land cover, vegetation cover changes estimated using multi-source satellite data

International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

Contents lists available at ScienceDirect

International Journal of Applied Earth Observation and Geoinformation journal homepage: www.elsevier.com/locate/jag

Changing landscape in the Three Gorges Reservoir Area of Yangtze River from 1977 to 2005: Land use/land cover, vegetation cover changes estimated using multi-source satellite data Jixian Zhang a,*, Liu Zhengjun a, Sun Xiaoxia b a b

Key Laboratory of Mapping from Space of State Bureau of Surveying and Mapping, Chinese Academy of Surveying and Mapping, 16 Beitaiping Road, Haidian District, Beijing, PR China China University of Mining and Technology, Sanhuannan Road, Xuzhou, PR China

A R T I C L E I N F O

A B S T R A C T

Article history: Received 14 October 2008 Received in revised form 15 April 2009 Accepted 9 July 2009

The eco-environment in the Three Gorges Reservoir Area (TGRA) in China has received much attention due to the construction of the Three Gorges Hydropower Station. Land use/land cover changes (LUCC) are a major cause of ecological environmental changes. In this paper, the spatial landscape dynamics from 1978 to 2005 in this area are monitored and recent changes are analyzed, using the Landsat TM (MSS) images of 1978, 1988, 1995, 2000 and 2005. Vegetation cover fractions for a vegetation cover analysis are retrieved from MODIS/Terra imagery from 2000 to 2006, being the period before and after the rising water level of the reservoir. Several analytical indices have been used to analyze spatial and temporal changes. Results indicate that cropland, woodland, and grassland areas reduced continuously over the past 30 years, while river and built-up area increased by 2.79% and 4.45% from 2000 to 2005, respectively. The built-up area increased at the cost of decreased cropland, woodland and grassland. The vegetation cover fraction increased slightly. We conclude that significant changes in land use/land cover have occurred in the Three Gorges Reservoir Area. The main cause is a continuous economic and urban/rural development, followed by environmental management policies after construction of the Three Gorges Dam. ß 2009 Elsevier B.V. All rights reserved.

Keywords: Remote sensing Land use and land cover Vegetation cover fraction Change analysis Three Gorges Reservoir Area

1. Introduction In recent years, to satisfy the hydrological energy and water resources consumption demand from the rapid development of economy and society, many large-scale water conservation projects have been undertaken at the global level. These major projects have brought certain economic benefits, but have also had adverse effects on the ecological environment. For example, the influence of Egypt’s Aswan High Dam, on water and soil quality, human health 20 years after its completion (White, 1988; Moussa et al., 2001); the influence of the Itaipu Project located along the border river between Brazil and Paraguay, on vegetation, animals, water quality and soil pollution (Murphy, 1976; Strand et al., 2007). Therefore, for many years using new technologies to monitor the impacts of these activities on the ecological environment has been a focus of attention around the world (Ivits et al., 2009; Liao, 2004; Liu et al., 2002; Veldkamp and Lambin, 2001). As the largest water conservation project in the world, China’s Three Gorges Project has attracted worldwide attention. This

* Corresponding author. Tel.: +86 10 88229375; fax: +86 10 68218654. E-mail address: [email protected] (J. Zhang). 0303-2434/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jag.2009.07.004

attention has not been only for its comprehensive social and economic benefits such as flood prevention, hydropower generation, and shipping capacity, but also for the potential security impacts on the natural environment, potential geological disasters, as well as on the biological diversity imposed on the surrounding reservoir area. Specifically, the major impacts include land cover changes caused by population migration, potential water pollution and soil erosion following the construction of the Three Gorges Dam and the immigration towns, etc. The Chinese Government and the environmental management professionals have long been aware of these problems, and have gradually formulated and implemented a series of relevant policies (Luo and Shen, 1994; Tullos, 2009). Among these impacts, land use/land cover change (LUCC), as well as the vegetation cover change, have been well recognized as some of the most important indicators for global and regional environmental changes (Meyer and Turner, 1994; Lindquist et al., 2008). Therefore, quantifying the LUCC and vegetation cover change is crucial for assessing the effect of land management policies and environment protection decisions (Opoku, 2007). Many studies have been carried out about land use mapping, change detection, as well as vegetation monitoring using multitemporal satellite data for regional ecological and environmental

404

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

change research (Veldkamp and Lambin, 2001; Peng et al., 2006; Pouliot et al., 2009; Berberoglu and Akin, 2009). Various techniques have been successfully used in the land use/land cover classification and change detection, e.g., pixel based classification (Foody, 1996; Duda et al., 2001), object oriented classification (Geneletti and Gorte, 2003; Elmqvist et al., 2008), artificial neural network classification (Kanellopoulos et al., 1992; Liu et al., 2004), post-classification comparison change detection (Serra et al., 2003), and visual interpretation (Liu et al., 2005). For vegetation monitoring, some biophysical parameters, e.g., vegetation leaf area index (LAI), fraction of photosynthetically active radiation (fPAR), and vegetation cover fraction, have been recommended for monitoring its long-term changes (Ganguly et al., 2008; North, 2002; Sun et al., 2008; Mostovoy et al., 2008). For vegetation changes estimation over large areas, the retrieval of the fraction of vegetation cover (FVC) from remotely sensed data has been an effective method (Carlson and Ripley, 1997; Zhang et al., 2006). In this paper, the multi-temporal satellite dataset in the Three Gorges Reservoir Area has been analyzed to understand LUCC as a consequence of driving factors. Our study focused on the following two aspects: (1) to estimate LUCC from 1977 to 2005 in the TGRA, and to obtain vegetation cover changes from 2000 to 2006, being the time before and after the water line rising of the reservoir, and (2) to incorporate and analyze landscape changes in the TGRA using these estimated results. The remaining sections of this paper are organized as follows. Section 2 introduces the background of the study area. Section 3 describes the data and method used in this article. In Section 4, the LUCC and vegetation change results are presented, followed by a discussion of the results in Section 5. The conclusions of this research are given in Section 6. 2. Study area The Three Gorges Reservoir Area (TGRA) is located between latitude 288560 N–318440 N and longitude 1068160 E–1118280 E, covering the lower section of the upper reaches of the Yangtze River, with an area of 58,000 km2 and with a population of almost 20 million (Meng et al., 2005). It consists of 21 counties or cities of Chongqing municipalities and Hubei province (see Fig. 1), with various geographic conditions, 74% of the region is mountainous, 4.3% of the region is plain area and 21.7% hilly area (Peng et al., 2004). The climate in the TGRA is a subtropical monsoon climate, and vegetation in this area is abundant and diverse.

