Environ Monit Assess (2008) 138:139–147 DOI 10.1007/s10661-007-9751-x
A landscape approach to quantifying land cover changes in Yulin, Northwest China Yong Zha & Yansui Liu & Xiangzheng Deng
Received: 18 November 2006 / Accepted: 11 April 2007 / Published online: 11 May 2007 # Springer Science + Business Media B.V. 2007
Abstract In this study we quantified land cover changes in the arid region of Yulin City, Northwest China between 1985 and 2000 using remote sensing and GIS in conjunction with landscape modeling. Land covers were mapped into 20 categories from multitemporal Landsat TM images. Five landscape indices were calculated from these maps at the land cover patches level. It was found that fallow land decreased by 125,148 ha while grassland and woodland increased by 107,975 and 17,157 ha, respectively. Landscape heterogeneity, dominance and fractal dimension changed little during the 15-year period while landscape became more fragmented, with an index rising from 0.56 to 0.58. The major factors responsible for these changes are identified as the change in the government policy on preserving the environment, continued growth in mining, and urbanization. Y. Zha (*) College of Geographic Science, Nanjing Normal University, Nanjing 210097, China e-mail:
[email protected] Y. Liu : X. Deng Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China Y. Liu e-mail:
[email protected] X. Deng e-mail:
[email protected]
Keywords Land cover . Landscape . TM . Northwest China
Introduction Arid and semiarid China is extensively distributed in the north and northwest of the country, encompassing an area of about 690,000 km2. Landscape in this region is extremely vulnerable to change. Due to natural fluctuations and increased human activities, it tends to change more frequently than elsewhere. Within this region lies Yulin City in Northwest China. Its typical arid landscape has been shaped through mutual interactions among natural, biological, and social factors over a long period. Centuries ago this landscape was characterised by sparsely populated forests juxtaposed with fertile grassland. Massive land reclamation for farming during the 13th–19th centuries had gradually transformed its landscape into farmland intertwined with grazing land. In modern history population explosion coupled with abusive land use practices accelerated the depletion of natural resources and destruction of the ecosystem. Accompanied with these changes were land degradation and desertification. Since landscape changes result essentially from the interactions between human activities and the land in a specific spatial and temporal context, the study of the driving forces behind these changes is conducive
140
Environ Monit Assess (2008) 138:139–147
to revelation of the relationship between socioeconomic factors and the natural environment (Zhou 2000; Burgi and Russell 2001). Thanks to the emergence of modern technologies such as remote sensing, geographic information system (GIS), and mathematical modelling, the spatial pattern of landscape and its change can be characterised and quantified with an increasing ease (Wang et al. 2001; Yang and Lo 2003). Landscape changes have been quantified widely with the assistance of remote sensing. Cushman and Wallin (2000) quantified the rates and patterns of landscape change between 1972 and 1992 for a forested area in the central Sikhote-alin Mountains of the Russian Far East using Landsat imagery and GIS. Lu et al. (2003) investigated landscape evolution in the middle Heihe River basin during the last decade through analysis of the changes of various landscape metrics. Wu and Ci (2002) assessed landscape change in the Mu Us Sandland in northern China from the 1950s to the 1990s through analysis of satellite images, historical maps, meteorological and socioeconomic data. These studies are concerned with detection of the changes from multitemporal satellite data. Guan et al. (2003) explored the process of change, and Lo and Yang (2002)
Study area Yulin City is located in northern Shaanxi Province, China, with a spatial extent of 36°57′–39°34′ N and 107°28′–115°15′ E. Its terrain descends from east to west, with an elevation ranging from 1,907 to 585 m above sea level. It has an elongated shape, measuring 385 km in east–west by 263 km in south–north. Encompassed within its administrative boundary are 12 counties at a combined territory of 43,417 km2 (Fig. 1).
N
LEGEND
Yulin City
Prefectural boundary River
Fugu
Study area Ku ye
li u
Mizhi
er
H ong
Jiaxian
X
I
W u d in g Hengshan
N
Gr
ea
t
Yulin
A
GO ON
IN N E R M
l
R iv
NINGXIA
al
Yellow
Mu Us Desert
W
r
N
R iv e
800 km
SH
LI
AN
Shenmu r Yu xi R i v e
0
Zizhou
Dingbian Jingbian
Suide
B
y ai
Yan'an City
u
M
Rive r
SU
Wubu
GAN
Fig. 1 Location of the study area and its vicinity
analyzed the driving forces behind land use/cover changes and even attempted to model the dynamics for metropolitan Atlanta, Georgia. This study focuses on a representative area in arid China to quantify the process of landscape evolution, and the pattern of land cover changes using multitemporal remotely sensed data. This quantification was made possible through categorization of the study area into different land cover types, and analysis of their spatial pattern based on landscape indices. Such knowledge is essential to establish a harmonic relationship between human beings and the land.
