Vegetation Change Detection of Neka River in Iran by Using Remote-sensing and GIS

Vegetation Change Detection of Neka River in Iran by Using Remote-sensing and GIS Hassan Ahmadi (Corresponding author) Department of studies in Geography University of Mysore, Mysore, India E-mail: ahamdineka@yahoo.com Asima Nusrath Reader at Department of Studies in Geography University of Mysore, Mysore, India Abstract Land use change has transformed a vast part of the natural landscapes of the developing world for the last 50 years. Land is a fundamental factor of production, and through much of the course of human history, it has been tightly coupled with economic growth. On the other hand, bare soil has recently increased and has become one of the most important land degradation processes in the Mediterranean basins. The land use has changed rapidly within and near Neka River which is a fast growing agricultural river Basin. The land use changes in this region were analyzed based on Landsat data from 1977 to 2001. Supervised/unsupervised classi cation approach, coupled with RS and GIS analyses, was employed to generate the change over land use/cover maps. In order to analyze landscape fragmentation, land-use change was calculated using normalized different vegetation index( NDVI). Based on the results of the analysis, the range of NDVI has changed from the reflection of 0.9597 to 0.2876 in 1977 to the reflectin of 0.6420 to 0.187 in 2001, which indicates that an increase in bare lands led to a decrease in forest lands. Keywords: Land use change, NDVI, Neka, Classification, RS, GIS 1. Introduction In developing countries where a large proportion of human population depends almost entirely on natural resources for their livelihoods, resulting in land-use and cover changes. There are increasing concerns for sustainable management of the land resources (e.g., natural vegetation). The political and economic condition in Iran after 1970s, favored a sudden increase in the application of scientific methods in agriculture and forestry. The exports of timber to Romania had taken a leap forward and there was a need for the application of quick and speedy production and exploitation of natural resources. It was at this time that the introduction of tractors and the use of machines in cutting trees were introduced in Iran. Due to the mechanization there was a rapid increase in the cutting of trees and disappearance of forest and reduction of the area under forest. The change in the temperature and rainfall began to be noticed from 2000 onwards, which the temperature and rains between 1977 and 2001 has remained constant. The vegetation has taken a noticeable decline, and the bare soil has suddenly increased. Due to these changes, the incidence of floods has also increased. The flood which occurred in 1998 was one of the most destructive causing a great damage to life and property. The impact of deforestation has led to an increase of barren land and soil erosion. These impacts have affected the water permeability of the soil. The rate of percolation has diminished, leading to the gushing of water into the Neka valley causing floods. It is in this context that the present study will aim to analyze the change in the forest area for a period of 25 years. In this study, some change detection techniques, radiance/reflectance band differencing, NDVI differencing, tasseled cap (KT), change vector differencing and NDVI Rationing changes the north of Iran environment used image differencing, image rationing and normalized difference vegetation index (NDVI) differencing to detect land-use changes in a coral mining area of India and found that no significant difference existed among these methods in detecting land use change. Dhaka et al. (2002) Maximum value compositing (MVC) is the most common form of NDVI compositing used to produce NDVI time-series data sets with minimal effects from 58 ISSN 1916-9779 E-ISSN 1916-9787

