CC BY
20
АРИДНЫЕ ЭКОСИСТЕМЫ, 2004, том 10, № 24-25
================= ОТРАСЛЕВЫЕ ПРОБЛЕМЫ ЗАСУШЛИВЫХ ЗЕМЕЛЬ=============
УДК 631.445.57:577.4
ОЦЕНКА ИЗМЕНЕНИЙ ЗЕМЛЕПОЛЬЗОВАНИЯ И РАСТИТЕЛЬНОГО ПОКРОВА В СОМОНЕ БУЛГАН С ИСПОЛЬЗОВАНИЕМ ДИСТАНЦИОННЫХ МЕТОДОВ ЗОНДИРОВАНИЯ
© 2004. Ю. Баяржаргал, А. Карниели
Лаборатория дистанционных методов, Институт пустынь Университета Бен-Гуриона, Седе Бокер Кампус,
84990, Израиль
В статье излагаются результаты использования методов дистанционной оценки изменения в характере землепользования и покрытия земной поверхности растительностью за 12-летний период в сомоне Булган (Южно-Гобийский аймак, Монголия), в пределах N 43.75°-44.85° и E 102.85°-104.12°. В целях оценки динамики растительного покрова и его классификации, а также изучения взаимосвязей между растительностью и vegetation indices и мониторинга процессов опустынивания, использовались либо крупномасштабные снимки, либо данные наземных измерений.
Целью работы было определить и выразить количественно эти изменения, произошедшие с 1990 по 2002 гг. Предполагалось ответить на следующие вопросы: 1) Произошли ли какие-либо изменения растительного покрова? 2) Какие элементы земной поверхности подверглись воздействию? 3) Чем были вызваны изменения: антропогенным воздействием или естественными причинами?
Изменения определялись с использованием индекса vegetation index differencing. Наиболее широко применяемый индекс - the Normalized Difference Vegetation Index (NDVI). Значения NDVI получали на основе изображений Landsat-5 TM and Landsat-7 ETM+ 1990 и 2002 гг. по уравнению NDVI = (pNIR - pRed )/(Pnir + PRed) на основе значений отражающей способности земной поверхности. Обнаруженные различия легли в основу дифференцированного изображения, показывающего изменения NDVI за указанный период. Стандартное отклонение от среднего обычно берется как порог между пикселями «изменение» и «не изменение». В нашем исследовании, была проанализирована гистограмма, отражающая NDVI, и вычислены стандартное и среднее отклонения. В качестве порога, определяющего изменения земной поверхности в дифференцированном изображении, были выбраны 2 стандартных отклонения выше и ниже средней на гистограмме NDVI. Нижняя область в пределах двух стандартных отклонений (высокие значения NDVI в 1990 и низкие -в 2002 гг.) представляет негативные изменения земной поверхности, тогда как верхняя область плюс два стандартных отклонения (высокие значения NDVI в 2002 и низкие - в 1992 гг.) представляют позитивные изменения.
Установлено, что произошли разноуровневые изменения в землепользовании и растительности сомона, включающие возобновление растительного покрова в различных частях исследованной территории и пересыхание водоема. Эти изменения были вызваны климатическими флуктуациями (режим осадков), наряду с антропогенной деятельностью, такой как увеличение поголовья скота, перевыпас и переэксплуатация водных источников.
На сравнительно больших территориях опустыненных степей индекс NDVI увеличился. Этот феномен можно объяснить неэффективной организацией пастбищеоборота в сомоне с тех пор, как выпас интенсифицировался в течение периода, охваченного исследованиями.
Исчерпание водных ресурсов озер вследствие нерациональной человеческой деятельности создало возможность для непоедаемых скотом растений развиваться в высохших озерных котловинах. Как следствие, эти виды растений вызвали более высокие значения NDVI в 2002 г. по сравнению с 1990 г.
Исследование демонстрирует возможность использования дистанционных методов для идентификации изменений характера землепользования и поверхности земли на выбранных территориях в условиях пустынных степей Монголии. Получаемые при этом результаты могут служить важным инструментом в процессе принятия решений.
