Ecological Indicators 46 (2014) 240–252

Analysis of landscape fragmentation processes and driving forces inwetlands in arid areas: A case study of the middle reaches of the HeiheRiver, China

Penghui Jiang a,c, Liang Cheng a,b,c,*, Manchun Li a,b,c,**, Ruifeng Zhao d,Qiuhao Huang a,b,c

a Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing University, Nanjing, 210093, ChinabCollaborative Innovation Center for the South Sea Studies, Nanjing University, Nanjing 210093, ChinacDepartment of Geographic Information Science, Nanjing University, Nanjing 210093, ChinadCollege of Geography and Environment, Northwest Normal University, Lanzhou, Gansu Province 730070, China

A R T I C L E I N F O

Article history:Received 3 September 2013Received in revised form 30 April 2014Accepted 18 June 2014

Keywords:WetlandsFragmentationDriving forcesHeihe RiverLandscape indices

A B S T R A C T

Landscape fragmentation in wetlands usually implies degradation of its ecological functions. Landscapefragmentation divides wetlands into isolated islands, which disturbs the energy flow and nutrient cyclingwithin the wetland. Research into the development and causes of landscape fragmentation in wetlands isurgently needed for effective monitoring and protection of wetlands. We use a combination of techniquesincluding remote sensing, a landscape index model, and redundancy analysis to analyze landscapefragmentation and its driving forces in the middle reaches of the Heihe River from both temporal andspatial perspectives. A new mathematic morphological method that enhances the credibility oflandscape fragmentation analysis without changing the original pixel size of the interpreted data isproposed for the calculation of landscape indices. The combination of this new mathematicmorphological method and traditional landscape pattern indices enhances the evaluation strategy forlandscape fragmentation. Our results demonstrate that the fragmentation processes that affect thewetland landscape of the study area are primarily represented as the shrinking of core wetland area anddecrease in mean size of wetland patches. Our results also show an increase in the fragmentation index(FS) of the landscape in recent decades. The impacts of natural factors on wetland landscapefragmentation processes are typically reflected in changes in climate and hydrology. In the study area,temperature, which is more important than precipitation in driving wetland landscape fragmentationprocesses, cannot be omitted. In addition, our analysis proves that unnecessary human activity is a majorthreat for sustainable development and maintenance of wetlands.

ã 2014 Elsevier Ltd. All rights reserved.

Contents lists available at ScienceDirect

Ecological Indicators

journa l home page : www.e l sev ier .com/ loca te /eco l ind

1. Introduction

Wetlands are unique ecosystems formed as a result ofinteractions between the forces affecting land and water (Zhanget al., 2010). As key parts of the global ecological system and carbonpool, wetlands offer important ecological functions and effects thatcannot be replaced: they mitigate pollution, provide habitats forwildlife, regulate climate, and preserve biodiversity, among other

* Corresponding autho at: Department of Geographic Information Science,Nanjing University. Tel.: +86 25 83597359; fax: +86 25 83597359.** Corresponding author at: Department of Geographic Information Science,Nanjing University. Tel.: +86 25 83593660; fax: +86 25 83593660.

E-mail addresses: lcheng@nju.edu.cn (L. Cheng), limanchun_nju@126.com(M. Li).

http://dx.doi.org/10.1016/j.ecolind.2014.06.0261470-160X/ã 2014 Elsevier Ltd. All rights reserved.

things (Mander and Mitsch, 2009, Copeland et al., 2010; Sulmanet al., 2013). However, the Organization of Economic Cooperationand Development has estimated that approximately 50% of globalwetlands could have been lost since 1900, with the remainderexperiencing increased fragmentation in recent years (Lienertet al., 2002; Erwin, 2008; Khaznadar et al., 2009; Zhang et al., 2010;Wang et al., 2011). Similarly, incomplete statistics suggest that twothirds of wetlands in France were lost during 1900–1993 (West-erberg et al., 2010). In general, climatic change accompanied byincreased disturbance caused by human activity has placedwetlands at greater risk, particularly in arid zones (Zhao et al.,2009). Wetlands in arid zones are key nodes and represent criticalareas in the landscape patterning that is crucial to the functioningof arid environments; they also play a key role in providing water,energy, and other resources to sustain human life (Zhou, 2005;

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 241

Mwakaje, 2009). However, owing to a lack of consideration of theirrole in sustainable resource use, wetlands in arid areas continue tofragment, resulting in decreased wetland biodiversity (Soomerset al., 2013; Walz and Syrbe, 2013).

Wetland landscape fragmentation plays a major role in thedegradation of ecological systems (Liu et al., 2014) and thereduction of wetland biodiversity; in particular, this fragmentationresults in complex spatial distributions of vegetation and reducedfunctioning of wetlands (Alexandridis et al., 2009; Liu et al., 2009;Song et al., 2012). As such, fragmentation can be considered as oneof the most significant expressions of wetland degradation and hasbeen recognized as a core component of landscape ecology andlandscape conservation because it affects cycles of material andenergy directly (Wiens, 1994; Cerian and Keys, 2009). As a keyfactor exerting pressure on biodiversity, wetland fragmentationalso has a negative effect on the functioning of wetland ecosystems(Opdam and Wascher, 2004). In most cases, fragmentationrepresents the first step in the process of wetland degradation.Accordingly, research into the processes underlying wetlandlandscape fragmentation should improve our understanding oftypical wetland evolution and help advance techniques to protectwetlands. The underlying processes and driving forces of wetlandlandscapes are core research topics in current wetland science (Liuet al., 2009).