clouds and atmospheric scattering (Eidenshink and Faundeen, 1994; Holben, 1986). Frequent cloud cover in Arctic locations, a 14-day or half-monthly compositing period has been shown by Hope, Boynton, Stow and Douglas (2003) to produce NDVI images with little apparent bias or contamination due to clouds. A time series of these composited NDVI images is the basis for estimating vegetation production. Seasonally integrated NDVI values (SINDVI) correspond closely to vegetation above ground production (Goward and Dye, 1987). The objectives of this study are to provide a recent perspective for land cover types and land cover changes that have taken place in the last 24 years; to integrate visual interpretation with supervised and unsupervised classification using Remote sensing; to examine the capabilities of integrating remote sensing and GIS in studying the spatial distribution of different lands and to assess the land affected by soil erosion and the barren land. 2. Study Area Neka river basin, one of the largest watershed in Mazandaran Province is draining the northern flank of Alborz range to the Caspian Sea which divides Neka city as eastern and western parts. The tributaries of this watershed originated from mountainous and forest upland and geologically covered by Shemshak and Quaternary materials. Climate is temperate and seasonal; original land-cover was temperate hyrcanian mixed forest. Major land-uses in the area are rain-fed agriculture and cattle-grazing. The geographical location of the Southern Neka basin based on Neka topography map published by the Iranian Geographical Organization is located at 530, 07/, 57//E.-540, 09/, 03//E. and 360,19/,50//N.-360,32/,42//N. (Figure 1). <<Figure 1. Geographic location of southern Neka basin.>> 3. Methodology 3.1 Data source and materials Three sets of material were used here. First the Landsat imagery, MSS, WRS 82; PATH 175 row35 dated June 06 1977; all four bands have pixel size 57 m and a Landsat ETM WRS path 163 row 35 was in July 30, 2001 have eight bands. Band 1 is high resolution and the pixel size 14 m bands 1, 2, 3, 4, 5, 7 are 28.5 and band 6 is 57mts pixel size. Both 1977 and 2001 were obtained from Global Land Cover Facility (GLCF) used to classify land use/cover in the study area. Geo referenced to the UTM map projection zone 39 projection based on the WGS84 datum (this was used to rectify the 1977 and 2001 Landsat images using 40 45 ground control points (GCPs) with a root mean square error (RMSE) ranging from 0.5 to 0.7 pixels). Second, digital topographic maps digitized from hardcopy topographic maps with scale of 1:20,000 were used mainly for geometric correction of the satellite images and for some ground truth information. Finally, ground information was collected between 1977 until 2002 for the purpose of supervised classification and unsupervised classification and accuracy assessment (Table 1; Figure 2). <<Table 1. Data source and materials>> <<Figure 2: Flowchart of the land use and NDVI classification.>> 3.2 Procedure 3.2.1 Geometric Correction Accurate per-pixel registration of multi-temporal remote sensing data is essential for change detection Change detection analysis is performed on a pixel-by-pixel basis; therefore any misregistration greater than 1 pixel will provide an anomalous result of that pixel. To overcome this problem, the RMSE between any two dates should not exceed 0.5 pixels (Lunetta and Elvidge, 1998). In this study geometric correction was carried out using ground control points from topographic maps with scale of 1:20000 produced in 1994 to geocode the image of 2001. This image was used to register the image of 1977. The Root Mean Square Error (RMSE) between the two images was less than 0.4 pixel which is acceptable. The RMSE could be defined as the deviations between GCP and GP location as predicted by the fitted-polynomial and their actual locations. The rectified Landsat images were resampled to a 100 mts pixel size using the nearest neighbor resampling. The change of resolution was necessary in order to gain better results. Radiometric enhancement was done using low-pass filter because the sum of mean and standard deviation are equal to the original image in this study, while this is not the case with other filters. In other words, the sum of mean and standard deviation were not equal. This means that the number of image pixels is reduced in this filter (Saleh, 2004). The color composites generated from bands 2, 3 and 4 (Figures 3 and 4) were visually interpreted through a screen digitizing. The visual interpretation gave a general idea about the forms of land cover changes over the Published by Canadian Center of Science and Education 59