The construction period of the Three Gorges Project lasted from 1993 to 2009 with the final water level at 175 m, the total storage capacity of the reservoir being 39.3 billion m3. The water level was 135 m in June 2003. With its continuous rising, 1.13 million immigrants will be resettled. For instance, from 2000 to 2004, about 96,000 immigrants from the Chongqing reservoir region along the upper reach of the Three Gorges Reservoir Area have been moved and settled outside the region. As the water in the reservoir increased up to 175 m height, about 240 km2 citrus and farmland will be submerged. Along with the rapid population growth and economic development in this region since 1980s, the ecological environment has changed rapidly because of excessive cultivation and over-felling, e.g., the woodland area and the diversity of vegetation reduced quickly (Zhou et al., 2004). As a consequence, soil erosion in certain parts of the TGRA is becoming more and more serious. To protect and reverse the deterioration of the ecological environment of the Yangtze River Valley, several policies and management projects have been carried out in this region. For example, the Shelterbelt Construction Project in the upper reaches of Yangtze River was started by the central government in 1989 (State Forestry Administration, 2006). This project included tree planting, aerial planting, closing hillsides to facilitate afforestation and raising seedlings, through which, a net increase of 9.6 percentage points of the forest cover was gained by the end of 2000, the close of the first phase of the Shelterbelt Construction Project. After that, the second phase construction of shelter forest system was carried out. At the same time, to strengthen the ecological and environmental construction in the TGRA and the surrounding area, the State Council approved the ‘‘Greenbelt Around the Three Gorges Reservoir Construction Project Planning’’ in July 2004 with the construction period from 2004 to 2007. The work includes the implementation of returning farmland to forest, planting tree on barren hills and wasteland, closing hillsides to facilitate afforestation and construction of basic farmland, and other measures to protect the existing forest resources, and create a water conservation forest and soil conservation forests, ensuring the ecological safety of the Three Gorges Reservoir. The project involved 26 districts and counties and 204 townships of Hubei and Chongqing, with the area of returning farmland to forest and grass being 488.7 km2, afforestation of barren hills and wasteland being 119.3 km2, closing hillsides to facilitate afforestation being 120.0 km2, and

Fig. 1. Location of the study area.

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

509.3 km2 of basic farmland construction (NDRC, 2005). Additionally, some terrestrial plant rescue projects such as Hubei Yichang Dalaoling Plant Protection Zone, Hubei Xingshan Longmen River Natural Evergreen Broad-leaved Forest Protection Project, and the Protection of Ancient Large Trees Projects, were also carried out simultaneously. 3. Data and methods This study adopts remote sensing techniques for ecological environmental monitoring, and image analysis approaches as well as geographic information techniques to extract the land use/land cover change during the past 30 years in the reservoir area using Landsat MSS/TM/ETM + data, and its recent vegetation cover changes during the last 7 years using MODIS data. Based on the above results, we analyzed the spatial and temporal changing patterns of land use/land cover and fractional vegetation cover as several indices. The technical flowchart of this research is given in Fig. 2. 3.1. Land use/land cover change monitoring Because of the long time span and the large area of the study site, five time series of Landsat satellite images were selected for land use/land cover change monitoring. These are MSS images acquired between 1977 and 1979 (for short 1977), with a spatial resolution of 80 m, and TM images with the spatial resolution 30 m acquired from 1987 to 1988 (for short 1987), from 1995 to 1996 (for short 1995), in 2000 and in 2005 respectively. Each temporal dataset includes 9 scenes of images covering the entire TGRA. According to the land use/land cover classification capability of Landsat images in the TGRA, the National Land Use Change Database Hierarchical Classification System was adopted in this

405

research (Liu et al., 2005). This classification system is divided into 3 levels. At first level, there are 6 classes, which are croplands, woodlands, grassland, water bodies, built-up areas and unused land. These 6 classes are subdivided into 25 classes at second level, which are paddy field and dry land; woodland, shrub forest, low canopy density forest, and other wooded land (e.g., nursery for trees, garden); high coverage grassland, medium coverage grassland and low coverage grassland; river, lake, reservoir, permanent glaciers snow, sand beach, tideland; urban, rural, other built-up areas; sandy, gobi, saline-alkali soil, wetlands, bare land, and uncovered rock gravel, and other unused lands (e.g., alpine desert, tundra). At third level, only paddy field and dry land are subdivided into 4 classes according to the topography and the slope, which are mountains paddy field (or dry land), hills paddy field (or dry land), plains paddy field (or dry land), and slopes greater than 258 paddy field (or dry land). In this research, in order to facilitate the analysis of problem, the third level classes were regrouped into slopes lower than 258 paddy field (or dry land), and slopes greater than 258 paddy field (or dry land). Due to the large geographic extent of the research area, substantial time difference of the dataset, and especially the complicate land use structure and steep terrain in this area, automatic classification methods will be difficult to meet the accuracy requirements. Therefore, the strategy that combines visual interpretation techniques with prior knowledge and historical data from a GIS database, together with ground truth from field investigation, were implemented to detect the LUCC information and produce the land use and land cover map. Firstly, The Landsat TM (MSS) images were geo-referenced and orthorectified, using ground control points and a 25 m Digital Elevation Model (DEM). The mean location errors for geometric rectification are less than 1 pixel (i.e., 30 m for Landsat TM images). Image registration, color balancing, and mosaicking were then performed. Secondly, in order to improve the accuracy of image

Fig. 2. Flowchart of methodology.

