Spatio-temporal dynamic analysis of an island city landscape: a case study of Xiamen Island, China

Lizhong Huaa,b, Xuanqi Lia, Lina Tanga*, Kai Yina and Yu Zhaoa

aKey Laboratory of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, 1799 Jimei Road,Xiamen 361021, China; bInstitute of Spatial Information Technology, Xiamen University of Technology, Xiamen, China

The spatio-temporal dynamic landscape pattern of Xiamen throughout its rapid urbanisation from 1987 to 2007 was studiedusing analyses of landscape indices and fractal methods (box counting method, BCM). Based on remote sensing (RS) and ageographic information system (GIS), five Landsat TM5 images were used to construct a land-use database for the studied area.BCM was programmed in Microsoft Visual Basic 6.0 and MapObject. Over the last 20 years, rapid urbanisation of XiamenIsland has induced enormous conversions of the landscape pattern. Anthropogenic disturbance after 1992 has led to a man-made landscape gradually replacing the farmland and forestland landscape and the former became the dominant landscape typein 2007, leading to a decrease in diversity and heterogeneity in Xiamen Island. The urban land area increased rapidly from 39.2to 94.4 km2 over these 20 years. The urban expansion mainly occupied forestland, farmland, water and sea. Farmland area haddisappeared completely by 2007. Besides urban expansion, the Grain to Green Programme was also an important reason forfarmland decrease. The size and spatial distribution of coastal reclamation was closely related to policies implemented invarious phases. The island’s coastline has become simplified and smoothed, with fractal dimensions decreasing from 1.157 to1.124, and the shape approaching a circle. The new policy of transforming Xiamen from an island city to a bay city has hadimportant effects on coastal reclamation, with reclamation of 7.89 km2 between 2002 and 2007, more than the total areareclaimed during the previous 15 years.

Keywords: landscape pattern; urban expansion; coastal reclamation; fractal characteristics

Introduction

Urban sprawl can cause larger changes in environmentalconditions and ecosystem functioning than any other landuse (Grimm et al. 2008). The need to understand urbanevolution and preserve resources has culminated in analysisof urban processes over the medium or long term (Wardet al. 2000). In order to trace the urban development trend inspace, landscape metrics can be applied to measure spatialcharacteristics of the landscape pattern through time(Dietzel et al. 2005). Much research has focused on regionaland urban landscape pattern dynamic changes on multi-temporal and spatial scales (Dietzel et al. 2005; Li et al.2005; Weng 2007). But the study of the landscape pattern ofan island city is rare. This is an emergent issue that needs tobe addressed to achieve coordinated development betweenurban construction and the ecological environment of anisland city. Xiamen, China, is one of five special economiczones and is representative of coastal port cities in China.

This research aimed at understanding landscape patternevolution of Xiamen Island and policy analysis for coastalreclamation in the urbanisation process. The results will alsoprovide basic data and decision-making reference for the lay-out of urban space and landmanagement for other island cities.

Materials and methods

Study area

The study area is Xiamen Island (Figure 1; area of 133 km2),which is the mainly urbanised area and commercial centre

of Xiamen City. The island has a typical subtropical mon-soon climate, with annual average temperatures of around21�C and annual rainfall of 1143.5 mm.

Data sources

Five Landsat TM images were used in this study. The multi-data sets, taken under very clear atmospheric conditions,were acquired for 15 January 1987, 15 January 1992, 12January 1999, 2 January 2002 and 8 January 2007. AllLandsat images were rectified to a common UniversalTransverse Mercator coordinate system based on topo-graphic maps. The resultant root mean square error(RMSE) was less than 0.5 pixels. The land-use datasetwas classified into six types: urban land, bare land, forest-land, farmland, water (reservoirs and ponds) and coastalwetlands. Visual interpretation from the TM images wascarried out to form the six land-use classes using vectortools in Erdas Image, with the help of ancillary data, includ-ing the topographic map and ground survey information.

