Journal of Arid Environments (2003) 53: 395–408doi:10.1006/jare.2002.1032

Landscape evolution in the middle Heihe River Basin ofnorth-west China during the last decade

Ling Lu%w*, Xin Li w & Guodong Cheng%w

%State Key Laboratory of Frozen Soil Engineering, Cold and Arid RegionsEnvironmental and Engineering Research Institute, Chinese Academy of

Sciences, Lanzhou 730000, People’s Republic of ChinawCold and Arid Regions Environmental and Engineering Research Institute,Chinese Academy of Sciences, Lanzhou 730000, People’s Republic of China

(Received 9 January 2002, accepted 23 April 2002)

Landscape evolution in the middle Heihe River Basin during the last decadewas investigated by two methods: (a) the changes of various landscapemetrics were analysed using the landscape structure analysis programFRAGSTATS; (b) the transition matrix of landscape patch types wascalculated with the help of the GIS software, IDRISI. The results showedthat from early 1980s to late 1990s, the landscape within the study area hasundertaken a complicated evolution in landscape structure and composition,but as a whole it still displayed a pattern of sharp contrast between oasis anddesert landscapes. The human activities have significantly changed thedistribution and allocation of limited water resource in the basin, leading to acontradiction between desertification and expansion of oases. The decreaseof Shannon’s diversity index and evenness index manifested intensivemanagement and reconstruction of landscape by human beings. While thisimproved the socio economic benefits of the region, it reduced landscapeheterogeneity and landscape diversity, leading to the decrease of environ-mental benefits of some areas in the basin.

# 2002 Elsevier Science Ltd.

Keywords: landscape evolution; arid region; landscape metrics; landscapestructure; geographic information system; FRAGSTATS

Introduction

Landscape structure and composition evolves continuously in space and time. Theseevolutions are attributable to the complex interaction between natural environment,various organisms, and human activities, resulting in the change of the stability ofindividual elements in the landscape system and the spatial structure of the landscape(Xiao et al., 1990; Li, 1997).

*Corresponding author. State Key Laboratory of Frozen Soil Engineering, Cold and Arid RegionsEnvironmental and Engineering Research Institute, Chinese Academy of Sciences, Lanzhou 730000,People’s Republic of China.

0140-1963/02/030395 + 14 $35.00/0 # 2002 Elsevier Science Ltd.

396 L. LU ET AL.

In this paper, the landscape evolution in the middle Heihe River Basin during thelast decade is investigated. This area is located in the piedmont plain of the QilianMountains. It has been a densely populated region from ancient times and has anagricultural development history of over 2000 years since this area owns flat land,adequate sunlight, and convenient water diversion condition (Cheng & Qu, 1992).Especially over the past decades, local people constructed water conservancy projects,reclaimed wasteland, improved the soils, and thus formed today’s large area ofirrigation oases. These oases have become an important commodity grain base innorth-west China and hold an increasingly important social and economic position.Accordingly, the study on the landscape changes in this active area can improve ourunderstanding of the relationships between landscape structure and naturalenvironment, and human activities, help to define the magnitude and direction oflandscape changes caused by human interfering activities, and thus provide animportant scientific basis for the sustainable development of the region.

The landscape evolution in the middle Heihe River Basin was qualitatively analysedby some investigators (Cheng et al., 1999; Wang et al., 1998). It is characterized bygradual differentiation of desertification processes and expansion of oases. The drivingforces are firstly attributable to the most persistent natural factor in the area, the aridclimate; and then attributable to the leading factor, the quantity and distribution ofwater resource. With continuous socio-economic development and increase inpopulation density, human activities have become the much more active forcepromoting the landscape changes in the area. That is, the oases in the region are nowpredominantly man-made. They have greater variability than natural ones. Theseartificial oases receive significant influences from the interference of human activitiesand other socio-economic conditions. Particularly, even the short-term political andsocio-economic policies might produce decisive influences.

In this paper, two kinds of quantitative methods are used to study the landscapeevolution in the middle Heihe River Basin during the last decade. One method is toanalyse landscape structure change based on various landscape metrics by using thelandscape structure analysis program FRAGSTATS; the other method is to define themutual transition regimes of various landscape patch types by establishing a transitionmatrix with the help of the GIS software, IDRISI.