406

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

interpretation, field investigation was carried out in Kaixian and Wanzhou County in April 2005 and September 2006, roughly the same growing season for most TM images acquired. Sample photos are obtained according to different categories of land cover and different slopes for the same category. Using the sample photos, together with collected land use historical data during the whole period and the corresponding SPOT5 images in these two counties, the image interpretation sample database was established. Thirdly, each two adjacent temporal images were overlaid in the GIS software, the changes of land use/land cover was then identified and digitized by visual interpretation, with the help of the sample database. In this way, the change information of land use/land cover of four periods, which are from 1977 to 1988, from 1988 to 1995, from 1995 to 2000, and from 2000 to 2005, was extracted. Finally, to get the full coverage land use and land cover map, the TM mosaic image in 2000 was selected for land cover/land use classification by visual interpretation. By superimposing the change information obtained with the land use/land cover map in 2000, the land use/land cover maps of the other four periods were also reconstructed using GIS overlay analysis (Fig. 3). For

instance, the land use/land cover map of 2005 can been obtained using the land use/land cover vector data in 2000, and the change detection result from 2000 to 2005. 3.2. Vegetation cover fraction retrieval The fraction of vegetation cover (FVC) can be defined as the vertical projection of the crown or shoots area of vegetation to the ground surface in a unit area, expressed as the fraction or percentage (Abdelaziz et al., 2007; Felix et al., 2006; Purevdor et al., 1998). The dataset used in this study were MODIS 16-day composite NDVI time series data products from 2000 to 2006 provided by Earth Resources Observation Systems (EROS) Data Center, with the spatial resolution 250 m. The date of data acquisition is from April to October, which is in the growing season of vegetation and can better reflect the ground vegetation growth and changes. To reduce the image noise from clouds, shadows, etc., we synthesized two 16-day composite NDVI images to one 32-day composite NDVI image in sequence by using the maximum value

Fig. 3. Illustration of the changes of the land use and land cover (LULC) of the Three Gorges Reservoir Area in 1977 and in 2005. (a) LULC map in late 1977; (b) LULC map in 2005.

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

composite (MVC) method (Eklundh, 1995), as can be formulated below: NDVIi ¼ maxðNDVI j Þ

(1)

where NDVIi (i = 4, . . ., 10) is the NDVI image of ith month, NDVIj (j = 1, 2) is the two 16-day NDVI images to be composited. The implementation of this method is to choose the largest pixel value of the two index images for each pixel; therefore, some noise was reduced at the same time. A small amount of cloud pollution, however, was still presented after the MVC synthesis. The best index slope extraction (BISE) method (Viovy et al., 1992; Wang et al., 2005) was then used to remove the remaining cloud noise. dNDVIt1;t ¼

ðNDVIt1  NDVIt Þ  100% NDVIt1

(2)

dNDVIt;tþ1 ¼

ðNDVItþ1  NDVIt Þ  100% NDVItþ1

(3)

Table 1 Changes of land use/land cover in the TGRA from 1977 to 2005 (Unit: km2). Categories

1977

1987

1995

2000

2005

Cropland Cropland (gradient <258) Cropland (gradient >258)

22,327 21,690 637

22,308 21,671 637

22,293 21,656 637

22,289 21,651 638

22,206 21,570 636

Woodland Grassland

27,368 7,330

27,361 7,325

27,339 7,323

27,291 7,313

27,229 7,306

Water body River Other Water bodies

818 601 217

822 602 221

823 602 221

819 603 216

899 687 211

Built-up area Urban Rural Other built-up

417 240 124 53

447 260 129 59

483 274 130 80

527 308 137 82

613 376 151 86

10

10

10

10

9

Unused land

where NDVIt  1 and NDVIt + 1 denote the NDVI values of time t  1 and t respectively; dNDVIt  1,t and dNDVIt,t + 1 show the variation rate from t  1 to t and from t + 1 to t respectively. It is assumed that the pixel at time t is affected by clouds if dNDVIt  1,t and dNDVIt,t + 1 are both surpass 20%, then the t time pixel value is corrected by the average of time t  1 and time t + 1. We applied the algorithm to detect the contaminated position point and smooth the NDVI time series data for all the pixels throughout our study periods excluding the starting and ending points. For the first and end pixels, the improved BISE method is adopted: if dNDVI1,2 is

407

more than 20%, the first pixel is substituted by the average value of time 1 and time 2. The last pixel value is processed in the same way. Based on the resultant monthly NDVI images, the vegetation fractions from April to October in each year were calculated using the ‘‘Dense Vegetation Model’’ approach (Gutman and Ignatov, 1998), which can be formulated as: fg ¼

NDVI  NDVI0 NDVI1  NDVI0

(4)

Fig. 4. Change trend maps of different land use/land cover types from 1977 to 2005 in the Three Gorges Reservoir Area. (a) Cropland change trend map; (b) woodland change trend map; (c) grassland change trend map; (d) water body change trend map; (e) built-up area change trend map.