ASSESSING LAND-USE AND LAND-COVER CHANGE IN BULGAN SOUM BY REMOTE SENSING
CHANGE DETECTION TECHNIQUE
© 2004. Yu. Bayarjargal and A. Karnieli
The Remote Sensing Laboratory, Jacob Blaustein Institute for Desert Research, Ben Gurion University of the Negev,
Sede Boker Campus 84990, Israel
Introduction
The current paper reports a practical application of remotely sensed data coupled with change detection technique to assess land-use and land-cover change (LULCC) for a 12-year period in Bulgan Soum1, South-Gobi Aimag2, Mongolia. The Bulgan Soum is geographically located in the northern edge of the Mongolia's Gobi desert, in the desert-steppe environment. LULCC that exists within the Bulgan Soum are highly affected by natural factors as a consequence of seasonal and interannual fluctuations of precipitation, prolonged and frequent drought events, strong windstorms, etc. In addition, the area is also affected by pressure of human activities such as intensifying grazing by domestic livestock, overexploiting natural recourses (e.g., collection of firewood), developing irrigate agriculture, and unlimited use of the ground for off-road driving.
Methods for assessing LULCC range from a plot-level in-situ sampling to wide-ranging analysis of remotely sensed data. Although aerial photography can detect LULCC over relatively wide area at a reasonable cost, satellite imagery has proven as more cost-effective in terms of space and time. Enormous number of studies, utilizing remote sensing data, derived from various satellites with specific characteristics, over different ecosystems have been well documented during the past two-decades (e.g., Jensen 1986; Lunetta and Elvidge, 1998). Moreover, a large variety of change detection techniques have been formulated, applied, and evaluated (Singh 1989; Lambin and Strahler, 1994; Collins and Woodcock, 1996; Lunetta and Elvidge, 1998; Sohl, 1999). Multitemporal Landsat series data (Multispectral Scanner (MSS), Thematic Mapper (TM), and Enhanced Thematic Mapper Plus (ETM+)) in association with numerous change detection techniques have been demonstrated their usefulness for studying LULCC over semi-arid environments (e.g. Pilon et al., 1988; Ram and Kolarkar, 1993; Chavez and MacKinnon, 1994; Knick et al., 1997; Lunetta et al., 1998; Abuelgasim et al., 1999; Elmore et al., 2000; Koch, 2000; Rogan and Yool, 2001; Helmschrot and Flugel, 2002). The Landsat imagery that has high spatial resolution imagery of 30 m provides complementary information to the field survey in attempting to detect and evaluate LULCC. However, practical utilization of such data and technique is limited in Mongolia. Here, either coarse resolution images or ground measured information were used for assessing land cover change (Chuluun et al., 1999; Erdenetuya and Khudulmur, 2002) and classification (Tateishi et al., 1997), for studying relationships between vegetation cover and vegetation indices (Purevdorj et al., 1998), and for monitoring vegetation state with respect drought events (Adyasuren and Bayarjargal, 1995; Bayarjargal et al., 2000; Bayarjargal and Karnieli, 2004).
Consequently, the objective of this study is to detect and quantify land cover changes over more than a decadal period (1990 - 2002) by using remote sensing change detection technique. The study intends to reply the following three questions: (1) Were there changes of land cover? (2) Which ground features had been affected? And (3) Were the changes derived by human activity or natural phenomena?
Study area
The study area, Bulgan Soum of South-Gobi Aimag, is located between 43.75°-44.85°N and 102.85°-104.12°E (Appendix 10). The total area of Bulgan Soum is 747,907 ha and it belongs to the desert-steppe environment. In spit of that, a more arid-desert environment is found in the central west and eastern regions of the Soum while the southwestern- and western-edges are characterized as mountain steppe environment. A harsh seasonal climate and limiting resources of water and nutrients for domestic and wild herbivory dominate the study area. The desert-steppe part of the study area is characterized by relatively flat terrain with high-erodible light-chestnut and stony soils, and sparsely distributed perennial grasses along with semi-shrubs. The desert environment is characterized by sandy soil with eolian deposits and desert woody shrubs. The mountain steppe area is characterized by rocky terrain, rolling topography with broad ridges and sharply
1 Soum (district) - a local administrative unit.
2 Aimag (province) is a largest administrative unite that has including several Soums.
128 БАЯРЖАРГАЛ, КАРНИЕЛИ
indented valleys, and higher plants, wheatgrass herbs joining in sage shrub. Elevations within the Bulgan Soum range from 1030-1700 m AMSL in the desert and desert-steppe plateau and up to 2600 m in the mountain steppe area. The climate of the Bulgan Soum is harsh; cold winter, and dry and hot summer. Precipitation, which is the most important climatologic factor for vegetation growth in this area, varies considerably from year to year and from month to month. Mean annual precipitation is 120 mm, varying from 54 to 195 mm during the 42-year long-term data from 1961 to 2002 (Figure 1a). 82% of the total precipitation is mainly concentrated in the vegetation-growing season, from May to September, with a peak in July-August (Figure 1b). Mean monthly temperature varies from -13°C in January to 22.3°C in July.