Background of the middle Heihe River Basin

The Heihe River basin is the second largest inland river basin in the arid region ofnorth-west China. It is located between 961420–1021000E and 371410–421420N andcovers an area of approximately 128,000 km2. The middle part of the Heihe RiverBasin constitutes the primary part of the Hexi corridor plain in Gansu Province, withan area of 17,000 km2 (Fig. 1).

Sandwiched between the southern Qilian Mountains and the northern MazongMountains, the middle Heihe River Basin receives an annual precipitation varyingfrom 250 mm in the south mountainous area to less than 100 mm in the north high-plain area. Zonal soil types in the area include gray-brown desert soil and gray desertsoil. Azonal soil types include irrigation-warping soil, saline soil, meadow soil, bogsoil, and brown sand soil. Most common vegetation encountered in the area istemperate dwarf shrub and subshurb desert vegetation dominated by Chenopodiaceae,Zygophyllaceae, Ephedranceae, Asteraceae, Poaceae, and Leguminosae. Under theinfluence of the water resource distribution and human activities, there are cropsand afforested forest distributed on the piedmont lower alluvial fan and fluvial plain inthe middle part of the reaches of the river.

98 100 102

38

40

42

40

38

10210098

Gaxun LakeSogo Lake

Mazong Mountain

Beishan MountainQilian Mountains

Middle Heihe River Basin

42

N

Figure 1. Study area and location of the middle Heihe River Basin in north-west China.

LANDSCAPE EVOLUTION 397

Data and Methods

Landscape maps of two periods in the study area

Data used in this study were mainly obtained from two land-use maps of differentperiods in the Heihe River basin. One is a 1:500,000 land-use map of the HexiCorridor published in 1985, which was compiled with the help of rectified aerialphotographs and 1:100,000 topographic maps in late 1970s and field investigationsfrom 1980 to 1984 (Cheng & Qu, 1992). Another map used is the 1:500,000 land-usemap of the Heihe River Basin (Xiao, 1999), which was compiled based on a hard-copymosaic of 11 Landsat Thematic Mapper (TM) images during the summer season of1997. Because these two thematic maps used different map projections, we used theGIS software ARC/INFO to make co-ordinate transformations and therebytransformed them into the same geographic co-ordinate system defined by the Albersequal-area map projection. The registration results of these two maps are ideal andhave a system error within 100 m.The compilation of landscape maps was based on a map generalization method. Wefirst formulated a classification system of the landscape patch types in the middleHeihe River Basin, within which a total of 15 types were identified (Table 1). Then,with the support of the GIS software ArcView 3?0a, we compiled the landscape mapsof the Heihe River Basin for the two periods by reclassifying the 66 land-use types into15 patch types and dissolving the polygons with the same type. The detailedprocedures were described by Li et al. (2001). Finally, we clipped the two landscape

Table 1. Landscape patch types in the middle Heihe River Basin and theirdefinitions

Patch ID Patch type Definition

1 Steppe Stipa purpurea, Agropyron cristatum,Carex montana, Allium polyneura

2 Middle-mountain meadow Carex montana, Kobresia spp., weeds3 Desert steppe Stipa spp., Salsola passerina, Artemisia

spp.4 Nonirrigated farmland Nonirrigated crops5 Irrigated farmland Irrigated crops6 Residential area City or town7 Artificial forest Fruit trees, Elaeagnus angustifolia, Po-

pulus spp., Populus euphratica,Elaeagnus angustifolia, Tamarix spp.

8 Meadow Phragmites communis, Achnatherumsplendens, Aneurolepidium dasystachys,Calamagrogstis spp.

9 Desert Stipa spp., Sympegma regelii, Reaumur-ia soongorica, Salsola passerina,Ephedra przewalskii

10 Sandy desert Stabilized dune, semi-stabilized dune,mobile dune

11 Bare gobi Hap-Orthic Aridisols F loam flat orhill

12 water area Reservoir, lake, wandering river bed13 Salinized meadow Karelinia caspica, Glycyrrhiza inflata,

Tamarix spp., Phragmites communis14 Salt desert Ari-Orthic Halosols F bottomland15 Marsh Phragmites communis, Carex Montana,

Potentilla spp. F wet bottomland

398 L. LU ET AL.

maps with the map of landscape zones of the middle Heihe River Basin (Li et al.,2001), and therefore obtained the landscape maps of the middle Heihe River Basin inthe two periods. An important principle used in the delineation of landscape zones isthat the patch integrity must be ensured, i.e. an entire patch polygon should not bedivided into several polygons when performing map clipping. By doing so, the edge oflandscape zones is ensured to be the edge of patches thus ensuring the accuracy oflandscape indexes. Figure 2(a, b) represents the landscape maps of the middle HeiheRiver Basin in 1985 and 1997. In Fig. 2(a) there are only 12 patch types. This isbecause on the land-use map in 1985, there were no middle-mountain meadow anddesert steppe. In addition, the residential area was not a class since cities and townswere not well developed at that time.