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

408

where NDVI0 and NDVI1 correspond to the values of NDVI for bare soil and for dense vegetation cover, respectively. The vegetation fractions of seven months from April to October were then averaged to create the annual vegetation cover. This was used to analyze the interannual change of vegetation cover. In order to study the change between different level of vegetation cover, 5 classes were divided by the threshold value of FVC data, which were no vegetation cover (NC) (FVC  0.2), low vegetation cover (LC) (0.2 < FVC  0.45), medium vegetation cover (MC) (0.45 < FVC  0.75), high vegetation cover (HC) (0.75 < FVC  0.90, and full vegetation cover (FC) (FVC > 0.9). 4. Results 4.1. Land use/land cover change monitoring According to the method described in Section 3.1, a qualitative classification and change information extraction about the land use/land cover in the TGRA was conducted, using both Landsat MSS/TM/ETM + images. Then the land use/land cover and its accompanied change information for the 5 different times over the recent 30 years were obtained. The results are shown in Fig. 3 and Table 1. From Fig. 3, and Table 1, obvious land use/land cover changes in the TGRA during the research period can be observed. This can be summarized as a continuous decrease of the area of croplands, woodlands and grasslands, and a significant increase in the area of built-up lands and water bodies. Change trends of the main land cover classes are shown in Fig. 4. Specifically, the area of croplands decreased with 121 km2, from 22,327 km2 in 1977 to 22,206 km2 in 2005. Croplands (gradient >258) in mountainous areas showed little changes, whereas croplands (gradient <258) showed more severe changes. From 1977 to 2000, the area of croplands (gradient <258) reduced steadily, with 39 km2 of croplands (gradient <258) being converted to other types of land cover, about 2 km2/year; a rapid drop of 81 km2 occurred from 2000 to 2005, about 16 km2/year. From 1977 to 2005, the woodland and grassland areas reduced by 139 km2 and 23 km2, respectively. Approximately 70% of the changes occurred between 1995 and 2005, when woodland and grassland areas decreased by 110 km2 and 16 km2, respectively. The grassland areas reduced relatively slowly over the different study periods, as these are mainly located in the mountainous areas. The water body area increased about 1 km2 between 1977 and 2000; while it increased 80 km2 from 2000 to 2005, when the water level rose to 135 m high. From Table 1 it is clearly shown that the change of water body area is mainly contributed from the increased area of river. Compared to other land cover types, the built-up area increased notably from 417 km2 in 1977, to 613 km2 in 2005. Fig. 4 shows that all built-up classes increased steadily, with

urban areas showing an increase of 136 km2 between 1977 and 2005, accounting for 69% of the increasing total area of built-up land. Between 2000 and 2005, urban areas changed most rapidly with an increase of 69 km2, about 14 km2 per year. Compared with urban area, changes of rural and other types of built-up areas were at a normal level, increasing by 27 km2 and 33 km2 respectively. To identify the pattern of land use changes from one land use class to another, we created land use transformation matrices between each subsequent period. It is shown that the main conversion direction was almost consistent for each period. For example, the land use transformation matrix during 2000–2005 demonstrates that the major land use changes occurred in this period, as shown in Table 2. In Table 2, it is shown that the conversion of land cover classes mainly took place from cropland, woodland and grassland to builtup land and water body, with the dominant conversion being from cropland to urban land. This implies that urban expansion occupied mainly croplands. In addition, some urban lands were converted to water areas due to the towns’ inundation caused by the rising water level after completion of the dam construction. Increase of construction lands and water areas was largely at the expense of croplands, woodlands and grasslands. 4.2. Vegetation cover change estimation Similarly, according to the method described in Section 3.2, the vegetation cover fractions from 2000 to 2006 in the TGRA were retrieved using MODIS NDVI data, and 7 periods of FVC maps were obtained. Fig. 5 shows the FVC maps of 2000 and 2006, from which a small increase of vegetation cover can be observed. To analyze the change further, the feature values of these 7 FVC images were calculated in Table 3, and the areas for different levels of vegetation cover are given in Table 4. In Table 3, an overall small increase of the vegetation cover during these 7 years is noticed, despite a slight decrease in 2001 and in 2005. The mean vegetation cover fraction was equal to 0.761 in 2000 and to 0.769 in 2006, showing an increase of 0.8%. Regions with vegetation cover changes were usually built-up areas and areas under rapid economic development, such as the dam construction area, Fengjie County, Yunyang County, Kai County, the northern part of Fuling, the middle part of Chongqing City, and the southern part of Jiangjin City. The vegetation cover in TGRA changed from NC to LC or from MC to HC (see Fig. 5 and Table 4), e.g., the ratio of HC area increased from 68.7% in 2000 to 72.5% in 2006. 4.3. Accuracy assessment The land use/land cover classification map from 2.5 fused SPOT5 imageries was used in this study to facilitate the evaluation of classification results from Landsat imageries (Jensen, 2005). The

Table 2 Land use transformation matrix in the Three Gorges Reservoir Area during 2000–2005 (km2). From

To Cropland

Cropland Woodland Grassland Urban land Village land Other built-up Water body Unused land

Woodland 0

0 0 0 0 0.2 0 0

0 0 0 0 0 0

Grassland 0 0 0 0 11.8 1.5 0

Urban land

Village land

Other built-up

Water body

Unused land

60.0 9.7 3.5

5.3 9.1 4.9 0

5.4 4.9 5.7 0 0

0 3.8 7.3 28.1 0 0

0 0 0 0 0 0 0

0 0 0 0

0 0 0

1.8 0

0

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

409

Fig. 5. The vegetation cover fraction maps of the Three Gorges Reservoir Area in 2000 and in 2006. (a) Vegetation cover fraction map in 2000; (b) vegetation cover fraction map in 2006.