The wetlands distributed along the middle reaches of the HeiheRiver have great significance for the eco-environmental protectionof northwestern China. In particular, as the primary westwardroute for migratory birds of China, the wetlands in the middlereaches of the Heihe River also act as a staging point along the EastAsia–India section of the eight global migratory channels.Additionally, these wetlands act as an environmental barrier thatprevents the southward invasion of the Badain Jaran desert andplay a key role in maintaining the ecosystem balance of the HeiheRiver basin. However, the wetlands have been partially destroyedby human activity and climate change in recent years, withfragmentation and ecosystem degeneration of these wetlands

Fig. 1. Location of

becoming increasingly pronounced over time. Such degenerationhas induced the southward expansion of the Badain Jaran desert,exposing large expanses of grassland and oasis to the threat ofdesertification, and endangering the survival and development ofthe Heihe River basin. At worst, such changes could have aprofound influence on the ecological security of the Hexi Corridorand northwestern China as a whole. Consequently, research intothe wetlands of the Heihe River basin is urgently required. Previousstudies of the Heihe River focused on various aspects of theenvironment, including landscape evolution, hydrological changeand its ecological effects, water consumption, evapotranspiration,vegetation cover, and carbon sequestration (Lu et al., 2003; Kanget al., 2007; Kong et al., 2009; Wang et al., 2009; Zhao et al., 2010; Liet al., 2012; Wang et al., 2012; Li et al., 2013). However, manyfeatures of these wetlands remain unknown; in particular,knowledge of landscape fragmentation in the region remainsinsufficient (Li and Zhao, 2010).

Here, we adopt remote sensing (RS), geographical informationsystem (GIS), redundancy analysis (RDA), and techniques ofmorphological image processing to explore the processes anddriving forces of wetland landscape fragmentation in the HeiheRiver basin. The primary aims of this study are: (1) to characterizethe processes of wetland fragmentation in the middle reaches ofthe Heihe River during 1975–2010 at the landscape and pixellevels; and (2) to analyze the driving forces that promote wetlandlandscape fragmentation, distinguishing between the effects ofhuman activity and natural environmental changes in the middlereaches of the Heihe River.

2. Study area

The middle reaches of the Heihe River (98�570–100�52 0E,38�390–39�59 0N) are located in the western part of Gansu Province,northwestern China, spanning three counties (Ganzhou, Gaotai,and Linze) and covering an area of 10,753 km2 (Fig. 1). The climateof the region is characterized by cold winters, hot summers, and

the study area.

Fig. 2. Distribution of wetlands in the middle reaches of the Heihe River in1975 and2010.

242 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

generally dry weather. The mean annual precipitation andevaporation are 50–250 mm and 2000–3500 mm, respectively.The terrain of the study area can be divided primarily intomountain and plain areas and includes the Hexi Corridor plain andthe Qilian and Heli mountains that are located to the south andnorth of the Heihe River, respectively. The study area lies ataltitudes of 1234–3633 m above sea level, with maximum altitudesin the Qilian Mountains. As a result of human activity and thepresence of the Heihe River, cropland and forest cover much of thepiedmont alluvial fan and floodplain, with farmland, grassland,and the Gobi desert constituting the primary land use types. Thevegetation in the middle reaches of the Heihe River consistsprimarily of temperate shrubs and desert vegetation.

In recent decades, the social and economic landscapes ofGanzhou, Gaotai, and Linze have changed considerably. Grainoutput and the income of farmers have continued to rise, resultingin corresponding increases in gross domestic product and the totalpopulation of the study area (to 14.7 � 109 yuan and 82.55 �104,respectively, by 2010).

In the study area, wetlands are distributed mainly along theHeihe River and occur as marshes, lakes, and bottomlands, amongothers. Wetland flora and fauna such as Elsholtzia ciliata, Scirpusvalidus, and the black stork occur extensively in these wetlands. Toprotect these wetland resources and ensure that this uniqueecosystem is protected from disturbance by desertification orclimate change, the Chinese State Council established the HeiheRiver Basin Nature Reserve in 2011. The reserve covers a total areaof 41,164.56 ha.

3. Materials and methods

3.1. Data sources and processing

We extracted all data relating to the wetlands in the middlereaches of the Heihe River for the period from 1975 to 2010 fromremote sensing images (Table 1). We combined the extracted datawith land use/land cover data for the study area (acquired at a scaleof 1:100,000 for 1985; and 1996) obtained from the ChineseAcademy of Science (http://westdc.westgis.ac.cn/) and a topo-graphic map at the 1:50,000 scale (obtained by a field study in1972). The selection of yearly images was conducted based onhistorical events (e.g., the initiation of reform and openness in the1980s, the influx of western development in the 1990s), policychanges (e.g., water diversion work in the middle reaches of theHeihe River that began in 2001), and image availability. Conversely,the selection of monthly images was conducted based on theoccurrence of major changes in vegetation, such as that in summerand at the beginning of autumn.

Image processing was performed using EDRAS Imagine (version9.2), and data preprocessing involved several steps, includingimage adjustment, image enhancement, and image cutting.Moreover, MSS (58 m), TM (30 m), and ETM+ (30 m) images wereresampled to a 60 m � 60 m pixel size using the nearest-neighborresampling method. Then, the wetland areas were categorized intorivers, lakes, marshes, bottomlands, reservoirs, and ponds based onthe China Wetland Classification System (Zhao et al., 2009) and the

Table 1Data sources of landscape information for the study area.

Year Data source

1975 MSS

1987 TM

1992 TM

2001 ETM+

2010 ETM+

landscape characteristics of the study area (Fig. 2). Five land use/land cover types were also identified according to the classificationcriteria of Liu et al. (2010).

The extraction of wetlands and other land use and land coverinformation for 1975, 1987, and 1992 required the use of thetopographic map at scale 1:50,000 (produced in 1972) and land use

Remote sensing image

143/33, 1975.10.07; 145/32, 1975.10.09; 144/33, 1976.07.04133/33,1987.08.15; 134/33, 1987.10.09; 134/32,1989.09.28133/33, 1992.09.05; 134/33, 1992.08.27; 134/32, 1991.09.02133/33, 1999.07.07; 134/33, 2001.07.03; 134/32, 2001.08.20134/33, 2010.08.05; 133/33, 2010.08.14; 134/32, 2010.08.21

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 243

and land cover data from 1985 to 1996 acquired from the ChineseAcademy of Science (http://westdc.westgis.ac.cn/); all theseproducts have high accuracy. Therefore, the present study focusesprimarily on assessing the accuracy of the classification of the 2001and 2010 data based on field surveys and high-resolution images.We calculated the kappa coefficient as follows:

K ¼NXm

i¼1

Pli �Xm

i¼1

ðPpi � Pl iÞ

N2 �Xm

i¼1

ðPpi � Pl iÞ(1)

where K is the kappa coefficient, N is the total number of samples,and Ppi and Pli are the total number of samples in the row and rank,respectively, of a given land use/land cover type. Our results showthat the accuracies of the classified images for 2001 and 2010 are84.06% and 81.18%, respectively (Table 2).