Establishment of the transition matrix of landscape patch types

To define the transition of landscape patch types in the middle Heihe River Basin over thelast decade, we used the Crosstab module in the GIS software IDRISI to calculate boththe transition matrix and the cross-classification map. The procedures are as follows:

(1) Convert the landscape maps in vector format into gridded maps. The

resolution of the gridded maps was chosen as 250 m to ensure that the

Figure 2. Landscape maps of the middle Heihe River Basin in 1985 and 1997.

LANDSCAPE EVOLUTION 399

smallest patch can be considered and computation efficiency is high. The

RMS error of the co-registration of the two maps (o100 m) is less than one

grid cell.

400 L. LU ET AL.

(2) Use the Crosstab module in IDRISI to calculate the cross-tabulation table,

which was output as a transition matrix (Table 2). The element Pij in the

matrix represents the transition ratio from the ith patch type in 1985 to the

jth patch type in 1997.

(3) While calculating the transition matrix, the Crosstab module can also

produce a cross-classification map, which expresses the where and how of

the transition. Among the 180 possible transitions (12 types in 1985� 15

types in 1997), 116 were effective transitions and others were just 0.

Obviously, it was difficult to display and analyse this very complex

transition map, so we generalized the 116 transitions into 15 types of major

and important transitions in accordance with two principles: (a) The new

map should illustrate the intensive differentiation of two dominant patch

types, the farmland and the desert; (b) it should be favorable to analyse the

change of two basic contradiction processes, i.e. the expansion of oasis and

the desertification.

(4) For easy mapping, the transition map was exported to the ArcView 3.0a

software (Fig. 3).

Calculation of landscape metrics

FRAGSTATS, developed by the Forest Science Department, Oregon StateUniversity, U.S.A., is a program for quantifying landscape structure. There are twoversions of FRAGSTATS, Vector (ARC/INFO) & Raster (Image Maps) versions(McGarigal and Marks, 1994). We used the raster version in this study to ensure thata great number of metrics can be calculated. In addition, we computed two groups ofmetrics: (1) the class level, which means each patch type in the landscape mosaic; (2)the landscape level, which means the landscape mosaic as a whole.

FRAGSTATS is capable of calculating more than 40 landscape metrics. However,many of them can be highly correlated (Environmental Protection Agency, 1994;Riitters et al., 1995) so that an important principle is to select uncorrelated metrics. Inour analysis of the landscape structure of the middle Heihe River Basin at the classlevel (Table 3), nine indices were selected: class area, percent of landscape, number ofpatches, mean patch size, area-weighted mean shape index, area-weighted mean patchfractal dimension, mean nearest-neighbor distance, mean proximity index, andinterspersion and juxtaposition index. For the analysis of landscape structure at thelandscape level (Table 4), 13 indices were selected: total landscape area, largest patchindex, number of patches, mean patch size, area-weighted mean shape index, area-weighted mean patch fractal dimension, mean nearest-neighbor distance, meanproximity index, patch richness, Shannon’s diversity index, Shannon’s evenness index,interspersion and juxtaposition index, and contagion index. The definition anddescription of these indices in FRAGSTATS are given by the FRAGSTATS user’sguide (McGarigal & Marks, 1994).

Results and analysis

Comparison of landscape metrics

Comparison of landscape metrics at class level

In Table 3 we compare the change of landscape metrics at class level. Landscapeevolution is also illustrated in Fig. 2. These results showed the following.