Kai County with an area of 3959 km2 was selected as a test area. Reference locations used for accuracy assessment were selected using a stratified random sampling method, keeping the sample size for each class proportional to its area. The sample pixels were then compared with their corresponding pixels on the Landsat classification maps. User’s and producers’ accuracies as well as kappa statistics for each class are reported in Table 5. The overall accuracy and overall kappa statistics are 82.1% and 0.77%, respectively. To assess the change detection accuracy, the 1:250,000 land use maps in Chongqing in 1999 and 2005 were used. The class changes of the randomly selected sample points were obtained from the land use maps. This reference information was then compared

with the change detection results between 2000 and 2005. It is shown that the consistency of the evaluated pixels is over 89%. To assess the vegetation cover fraction retrieval accuracy, the SPOT 5 land cover classification results were used. All the land use/ land cover classes of SPOT5 classification data were firstly rasterized to a 2.5 m resolution image, and reclassified as ‘‘vegetation’’ for forest, shrub, and grass, or ‘‘non-vegetation’’ for other classes. Secondly, the reclassified data was then resampled to 250 m for each pixel, and the digital value of the new pixel was the percentage of pixels classified as vegetated land cover. This resampled raster data was finally used for the accuracy assessment and compared with corresponding pixels on the FVC image. Results show that the FVC retrieval accuracy is 84%.

Table 3 Feature values of the fraction images of vegetation cover from 2000 to 2006.

5. Discussions

Time (year)

Minimum

Maximum

Average

Mean square error

2000 2001 2002 2003 2004 2005 2006

0.074 0.100 0.065 0.058 0.068 0.085 0.097

0.907 0.915 0.913 0.930 0.922 0.903 0.916

0.761 0.749 0.771 0.779 0.783 0.764 0.769

0.067 0.071 0.065 0.068 0.074 0.073 0.073

5.1. Dynamic degree analysis of LUCC To determine the change rate of land use categories in different study periods, and assess the influence of the construction of the Three Gorges Project on the changing trend and speed of land use and land cover in the TGRA, the single land use dynamic degree and the synthetic land use dynamic degree were adopted (Wang and Bao, 1999).

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

410

Table 4 Area of vegetation cover changes in the Three Gorges Reservoir Area. Vegetation cover levels

Area (km2)

2000–2006

2000

2001

2002

2003

2004

2005

2006

Change area (km2)

Change ratio (%)

No vegetation cover Low vegetation cover Medium vegetation cover High vegetation cover Full vegetation cover

44 619 17,492 39,835 0

34 651 23,896 33,408 1

30 565 14,518 42,876 1

31 636 10,742 46,579 1

52 670 10,781 46,483 4

33 684 17,774 39,499 0

22 728 15,186 42,029 1

22 109 2,306 2,194 1

49 18 13 6

Total

57,990

57,990

57,990

57,990

57,990

57,990

57,966

24

0

Table 5 Accuracy assessment of the six main classes for the classified Landsat TM images. Reference

Cropland

Forest

Grass

Water

Built-up

Unused land

Total

Classified as Cropland Forest Grass Water Built-up Unused land Total

82 7 10 1 0 0 100

7 82 9 1 0 1 100

7 7 84 0 1 1 100

2 0 1 45 2 0 50

3 1 1 0 45 0 50

7 4 2 0 0 7 20

108 101 107 47 48 9 420

90.0 95.7

90.0 93.8

35.0 77.8

Producer’s accuracy (%) User’s accuracy (%) Overall classification accuracy (%) Kappa value

82.0 75.9 82.1 0.77

82.0 81.2

84.0 78.5

Table 6 Dynamic degree of single land use in the Three Gorges Reservoir Area. Land use types

1977–1987

1987–1995

1995–2000

2000–2005

Cropland (gradient <25˚) Woodland Grassland River Urban built-up Unused land

0.007

0.009

0.005

0.075

0.002 0.005 0.015 0.684 0

0.010 0.004 0 0.680 0

0.035 0.027 0.047 2.474 0.262

0.046 0.017 2.792 4.453 1.472

The single land use dynamic degree can be defined as: Ds ¼

At2  At1 1   100% t2  t1 At1

(5)

where At1 is the area of the land use type in time t1; At2 is the area of the land use type in time t2. The results are shown in Table 6. The synthetic land use dynamic degree can be defined as: Pn Dc ¼

1 i¼1 jDAini  DAouti j P   100% t2  t1 2 ni¼1 Ai

(6)

In which Ai is the area of class i in time t1, DAin  i is the total area converted from other classes to class i, and DAout  i is the total area converted from class i to other classes. Using Eq. (6) combined with land use transformation matrix, the synthetic land use dynamic degrees were calculated for all the periods in the study area (see Fig. 6). From Table 6 it can be seen that during each study period, cropland (gradient <258), woodland, and grassland reduced at a low speed, whereas urban built-up and river increased at a relatively high speed. For the urban built-up area, the change trend accelerated continuously during almost all periods, reaching the highest dynamic degree of 4.453% between 2000 and 2005. Moreover, the most rapid dynamic change rates of several land types all happened either between 1995 and 2000, or between 2000 and 2005, i.e., just before or just after the Three Gorges Reservoir water storage respectively. The construction of Three

Gorges Dam therefore did not only cause a rapid expansion of water body and built-up areas, but also had an influence on land use changes in the immigration regions because of a large amount of resettlement and population migration. From Table 6 we can also see a continuous increase trend of the land use change rate, from 0.015% between 1977 and 1988, to 0.112% between 2000 and 2005. Clearly, the rate of change of the land use in the TGRA over the past 30 years has accelerated between the different time periods. It seems that the result is most probably caused by the growing population and urbanization in this area. The population in this area has grown from 14.8 million in 1992 to 20.0 million in 2004. The growth in population and urbanization has increased the demand for food, houses, and factories (Hunter et al., 2003; Semwal et al., 2004). Furthermore, the Three Gorges Project accelerated the regional economy, which normally leads to the rapid rise of built-up area. 5.2. The change extent analysis of land use and land cover The extent of land-use changes can be used to better understand the trend of land-use change and its driving forces (Chen et al., 2006). In general, some indices, e.g., the rate of land utilized

Fig. 6. Integrated dynamic degree in Three Gorges Reservoir Area.