Landscape pattern analysis

Landscape metrics are increasingly used to study the urbanenvironment (Gustafson 1998) and can help improve under-standing and representation of urban spatial structure andurban dynamics. Fragstats, a spatial pattern analysis pro-gram (McGarigal et al. 2002), calculates landscape indicesseparately for individual patches (patch-level indices),

International Journal of Sustainable Development & World EcologyVol. 17, No. 4, August 2010, 273–278

*Corresponding author. Email: lntang@iue.ac.cn

ISSN 1350-4509 print/ISSN 1745-2627 online# 2010 Taylor & Francis

DOI: 10.1080/13504509.2010.487410http://www.informaworld.com

land-use type (class-level indices) and for the entire land-scape (landscape-level indices). In this study, we selected asubset of landscape-level indices for all land-use types outof the entire set of landscape indices in Fragstats 3.3.Variable names and descriptions of all indices included inthis analysis are listed in Table 1. Further details, explana-tion and mathematical equations for all the metrics can befound in McGarigal et al. (2002).

Island morphology evolution

The shape of a naturally formed island is usually verycomplex, but human development on the island often

undermines the natural complexity and the coastline tendsto be simplified and straightened. To explore the morphol-ogy of such evolution, an island shape index (Equation 1)and fractal analysis can be used:

c ¼ 2ffiffiffiffiffiffi

pAp

P; (1)

where A and P are the area and perimeter of the island, c iscompactness, ranging from 0 to 1, where a larger valueindicates that the shape approximates a circle.

Ever since the ‘fractal’ concept was proposed byMandelbrot (1967), it has been applied extensively to

1987 1992 1997

2002 2007

N

LegendFarmland

AB

Forestland

Urban landBare landSea

0 1 2 4Km

Water

Figure 1. Landscape changes in Xiamen Island over 20 years. A and B represent the Zhongzai reservoir in 2002 and 2007.

Table 1. Landscape-level indices used in the analysis.

Index Range Description

NP [1, þ1] A measure to describe landscape heterogeneityLPI (0, 100] Percentage of the landscape that the largest patch comprisesPD [0, þ1] Amount of patches in the landscapeSHDI [0, þ1] Basic components of heterogeneity, e.g. number and proportion of classes and their spatial distribution (Li and

Reynolds 1995). An increase of SHDI value indicates the number of different patch types increasedSHEI [0, 1] Distribution of area among different patch types. The smaller the value, the more uneven the distribution

becomes and the landscape is dominated by one or a few patch typesCONTAG [0, 100] Level of land connectivity. A higher CONTAG value means the corresponding patches are equally adjacent to

all other patch types

Notes: NP, number of patches; PD, patch density; LPI, largest patch index; SHDI, Shannon’s diversity index; SHEI, Shannon’s evenness index; CONTAG,contagion index.

274 L. Hua et al.

describe many physical phenomena (e.g. Benguigui et al.2000). To determine the fractal dimension of an islandcoastline, the box counting method (BCM) (Benguiguiet al. 2000) was used. BCM obtains a box dimension bypartitioning an object using a square grid of size r and thenumber N(r) of grids that cover the object is counted forvarious r values (Tay and Teng 2008). The estimation of Dis based on the following equation:

LnNðrÞ ¼ A� D lnðrÞ; (2)

where N(r) is the total number of the grid squares, A is aconstant representing the intercept of the linear regressionline, and D, the slope of the linear regression, is the fractaldimension, which is a measure of the complexity of anobject in nature. For coastlines, D ranges from 1 to 2,where the larger the value, the more complex the coastline.

In this study, fractal software was developed usingMapObject and Microsoft Visual Basic based on BCM,which integrated a fractal dimension calculator with linearregression and greatly improved ability to carry out theanalysis.

Results and discussion

General trend of landscape change

Xiamen Island experienced a remarkable landscape trans-formation, characterised by the landscape main componentschanging from natural forestland and farmland to an artifi-cial landscape during the fast urbanisation process between1987 and 2007 (Figures 1 and 2). Urban land area increasedfrom 39.2 to 94.4 km2, which represents a 1.4-fold increaseduring the past 20 years. The discrete urban lands extendedgradually and then joined together. Progress in urban landextension was mainly caused by occupation of farmland,

forestland and water bodies, as well as extensive coastalreclamation. The changing curve of urban land expansionand coastal reclamation had two high peaks and two lowpeaks (Figure 3). The periods 1987–1992 and 1997–2002were slow extension periods, with an extension speed of1.17 and 1.75 km2/yr, respectively. The periods 1992–1997and 2002–2007 were fast extension periods, with speeds of4.76 and 3.35 km2/yr, respectively.