Table 2. The transition matrix of landscape patch types in the middle Heihe River Basin from 1985 to 1997 (%)

1997

1985 Steppe Middle-mountainmeadow

Desertsteppe

Nonirrigatedfarmland

Irrigatedfarmland

Residentialarea

Artificialforest

Meadow Desert Sandydesert

Baregobi

Waterarea

Salinizedmeadow

Saltdesert

Marsh Total

Steppe 25?5 0?6 0?0 22?3 27?5 0?2 0?0 0?0 23?8 0?0 0?0 0?1 0?0 0?0 0?0 100?0Nonirrigated farmland 8?3 0?0 0?0 26?4 61?3 0?1 0?0 0?0 4?0 0?0 0?0 0?0 0?0 0?0 0?0 100?0Irrigated farmland 1?0 0?0 0?1 3?2 72?3 1?0 1?0 3?7 14?0 0?5 1?8 0?6 0?8 0?0 0?1 100?0Artificial forest 0?0 0?0 0?0 0?0 51?7 2?1 12?6 22?0 2?5 4?5 2?1 0?5 2?1 0?0 0?0 100?0Meadow 0?0 0?0 0?0 0?0 16?8 0?2 1?6 32?4 4?6 4?5 7?6 5?5 8?8 12?0 6?1 100?0Desert 0?3 0?2 0?0 0?3 9?3 0?0 0?1 2?0 70?0 1?6 14?9 0?1 0?3 0?9 0?0 100?0Sandy desert 0?0 0?0 0?0 0?0 4?6 0?0 0?0 9?3 11?8 62?7 8?4 0?2 2?6 0?5 0?0 100?0Bare gobi 0?0 0?0 0?0 0?0 7?1 2?7 0?0 15?6 16?0 9?7 46?5 0?0 1?2 1?1 0?0 100?0Water area 0?0 0?0 0?0 0?0 34?0 0?0 0?0 21?1 5?2 3?4 3?1 18?4 12?5 2?3 0?0 100?0Salinized meadow 0?0 0?0 0?0 0?0 13?0 1?5 0?6 41?2 7?8 2?3 1?8 0?5 18?9 7?2 5?3 100?0Salt desert 0?0 0?0 0?0 0?0 9?3 0?0 0?0 35?3 31?6 0?0 0?3 0?3 13?4 7?7 2?1 100.0Marsh 0?9 0?0 0?0 0?9 33?8 0?2 6?9 9?5 33?7 5?6 1?5 0?5 5?9 0?0 0?6 100.0

LA

ND

SC

AP

EE

VO

LU

TIO

N401

#�

�

#�

#�

#�

#�

#�

Jiayuguan

Qiuquan

Minghua Salt Pond

Gaotai

Zhangye

Shandan

Mingle

LegendDesert (no changes)Farmland (no changes)Other no changesSteppe to FarmlandMeadow to FarmlandDesert to FarmlandMarsh to FarmlandFarmland to MeadowDesert to MeadowSalinized meadow to MeadowFarmland to DesertDesertificationDesert to Bare gobiOthers to Water areaOther changes

City or town

N

50 0 50 100 150 Kilometers

�

Figure 3. Patch type transition map of the middle Heihe River Basin during the last decade.

402 L. LU ET AL.

Irrigated farmland is the highest variable landscape patch type: its area increasedfrom 407,981 ha in 1985 to 505,150 ha in 1997, whereas its number of patchesdecreased from 132 to 84 and the mean patch size was almost twice the original value(Table 3). The mean proximity index of the patches increased, but the interspersionand juxtaposition index decreased. The changes of these metrics reflect that thedistribution of irrigated farmland in the middle Heihe River Basin was morefragmentary in the early 1980s: the mean patch grain size was smaller but the discretedegree among patches was higher than today. Through more than 10 years ofreclamation and irrigation not only did the irrigated farmland expand by nearly100,000 ha but also the newly reclaimed farmland tended to surround the oldfarmland and finally joined together to form strips. Consequently, the patch grain sizebecame larger and fragmentation degree evidently declined, thus resulting in theleading position of irrigated farmland in the whole landscape mosaic. All thesechanges showed that agricultural development in the area was rather intense.

Desert is the patch type with the largest area and the coarsest patch grain size in theregion. An interesting feature is that the changes of various metrics showed anopposite tendency as compared with the irrigated farmland. For instance, the desertarea percentage in the total landscape area decreased from 42?0% in 1985 to 37?9% in1997. Patch number increased, mean patch grain size decreased by two times, meanproximity index of the patches was less than one-third of the corresponding value in1985, but the interspersion and juxtaposition index increased. Such growth anddecline changes in irrigated farmland and desert areas resulted from building waterconservancy projects and transforming desert areas into cultivated land in the lastdecade.