period. Red colour represents forest and blue colour represents bare soil and low vegetation in Figure 4. In figures 4 and 5, forest area remains the same but the bare soil area changes. <<Figure 3. False color composite image of Landsat MSS (June 1977) bands 2, 3 and 4.>> <<Figure 4. False color composite image of Landsat ETM (July 2001) bands 2, 3 and 4.>> The images in figure 3, and 4 vis-à-vis ETM and MSS data were used to produce Normalized Difference Vegetation Index (NDVI); which is defined as NDVI=(band4+band3)(band4 band3) where band4 and band3 are channels in the near infrared (NIR) and the red wavelengths of Landsat data, respectively. However, the three NDVI data sets were derived from the same season and field work was performed to calibrate the NDVI classification. Saleh emphasized that gray tone should be considered as microstructure of an image and texture as the macrostructure, image with high-spatial resolution is having the disadvantage of low-spatial coverage. To overcome this problem that indeed affects the accuracy of land cover, image must be resample to suitable low-spatial resolution (Saleh, 2004). In the current paper; landsat images with pixel sizes 14, 28, 57 m were resemple to 100 m. 4. Results The result of NDVI has ranged from +0.9597 to 0.2876 and +0.6420 to 0.187 in 1977 and 2001 respectively. This amount of change indicated the fact that the amount of bare soil has increased in the years 1977 and 2001. Based on the classification table, there has been an increment in the amount of bare soil while a decrease in the amount of land occurred in other classes. Further, as it is evident from Tables 2 and 3. <<Table 2. Range of NDVI: MSS image of 1977 Neka basin of Iran>> <<Table 3. Range of NDVI: MSS image of 2001 Neka basin of Iran.>> As it is evident from Figure 5, in subtraction regions, those regions which have experienced more changes are shown by black color and those experienced little changes are specified as bright color. <<Figure 5: NDVI change detection during 1977 2001.>> It can be conclude that some factors may have a role in increasing the amount of discharge including an increase in dry-land and bare soil, and a decrease in forest size. Besides, there are other factors such as over grazing in the valleys which can also promote greater volume of discharge. However Change detection technique for 1977 and 2001 images has helped to identify change image (Figure 6) in the NDVI. Most of the sections in the study area have experienced a vegetation change (100%). In Figure 6, it is observed that there is a drastic change in the land use. <<Figure 6. NDVI change detection between 1977 2001>>. After calculating NDVI, two images of supervised and unsupervised classification using bands 3 and 4 were acquired on 6 June, 1977 and 30 July, 2001.Figs. 7 and 8 show the result of the classification <<Figure 7. Land use/cover classification map of Neka Iran, using Landsat MSS1977.>> <<Figure 8. Land use/cover classification map of Neka Iran, using Landsat ETM+ 2001>> In order to increase the accuracy of land cover mapping of the two images, ancillary data such as data from Google earth, and the result of visual interpretation was integrated. The area under the land use changes for different land cover classes are given in Table 4. Low Graze land, bare soil and dry land occupy 63532 ha in 1977 and 77905 ha in 2001. Grassland was 38333 ha in 1977 and 28053 ha in 2001. Low-density forest and vegetation was 25592 ha in 1977 and 20339 ha in 2001. Dense forest was 58261 ha in 1977 and 5940 ha in 2001. <<Table 4. Area and percentage of land cover change of during 1977 and 2001 classified images Neka Iran >> A standard overall accuracy for land-cover and land-use maps is set between 85% (Anderson, Hardy, Roach and Witmer, 1976) and 90% (Lins and Kleckner, 1996). In this study, the overall classification accuracy was found to be 99% for 1977 and 99.12% for 2001. Details of single class accuracy for both images of 1977 and 2001 can also be found in Table 4. This table shows that the highest accuracy is for all classes; this is due to the integration of the visual interpretation with the classified image in addition to the information obtained from the topographic map <<Table 5. Accuracy statistics for the classification result>> 60 ISSN 1916-9779 E-ISSN 1916-9787