100
£ 30
E
| 60 »
«
% 40 'o
О
□I 20 0
Figure 1. (a) Inter-annual variation of accumulated precipitation during the vegetation-growing period over the study area. Solid line shows the long-term average precipitation; (b) Multi-year average of monthly precipitations for the years 1990 through 2002. Arrows mark acquisitions dates of the Landsat images. Рисунок 1. Динамика межсезонных колебаний аккумулированных осадков за период вегетации на исследуемом районе. Сплошной линией показана динамика многолетних среднегодовых осадков. (b) Средняя многолетняя величина ежемесячных осадков за 1990-2002 гг. Стрелками показаны даты сравнения изображений Ландсат.
A major land-use practice in the Bulgan is domestic-livestock grazing that is still a natural-dependent and semi-nomadic. Five kinds of animals (sheep, goats, cattle, horses, and camels) have been herding all year-round, and an utilization rate of pasture capacity for the Soum was below fifty percent in 1990 when 63.7 thousand numbers of livestock had been grazed over some 728100 ha area. Due to the collapse of state-owned support at the beginning of 1990s and herders' interest to increase the number of privatized livestock, the rate was doubled in 1999. The numbers of livestock, especially number of goats, in the Bulgan Soum has increased from 1990, and currently goats are 48% of the total animals. One reason for increasing the number of the animals can be related to migration of joblessness from the Soum and Aimag centers to the countryside. While less than 1000 people were living in countryside in 1991, this number increased to 1671
(b) Landsat ETM+, 2002
' \ Landsat TM, 1990
1 I Mean -э- 1990 -*- 2002 1 / / 1 1 1 t \ \ \ \ \ \ * \
я / s s ... - / <t> \ \ <3-
' I 'W'l I I I 1
1 2 3 4 5 6 7 8 9 10 11 12
Month
in 2000. In addition to a direct influence of livestock grazing on the pastureland, land degradation processes were taken place around the watering points (e.g., natural spring, drilled wells, and lake), cattle-breeding camps (winter shelters, folds), and local administrative centers. Irrigated agriculture that is potatoes and vegetable planting by the rural residents in a relatively small area became an intensive mainly in the rural centers during the last a few years. Also, few herders who are living in or near the settlements with a smallest-numbers of animals grew vegetables as supplementary for their income. In addition, considerable cover of disturbance and land degradation was caused by unlimited use of the ground for off-road driving.
Dataset
The basic requirement for detecting LULCC by using remote sensing techniques is the availability of images for (at least) two dates on which a same area can be observed (Yuan et al., 1998). Since the geographic location of the Bulgan Soum is larger than a single Landsat scene (185X185 km), six scenes of the Landsat imagery, three of Landsat-5 TM (1990) and three of Landsat-7 ETM+ (2002), were acquired. In order to reduce scene-to-scene variations that are related to sun angle, differences in atmospheric condition, and vegetation phenology, all images were selected during the vegetation growing season, when the possibility to discriminate between vegetation and soil covers is the highest. The main specifications of the Landsat scenes are listed in Table 1. Also, cloud covers were minimal for these images. However, the study years have different precipitation regimes. 1990 was extremely wet year with 195 mm of rainfall, higher than the long-term average (120 mm). 107 mm of rainfall occurred between July and September. Opposing, 2002 was a dry year with 79 mm of rainfall. 39 mm rainfall occurred in July and September (Figure 1b).
Table 1. Specifications of the Landsat-5 TM and Landsat-7 ETM+ scenes used in this research. Таблица 1. Список фрагментов Ландсат-5 ТМ и Ландсат-7 ЕТМ+, использованных в данном исследовании.