To increase our understanding of the dynamics of wetlandlandscape fragmentation, we considered land use and land coverchange, socioeconomic development, hydrology, and climatechange in our analysis of the driving forces of fragmentation.We adopted ten socioeconomic indices: gross domestic product,primary industrial output, secondary industrial output, revenue,expenditure, total population, grain yield, total investment, totalretail sales, and rural per capita net income. The data for theseindices were obtained from the 1984–2011 yearbooks from GansuProvince, and we selected data from a 60-year period (1949–2008)to illustrate the effects of socioeconomic development on wetlandfragmentation. Similarly, we used annual precipitation and annualtemperature data for 1967–2010 obtained from the ChinaMeteorological Data Sharing Service System (http://cdc.cma.gov.cn/) to investigate the effects of climate change on fragmentation.Finally, we used runoff data acquired from monitoring at theYingluoxia and Zhengyixia stations between 1956 and 2005 toinvestigate the effects of changes in hydrology on wetlandlandscape fragmentation.

3.2. Analysis

3.2.1. Fragmentation processes at landscape levelLandscape fragmentation studies typically adopt a selection of

landscape indices. In this study, we selected the mean patch size(MPS), patch density (PD), mean patch shape fragmentation index(FS1), and area-weighted mean shape fragmentation index (FS2) foruse in quantitative analysis of the processes involved in wetlandlandscape fragmentation. Further details of these indices can befound in previous studies (Lu et al., 2003; Zhao et al., 2009; Wang

Table 2Error matrix for the interpretation of 2010/2001 data.

GPSand googleearth samples

Classification

A1 A2 A3 A4 A5

A1 21/1 2/1 2/0 0 0/5

A2 1/0 61/8 2/0 6/0 0

A3 2/0 7/1 18/2 1/0 0

A4 1/0 1/0 1/0 39/5 0

A5 0 0 0 0 16/191A6 0 0 0 0 0

A7 0 2/0 0 3/0 0

A8 0 0 0 0 0

A9 1/0 0 0 0 0

A10 0 0 0 0 0

Total 26/1 73/10 23/2 49/5 16/196

Notes: A1: farmland; A2: grassland; A3: forest; A4: desert; A5: construction; A6: river; A7

sample number of 2010 and the right is the sample number of 2001.

et al., 2011). MPS and PD were calculated using FRAGSTATS version3.3 and ArcGIS version 9.3. FS1 and FS2were calculated according toEqs. (2)–(4).

FS1 ¼ 1 � 1MSI

FS2 ¼ 1 � 1ASI

(2)

MSI ¼XN

i¼1

SIðiÞ=N; SI ið Þ ¼ P ið ÞffiffiffiffiffiffiffiffiA ið Þ4

p (3)

ASI ¼XN

i¼1

AðiÞSIðiÞ=A; A ¼XN

i¼1

A ið Þ (4)

Here, MSI is the mean shape index, ASI is the area-weightedshape index, and SI (i), A (i), and P (i) are the shape index, area, andperimeter, respectively, of patch i. SI (i) is typically calculated for asquare shape index; accordingly, the square patch shape index is 1and other shape indices are greater than 1.

3.2.2. Fragmentation processes at pixel levelWe developed four area indices using morphological image

processing to explore the process of wetland fragmentation. First,two 3 � 3 SEs were defined according to the research of Vogt et al.(2006) and Riitters et al. (2009). Then, four wetland types (Core,Edge, Patch, and perforated wetlands) were obtained from thebinary images (Fig. 3). ‘Core wetland’ is the wetland that isseparated from non-wetlands by a buffer zone. Spatially, they aregenerally expressed as wetlands with a large area and highconcentration. ‘Patch wetland’ refers to wetlands that are muchsmaller than ‘core wetlands’ and they are surrounded by non-wetlands. ‘Edge wetlands’ and ‘perforated wetlands’ are the bufferzones between ‘core wetlands’ and non-wetlands. ‘Perforatedwetlands’ are located in the interior of core wetlands and ‘edgewetlands’ in the exterior. In most cases, increases in ‘patch wetland’and ‘perforated wetland’ indicate an increase in fragmentation. Thespecific steps undertaken for extraction of the different wetlandsfrom the binary image are as follows (Fig. 4):

Step 1: The extraction of core wetland: SE1 was used to performan operation of erosion on wetland pixels. If all the neighbors arewetlands, we define the center pixel as the core wetland. The otherpixels are then taken to be different from core wetlands and non-wetlands and they are the patch, edge, and perforated wetlands,which are distinguished by the following operations.

Step 2: The extraction of patch wetland: SE2 was used toperform an operation of dilation on core wetlands. Since patch

A6 A7 A8 A9 A10 Total

0 0 0 0 0 25/70 0 0 0 0 70/80 0 0 0 0 28/30 0 0 0 0 42/5

0 0 0 0 0 16/1913/0 0 0 0 0 3/00 13/2 0 0 0 18/20 0 3/0 1/0 0 4/00 0 0 2/0 0 3/00 0 0 0 8/1 8/1

3/0 13/2 3/0 3/0 8/1 217

: bottomland; A8: lake; A9: marsh; A10: reservoir and pond. The left number is the

Fig. 3. Four classes of wetland patterns defined in this study.

Fig. 4. Extraction process of the four classes

244 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

wetland pixels are indirectly connected with the core wetland, weapplied a repeated dilation on core wetlands with SE2. Only whenall indirectly connected wetland pixels are identified can thedilation be stopped.

Step 3: The extraction of edge wetland: In contrast to theextraction of core and patch wetlands, edge wetlands wereidentified by set non-wetland pixels to perform the dilationoperation.

Step 4: The extraction of perforated wetland: Since the core,patch, and edge wetlands were subtracted, the remaining unde-fined wetland pixels were classified as perforated wetlands.