5. Conclusion It is concluded from satellite images that most of the land use has changed from 1977 to 2001. If this process of change in land use continues at the same rate, there is a possibility that within no longer period of time there will be a severe mismanagement of land, and the impact could become severe especially on agriculture, and the environment. The problem increasing dry land and bare soil can lead to more destructive events such as floods and devastation. References Anderson JR, Hardy EE, Roach JT and Witmer RE. (1976). A land-use and land-cover classification system for use with remote sensor data. US Geological Survey Professional Paper 964, Washington, DC. Berberoglu S and Akin A. Assessing different remote sensing techniques to detect land use/cover changes in the eastern Mediterranean. Int J Appl Earth Observation Geoinformation [journal homepage: www.elsevier.com/locate/jag]. Brandon RB. (2001). Mapping rural land use and land cover change in Carroll County, Arkansas Utilizing Multi-Temporal Landsat Thematic Mapper Satellite Imagery. Thesis University of Arkansas. [Online] Available: http://www.cast.uark.edu/local/brandon_thesis/index.html. Dale VH, Brown S, Haeuber RA, Hobbs NT, Huntly N, Naiman RJ, Riebsame WE, Turner MG, Valone TJ. (2000). Ecological principles and guidelines for managing the use of land. Ecological Applications 10: 639 670. Dall Olmo G and Karnieli A. (2002). Monitoring phenological cycles of desert ecosystems using NDVI and LST data derived from NOAA-AVHRR imagery. Int J Remote Sensing, 23, 4055 4071. Eidenshink JC and Faundeen JL. (1994). The 1 km AVHRR global land data set: first stages of implementation. Int J Remote Sensing, 15(17), 3443 3462. Gobin A, Campling P, Feyen J. (2001). Spatial analysis of rural land ownership. Landscape Urban Planning 55: 185 194. Goward S and Dye D. (1987). Evaluating North America net primary productivity with satellite observations. Advances in Space Research, 7, 165 174. Grime JP. (1997). Climate change and vegetation. In: Plant Ecology, 2nd edition, Crawley MJ (ed.). Blackwell Science: Oxford, UK; 582 594. Hope A et al. (2003). Interannual growth dynamics of vegetation in the Kuparuk River watershed based on the normalized difference vegetation index. Int J Remote Sensing, 24(17): 3413 3425. Kamusoko C and Aniya M. (2007). Land degradation and development Land Degrad Develop 18: 221 233 (2007). Published online 8 August 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ldr.761. Kamusoko C and Aniya M. (2009). Land use/cover change and landscape fragmentation analysis in the Bindura District, Zimbabwe. Asian J Appl Sci, 2 (6): 464 474. Karim Solaimani et al. (2009). Soil erosion prediction based on land use changes (A Case in Neka Watershed). Am J Agri Biolog Sci, 4 (2): 97 104. Lunetta RS and Elvidge CD. (1998). Remote sensing change detection. MI: Ann Arbor Press. MEA (Millennium Ecosystem Assessment). (2005a). Ecosystems and Human Well-being: Synthesis. Island Press: Washington, DC. MEA (Millennium Ecosystem Assessment). (2005b). Ecosystems and Human Well-Being: Current State and Trend, Vol. 1. Island Press: Washington, DC; 585 621. Mwavu EN and Witkowski ETF. (2008). Land-use and cover changes (1988 2002) around Budongo forest reserve, NW Uganda: implications for forest and woodland sustainability, Land Degrad Develop 19: 606 622. Saleh N.M. (2004). Automated oil spill detection with ship borne radar (MS Thesis). International Institute for Geo-informational Science and Earth Observation. ITC, The Netherland. Smailpour et al. (2007). Forest change detection in the north of Iran using TM/ETM+Imagery. Appl Geography 27: 28 41. Stow DA. (1999). Reducing mis-registration effects for pixel-level analysis of land-cover change. Int J Remote Sensing, 20, 2477 2483. Published by Canadian Center of Science and Education 61

Tommervik H et al. (2003). Monitoring vegetation changes in Pasvik (Norway) and Pechenga in Kola Peninsula (Russia) using multitemporal Landsat MSS/TM data. Remote Sensing of Environment, 85: 370 388. Turner BL II and Meyer BL. (1994). Global land use and land cover change: an overview. In: Changes in Land Use and Land Cover: A Global Perspective, Meyer WB, Turner BL II (eds). Cambridge University Press: Cambridge; 3 10. UN/ECE (United Nations Economic Commission for Europe). (2002). Forest conditions in Europe, 2001 Executive Report. Geneva and Brussels, 29. Table 1. Data source and materials Satellite images Date Spatial resolution Source Landsat MSS 1977 57 m GLCF Landsat ETM 2001 28 m GLCF Topographical map 1955 1;55000 The Iranian Geographical Organization Topographical map 1965 1:50000 The Iranian Geographical Organization Topographical map 1994 1:20000 The Iranian Geographical Organization Table 2. Range of NDVI: MSS image of 1977 Neka basin of Iran 200000 NDVI MSS 1977 NEKA basin of iran 180000 174875 160000 140000 D N 120000 100000 1977 80000 60000 40000 20000 0 0000000025 000000000025 0050 00125 125 175 150 225 400 275 375 450 525 800 575 725 1000 800 1025 1075 1725 1600 1325 1550 2075 1975 2325 50 1275 2300 1775 2225 2150 2400 2125 1850 1900 2275 2425 2000 1800 1700 1750 1725 1575 1325 2050 1475 1700 1625 1550 1350 900 1425 1150 1125 1375 1175 850 1350 1300 1450 650 875 675 725 1025 700 825 800 925 1075 700 825 650 850 475 925 825 475 775 525 750 675 725 875 50 925 250 1050 800 450 600 550 675 750 475 850 550 775 575 475 450 350 675 550 475 525 550 725 400 250 550 525 350 825 550 275 650 600 675 800 750 550 850 600 1000 750 1025 1200 1000 1075 1475 2275 2100 2675 2400 2900 3275 2925 3650 2225 2600 2325 2250 1275 925 675 875 425 200 150 175 025 000000000000000000000000000000000000000000000000000000050-0.3-0.3-0.2-0.2-0.1-0.1-0.1-0 -0 0.03 0.06 0.1 0.13 0.17 0.21 0.24 0.28 0.31 0.35 0.38 0.42 0.45 0.49 0.52 0.56 0.59 0.63 0.66 0.7 0.73 0.77 0.8 0.84 0.87 0.91 0.94 0.98 Reflection 62 ISSN 1916-9779 E-ISSN 1916-9787