Dates DD/MM/YYYY (Sensor) Path/Row Sun azimuth (o) Sun elevation (o) Day of year
17/09/1990 (TM) 132/029 142.0 41.0 260
17/09/1990 (TM) 132/030 141.0 42.0 260
30/07/1990 (TM) 133/029 128.0 60.0 283
24/07/2002 (ETM+) 132/029 134.2 58.6 205
24/07/2002 (ETM+) 132/030 131.9 59.3 205
16/08/2002 (ETM+) 133/029 140.4 53.5 228
Fieldwork
The objective of the fieldwork was to collect ground reference data that could be used for satellite data analysis. Although unbiased ground reference information at the time of each remote sensing data acquisition needs to be collected in order to aid satellite image interpretation, this is rarely possible when historical images have to be used. Hence, we collected ground information for the most recent imagery, Landsat ETM+ in 2002. Fieldwork took place from late July to mid August in 2002. Ground information was collected at fifty-five sites. Sites were selected throughout the north-south and east-west extent of the Bulgan Soum based on an erratic network of major and minor roads. The locations of sites, which were considered enough homogenous or significantly heterogeneous, were recorded by using a Global Positioning System (GPS) and mapped directly onto the corresponding topographic map sheets with scale 1:100,000 and satellite images. A color printout image of the Landsat TM in 1990 was used during the fieldwork. Descriptive information, which includes identification of plant species, percentage cover of green vegetation, plant phenological stages and heights, and spreading of bare soil and sand dunes, was noted for each site. These records were compared to the soil type and vegetation maps and then used as reference and validation information for satellite image processing.
Satellite image pre-processing
Prior to applying change detection analysis, pre-processing operations were performed on the remotely sensed data in order to improve image quality, adjust digital values for the difference of the atmospheric conditions, and bring images into registration with the same projection. The most important procedures in the performances of the pre-processing are radiometric calibration, atmospheric correction, and geometric rectification or registration. The thermal bands of the TM and ETM+ sensors were not included in the processing and further analyzing. All other six bands were individually subject to calibration and correction.
130
EA^P^APrAH, KAPHHEHH
This process included the conversions of digital number (DN) values to at-satellite radiance values (Landsat 7 Users Handbook, Markham and Barker, 1987).
Atmospheric correction of the radiance values at satellite for each Landsat images were carried out using the Second Simulation of the Satellite Signal in the Solar Spectrum (6S) algorithm (Vermote et al., 1997a). The 6S computer program can correct satellite-derived solar radiation that was backscattered from the Earth-surface-atmosphere system. Due to the absence of in-situ or ground measured information of the atmospheric condition at the time of satellite data acquisition, the Landsat TM images were corrected by using a default midlatitude-summer atmospheric model. In addition, because there was no correction model for Landsat-7 ETM+ in the 6S code (Vermote et al., 1997b), the program was upgraded for ETM+ during in framework of the current research. Data about the atmospheric condition such as aerosol contents and total precipitable water that are required for atmospheric correction during the images acquisition times were obtained from an automatic tracking sunphotometer (CIMEL), which is installed in Dalanzadgad, center of South-Gobi Aimag, about 60 km southeast of the study area (Appendix 11). Atmospheric correction procedure ended with surface reflectance values.
In order to obtain cartographic uniformity of the scenes, a geometric rectification to a unique geodetic system (UTM, Zone 48, Spheroid WGS 84) was applied to the Landsat-7 ETM+ images of 2002 based on 49 ground control points that had been collected by using GPS during the fieldwork in summer 2002. A polynomial second-order transformation with cubic convolution resampling method in ERDAS image-processing package was adopted in this study (ERDAS, 1997). Mean root-mean square errors (RMSE) in geometric rectification approach were less than half pixel for each scene, and the images' pixels were carried out with 28.5-meter pixel size. After applying the geometric rectification, two merged images were created for 1990 and 2002 based on three-registered images (RMSEs were less than 0.3 pixel for every registration set) for each year. Two mosaic images in 1990 and 2002 were registered one to another with the RMSE of less than 0.2 pixels. The study area was subseted on the inter-registered images by using vector file of ARC-GIS. Thus, the ground reflectance values on the subseted images for 1990 and 2000 over the Bulgan Soum area (Figure 1) are used simultaneously in the further change detection analysis.
Change detection analysis
The change detection method, named vegetation index differencing (Lyon et al., 1998) was used. Main assumption was that LULCC could be detected in variation of vegetation index images on two different dates, 1990 and 2002. Vegetation indices are algorithms aimed at simplifying and reducing data from multiple reflectance bands to a single value correlated to physical vegetation parameters, such as biomass, productivity, leaf area index, or percent vegetation ground cover. The majority indices are based on intensive chlorophyll absorption in the visible red part of the electromagnetic spectrum used for photosynthesis and strong reflection in the near-infrared (NIR) part of the spectrum due to scattering caused by internal leaf structure (Tucker, 1979). The most widely used and well-known vegetation index - the Normalized Difference Vegetation Index (NDVI) is formulated as:
NDVI = (PnIR - pRed )/(PNm + pRed ) (1)
where pNIR and pRed are the reflectance in the near-infrared and red spectral bands, respectively. Thus, denser and/or healthier vegetation having higher NDVI values, while lower values response to sparse or stressed vegetation (Sellers, 1985). Soil spectra typically do not show such dramatic spectral difference as vegetation. The ease in calculating NDVI from a variety of sensors and the success of the NDVI in detecting vegetation and vegetation change has made it a popular index. The NDVI was used as vegetation index differencing method in numerous studies for analyzing vegetation conditions and changes (Briggs and Nells, 1991; Chilar et al,. 1991; Lyon et al., 1998). This technique also been used to detect changes in canopy or vegetation biomass (Hayes and Sader, 2001). Singh (1989) compared several methods of change detection and found that the NDVI ratio is one of the most accurate techniques. Lyon et al. (1998) found this method to be less affected by topographic features than other change detection techniques. Moreover, Yuan and Elvidge [1998] concluded that NDVI differencing do better than other tested change-detection techniques.
