375 - 386. <pub-id pub-id-type="doi">10.1016/j.rse.2006.09.003</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0034425706003191">http://www.sciencedirect.com/science/article/pii/S0034425706003191</ext-link></mixed-citation></ref>
<ref id="R248"><mixed-citation>Zhang Q., Wang J., Peng X., Gong P. and Shi P., 2002. Urban build-up land change detection with road density and spectral information from multitemporal Landsat TM data. <italic>International Journal of Remote Sensing</italic>, 23, 3057–3078. <pub-id pub-id-type="doi">10.1080/01431160110104728</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.tandfonline.com/doi/abs/10.1080/01431160110104728">http://www.tandfonline.com/doi/abs/10.1080/01431160110104728</ext-link></mixed-citation></ref>
<ref id="R249"><mixed-citation>Zolekar R. B. and Bhagat V. S., 2015. Multi-criteria land suitability analysis for agriculture in hilly zone: Remote sensing and GIS approach. <italic>Computers and Electronics in Agriculture</italic>, 118, 300-321. <pub-id pub-id-type="doi">10.1016/j.compag.2015.09.016</pub-id> <ext-link ext-link-type="uri" xlink:href="http://www.sciencedirect.com/science/article/pii/S0168169915002914">http://www.sciencedirect.com/science/article/pii/S0168169915002914</ext-link></mixed-citation></ref>
</ref-list>
</back>
<floats-group>
<fig id="F39"><label>Figure 6</label><caption><title>Approach-II: Schematic preparation</title></caption><graphic xlink:href="https://gathacognition.com/uploads/admin/articlefigure/6b1427f7ab29edf58110a37038e24335.jpg" mimetype="image" mime-subtype="jpeg"/></fig>
</floats-group>
</article>