Table 3. Range of NDVI: MSS image of 2001 Neka basin of Iran 200000 NDVI ETM 2001NEKA basin of iran 180000 174650 160000 140000 D N 120000 100000 2001 80000 60000 40000 20000 0-0.1807229-0.1541824 0000025 50 075 50 75 100 200 175 250 325 450 775 375 625 1175 1125 1225 1100 1700 1375 2125 2350 2325 2075 2850 2250 2375 2525 2575 3250 3425 2525 2625 2450 2375 2000 2350 2575 2125 2525 1350 3275 25 25 2700 950 1525 1400 1075 775 1250 1150 1000 1025 950 750 775 850 725 625 550 450 500 600 725 575 500 450 475 500 400 525 175 425 300 450 425 325 400 475 575 375 325 350 450 325 275 125 300 550 250 350 325 300 350 200 550 350 400 225 150 325 250 300 225 275 300 625 425 350 200 250 275 300 250 275 125 200 225 250 200 300 275 175 250 200 425 250 550 0450 300 400 425 275 400 250 275 300 225 350 275 200 400 550 500 425 325 350 475 325 400 425 375 600 225 400 700 550 800 675 800 575 1100 900 800 700 1025 1250 750 1100 1575 1900 2075 2025 875 1500 1850 1750 1125 1675 1575 1550 1050 1400 1250 1175 1800 1125 750 850 525 650 700 475 400 500 350 325 275 200 225 250 75 75 25 75 50 25 0050-0.127642-0.1011016-0.0745611-0.0480207-0.0214802 0.0050602 0.0316007 0.0581411 0.0846816 0.111222 0.1377625 0.1643029 0.1908434 0.2173838 Reflection 0.2439243 0.2704647 0.2970052 0.3235456 0.350086 0.3766265 0.4031669 0.4297074 0.4562478 0.4827883 0.5093287 0.5358692 0.5624096 0.5889501 0.6154905 0.642031 Table 4. Area and percentage of land cover change of during 1977 and 2001 classified images Neka Iran Class name 1977 (area in ha) Percentage 2001 (area in ha) Percentage Low grassland and bare soil and dry land 63532 34.20 77905 41.95 Graze land 38333 20.64 28053 15.10 Low forest and vegetation 25592 13.78 20339 10.95 Dense forest 58261 31.37 59401 31.98 Table 5. Accuracy statistics for the classification result class name Producer s accuracy (%) Reference total User accuracy (%) Kappa statistic (a) 1977 MSS image Low Graz land and bare soil, and dry land 100 26 96.67 0.96 Graze land 96.77 31 100 1 Low forest and vegetation 100 23 100 1 Density forest 100 91 100 1 (b) 2001 ETM Image Low Graz land and bare soil and dry land 100 31 100 1 Graze land 100 23 96 0.95 Low forest and vegetation 96 24 100 1 density forest 100 93 100 1 Published by Canadian Center of Science and Education 63

Figure 1. Geographic location of southern Neka basin 64 ISSN 1916-9779 E-ISSN 1916-9787

Landsat MSS and ETM+ Topography map and ground data NDVI Supervised and unsupervised classification Accuracy assessment Changed detection Figure 2. Flowchart of the land use and NDVI classification. Figure 3. False color composite image of Landsat MSS (June 1977) bands 2, 3 and 4 Published by Canadian Center of Science and Education 65

Figure 4. False color composite image of Landsat ETM (July 2001) bands 2, 3 and 4 Figure 5. NDVI change detection during 1977 2001 Figure 6. NDVI change detection between 1977 2001 66 ISSN 1916-9779 E-ISSN 1916-9787

Figure 7. Land use/cover classification map of Neka Iran, using Landsat MSS1977 Figure 8. Land use/cover classification map of Neka Iran, using Landsat ETM+ 2001 Published by Canadian Center of Science and Education 67