Except for irrigated farmland and desert, the bare gobi is another highly variablepatch type. In the last decade the bare gobi area expanded by 64,000 ha; its areapercentage increased dramatically from 5?8% in 1985 to 9?2% in 1997 in the totallandscape area. Patch number decreased from 22 to 8, but the mean patch area

Table 3. Metrics comparison at class level in the middle Heihe River Basin in 1985 and 1997

ID TYPE CA (ha) LAND (%) NP MPS (ha) AWMSI AWMPFD MNN (m) MPI IJI

19851 Steppe 76,850 4?8 10 7685 2?1 1?07 2381 400?5 47?34 Nonirrigated farmland 42,531 2?7 8 5316 2?8 1?1 2893 536?5 40?75 Irrigated farmland 407,981 25?6 132 3091 5?2 1?15 1392 2137?8 72?97 Artificial forest 9812 0?6 21 467 2?0 1?08 5430 11?8 62?78 Meadow 59,406 3?7 18 3300 4?1 1?15 5442 181?2 72?19 Desert 668,856 42?0 14 47,775 7?6 1?18 2964 9286?3 50?010 Sandy desert 78,231 4?9 26 3009 2?7 1?1 5593 78?5 67?311 Bare gobi 93,006 5?8 22 4228 2?6 1?1 5035 144?0 70?512 Water area 3881 0?2 19 204 1?6 1?06 10,900 3?7 66?013 Salinized meadow 60,606 3?8 24 2525 3?0 1?11 3110 102?0 75?714 Salt desert 61,606 3?9 4 15,402 3?1 1?11 1583 2049?4 51?315 Marsh 28,825 1?8 43 670 2?4 1?1 2227 53?6 59?419971 Steppe 43,113 2?5 9 4790 2?0 1?07 6891 63?1 36?72 Middle-mountain meadow 1588 0?1 2 794 2?7 1?12 95,707 0?0 32?53 Desert steppe 1269 0?1 1 1269 2?1 1?09 F 0?0 0?04 Nonirrigated farmland 60,394 3?5 12 5033 2?5 1?10 2762 77?9 31?05 Irrigated farmland 505,150 29?5 84 6014 6?1 1?17 1477 2391?2 68?36 Residential area 7844 0?5 9 872 1?7 1?06 12,396 16?0 49?37 Artificial forest 9419 0?6 4 2355 2?8 1?11 24,902 1?4 27?38 Meadow 119,706 7?0 47 2547 3?6 1?13 1464 359?9 65?69 Desert 649,088 37?9 27 24,040 6?9 1?17 2235 2834?6 54?910 Sandy desert 75,875 4?4 18 4215 2?4 1?09 1382 73?9 61?111 Bare gobi 157,319 9?2 8 19,665 3?4 1?12 6495 228?6 58?512 Water area 8131 0?5 9 903 4?8 1?17 24,730 21?9 43?413 Salinized meadow 40,319 2?4 34 1186 2?7 1?10 2056 15?2 59?614 Salt desert 24,738 1?4 8 3092 3?2 1?13 10,188 13?9 67?515 Marsh 8488 0?5 3 2829 2?2 1?09 20,190 17?3 52?6

ID: ID of landscape type; TYPE: patch type; CA: class area; LAND: percent of landscape; NP: number of patches; MPS: mean patch size; AWMSI: area-weighted meanshape index; AWMPFD: area-weighted mean patch fractal dimension; MNN: mean nearest-neighbor distance; MPI: mean proximity index; IJI: interspersion andjuxtaposition index.

LA

ND

SC

AP

EE

VO

LU

TIO

N403

Table 4. Metrics comparison at landscape level in the middle Heihe River Basin in 1985 and 1997

TIME TA (ha) LPI NP MPS (ha) AWMSI AWMPFD MNN (m) MPI PR SHDI SHEI IJI CONTAG

1985 1591593 26?8 341 4667 5?41 1?15 3297 1297 12 1?76 0?71 68?2 58?81997 1712437 22?8 275 6227 5?42 1?15 4607 1091 15 1?75 0?65 62?4 62?2

TA: total landscape area; LPI: largest patch index; NP: number of patches; MPS: mean patch size; AWMSI: area-weighted mean shape index; AWMPFD: area-weightedmean patch fractal dimension; MNN: mean nearest-neighbor distance; MPI: Mean proximity index; PR: patch richness; SHDI: Shannon’s diversity index; SHEI:Shannon’s evenness index; IJI: Interspersion and juxtaposition index; CONTAG: Contagion index.