3.2.3. Driving forces of wetland landscape fragmentationThe forces driving wetland landscape fragmentation can be

divided into two groups: factors related to human activity andnatural factors. We selected annual temperature and annualprecipitation as primary variables to investigate the effects ofnatural factors on fragmentation based on the characteristics of thestudy area. In addition, factors such as total population, rural percapita net income, gross domestic product, primary industrialoutput, secondary industrial output, total investment, total retailsales, and grain yield, were chosen to explore the effects ofsocioeconomic development on fragmentation. Then, we per-formed redundancy analysis (RDA) to calculate the cumulative

of wetland patterns defined in this study.

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 245

contribution ratio of socioeconomic development and climatechange. In this manner, the effects of socioeconomic developmentand natural variations on wetland landscape fragmentation couldbe assessed quantitatively.

Furthermore, we conducted other, more qualitative, analyses toderive the relationship between landscape fragmentation andchanges in hydrology and land use/cover, which are detailed in thediscussion section. In this manner, we identified and quantified theeffects of human activity and natural changes on wetlandlandscape fragmentation.

4. Results

4.1. Fragmentation processes in wetland landscapes

4.1.1. Fragmentation processes at landscape levelOur results demonstrate that the MPS of the wetlands in the

study area decreased each year during 1975–2010 (Fig. 5a and b),with particularly marked reductions during 1975–1987 and 2001–2010 (reductions of 22.56% and 19.35%, respectively). Conversely,PD, FS1, and FS2 in the study area increased considerably in recentyears, particularly during 1975–1987 and 1992–2010. PD increasedfrom 0.008 Ind/ha to 0.111 Ind/ha and from 0.012 Ind/ha to0.014 Ind/ha during 1975–1987 and 1992–2010, respectively.During these same intervals, FS1 increased from 0.565 to 0.575and from 0.270 to 0.275, and FS2 increased from 0.524 to 0.556 andfrom 0.269 to 0.276, respectively.

In general, the indices used here suggest increased fragmenta-tion of the wetland landscape, particularly in the 1990s and 2000s.The Ninth Five-Year Program was implemented during 1996–2000and the Western Development Program and a water diversionproject in the middle reaches of the Heihe River were initiated in

Fig. 5. Landscape indices of wetlands in t

2001. The resulting high-density human activity in the study arealikely resulted in accelerated fragmentation of the wetlandlandscape; this fragmentation would have been exacerbated byfactors such as the expansion of farmland, excessive use of waterresources, and wetland reclamation. Therefore, our results suggestthat, even with different evolutionary pathways or components,landscape fragmentation is the primary manifestation of land-scape pattern changes in the study area and will likely persist overrelatively long periods.

4.1.2. Fragmentation processes at pixel levelThe change in area of the core, patch, edge, and perforated

wetlands shows a significant process of fragmentation of thewetland landscape in the past decades. The area of core wetlandsdecreased by 42.54% from 1975 to 2010 (Fig. 5c, Fig. 5d).Correspondingly, the proportion of its pixels in the total wetlandpixels reduced from 49.17% to 36.83% during 1975–2010(Fig. 5d). Along with the reduction of core wetlands, edgewetlands also decreased. However, the area of patch andperforated wetlands increased by 586.8 ha and 34.20 ha, respec-tively (Fig. 5c). Thus, the core wetlands in the study area areobserved to have gradually fragmented into patch and perforatedwetlands.

Furthermore, we can directly observe that the spatial patternof wetlands in the Heihe River basin is characterized byfragmentation. As Fig. 6 shows, core wetlands were the dominantlandscape in 1975. Patch and perforated wetlands were scatteredand only covered a small area in the wetland landscape. However,in 2010, this situation had gradually changed. In this period, thearea of perforated wetlands had enlarged significantly and largenumbers of core wetlands had been replaced by patch and edgewetlands.

he middle reaches of the Heihe River.

Fig. 6. Spatial distribution of the four classes of wetland patterns in five periods from 1975 to 2010.

246 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

4.2. Driving forces of wetland landscape fragmentation

When performing RDA analysis, landscape indices, such as FS1,FS2, MPS, PD, and area of core, patch, edge, and perforated wetlands,were selected as dependent variables, and factors reflectingsocioeconomic and natural change were selected as independentvariables (Fig. 7). Prior to the RDA analysis, a significant test (MonteCarol Permutation Tests, Number of permutation: 499) for eachvariable was performed to primarily screen the important factors.As the results in Table 3 show, the PIO (primary industrial output), T(annual temperature) and GDP (gross domestic product) passed thehypothetical testing under the 0.05 significance level. Therefore,only the PIO, T, and GDP were selected for the RDA analysis.

According to the results of RDA analysis, the first threeeigenvalues reported above are canonical and the fourth is notsince only three independent constraints can be formed from theenvironmental variables. As Table 4 and Table 5 show, thecumulative percentage variance of species data and species-environment relations are high, at 84.6% and 100%, in the first threeaxes. This demonstrates that the environmental variables weselected interpret the process of wetland landscape fragmentationappropriately. Based on the angles between the landscape indices

and environmental variables, we found that the PIO and GDP have asignificant positive correlation with PD and perforated wetlands,but a significant negative correlation with MPS and core wetlands(Fig. 8). Furthermore, T also has a significant positive correlationwith FS1, FS2, PD, and patch wetlands, but a significant negativecorrelation with MPS and core and edge wetlands. Moreover, T-testresults for the influence of T, PIO, and GDP on the landscape indicesof wetlands also show that FS2 and patch wetlands have asignificant positive correlation with T, PIO, and GDP (Fig. 8).However, it is obvious that PIO and GDP contribute more towetland landscape fragmentation than T according to the length ofthe arrow. This further demonstrates that human activities are themost important driving forces in our study area as compared withnatural changes.

5. Discussion

5.1. Extraction of wetland information

Accurate extraction of wetland information is the most criticalstep in the analysis of long time series of wetland landscapes

Fig. 7. Changes of the environmental variables selected in this study for different periods.

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 247

(Wang et al., 2011). However, in the present study, the selectedlandscape pattern indices were all calculated from data obtainedfrom remote sensing images; this technique often results inconsiderable errors relating to incorrect classification during thecalculation of landscape pattern indices (Langford et al., 2006).