Farmland decreased the most of all the landscape types.The farmland area was 33.28 km2 in 1987 and then began todecrease rapidly, with the highest decrease from 1992 to 1997.Farmland had decreased by 16.02 km2 by 1992 and a further12.52 km2 of farmland was changed into urban land by 2002.From 2002 to 2007, almost all farmland was changed to otherlandscape types, including 5.49 km2 of urban land, 3.52 km2

of forestland and land zoned for future construction. Anotherreason for the farmland decrease was inter-transformationbetween farmland and forestland through the Grain forGreen Programme, which surpassed the area of forestlandchanged to farmland. Water body area decreased relativelyslowly from 1987 to 2002, with a loss of 0.94 km2, butdecreased rapidly by 1.23 km2 between 2002 and 2007.This was because the Xiamen Government proposed tobuild on Wuyuan Bay, which resulted in the majority of theZhongzhai Reservoir being reclaimed as urban land (Figure1). Forestland remained basically stable during 1987–2007,andwill remain stable for the long term because it is mostly inhilly areas that are not suitable for urban land use.

Characteristics of landscape pattern changes

Landscape pattern dynamics of Xiamen Island in the past 20years are shown in Table 2. (1) Landscape indices NP andPD had a distinct downward trend; LPI and AREA_MNincrease continuously, for example LPI increased from

Figure 2. Area of landscape types in Xiamen Island over 20 years.

International Journal of Sustainable Development & World Ecology 275

20.57 to 65.71, making the advantage of large patches moreobvious. Mean patch area increased and spatial distributionwas more concentrated, leading to gradual decreases inlandscape fragmentation. (2) Landscape diversity index(SHDI) decreased gradually from 1.48 in 1987 to 0.88 in2007. Large tracts of farmland were transformed into urbanland, and this rapid land-use change led to a continuousdecrease in landscape heterogeneity. (3) The landscapeevenness index decreased year on year. The larger SHEIof 0.82 means that there was no evident dominant landscapetype before 1992, when the distribution of farmland, forest-land and urban land was comparatively even. The SHEIdecreased sharply after 1992, which means that the formerbalanced distribution pattern was completely broken. Urbanland extended rapidly and became the dominant landscapetype after 1992, and overwhelmingly dominant by 2007.(4) The increasing CONTAG indicated that urban landareas, as the dominant landscape type, were connected toeach other. The concentration and connections of urban landare also shown in Figure 1.

Coastal reclamation processes and analysis of policyimplications

An overlay of the five time-series images of Xiamen Islandwas made to display coastal reclamation and coastline

changes in the region (Figure 4). The results show that inthe past 20 years, the total area of coastal reclamation inXiamen Island was 14.97 km2, with an average outwardexpansion speed of 0.75 km2/yr. The curve of sea reclama-tion is similar to that of urban expansion in four phases, andboth show the ‘two low, two high’ peaks form (Figure 3).Between 2002 and 2007, Xiamen Island significantlyincreased the reclamation process, and this area was morethan the total area of the previous 15 years.

The spatial pattern of Xiamen Island coastline wasmeasured for the past 20 years (Table 3) and showed thatcoastal reclamation had led to a decrease in coastline com-plexity, with the island tending to become more circular.The island coastline length had a high negative correlationwith compactness, y ¼ �0.0134x þ 1.483 (r2 ¼ 0.986, xis the length of the coastline, x ¼ 36.1 km). In addition toslight growth in 1992, the island coastline was shortened,and decreased 14.12% from 65.81 to 56.62 km. The fractaldimension, which reflects the complexity of the coastline,also decreased, especially in 2002–2007, and large-scalecoastal reclamation caused a sharp fractal dimensiondecrease to a minimum of 1.124.