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

411

Table 7 Extent indices of land use in the Three Gorges Reservoir Area during different periods (%). Time

The rate of land utilized

The rate of land reclaimed

The rate of slope land reclaimed

The rate of land used for building

The rate of land for forest and grass

1975 1987 1995 2000 2005

99.98 99.98 99.98 99.98 99.98

38.32 38.28 38.26 38.27 38.11

1.09 1.09 1.09 1.09 1.09

0.72 0.77 0.83 0.90 1.05

59.55 59.52 59.48 59.41 59.28

(the area of utilized land/total area), the rate of land reclaimed (the area of cultivated land/total area), the rate of slope land reclaimed (the area of cultivated land with slope larger than 258/total area), the rate of land used for built-up (the area of built-up/total area), the rate of land for forest and grass (the area of forest and grass land/total area), are usually used to measure the extent of land use. Table 7 shows the extent indices of land use in the TGRA during the different study periods. The rate of land utilized was 99.98%, far higher than the national average rate of 77.6% (Yu and Gao, 2003), which means the area of unused land was quite small in the TGRA. The rate of land reclaimed was decreasing, but still higher than the national average rate of 14.21% (Yu and Gao, 2003). For slope land, the rate remained steadily at 1.09%. The rate for built-up land continuously rises, whereas for forest and grass the rates decrease slowly from a relatively high value in this period. As can be seen over the 30-year period, the rate for built-up areas changed the most, followed by forest and grass (Cao et al., 2007). To understand the integrated land use extent in the TGRA, the synthetic index of land use extent was calculated. According to the natural equilibrium state of land natural complex under social factors, the extent indices of land use were divided into four levels and given an index to each level (Liu, 1992). The synthetic index of land use extent is represented by the following equation: Lj ¼

n X Ai C i  100;

L j 2 ½100; 400

(7)

i¼1

where Ai is the given index of level i, Ci is the percentage of the area of level i in the study area, n is the number of levels. The synthetic indices of land use extent in the TGRA over the different periods are shown in Table 8. The results indicate that the extent integrated of land use was increasing gradually during these five periods. From 1977 to 1987, the synthetic indices of land use extent increased slowly at 0.005/year. A rapid growth at 0.03/year occurred from 1995 to 2005, six times that of 1977 to 1987. It shows that the intensity of land use increased noticeably during the past 10 years. Compared with the national average value of 231.92 (Yu and Gao, 2003), the synthetic indices of land use extent shown in Table 8 were relatively higher. According to this analysis, it seems that human activities greatly affected the landscape in the TGRA over the past 30 years. 5.3. Integrated analysis of vegetation area and vegetation cover changes Monitoring results from Landsat and MODIS data indicates only small changes of the vegetation cover during the 7 years from 2000 to 2006 in this area. We also observed a slight decrease in the area of cropland, woody land and grassland, however, as compared to a small increase on the average per pixel vegetation cover. In many places, the vegetation area and vegetation cover fractions are correlative and consistent. If, for example, nonvegetation classes such as built-up areas and water body are transformed into vegetation, they both increase in area size. Such consistence does not exist, however, for changes within vegetation classes. In tree planting in woodlands, for example, the land use/

Table 8 Synthetic indices of land use extent in the Three Gorges Reservoir Area during different periods. Time

1975

1987

1995

2000

2005

Index

239.75

239.80

239.90

240.06

240.20

land cover remains the same, whereas the fraction of vegetation cover shows a significant change. Therefore, both indices are necessary to be used for understanding the per-pixel level environmental changes qualitatively and quantitatively. In our study, although the conversion from croplands, woodlands, grasslands to built-up lands and water areas resulted in the reduced vegetation area and vegetation cover fraction; the vegetation cover fraction in the vegetated area has been increased comparatively as a result of the implementation of some environment management policies, as have been described in Section 2. The implementation of all these measures increases the density of the vegetation cover in the original vegetation category, and which is the main consequences of the rising of vegetation cover fractions as we know from this study. It is clear that the administrative management of the ecological environment of the TGRA has apparently had a positive effect. 6. Conclusions In this paper, two indicators, i.e., vegetation area and vegetation cover fractions, were employed for regional vegetation assessment. In general we observed that the total vegetation area decreased in the study area during these periods while the vegetation cover fraction increased. Specifically, between 1977 and 2005, large changes occurred in the land use/land cover in the Three Gorges Reservoir Area. The main change is a continuous decrease of croplands and a continuous increase of built-up lands and water bodies. The main transformation trends are the conversion from croplands, woodlands, grasslands to urban lands and rivers. The speed and trend of land use/land cover changes differs between the different research periods. Before the Three Gorges Project construction, the land use/land cover changed gradually. The speed of LUCC was most rapid just before and just after the dam water level rising. The degree of change of water bodies and built-up areas was highest between 2000 and 2005, being equal to 2.792% and 4.453% respectively. Over the past 30 years, the rate of change from croplands, woodlands, grasslands to built-up lands and water bodies increased rapidly. The extent of land use increased gradually in the TGRA during the 30 years. The speed of land use extent enhanced was most rapid between 1995 and 2005. The rate for built-up areas changed the most, with an increasing trend, followed by forest and grass with a decreasing trend. Since 2000, vegetation cover remains relatively constant, because of the implementation of the pertinent policies. The change trends are conversions from NC to LC and from MC to HC. The rate of HC increased from 68.7% in 2000 to 72.5% in 2006.