Using the topographic maps and land use data described above,we were able to extract precise wetland information for 1975,1987,and 1992. Therefore, we considered only the accuracy ofinformation extracted for 2001 and 2010 explicitly in the presentstudy, placing more importance on the collection of historical atlasdata, high-resolution Google Earth images, and field survey datafor these periods. Moreover, wetlands can be distinguished moreeasily than other land cover types in arid regions on TM and ETM+images, and the primary error in our technique arises fromclassification. However, by selecting different band combinationsand collecting data from different channels, we were able toconfirm the high quality of our results through accuracyassessment.

5.2. High-density human activity as a primary cause of wetlandlandscape fragmentation

The relationship between wetland landscape pattern evolutionand human activity is currently the subject of considerable debate

Table 3P-value and F-value of the environmental variables

GY TRS PIO GDP T

P 0.068 0.052 0.042a 0.044a 0.044a

F 2.816 2.784 2.767 2.748 2.722

Notes: GY (grain yield), TRS (total retail sales), PIO (primary industrial output), GDP (grRPCIN (rural per capita net income), TI (total investment), TP (total population), P (ann

a P < 0.05.

in the field of landscape ecology. Many previous studies haveposited that high-density human activity is the primary factorcontrolling the evolution of wetland landscape pattern (Khaznadaret al., 2009; Li and Zhao, 2010; He et al., 2011; Song et al., 2012).Based on our results, we believe this to be the case in the middlereaches of the Heihe River.

Known for their oasis agriculture, reclaimed wetlands (Fig. 9b)have long been considered an important means by which toincrease available arable land (Table 6). During recent decades,5410.82 ha of wetland has been transferred into farmland(Table 6). Moreover, agricultural activities such as flood irrigationand application of insecticide and chemical fertilizers can alsodamage wetland ecosystems and alter their pattern. Statisticaldata have shown that agriculture accounts for 95% of the totalconsumption in the study area (Wang et al., 2009), althoughvarious forms of construction engineering (such as artificial sandexcavation; Fig. 9a), the Binhe New Zone, and the Second Railwaybetween Lanzhou and Urumqi in Ganzhou, are likely to placeadditional stress on the wetland landscape as the economy of theregion develops further. Throughout the study period, 434.24 haof wetlands were developed for construction and used asresidential land (Table 6); as the population in the study areascontinues to increase, yet more land will be required to meetdemand.

SIO RPCIN TI TP P

0.058 0.060 0.076 0.080 0.8982.699 2.698 2.607 2.536 0.440

oss domestic product), T (annual temperature), SIO (secondary industrial output),ual precipitation).

Table 4Correlation matrix between axes and environmental variables.

1. Correlations between axis and environmental variables

SPECAX1

SPECAX2

SPECAX3

SPECAX4

ENVIAX1

ENVIAX2

ENVIAX3

ENVIAX4

SPECAX1

1.00

SPECAX2

�0.07 1.00

SPECAX3

0.19 �0.19 1.00

SPECAX4

0.26 �0.27 0.72 1.00

ENVIAX1

0.96 0.00 0.00 0.00 1.00

ENVIAX2

0.00 0.96 0.00 0.00 0.00 1.00

ENVIAX3

0.00 0.00 0.69 0.00 0.00 0.00 1.00

ENVIAX4

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

PIO 0.87 �0.41 �0.04 0.00 0.90 �0.43 �0.05 0.00GDP 0.87 �0.39 �0.09 0.00 0.90 �0.41 �0.14 0.00T 0.90 �0.04 0.26 0.00 0.93 �0.05 0.37 0.00

2. Correlations between environmental variables

PIO GDP TPIO 1GDP 0.99 1T 0.84 0.81 1

Notes : GY (grain yield), TRS (total retail sales), PIO (primary industrial output), GDP(gross domestic product), T (annual temperature), SIO (secondary industrialoutput), RPCIN (rural per capita net income), TI (total invest), TP (total population),P (annual precipitation).

Table 5Summary of the results of the redundancy analysis (RDA).

Axes 1 2 3 4 Totalvariance

Eigenvalues 0.538 0.227 0.081 0.154 1.000Species-environmentcorrelations

0.965 0.964 0.694 0.000

Cumulative percentagevariance of species data (%)

53.8 76.5 84.6 100.0

Cumulative percentagevariance of species-environment relation (%)

63.6 90.4 100.0 0.0

Sum of all eigenvalues 1.000Sum of all canonicaleigenvalues

0.846

Notes: The first three eigenvalues reported above are canonical and the fourth is notsince only three independent constraints can be formed from the environmentalvariables.

248 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

5.3. Land use/land cover change and wetland fragmentation

Land use and land cover changes are closely linked tosocioeconomic development (Wang et al., 2011). In fact, in mostcases, such changes are the most appropriate proxies for humanactivity, and they have been shown to have a significant influenceon all aspects of river ecosystems (Strayer et al., 2003).

In the middle reaches of the Heihe River, land use and landcover changes play a key role in driving wetland landscapefragmentation. Our results demonstrate an overall loss of11,536.99 ha of wetlands (Table 6), which appears to have beenconverted primarily to farmland, desert, or grassland. In particular,during 1975–2010, about 5410.82 ha of wetlands were convertedinto farmland; this transition represents 46.89% of the totalwetland loss and is likely due primarily to the well-knownhistorical development of agriculture in the study area. Theconcept of farmland as a source of wealth is well established in theregion, and wetlands are widely known to offer the greatestpotential for cultivation owing to their fertile soils. Therefore, it islogical that wetlands were extensively cultivated. This cultivationresulted in the division of the wetlands into ever smaller patches,eventually resulting in fragmentation.

Despite the extensive transition from wetland to farmland,considerable amounts of wetland also transitioned to grasslandand desert. According to our transition matrix (Table 6), approxi-mately 3498.16 ha and 2028.16 ha of wetlands transitioned intograssland and desert, respectively, and was likely caused by over-exploitation of groundwater resources and water diversion in themiddle reaches of the Heihe River. Previous studies have shownthat the groundwater recharge rate decreased by 2.168 � 108m3/aduring the 1970s and 1980s, whereas the amount of groundwaterstorage decreased by 0.545 �108m3/a since 1986 (Wang et al.,2005). The water surplus also tended to decrease, in particular,

since the implementation of the water diversion policy in 2001.Ignorance of the dwindling water helped increase the farmlandarea in the study area by 7798.14 ha during 1975–2010. Accord-ingly, increasingly small amounts of water were available for thewetlands, causing those without adequate water to degenerate andbe replaced by desert or grassland. Accordingly, wetlands with asufficient water supply have formed vegetation mosaics withingrassland and desert landscapes.