404

L.

LU

ET

AL

.

LANDSCAPE EVOLUTION 405

increased by about five times. The mean proximity index of the patches increased, butinterspersion and juxtaposition index significantly reduced. The development of baregobi area, compared to the expansion of oases, displayed a contrary process in theregion, that is, the desertification.Other patch types also had notable variation. For instance, the area of meadowincreased by about 60,000 ha during the last decade. Its percentage in the totallandscape area increased from 3?7% to 7?0% and the patch number increased from 18to 47. The expansion of meadow mainly occurred around the Minghua Salt pond andalong the Heihe River (Fig. 2), this is closely related with the plentiful water resources.The marsh showed reversal changes. Its area decreased to about one-third of that in1985. Patch number decreased from 43 to 3 and the mean grain size was four times aslarge as the original one. It is interesting that no marsh existed in the Minghua SaltPond area in 1985, but two pieces of large marsh patches appeared in 1997. Thisdramatic change is because in recent years people diverted large quantity of water towash out soil salts and accordingly the accumulated water resulted in the developmentof new marshes. Furthermore, this also caused a significant decrease of halophytes onsalinized meadow soils and salt desert. The salinized meadow decreased by about one-third and salt desert decreased by nearly 40,000 ha.

Comparison of landscape metrics at landscape level

Comparison of the change of landscape metrics at landscape level (Table 4) shows thefollowing.

In the last decade the patch number in the middle Heihe River decreased from 341to 275 and the mean patch area expanded by about 1600 ha, showing that manyoriginal patches merged into larger ones and landscape heterogeneity declined. This isalso proven by the change of Shannon’s diversity and evenness indexes, both of whichbecame smaller. The spatial context of landscape patches also had significant changes.For instance, the mean nearest-neighbor distance extended by 1?3 km and the meanproximity index decreased, displaying that the spatial distribution of patches becamemore discrete. In addition, the interspersion and juxtaposition index became smallerand the contagion index became larger, illustrating that the spatial distribution ofvarious patches in the landscape became clumped, and the dominance of one orseveral patch types increased and had an even greater connectivity.

Results on the transition of landscape patch types

The magnitude and the direction of changes in landscape are the most importantfactors relating to landscape evolution (Antrop, 2000). From comprehensive analysisof Table 2 and Fig. 3, the following results were obtained.

Although the transitions among different patch types were quite complex in themiddle Heihe River Basin during the last decade (Table 2), the entire landscape stillremained a pronounced differentiation between desert landscape (desert, bare gobi,and sandy desert) and oasis landscape (irrigated farmland, nonirrigated farmland, andother natural oasis types). Desert and irrigated farmland were not only the types withthe highest predominance but also the types with the least transition frequencies.From 1985 to 1997, 70% of original desert kept unchanged, the rest mainly turnedinto bare gobi (14?9%) and irrigated farmland (9?3%). Among the irrigated farmland,unchanged area occupied 72?3%, whereas 14% turned into desert. Various types oftransition occurred in the oasis zone, demonstrating that the oasis environment in thearid region is variable.

The desert–oasis landscape is characterized by another phenomenon. There is along transitional zone between desert and farmland, where desert-transformed

406 L. LU ET AL.

farmland and farmland-transformed desert co-exist. This zone displays a remarkablefragmentary distribution and complex shape. We noticed that in this zone a total of59,825 ha of desert turned into farmland and 56,618 ha of farmland turned intodesert, having similar transformation intensities but in opposite directions. Thisindicated that the transitional zone is the most sensitive zone to the two contradictionprocesses, namely the desertification and expansion of oasis. The growth and declineof these two opposite processes in the transitional zone are dependent on thevariability of water resource.