T.
0
40
Qingjian 80 km
SHANXI
Environ Monit Assess (2008) 138:139–147
The city is bisected by a section of the Great Wall, totalling 700 km in length in the orientation of east– west. To the north of the Wall lies the gently rolling Mu Us Sandy Land that accounts for 42% of the study area. To the south of the Wall is the hilly Loess Plateau. Ecologically, this area typifies a transitional ecotone in both climate and landform. Its climate exhibits a distinct pattern of horizontal westward differentiation, changing from semiarid to arid. Associated with this change in climate is the gradual transition in vegetative cover from forested steppe to desert steppe and desert. This landscape results from both natural and human variables. Compared with their natural counterparts, human factors play a much more important role in the formation and evolution of landscape in Yulin (Liu et al. 2003). Human activities range from crop cultivation to construction of irrigation projects at various scales. Damming of rivers and construction of irrigation canals all exerted an impact on the composition of the landscape. Since the late 1980s the landscape has been diversified to include coalmines and natural gas fields. Rapid growth in population together with mining has exacerbated the negative impact of human activities on the fragile environment. Consequently, the landscape is increasingly fragmented while new ecological problems such as soil erosion and land desertification have emerged (Liu and Gao 2002). These changes have attracted much attention in the literature (Gao et al. 2001; Zhang et al. 2003).
Materials and method Data used Four kinds of data were used in this study, remotely sensed data, topographic data, socioeconomic data, and ancillary data. The remote sensing data are Landsat TM images recorded in 1985, 1995 and 2000. Two topographic maps had a scale of 1:100,000 and 1:50,000, respectively. Socioeconomic data were the annual statistical data released by the State Statistical Bureau. The data published in 1986, 1996 and 2001 were acquired. All of them contained a total of 45 socioeconomic indicators for the years of 1985, 1995 and 2000. Two types of ancillary data were used, a vegetation map at a scale of 1:500,000 (Lei
141
1999), and a current land use map supplied by Yulin Land and Resources Bureau (2000). Data processing The administrative boundary of the study area was digitized from the 1:100,000 topographic map in Arc/ Info. The acquired vector coverage was later rasterized at a cell size of 1 km2. After rasterisation all border grid cells were further examined to ensure that they fell entirely inside the study area. Otherwise, these border cells were eliminated from the raster layer. After elimination a total of 44,099 grid cells were retained in the raster layer. All the multitemporal satellite images were geometrically rectified and projected to the Albert coordinate system using ground control points selected from the 1:100,000 topographic map. These rectified images were clipped against the vector coverage of the administrative boundary that had been constructed already. Each of them was assigned a unique row and column number. All image processing was undertaken in the Image Analysis module in MicroStation. After consulting the Landsat TM images, the 1:50,000 topographic map, and the ancillary thematic maps of vegetation and land use, in conjunction with field reconnaissance, we decided that the landscape should be mapped into six categories of land cover at the first level and 20 subcategories at the second level (Table 1). Their properties on the false colour composite of bands 7 (red), 4 (green), and 3 (blue) were identified indoors and later verified during field visits at representative spots. Each patch of land cover on the composite image was visually delineated through on-screen digitisation. All digitised patches were saved in the dxf format and subsequently exported to Arc for editing. During editing, each interpreted land cover patch was assigned a code. After a coverage was built from an individual image, it was mosaicked with those from other images to form a large coverage for the entire study area. Calculation of landscape indices Five landscape indices were calculated for the quantification. They were heterogeneity (H), dominance (D), fragmentation (Fr), degree of separation (Bi), and fractal dimension (Fi). Landscape heteroge-
142
Environ Monit Assess (2008) 138:139–147
Table 1 Variables used to classify land cover types and their image evidences Level I
Code
Level II
Evidence of classification
Farmland
A1 A2 B1 B2 B3 C1 C2 C3 D1 D2 D3 D4 E1 E2 E3 E4 F1 F2 F3 F4
Paddy Nonirrigated Forest Shrub Sparse forest High density Medium density Low density Rivers and canals Lakes Reservoirs Beachland Towns Rural settlement Industrial and mining Transportation Sandy land Salinilised land Wetland Barren and gravel
Presence of irrigation facilities nearby Absence of irrigation facilities nearby Cover density>40% Cover density>30% Cover density 10–30% Cover density>50% Cover density 20–50% Cover density1.0) separation index (Table 3). As a kind of artificial landscape components and the products of human activities, most of them have a spatially uniform distribution with a simplistic shape. By comparison, nonirrigated farmland, grassland of medium and low densities, and sandy land all have a very low degree of separation (