5.4. Key of water diversion in the middle reaches of the Heihe River inwetland fragmentation during 2001–2010

Water is a critical factor controlling the formation anddevelopment of wetland vegetation and soil, particularly in aridenvironments (Kong et al., 2009). Accordingly, in most cases,changes in hydrology have direct effects on the energy cycle andmaterial flow in wetland systems; such systems are particularlysensitive to hydrological changes such as variations in the quantityand quality of their water sources (Wang et al., 2011).

Water diversion (Fig. 7f) in the middle reaches of the HeiheRiver was initiated in 2001 and was directly responsible for thedecrease in the water surplus during 2001–2010 (Table 7). Thiswater diversion also had a profound effect on the processesinvolved in the maintenance of the wetland landscape pattern.Wetlands are particularly sensitive to changes in water supply; inour study area, such changes affected the productivity of aquaticplants and animals and disrupted the regional water balance. Ourresults show that these intervals with decreased water availabilitycoincided with periods of obvious landscape fragmentation. Forexample, the water surplus during 1975–1987 was less than that atany other time, and FS1 and FS2 for wetlands as a whole increasedfrom 0.565 to 0.575 and from 0.270 to 0.275 (Fig. 5), respectively,during this period. Similarly, the water surplus in 2001–2005 wasonly 6.50 � 108m3 (Table 7) owing to the influence of waterdiversion; this period of low surplus corresponded to decreases inMPS (19.35%) and PD (from 0.012 Ind/ha to 0.015 Ind/ha) butincreases in FS1 and FS2 (from 0.272 to 0.276 and from 0.539 to0.556, respectively) (Fig. 5). Thus, our results demonstrate a directrelationship between hydrology and wetland landscape fragmen-tation.

Water demand in the region will continue to increase withincreases in industrialization and urbanization, resulting in tensionas competition grows for the available resources. Such humanintervention has altered the hydrological cycle in the wetlandecosystem, reducing the supply of water from upstream andendangering the balance between supply and demand anddisrupting the wetlands in the middle reaches of the Heihe River.Moreover, because the area requiring irrigation in the study area

Fig. 8. T-test results for GDP (a), PIO (b), and T (c) influencing landscape indices (red circle means negative correlation and grey circle means positive correlation); Redundancyanalysis results for landscape indices and environmental variables (d).

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 249

has been maintained since the water diversion, farmers havebegun to develop groundwater resources to meet irrigation needs(Table 8), resulting in reductions in exploitable groundwater andoverall groundwater level described above (Wei et al., 2007).

Groundwater flow has a considerable influence on the chemicalproperties of groundwater in the aquifer that feeds the study area,and it has been shown that river water is preferable for irrigation(Qin et al., 2011). Therefore, high-density exploitation of thegroundwater will likely induce soil salinization without maximiz-ing the potential of the regional water balance and may have aconsiderable negative effect on plant life and soil characteristics.Accordingly, we believe that decreasing groundwater levels may beprimarily responsible for the degradation of wetland landscapepatterning and that the wetlands in the middle reaches of theHeihe River will continue to fragment and even disappear withoutintervention.

5.5. Effects of climate change on wetlands

Climate is the primary factor controlling the dynamics ofwetland formation and material exchange (Sutula et al., 2003),

and the influence of climate change on wetlands is expressedprimarily through changes in precipitation and temperature,which determine wetland plant physiology (Dawson et al., 2003).Accordingly, changes in precipitation and temperature can inducechanges in wetland hydrology (Milzow et al., 2010), causingwetlands to expand or contract. Furthermore, as previous studieshave shown, temperature is the most fundamental factorcontrolling the dynamics of wetland landscape patterning (Songet al., 2012).

Wetland systems are particularly vulnerable to changes inclimate and water supply, particularly in arid regions, where suchchanges threaten the continued existence of wetlands. In particu-lar, climate change often affects the landscape pattern of wetlandsthrough its effects on wetland hydrology, and fluctuations inprecipitation or temperature are known to affect water availabilityand stream flow (Dawson et al., 2003).

Throughout our study period, a significant increase intemperature and fluctuations in precipitation occurred in themiddle reaches of the Heihe River: temperature increased from7.53 �C to 8.72 �C in recent decades, whereas precipitationdecreased from 121.78 mm to 118.74 mm between 1992–2001and 2001–2010 (Fig. 7e). Generally speaking, the region has

Fig. 9. Wetland degradation resulting from (a–c) human activity and (d) climate change.

Table 6Transition matrix of land use during 1975�2010 (units: ha)

2010

1975 Construction land Desert Forest Grassland Farmland Wetland Total

Construction land 15387.40 0.58 24.52 35.14 608.38 51.61 16107.63Desert 1951.64 624650.00 3749.15 14577.80 18284.10 1593.02 664805.71Forest 112.14 1174.38 11952.40 2140.37 4155.81 62.08 19597.18Grassland 632.58 16931.70 2588.06 128635.00 39704.00 1736.47 190227.81Farmland 2839.37 1445.90 391.13 2593.61 144148.00 528.13 151946.14Wetland 434.24 2028.16 165.62 3498.16 5410.82 21118.60 32655.59Total 21357.37 646230.72 18870.87 151480.08 212311.11 25089.92 1,075,340.06

250 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

become warmer and drier, and the wetlands have exhibited someevidence of fragmentation in response to these changes.

5.6. Recommendations for wetland protection

Wetlands in the middle reaches of the Heihe River are crucialfactors for the sustainable development of the area and itssurroundings. However, landscape fragmentation has caused thesewetlands to shrink and even disappear. It can be predicted thatthese wetlands will vanish in the near future if measures are nottaken to protect them. Therefore, measures to protect wetlandsfrom fragmentation are urgently needed.

Table 7Mean surplus of water during different time periods (� 108m3).

1975–1987 1987–1992 1992–2001 2001–2005

5.61 7.23 8.01 6.50

Notes: Water surplus = influx from Yingluoxia valley � outflow from Zhengyixiavalley.