The rapid increase in socio-economic demand for landled to coastal reclamation, but the spatial and temporaldistribution of this reclamation was closely related to policy(Figure 4) as summarised below:

5.0

4.0

3.0

2.0

1.0

0.0 0.0

0.3

0.6

0.9

1.2

1.5

1.8

1987–1992 1997–2002

Year

Construction land

Urb

an la

nd e

xpan

sion

(km

2 /yr

)

Coa

stal

rec

lam

atio

n (k

m2 /

yr)

Sea

2002–20071992–1997

Figure 3. Comparison of urban land expansion and coastal reclamation in Xiamen Island over 20 years.

Table 2. Landscape-level indices of Xiamen Island over 20 years.

Year NP PD AREA_MN LPI CONTAG SHDI SHEI

1987 194 1.36 73.62 20.57 51.59 1.48 0.821992 188 1.32 75.97 24.50 51.81 1.47 0.821997 180 1.26 79.35 47.13 55.88 1.35 0.762002 160 1.12 89.27 53.33 60.03 1.23 0.692007 148 1.04 96.51 65.71 67.56 0.88 0.55

Notes: NP, number of patches; PD, patch density; LPI, largest patch index; SHDI, Shannon’s diversity index; SHEI, Shannon’s evenness index; CONTAG,contagion index.

276 L. Hua et al.

(1) Between 1987 and 1992, only 0.86 km2 of sea wasreclaimed. This was because, in 1980, the StateCouncil approved a 2.5-km2 area in Huli District,Xiamen, as the Xiamen Special Economic Zone. In1984, the State Council approved extension of thiszone to the entire island, and a free port policy wasintroduced.

(2) Between 1992 and 1997, coastal reclamation areaincreased sharply to 3.58 km2 through the majorprojects of Gaoqi International Airport, Haitian andXiangyu Ports, Xiangyu BondedZone andYuandangLake. The State Council approved establishment ofXiangyu Bonded Zone in 1992, and the first phaseproject was implemented between 1993 and 2000. In1995, Xiamen City established a strategy of ‘prosper-ity with the port’ and, consequently, construction onthe entire west side of the island started, including thetwo main ports, Haitian and Xiangyu.

(3) Between 1997 and 2002, there was a slight decreasein coastal reclamation to 2.65 km2. The main pro-jects included Xiangyu Port and the International

Convention and Exhibition Centre. Based on thestrategy of ‘the city lives on the port’, XiamenGovernment changed strategy from an island cityto a bay city in 2001.

(4) Between 2002 and 2007, the coastal reclamationarea reached a peak of 7.89 km2, which is morethan the total area in the previous 15 years. Themain projects were Wuyuan Bay and Hecuo andLingdou residential area construction projects. Thisperiod was the start-up phase when Xiamen Citychanged from an island to a bay city, and there was asubstantial increase in coastal reclamation, espe-cially the Wuyuan Bay construction project, as thefuture Xiamen New Reception area and ‘NewHercynian Pearl’.

Conclusions

Due to intense disturbances and severe destruction throughhuman activities in the process of urbanisation from 1987 to

Figure 4. Process of coastal reclamation in Xiamen Island over 20 years. A, Airport; B, Wuyuan Bay; C, Exbition; D, Yundang Lake;E, Haitian Port; F, Xiangyu Port; G, Xianyu Bonded Zone; H, Xiamen Island terrene-area in 1987.

Table 3. Spatial pattern measures for Xiamen Island coastline over 20 years.

Year Terrene area (km2) Coastline length (km2) Compactness Fractal dimension

1987 121.54 65.81 0.594 1.1571992 122.40 68.29 0.574 1.1671997 125.98 67.52 0.589 1.1772002 128.62 62.02 0.648 1.1602007 136.51 56.52 0.733 1.124

International Journal of Sustainable Development & World Ecology 277

2007, Xiamen Island experienced an enormous landscapetransformation, resulting in a landscape pattern change fromforest/farmland in 1987 to an artificial landscape in 2007.The changes to urban land, an artificial landscape generatedby human directional disturbance, were not only the largestin scale but also the most rapid, with urban land areaincreasing from 39.2 to 94.4 km2; while farmland was33.28 km2 in 1987 and had almost disappeared in 2007.Landscape patch density and patch quantity decreased, thelargest patch index and mean patch area increased, the land-scape fragmentation index decreased, diversity and even-ness indices decreased gradually and landscapeheterogeneity also decreased. The formerly even landscapepattern was broken by continuous extension of urban landsafter 1992, which became the dominant landscape type by2007. The rapid changing of landscape pattern has created asevere challenge to ecosystem stability and urban sustain-ability. The administration must pay attention to the abovesituation and propose a scientific urban development plan.