The vegetation index differencing method consists of three processing levels: vegetation index transformation, vegetation index differencing, and evaluation of change statistics (Lyon et al., 1998). The
NDVI values calculated, individually for the 1990 and 2002 images, by using Equation 1 from the surface reflectance values. The NDVI image of 1990 was subtracted from those of 2002 to create the NDVI change differenced image. In order to identify actual LULCC, it is necessary to determine a threshold value that can be used to distinguish between pixels that were not changed and those that were significantly changed. Once a histogram of changed pixels was established, pixels that their values were not changed are distributed around the mean, while pixels with significant change are locating on the tails of the histogram (Jensen, 1986). A standard deviation from the mean of change is often selected as the threshold between "change" and "no change" pixels (Jensen, 1986). Higher degrees of change can be located beyond higher levels of standard deviation. In this study, NDVI differenced image's histogram was examined and the mean and standard deviation values were calculated.
Uncertainty of NDVI differencing method was accounted for a range of threshold values from one standard deviation (where 80% of the study area is assumed unchanged) up to three standard deviations where more than 99% of areas are probably unchanged. Through this study, we assumed that more than 80% of the study area has been unchanged during the monitoring period. In other words, we believed that there were no large-disturb effects for the study area; however, LULCC should be no larger than 20% of the study area if there were some disturbing impacts on land cover. Then, two standard deviations above and below the mean of the histogram of NDVI differenced image (mean ±2 standard deviations) were selected as a threshold to determine the changes of land cover in differenced image of two NDVI images in 1990 and 2002. Therefore, area that is located within two standard deviations from the mean of NDVI differenced image's histogram was considered to represent no-change (Appendix 12). While, the area downward within two standard deviations (higher NDVI in 1990 and lower NDVI in 2002) represents negative change of land cover, and area upwards plus two standard deviations (higher NDVI values in 2002 and lower in 1990) represents positive change of land cover. Moreover, to enhance and assess LULCC over the Bulgan Soum, NDVI differenced image was labeled from high decrease to high increase of NDVI changes during the 12-year period, with one standard deviation interval, as shown in Appendix 12.
Results and discussion
Appendix 11 presents results of NDVI differencing change detection superimposed on surface reflectance values of Band 3 in 1990 as gray-levels. Brighter tones in the background indicate a higher surface reflectance values while darker tones indicate lower ones. Significant changed values of NDVI were thresholded by two standard deviations from the mean of NDVI differenced image's histogram, and are highlighted in colors from red to blue. Positive changes are colored from green as low change to blue as high change through cyan as middle change of NDVI value. Negative changes are colored from yellow as low change to red as high through orange as middle change of NDVI. Positive and negative changes were subdivided in intervals of one standard deviation.
The quantified change information of NDVI between 2002 and 1990 is shown in Table 2. No-change of land cover or no difference of NDVI between the study years over the Bulgan Soum, within the threshold of plus and minus two standard deviations occupies 726856 ha or 97.2% of the study area. No-changes are not colored, shown as gray-level values. The desert-steppe rocky terrain in the north, the mountain steppe in the southern- and western-fringes, and the desert sand dunes with shrubs and semi shrubs in the central-eastern and -western parts of the Soum do not show significant changes. Contrary, LULCC in different levels are mostly occurred over the desert-steppe environment with herbaceous plants and semi-shrubs in the southeast and eastern portions of the study area.