The type transition of steppe is shocking. In the south-east corner of the region aswell as the outer edge of the Qilian Mountains were located some high-qualitygrasslands (steppe) with a total area of 75,033 ha in 1985, but only 25?5% was left by1997, 22?3% turned into nonirrigated farmland, 23?8% turned into irrigatedfarmland, and 27?5% turned into desert. These changes showed that in the lastdecade, people converted grassland into cropland in this region. Such detrimentalactivities reduced available rangeland and led to severe degradation of rangeland. Thisthreatened the sustainable development of forestry and grazing in the mountainousarea.

The Minghua Salt Pond region also displayed notable and complex transitions ofpatch types. For instances, about 41?2% of salinized meadow, 35?3% of salt desert,15?6% of bare gobi, 9?3% sandy desert, and 2% of desert turned into meadow (Fig.3). As a result, the meadow area increased from 59,406 ha in 1985 to 119,706 ha in1997. The new meadows were created by diverting water to wash out soil salts. On theother hand, opposite directional evolution occurred near this area (Fig. 3). There arethree large desertification patches around the Minghua Salt Pond, one patch is adesert transformed from meadow and salinized meadow, another is a salt deserttransformed from meadow, and the third was originally a marsh in 1985 but has nowbecame entirely engulfed by desert.

Another kind of significant transition occurred along the long and narrow farmingand grazing cross-bedding belt in the Jiuquan and Zhangye regions. There are sevenlarge tracts of desert-transformed bare gobi patches distributed in the region, with atotal area of 93,006 ha, showing that in the last decade the transformation rate of baregobi from desert has further been accelerated. These patches tend to be distributed incontinuous strips with coarse grain size as well as mean grain size becoming larger andmean proximity index enhanced. All these indicated that the connectivity and aaggregation degree among bare gobi patches increased. Apparently, such a strongexpansion trend of bare gobi will constitute potential environment pressure on theanimal husbandry development in the Sunan region.

Conclusion and discussion

The changes of landscape structure and composition in the middle Heihe River Basinwere investigated in this paper. The results showed that landscape evolution in theregion is characterized by two opposite processes, namely oasisification (expansion ofoasis) and desertification. On the one hand, people reclaimed about 100,000 ha ofnew irrigated farmland in the last decade and farmland patches joined together toform bigger oases. This led to a more important position of irrigated farmland in thewhole landscape mosaic. Oases also had significant variations in composition, 30% ofthem changed to other types. On the other hand, the area of desert decreased, but thisdoes not mean that desertification decelerated in the region because most of the desertpatches turned into more deserted type, the bare gobi. As a result, the aggregationdegree of bare gobi increased, threatening the development of the rangelands nearby.An interesting phenomenon is that in the transitional zone between oasis and desert,

LANDSCAPE EVOLUTION 407

landscape evolution was very active and the growth and decline of oases and desertswere maintaining a dynamic balance. Additionally, the heterogeneity of the wholelandscape declined. This was demonstrated by the change of various landscapemetrics in both the class and landscape levels.

The results also suggested that human activities have great impacts on the landscapeevolution in this region. We can conclude that: (1) the limited water resource (LIGG,1999) is a restriction for landscape evolution. For instance in the transitional zone,59,825 ha of desert was reclaimed into farmland but in the meanwhile 56,618 ha offarmland turned into desert. This can only be explained by the fact that the waterresource in the zone is limited. (2) The reclaim in the region had turned somegrassland into farmlands. This on the one hand improved economic benefits, but onthe other hand depressed landscape heterogeneity. (3) Improvement of salinized soilsresulted in landscape diversity in the Minghua Salt Pond area. New meadow andmarshes were brought out in the area.

From these conclusions we learned that agricultural development and waterresource utilization in this fragile arid region will soon lead to landscape changes,some of which as the evolution to desert landscape is uneasy to be restored. Therefore,in any development we should give special considerations to protect landscapediversity and to seek both economic and environmental benefits.