Coastal reclamation activities were related to the landneeds of an economy in rapid development. The area as wellas spatial and temporal distribution of coastal reclamationwas related to policy, especially the strategic plan of chan-ging from an island city to bay city, which promoted activecoastal reclamation of 14.97 km2. The coastline of XiamenIsland has been smoothed, and the fractal dimension hasdecreased from 1.157 in 1987 to 1.124 in 2007.

Just like other coastal cities with high population den-sity, Xiamen lacks space; thus, coastal reclamation is themain solution. While providing space for urban develop-ment, sea reclamation will permanently change or destroymarine and coastal ecosystems on which Xiamen depends(Peng et al. 2005). We suggest that environmental capacityevaluation and ecological risk assessment should beperformed before coastal reclamation projects are executed.In protected areas for rare or endangered ocean speciesand areas of ecological importance, coastal reclamationactivities must be forbidden. The construction period ofreclamation should be arranged to avoid affecting ordestroying rare coastal species, such as the Indo-Pacifichump-backed dolphin (Sousa chinensis).

AcknowledgementsThis research was supported by the Chinese Academy of Sciences(CAS) (KZCX2-YW-422; KZCX2-YW-450), National NaturalScience Foundation of China (40701059), One Hundred TalentsProgram and Academy-Locality Cooperation Program of CAS,Key Laboratory of Spatial Data Mining & Information Sharingof Ministry of Education (201009), Xiamen University ofTechnology (YKJ09011R) and the National Key TechnologiesR&D Program of China (2007BAH16B01-R1-1).

ReferencesBenguigui L, Czamanski D, MarimovM, Portugali Y. 2000. When

and where is a city fractal? Environ Plann B. 27:507–519.Dietzel C, Herold M, Hemphill JJ, Clarke KC. 2005. Spatio-

temporal dynamics in California’s Central Valley: empiricallinks to urban theory. Int J Geogr Inf Sci. 19:175–195.

Grimm NB, Faeth S, Golubiewski NE, Redman CL. 2008.Global change and the ecology of cities. Science.319:756–760.

Gustafson EJ. 1998. Quantifying landscape spatial pattern: what isthe state of the art? Ecosystems. 1:143–156.

Li H, Reynolds JF. 1995. On definition and quantification ofheterogeneity. Oikos. 73:280–284.

Li WF, Wang Y, Li G, Peng J. 2005. Landscape spatial changesassociated with rapid urbanization in Shenzhen, China. Int JSustain Dev World Ecol. 12:314–325.

Mandelbrot BB. 1967. How long is the coast of Britain? Statisticalself-similarity and fractional dimensions. Science.156:636–638.

McGarigal K, Cushman SA, Neel MC, Ene E. 2002.FRAGSTATS: spatial pattern analysis program for categoricalmaps [internet]. [cited 2009 October 12]. Available from:www.umass.edu/landeco/research/fragstats/fragstats.html.

Peng BR, Hong HS, Hong JM, ChenWQ, Xue XZ, Peng JP. 2005.Ecological damage appraisal of sea reclamation and its appli-cation to the establishment of usage charge standard for filledseas: case study of Xiamen, China. Environ Inform Arch.3:153–165.

Tay LT, Teng HT. 2008. Roughness index and fractal dimensionfor surface information extraction. Asian J Geoinform.8:21–24.

Ward D, Phinn SR, Murray AT. 2000. Monitoring growth inrapidly urbanizing areas using remotely sensed data. ProfGeogr. 52:371–386.

Weng YC. 2007. Spatiotemporal changes of landscape pattern inresponse to urbanization. Landsc Urban Plan. 1:341–353.

278 L. Hua et al.