412

J. Zhang et al. / International Journal of Applied Earth Observation and Geoinformation 11 (2009) 403–412

In conclusion, these research results show that the protection measures for ecological environment protection are, so far, effective in the study area. Acknowledgements This work has been supported by China’s National Social Development Research Program funded by the Ministry of Science and Technology (2004DIB3J107), National Key Basic Research and Development Program of China (2006CB701303). We would like to thank the anonymous reviewers for their very helpful comments and constructive feedback, and Professor John van Genderen at the ITC for his efforts on improving the article. References Abdelaziz, K., Sylvie, L., Hegarat, M., Catherine, O., Laurence, H.M., 2007. Determination of vegetation cover fraction by inversion of a four-parameter model based on isoline parametrization. Remote Sensing of Environment 111 (4), 553– 566. Berberoglu, S., Akin, A., 2009. Assessing different remote sensing techniques to detect land use/cover changes in the eastern Mediterranean. International Journal of Applied Earth Observation and Geoinformation 11 (1), 46–53. Cao, Y.G., Wang, J., Liu, Z.J., Cheng, Y., Liu, A.X., Xu, N., 2007. Analyses on temporalspatial traits of land use in Three Gorges during 30 Years. Science of Surveying and Mapping 32 (6), 167–170 (in Chinese). Carlson, T.N., Ripley, D.A., 1997. On the relation between NDVI, fractional vegetation cover, and leaf area index. Remote Sensing of Environment 62 (3), 241–252. Chen, W.B., Cui, L.J., Zhao, X.F., 2006. Temporal-spatial characteristics of land use in Xin jian County Jiangxi Province. Chinese Journal of Applied Ecology 17 (5), 873–877 (in Chinese). Duda, R.O., Hart, P.E., Stork, D.G., 2001. Pattern Classification. John Wiley & Sons, New York. Eklundh, L.R., 1995. Noise estimation in NOAA AVHRR maximum-value composite NDVI images. International Journal of Remote Sensing 16 (15), 2955–2962. Elmqvist, B., Ardo, J., Olsson, L., 2008. Land use studies in drylands: an evaluation of object-oriented classification of very high resolution panchromatic imagery. International Journal of Remote Sensing 29 (24), 7129–7140. Felix, Morsdorf, Benjamin, Kotz, Erich, Meier, Itten, K.I., Britta, Allgower, 2006. Estimation of LAI and fractional cover from small footprint airborne laser scanning data based on gap fraction. Remote Sensing of Environment 104 (1), 50–61. Foody, G.M., 1996. Approaches for the production and evaluation of fuzzy land cover classifications from remotely sensed data. International Journal of Remote Sensing 17 (7), 1317–1340. Ganguly, S., Schull, M.A., Samanta, A., Shabanov, Nikolay, V., Milesi, C., Nemani, R.R., Knyazikhin, Y., Myneni, R.B., 2008. Generating vegetation leaf area index earth system data record from multiple sensors Part 1: Theory. Remote Sensing of Environment 112 (12), 4333–4343. Geneletti, D., Gorte, B.G.H., 2003. A method for object-oriented land cover classification combing Landsat TM data and aerial photographs. International Journal of Remote Sensing 24 (6), 1273–1286. Gutman, G., Ignatov, A., 1998. The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction model. International Journal of Remote Sensing 19 (8), 1533–1543. Hunter, L.M., Manuel, D.J., Gonzalez, G., Matt, S., Kimberly, S.K., Richard, T., Thomas, C., Edwards, J., Robert, J.L., Mary, C., 2003. Population and land use change in the California Mojave: natural habitat implications of alternative futures. Population Research and Policy Review 22 (4), 373–397. Ivits, E., Cherlet, M., Mehl, W., Sommer, S., 2009. Estimating the ecological status and change of riparian zones in Andalusia assessed by multi-temporal AVHHR datasets. Ecological Indicators 9 (3), 422–431. Jensen, J.R., 2005. Introductory Digital Image Processing: A Remote Sensing Perspective, 3rd Ed. Prentice-Hall, New Jersey. Kanellopoulos, I., Varfis, A., Wilkinson, G.G., Megier, J., 1992. Land-cover discrimination in SPOT HRV imagery using an artificial neural network-a 20-class experiment. International Journal of Remote Sensing 13 (5), 917–924. Liao, K., 2004. Application of RS and GIS in ecological environmental dynamic monitoring and management information system. In: Proc. 12th Int. Conf. on Geoinformatics—Geospatial Information Research: Bridging the Pacific and Atlantic, University of Ga¨vle, Sweden. Lindquist, E.J., Hansen, M.C., Roy, D.P., Justice, C.O., 2008. The suitability of decadal image data sets for mapping tropical forest cover change in the Democratic Republic of Congo: implications for the global land survey. International Journal of Remote Sensing 29 (23–24), 7269–7275. Liu, J.Y., 1992. The Land Use in Xizang Autonomous Region. Science Press, Beijing (in Chinese). Liu, J.Y., Tian, H.Q., Liu, M.L., Zhuang, D.F., Melillo, J.M., Zhang, Z.X., 2005. China’s changing landscape during the 1990s: large-scale land transformations estimated with satellite data. Geophysical Research Letters 32, L02405.