About 16696 ha (2.2% of the study area) were indicated as decreased NDVI from 1990 (wet year) to 2002 (drought year). The red in the NDVI differenced image represents high or severe rate of decreased NDVI (533 ha or 0.1% of the study area). The moderate (1460 ha or 0.2% of the study area) and low (14703 hectares that about 1.9% of the study area) decreases of NDVI values are shown as orange and yellow colors, respectively, in differenced image. Decreased NDVI values are found in the desert-steppe plain environment by a decline or reduction of natural vegetation. This was happened in valleys, along temporal watering canals, and in an area of a lake (Ulaan-Nuur) that was dried up (Appendix 11). Two reasons can be pointed out: (1) natural effect caused by difference in rainfall regime between the two years (Figure 1b), since high correlation exists between the precipitation and NDVI (Tucker et al., 1991; Tucker and Nicholson, 1999); and (2) anthropogenic effect of grazing pressure on the study area of domestic livestock. During the
132
БАЯРЖАРГАЛ, КАРНИЕЛИ
fieldwork, it was surveyed that roaming patterns of grazing livestock have been distributed mostly around the centers of Soum and Bags. Number of animals almost doubled from 1990. Due to limitation of drinking water over the pastureland in the far north of the study area and absence of grazing plants in the rockymountain areas and sand dunes in the west and east of the study area, grazing of the increased number of animals is mostly located over the south-eastern part of the study area. Therefore, it should be noted that the NDVI differencing could give significant LULCC in the plain desert-steppe environment due to the combined effects of drought and grazing.
Table 2: Changes of land cover between 1990 and 2002 identified by NDVI differencing. Таблица 2. Изменение наземного покрытыя за период с 1990 по 2002 гг., установленное с помощью NDVI
NDVI changes Changed area Total area
Type Intensity (hectare) (%) (hectare) (%)
High 532.8 0.1
Decreased: Moderate 1460.0 0.2 16696.1 2.2
Low 14,703.3 1.9
Low 582.1 0.1
Increased: Moderate 333.7 0.0 4354.6 0.6
High 3438.7 0.5
Total NDVI change 21050.7 2.8
No change 726855.9 97.2
Total area 747906.6 100
Although there was less precipitation in 2002 (Figure 1b), positive changes of NDVI are also observed. Only 4,355 ha (0.6% of the total study area) were detected as increased NDVI from the wet to drought year (Table 2). This highest level of LULCC was found around the dried lake Ulaan-Nuur. This phenomenon is explained by disappearing of the brackish-water and growing of worm weed shrubs such as Artemisia frigida, which is not palatable for the animals during an intensive growth period of plants in early autumn and summer times. The Artemisia frigida is one of indicator-species of land degradation processes since it is more resistant to degradation. Increasing of such worn weed plant over the dried-table of brackish-water is proved by the fieldwork in 2002. Moreover, areas with increased NDVI values between the study years are found over the valleys and canals that are near the Soum and Bag centers. This can be explained by extensions of invader plants such as a dwarf shrub, Iris Bungei, which is also not palatable during its growth period. This plant can grow up as relatively large bunches with diameters range in 0.5-2 m and height about 30-40 cm tall. From the fieldwork, it was observed that the Iris Bungei does not mix with other dwarf shrubs and has a significant association to soil moisture in sandy soil. In addition, cultivated irrigated areas near the Soum and Bag centers are observed with increased values of NDVI. This type of LULCC is related to the irrigated-planting of vegetables that has become a new land use during the last a few years. Also, some increased NDVI values, which are observed in the floodplains of the Gobi Gurvan Saikhan Mountain in the west-south corner of the study area, were probably caused by flooding of local canals and ephemeral streams.
Summary and conclusions
Changes of land-use and land-cover over the Bulgan Soum during a 12-year period (1990-2002) were detected and quantified by using high spatial resolution imagery of Landsat TM and ETM+ coupled with change detection method based on normalized difference vegetation index differencing. LULCC was caused by combination of anthropogenic (i.e., livestock grazing) and natural (i.e., precipitation deficit) effects.
As change-detection analysis between two dates with different rainfall regimes, it was shown that:
(1) Relatively large areas have been detected over the desert-steppe plain environment as decreased in NDVI. This phenomenon can be explained by ineffective grazing-management in the Bulgan Soum since livestock grazing has been intensified during the study period.
(2) Vanishing of the lake's water due to mismanaged human activities in the study area caused a suitable opportunity to unpalatable plants to grow in the dried-lakebed. Consequently, these species caused higher NDVI values in 2002 than in 1990.