First,a general investigation of the wetland resource should becarried out to monitor its dynamic changes. Satellite remotesensing technology, GIS technology, and field investigation, amongother techniques, can be used for real time analysis of the processof landscape pattern change in the wetland from spatial andtemporal perspectives. In this manner, we can protect the wetlandresources from being fragmented. In addition, these investigationscan contribute to the scientific management and rational use ofwetland resources.

Second, artificial restoration measures need to be undertaken torestore fragmented wetlands. In wetland reserves, any reclamationof wetland must be forbidden and the reclaimed wetlands shouldbe returned. The water diversion that began in 2001 has caused thewetlands in the study area to face the threat of a lack of watersupply and, therefore, the critical artificial restoration measure isto guarantee sufficient ecological water supply. Thus, it isnecessary to undertake ecological water transport projects duringthe period of water diversion or in arid times. Water supplyengineering, such as water diversion works and impounded waterprojects, can also be undertaken to ensure sufficient water supplyto wetlands.

Table 8Changes in groundwater in the middle reaches of the Heihe River during 2001–2005 (based on Wei et al. (2007)).

County Study area(north–south � east–west, km2)

Mean change in groundwater level(m)

Change in groundwater storage(108m3)

Ganzhou 43.40 � 39.10 �2.68 �4.90Linze 43.23 � 25.68 �1.55 �1.72Gaotai 27.72 � 44.66 �1.30 �1.61

P. Jiang et al. / Ecological Indicators 46 (2014) 240–252 251

Laws for the protection of wetlands must be enacted andimproved in a timely manner in order for wetlands to come underthe unfailing aegis of law.

6. Conclusions

The results of our study show that the indices we chose revealthe process of landscape fragmentation in wetlands at landscapeand pixel levels well. A new mathematic morphological methodwas proposed for the calculation of four indices. Compared withthe landscape pattern index, the indices based on this new methodcan better support the analysis of landscape fragmentationwithout changing the original pixel size of the interpreted data;this enhances the credibility of the analysis. In addition, this newmethod is useful for researchers to evaluate landscape fragmenta-tion not only for wetlands but also for forests, farmlands, and otherland types.

Redundancy analysis was performed to analyze the drivingforces of landscape fragmentation in wetlands. In addition,qualitative analyses were also performed to explore the effectsof changes in hydrology and land use/cover on the landscapefragmentation in wetlands. By combining quantitative analysiswith qualitative analysis, we were able to comprehensively analyzethe cause of the landscape fragmentation in the wetlands in ourstudy area. The results show that PIO (primary industrial output), T(annual temperature), and GDP (gross domestic product) were themajor factors causing wetland fragmentation in our study area. Ourresults indirectly show that human activities are not the solereason for landscape fragmentation of wetlands in arid zones.Natural changes, such as increases in temperature, also play a keyrole and cannot be ignored. In addition, the driving forces analysisthat we propose in this study provides a reference for relatedstudies.

Acknowledgments

This work is supported by the National Natural ScienceFoundation of China (Grant No. 41371017, 41001238), the NationalKey Technology R&D Program of China (Grant No. 2012BAH28B02).Sincere thanks are given for the comments and contributions ofanonymous reviewers and members of the editorial team.

References

Alexandridis, T.K., Lazaridou, E., Tsirika, A., Zalidis, G.C., 2009. Using earthobservation to update a Natura 2000 habitat map for a wetland in Greece. J.Environ. Manage. 90, 2243–2251.

Cerian Gibbes, J.S., Keys, E., 2009. Wetland conservation: change and fragmentationin Trinidad's prtected areas. Geoforum 40, 91–104.

Copeland, H.E., Tessman, S.A., Girvetz, E.H., Roberts, L., Enquist, C., Orabona, A., Patla,S., Kiesecker, J., 2010. A geospatial assessment on the distribution, condition,and vulnerability of Wyoming's wetlands. Ecol. Indic. 10, 869–879.

Dawson, T.P., Berry, P.M., Kampa, E., 2003. Climate change impacts on freshwaterwetland habitats. J. Nat. Conserv. 11, 25–30.

Erwin, K.L., 2008. Wetlands and global climate change: the role of wetlandrestoration in a changing world. Wetl. Ecol. Manage 17, 71–84.

He, X.H., Gao, Y.J., Niu, J.X., Zhao, Y.F., 2011. Landscape pattern changes under theimpacts of urbanization in the yellow river wetland—king Zhengzhou as anexample. Procedia Environ. Sci. 10, 2165–2169.

Kang, E., Lu, L., Xu, Z., 2007. Vegetation and carbon sequestration and their relationto water resources in an inland river basin of Northwest China. J. Environ.Manage. 85, 702–710.

Khaznadar, M., Vogiatzakis, I.N., Griffiths, G.H., 2009. Land degradation andvegetation distribution in Chott El Beida wetland, Algeria. J. Arid. Environ. 73,369–377.

Kong, W., Sun, J., Xu, W., Chen, Y., 2009. Changes in vegetation and landscapepatterns with altered river water-flow in arid West China. J. Arid Environ. 73,306–313.

Langford, W., Gergel, S., Dietterich, T., Cohen, W., 2006. Map misclassification cancause large errors in landscape pattern indices: examples from habitatfragmentation. Ecosystems 9, 474–488.

Li, S.B., Zhao, W.Z., 2010. Landscape pattern changes of desert oasis wetlands in themiddle reach of the Heihe River, China. Arid. Land Res. Manage. 24, 253–262.

Li, X.M., Lu, L., Yang, W.F., Cheng, G.D., 2012. Estimation of evapotranspiration in anarid region by remote sensing—A case study in the middle reaches of the HeiheRiver Basin. Int. J. Appl. Earth Obs. Geoinf. 17, 85–93.

Li, X.Q., Gan, Y.Q., Zhou, A.G., Liu, Y.D., Wang, D., 2013. Hydrological controls on thesources of dissolved sulfate in the Heihe River, a large inland river in the aridnorthwestern China, inferred from S and O isotopes. Appl. Geochem. 35, 99–109.

Lienert, J., Diemer, M., Schmid, B., 2002. Effects of habitat fragmentation onpopulation structure and fitness components of the wetland specialist Swertiaperennis L. (Gentianaceae). Basic Appl. Ecol. 3, 101–114.