Liu, Zhengjun, Liu, Aixia, Wang, Changyao, Niu, Zheng, 2004. Evolving neural network using real coded genetic algorithm (GA) for multispectral image classification. Future Generation Computer Systems 20 (7), 1119–1129. Liu, S.Z., Zhang, J.P., Li, Y., Liu, G.C., 2002. Impacts of human activities on the land desertification in the Dry-Hot Valley of the Jinshajiang River. Archives of Agronomy and Soil Science 48 (4), 349–356. Luo, B.Z., Shen, H.T., 1994. Three Gorges Dam Project and Estuarine Ecological Environment. Science Press, Beijing (in Chinese). Meng, J.J., Shen, W.M., Wu, X.Q., 2005. Integrated landscape ecology evaluation based on RS/GIS of Three-Gorges Area. Acta Scientiarum Naturalium Universitatis Pekinensis 41 (2), 295–301 (in Chinese). Meyer, W.B., Turner, B.L., 1994. Changes in Land Use and Land Cover: A Global Perspective. Cambridge University Press, Cambridge. Mostovoy, G.V., Anantharaj, V., King, R.L., Filippova, M.G., 2008. Interpretation of the relationship between skin temperature and vegetation fraction: effect of subpixel soil temperature variability. International Journal of Remote Sensing 29 (20), 2819–2831. Moussa, A., Soliman, M., Aziz, M., 2001. Environmental evaluation for High Aswan Dam since its construction until present. In: Sixth International Water Technology Conference, IWTC, Alexandria, Egypt. Murphy, N.L., 1976. Itaipu hydroelectric project. In: Earthmoving Industry Conference, Peoria, IL, USA. National Development and Reform Commission (NDRC), 2005. The greenshelt construction project around the Three Gorges Reservoir (in Chinese). http:// www.sdpc.gov.cn/zjgx/t20050715_36863.htm. North, P.R.J., 2002. Estimation of fAPAR, LAI, and vegetation fractional cover from ATSR-2 imagery. Remote Sensing of Environment 80 (1), 114–121. Opoku, P., 2007. Understanding land-use/cover change process for land and environmental resources use management policy in Ghana. GeoJournal 68 (4), 369– 383. Peng, J., Wang, Y.L., Wu, J.S., Yue, J., Zhang, Y., Li, W.F., 2006. Ecological effects associated with land-use change in China’s southwest agricultural landscape. International Journal of Sustainable Development and World Ecology 13 (4), 315–325. Peng, T., Xu, G., Xia, D.Q., 2004. Trend of geological hazards and countermeasure of disaster. Journal of Mountain Science 22 (6), 719–724 (in Chinese). Pouliot, D., Latifovic, R., Olthof, I., 2009. Trends in vegetation NDVI from 1 km AVHRR data over Canada for the period 1985–2006. International Journal of Remote Sensing 30 (1), 149–168. Purevdor, T.s., Tateishi, R., Ishiyama, T., 1998. Relationships between percent vegetation cover and vegetation indices. International Journal of Remote Sensing 18, 3519. Semwal, R.L., Nautiyal, S., Sen, K.K., Rana, U., Maikhuri, R.K., Rao, K.S., Saxena, K.G., 2004. Patterns and ecological implications of agricultural land-use changes: a case study from central Himalaya, India. Agriculture, Ecosystems and Environment 102, 81–92. Serra, P., Pons, X., Saurr, D., 2003. Post-classification change detection with data from different sensors: some accuracy considerations. International Journal of Remote Sensing 24 (16), 3311–3340. State Forestry Administration (SFA), 2006. The Shelterbelt Construction Project in the upper reaches of Yangtze River (in Chinese). http://www.forestry.gov.cn/ distribution/2006/12/27/zllh-2006-12-27-1463.html. Strand, H., Ho¨ft, R., Strittholt, J., Miles, L., Horning, N., Fosnight, E., 2007. Sourcebook on remote sensing and biodiversity indicators. NASA-NGO Biodiversity Working Group and UNEP-WCMC. Sun, Z.G., Wang, Q.X., Matsushita, B., Fukushima, T., Ouyang, Z., Watanabe, M., 2008. A new method to define the VI-Ts diagram using subpixel vegetation and soil information: a case study over a semiarid agricultural region in the North China Plain. Sensors 8, 6260–6279. Tullos, D., 2009. Assessing the influence of environmental impact assessments on science and policy: an analysis of the Three Gorges Project. Journal of Environmental Management 90 (3), 208–223. Veldkamp, A., Lambin, E.F., 2001. Predicting land-use change. Agriculture, Ecosystems and Environment 85, 1–6. Viovy, N., Arino, O., Belward, A.S., 1992. The best index slope extraction (BISE): a method for reducing noise in NVDI time series. International Journal of Remote Sensing 13, 1585–1590. Wang, J., Zhang, J.X., Liu, Z.J., 2005. Image sequence analysis on Modis/Terra NDVI data for vegetation cover changes of 3-gorge Reservoir Area. In: Proc. of IEEE 2005 International Geoscience and Remote Sensing Symposium, Seoul, Korea. Wang, X.L., Bao, Y.H., 1999. Study on the methods of land use dynamic change research. Progress in Geography 10 (5), 51–54 (in Chinese). White, G.F., 1988. Environmental effects of the High Dam at Aswan. Environment 30 (7), 4–40. Yu, X.X., Gao, H.Z., 2003. Study on land use/cover change in urban and its edges—a case study of Linyi City. Areal Research and Development 22 (2), 47–51 (in Chinese). Zhang, X., Yan, G., Li, Q., Li, Z.-L., Wan, H., Guo, Z., 2006. Evaluating the fraction of vegetation cover based on NDVI spatial scale correction model. International Journal of Remote Sensing 27 (24), 5359–5372. Zhou, Lequn, Sun, Chang’an Hu Jiajun, Gao, Gaiping, Hu, Zongyu, Yang, Lan, 2004. Dynamic monitoring and GIS system development for soil and water conservation in Three Gorges Project Reservoir Area of Yangtze River. Bulletin of Soil and Water Conservation 24 (5), 49–53 (in Chinese).