This study demonstrates the potential for using remote sensing data and change detection method to identify LULCC in a selected area of Mongolia's desert steppe environment. Moreover, practical methodology of this study can probably be used as an example for regional level land cover monitoring system of the country. It is suggested that such a research on LULCC will be conducted on a regular interval, so that the information can be updated periodically. A further detailed investigation of LULCC in study area is necessary to use different change-detection techniques with integration of ancillary GIS datasets (e.g., environmental parameters such as elevation, slope, and aspect of the landscape and socio-economic information such as population and animal number) or sequential aerial photographs of the test area. Results of change detection analysis should be used as a tool for decision makers.
REFERENCES
1. Abuelgasim A.A., Ross W.D., Gopal S., and Woodcock C.E. Change Detection Using Adaptive Fuzzy Neural Networks: Environmental Damage Assessment after the Gulf War. Remote Sensing of Environmen., 1999. 70. Р. 208-223.
2. Adyasuren Ts., and Bayarjargal Yu. Vegetation and Drought Monitoring Using Satellite and Ground Data. In Space Informatics for Grassland Sustainable Development. Proceedings of the First International Seminar on Space Informatics for Sustainable Development: Grassland Monitoring and Management. Ulaanbaatar, Mongolia. 1995. Р. 213-230.
3. Bayarjargal Yu., and Karnieli A. Comparison Study of NOAA-AVHRR Derived Drought Indices using Change Vector Analysis Technique. 2004. In processing to press.
4. Bayarjargal Yu., Adyasuren Ts., and Munkhtuya Sh, Drought and Vegetation Stress Monitoring in the Arid and Semi-Arid Regions of the Mongolia using Remote Sensing and Ground Data. Proceedings of the 21st Asian Conference on Remote Sensing. Taipei, Taiwan. 2000. V1. Р. 372-377.
5. Briggs J., and Nellis D. Seasonal variation of heterogeneity in the tallgrass prairie: A quantitative measure using remotely sensing. Photogrammetric Engineering and Remote Sensing. 1991. 57. Р. 407411.
6. Chavez P.S., and MacKinnon D.J. Automatic Detection of Vegetation Changes in the Southwestern United States Using Remotely Sensed Images, Photogrammetric Engineering and Remote Sensing, 1994. 60. Р. 571-583.
7. Chilar J., Laurent L.S., and Dyer J.A. Relation between the normalized difference vegetation index and ecological variables. Remote Sensing of Environment. 1991. 35. Р. 279-289.
8. Chuluun T., Tieszen L.L., and Ojima D. Land Use Impact on C4 Plant Cover of Temperate East Asian Grasslands. In Proceedings of NIES Workshop on Information Bases and Modeling for Land-use and Land-cover Changes in East Asia. 1999. Р. 103-109.
9. Collins J.B., and Woodcock C.E. An Assessment of Several Linear Change Detection Techniques for Mapping Forest Mortality Using Multitemporal Landsat TM Data. Remote Sensing of Environment. 1996.56. Р. 66-77.
10. Elmore A.J., Mustard J.F., Manning S.J., and Lobell D.B. Quantifying Vegetation Change in Semiarid Environments: Precision and Accuracy of Spectral Mixture Analysis and the Normalized Difference Vegetation Index. Remote Sensing of Environmen., 2000. 73. Р. 87-102.
11. ERDAS. ERDASField Guide. ERDAS Inc. Atlanta. 1997. 655 р.
12. ErdenetuyaM., andKhudulmur S. Land cover change and pasture estimation of Mongolia from space. Proceedings of Map Asia - 2002. Bangkok, 2002. Р. 46-49.
13. Hayes J.D., andSader S.A. Change detection techniques for monitoring forest clearing and regrowth in a tropical moist forest. Photogrammetric Engineering and Remote Sensing. 2001. 67. Р. 1067-1075.
14. Helmschrot J., andFlugel W-A. Land use characterization and change detection analysis for hydrological model parameterisation of large scale afforested areas using remote sensing. Physics and Chemistry of the Earth. 2002. 27. Р. 711-718.
15. Jensen J.R. Introductory Digital Image Processing: A Remote Sensing Perspective. Prentice-Hall, Englewood Cliffs, New Jersey. 1986. 379 р.
134
EA^P^APrAH, KAPHHEHH
16. Knick S.T., Rotenberry J.T., and Zarriello T.J. Supervised classification of Landsat thematic mapper imagery in a semi-arid rangeland by nonparametric discriminant analysis. Photogrammetric Engineering and Remote Sensing. 1997. 63. P. 79-86.
17. Koch M. Geological controls of land degradation as detected by remote sensing: a case study in Los Monegros, northeast Spain. International Journal of Remote Sensing. 2000. 21. P. 457-473.