Liu, H.Y., Li, F., Cao, X., Hao, F., Hu, N., N, Zheng., 2009. The Current Problems andPerspectives of Landscape Research of Wetlands in China. Acta. Geogr. Sin. 64,1394–1401 (in Chinese).

Liu, J.Y., Zhang, Z.X., Xu, X.L., Kuang, W.H., Zhou, W.C., Zhang, S.W., Li, R.D., Yan, C.Z.,Yu, D.S., Wu, S.X., Jiang, N., 2010. Spatial patterns and driving forces of land usechange in China during the early 21st century. J. Geogr. Sci. 20, 483–494.

Liu, S., Dong, Y., Deng, L., Liu, Q., Zhao, H., Dong, S., 2014. Forest fragmentation andlandscape connectivity change associated with road network extension and cityexpansion: a case study in the Lancang River Valley. Ecol. Indic. 36, 160–168.

Lu, L., Li, X., Cheng, G.D., 2003. Landscape evolution in the middle Heihe River Basinof north-west China during the last decade. J. Arid. Environ. 53, 395–408.

Mander, Ü., Mitsch, J., 2009. Pollution control by wetlands. Ecol. Eng. 35, 153–158.Milzow, C., Burg, V., Kinzelbach, W., 2010. Estimating future ecoregion distributions

within the Okavango Delta Wetlands based on hydrological simulations andfuture climate and development scenarios. J. Hydrol. 381, 89–100.

Mwakaje, A.G., 2009. Wetlands, livelihoods and sustainability in Tanzania. Afr. J.Ecol. 47, 179–184.

Opdam, P., Wascher, D., 2004. Climate change meets habitat fragmentation: linkinglandscape and biogeographical scale levels in research and conservation. Biol.Conserv. 117, 285–297.

Qin, D.J., Qian, Y.P., Han, L.F., Wang, Z.M., Li, C., Zhao, Z.F., 2011. Assessing impact ofirrigation water on groundwater recharge and quality in arid environment usingCFCs, tritium and stable isotopes, in the Zhangye Basin, Northwest China. J.Hydrol. 405, 194–208.

Riitters, K., Vogt, P., Soille, P., Estreguil, C., 2009. Landscape patterns frommathematical morphology on maps with contagion. Landsc. Ecol. 24, 699–709.

Song, K.S., Wang, Z.M., Li, L., Tedesco, L., Li, F., Jin, C., Du, J., 2012. Wetlands shrinkage,fragmentation and their links to agriculture in the Muleng-Xingkai Plain, China.J. Environ.Manage. 111, 120–132.

Soomers, H., Karssenberg, D., Soons, M., Verweij, P., Verhoeven, J.A., Wassen, M.,2013. Wind and water dispersal of wetland plants across fragmentedlandscapes. Ecosystems 16, 434–451.

Strayer, D.L., Beighley, R.E., Thompson, L.C., Brooks, S., Nilsson, C., Pinay, G., Naiman,R.J., 2003. Effects of land cover on stream ecosystems: roles of empirical modelsand scaling issues. Ecosystems 6, 407–423.

Sulman, B., Desai, A., Mladenoff, D., 2013. Modeling soil and biomass carbonresponses to declining water table in a wetland-rich landscape. Ecosystems 16,491–507.

Sutula, M.A., Perez, B.C., Reyes, E., Childers, D.L., Davis, S., Day, J.W., Rudnick, D., Sklar,F., 2003. Factors affecting spatial and temporal variability in material exchangebetween the Southern Everglades wetlands and Florida Bay (USA). Estuar.,Coast. Shelf Sci. 57, 757–781.

Vogt, P., Riitters, K.H., Estreguil, C., Kozak, J., Wade, T.G., Wickham, J.D., 2006.Mapping spatial patterns with morphological image processing. Landsc. Ecol.22, 171–177.

Walz, U., Syrbe, R.-U., 2013. Linking landscape structure and biodiversity. Ecol. Indic.31, 1–5.

Wang, G.X., Yang, Y.L., Chen, L., Jumpei, K., 2005. Impacts of land use changes ongroundwater resources in the Heihe River Basin. Acta. Geogr. Sin. 60, 456–466(in Chinese).

252 P. Jiang et al. / Ecological Indicators 46 (2014) 240–252

Wang, X.F., Ma, M.G., Huang, G.H., Veroustraete, F., Zhang, Z.H., Song, Y., Tan, J.L.,2012. Vegetation primary production estimation at maize and alpine meadowover the Heihe River Basin, China. Int. J. Appl. Earth Obs. Geoinf. 17, 94–101.

Wang, Y., Xiao, L., F, Lu., 2009. Analysis of water consumption using a regional input–output model: model development and application to Zhangye City,Northwestern China. J. Arid. Environ. 73, 894–900.

Wang, Z.M., Huang, N., Luo, L., Li, X.Y., Ren, C.Y., Song, K.S., Chen, J.M., 2011. Shrinkageand fragmentation of marshes in the West Songnen Plain, China, from 1954 toand its possible causes. Int. J. Appl. Earth Obs. Geoinf. 13, 477–486.

Wei, Z., Jin, H.J., Nan, Y.C., Yang, S.Z., Wu, J.K., Wang, S.Z., 2007. The effect on waterenvironment in middle and downstream area of Heihe River after waterallocation. J. Irrig. Drain. 26, 34–37 (in Chinese).

Westerberg, V.H., Lifran, R., Olsen, S.B., 2010. To restore or not? A valuation of socialand ecological functions of the Marais des Baux wetland in Southern FranceEcol. Econ. 69, 2383–2393.

Wiens, J.A., 1994. Habitat fragmentation island landscape perspectives on birdconservation. Ibis 137, S97–S104.

Zhao, R.F., Chen, Y.N., Zhou, H.R., Li, Y.Q., Qian, Y.B., Zhang, L.H., 2009. Assessment ofwetland fragmentation in the Tarim River basin, western China. Environ. Geol.57, 455–464.

Zhao, W.Z., Liu, B., Zhang, Z.H., 2010. Water requirements of maize in the middleHeihe River basin, China. Agric. Water Manage. 97, 215–223.

Zhou, H.R., 2005. Prospect on Multifunctional Landscapes of Marshes in Arid Areas.Arid. Land Geogr. 16–20 (in Chinese).