18. Lambin E.F., and. Strahler A.H. Change-vector analysis in multitemporal space: A tool to detect and categorize land-cover change processes using high temporal-resolution satellite data. Remote Sensing of Environment. 1994. 48. P. 231-244.
19. Lunetta R.S. and Elvidge C.D. Remote Sensing Change Detection: Environmental Monitoring Methods and Applications. Ann Arbor Press. Michigan. USA. 1998. 318 p.
20. Lunetta R.S., Lyon J.G., Guindon B., and Elvidge C.D. North American Landscape Characterization: "Triplicate" Data Sets and Data Fusion Products. In R.S.Lunetta and C.D.Elvidge (eds.) Remote Sensing Change Detection: Environmental Monitoring Methods and Applications. Ann Arbor Press. Michigan. USA. 1998. P. 41-50.
21. Lyon J.G., Lunetta R.S., and Elvidge C.D. A change detection experiment using vegetation indices. Photogrammetric Engineering and Remote Sensing. 1998. 64. P. 143-150.
22. Markham B.L., and Barker J.L. Radiometric Properties of U.S. Processed Landsat MSS data. Remote Sensing of Environment. 1987. 22. P. 39-71.
23. Pilon P.G., Howarth P.J., BullockR.A., andAdeniyi P.O. An enhanced classification approach to change detection in semi-arid environments. Photogrammetric Engineering and Remote Sensing. 1988. 54. P. 1709-1716.
24. Purevdorj Ts., Tateishi R., Ishiyama T., and Honda Y. Relationships between percent vegetation cover and vegetation indices. International Journal of Remote Sensing. 1998. 19. P. 3519-3535.
25. Ram B., and Kolarker A.S. Remote sensing application in monitoring land-use changes in arid Rajasthan. International Journal of Remote Sensing. 1993. 14. P. 3191-3200.
26. Rogan J., and Yool S.R. Mapping fire-induced vegetation depletion in the Peloncillo Mountains, Arizona and New Mexico. International Journal of Remote Sensing. 2001. 22. P. 3101-3121.
27. Sellers P.J. Canopy reflectance, photosynthesis, and transpiration. International Journal of Remote Sensing. 1985. 6. P. 1335-1372.
28. Singh A. Digital change detection techniques using remotely sensed data. International Journal of Remote Sensing. 1989. 10. P. 989-1003.
29. Sohl T.L. Change analysis in the United Arab Emirates: An investigation of techniques. Photogrammetric Engineering and Remote Sensing. 1999. 65. P. 475-484.
30. Tateishi R., Wen C.G., and Perera L.K. Global four-minute land cover data set. Journal of the Japan Society of Photogrammetry and Remote Sensing. 1997. 36. P. 66-78.
31. The Landsat-7 Science Data User's Handbook. The Landsat Project Office, NASA's Goddard Space Flight Center, Greenbelt, Maryland. http://ltpwww.gsfc.nasa.gov/ IAS/handbook/handbook_toc.html
32. Tucker C.J. Red and Photographic Infrared Linear Combinations for Monitoring Vegetation. Remote Sensing of the Environment. 1979. 8. P. 127-150.
33. Tucker C.J., Nicholson S,. Variations in the size of the Sahara desert from 1980-1997. Ambio. 1999. 28. P. 587-591.
34. Tucker C.J., Dregne H.E., and Newcomb W.W. Expansion and contraction of the Sahara Desert from 1980 to 1990. Science. 1991. 253. P. 299-301.
35. Vermote E., Tanre D., Deuze J.L., Herman M., and Morcette J.J. Second Simulation of the Satellite Signal in the Solar Spectrum (6S), IEEE Transactions on Geoscience and Remote Sensing. 1997a. 35. P. 675-685.
36. Vermote E., Tanre D., Deuze J.L., Herman M., and Morcette J.J. Second Simulation of the Satellite Signal in the Solar Spectrum (6S), 6S User Guide Version 2. Department of Geography, University of Maryland. 1997b. 216 p.
37. Yuan D., Elvidge C.D. NALC Land Cover Change Detection Pilot Study: Washington D.C. Area Experiments. Remote Sensing of Environment. 1998. 66. P. 166-178.
38. Yuan D., Elvidge C.D., and Lunetta R.S. Survey of Multispectral Methods for Land Cover Change Analysis. In R.S. Lunetta and C.D. Elvidge (eds.) Remote Sensing Change Detection: Environmental Monitoring Methods and Applications. Ann Arbor Press. Michigan. USA. 1998. P. 21-39.