Viewed methodologically, the results showed that the landscape metrics in theFRAGSTATS are effective for characterizing landscape vulnerability and disturbanceassociated with human-induced and natural impacts (Environmental ProtectionAgency, 2000). Meanwhile, the transition matrix and transition map of landscapepatch types are helpful tools to analyse the changes of landscape structure andcomposition. However, some problems should be paid attention to. One problem isthe scale: some investigations showed that the analysis of landscape structure andchange are closely related to scale (Wu & Loucks, 1995; Chang & Wu, 1998).Landscape range and patch grain size define the maximum and minimum resolutionof the study (Turner et al., 1989; Wiens, 1989; McGarigal and Marks, 1994; Dobsonet al., 1997). Different scales and grid resolutions may produce different calculationresults of landscape indexes and further affect the analytical results of landscapestructure. Hence, caution should be taken in comparing landscapes in a variety ofscales. Another problem is the data sources. The data used in this paper came fromtwo different periods of land-use maps in the Heihe River basin, which were compiledbased on TM hard-copy images, topographic maps, and field investigations.Therefore, landscape maps and analytical results of landscape structure and changeused in this paper were directly restricted by the resolution of various data sources andthe classification system of the thematic maps. In future studies, we should directly useremote-sensing digital images from the same sources and in the same season. By usingthis kind of data, spatial co-ordinates can be strictly registered and classificationsystem of landscape patch types is more consistent (establish automatic remote-sensing classification), so we can calculate and analyse the landscape structure indexesand study or predict the dynamical changes of landscape more precisely.

This work was supported by the Innovation project of the Cold and Arid RegionsEnvironmental and Engineering Research Institute of CAS (CACX210018) and by theInnovation project of CAS (KZCX1-10-06).

Reference

Antrop, M. (2000). Background concepts for integrated landscape analysis. Agriculture,Ecosystems and Environment, 77: 17–28.

408 L. LU ET AL.

Chang, X.-L. & Wu, J.-G. (1998). Spatial analysis of pattern of sandy landscapes in Kerqin,Inner Mongolia. Acta Ecologica Sinica, 18: 225–232. (in Chinese).

Cheng, G.-D., Xiao, D.-N. & Wang, G.-X. (1999). On the characteristics and building oflandscape ecology in arid areas. Advances in Earth Sciences, 14: 11–15. (in Chinese).

Cheng, L.-H. & Qu, Y.-G. (1992). Water and Land Resources and Their Rational Development in theHexi Region. Beijing: Science Press. 263 pp. (in Chinese).

Dobson, A.P., Bradshaw, A.D. & Baker, A.J.M. (1997). Hopes for the future F restorationecology and conservation biology. Science, 277: 515–522.

Environmental Protection Agency (EPA). (1994). Landscape Monitoring and Assessment ResearchPlan. EPA 620/R-94/009, Office of Research and Development, Washington, DC.

Environmental Protection Agency (EPA). (2000). A Landscape Approach for Detecting andEvaluating Change in a Semi-arid Environment. http://www.epa.gov/crdlvweb/land-sci/san-pedro.htm.

Lanzhou Institute of Glaciology and Geocryology, Chinese Academy of Sciences (LIGG).(1999). Report of the Researches on Water Resource Allocation and Utilization and CoordinatedGrowth of Community, Economy and Ecosystems in the Heihe River Basin, Northwest China. LIGG(in Chinese).

Li, F. (1997). The theoretical analysis on the landscape ecological method application indesertification monitor. Arid Zone Research, 14: 69–73 (in Chinese).

Li, X., Lu, L., Cheng, G.-D.Xiao, H.-L. (2001). Quantifying landscape structure of the HeiheRiver Basin, northwest China using FRAGSTATS. Journal of Arid Environments, 48: 521–535.

McGarigal, K. & Marks, B. (1994). FRAGSTATS: Spatial Pattern Analysis Program forQuantifying Landscape Structure. Reference manual. Forest Science Department, OregonState University, Corvallis Oregon, March 1994. 62 pp.+Append.

Riitters, K.H., O’Neill, R.V., Hunsaker, C.T., Wickham, J.D., Yankee, D.H., Timmins, S.P.,Jones, K.B. & Jackson, B.L. (1995). A factor analysis of landscape pattern and structuremetrics. Landscape Ecology, 10: 23–39.

Turner, M., O’Neill, R., Gardner, R. & Milne, B. (1989). Effects of changing spatial scale on theanalysis of landscape pattern. Landscape Ecology, 3: 153–162.

Wang, G.-X., An, L.-Z., Xiao, H.-L. & Shi, D. (1998). Present Status, Development Tendency, andCountermeasures of the Eco-environment in the Heihe River Basin. Lanzhou Institute of Glaciologyand Geocryology, Chinese Academy of Sciences (LIGG) (in Chinese).

Wiens, J. (1989). Spatial scaling in ecology. Functional Ecology, 3: 385–397.Wu, J.-G. & Loucks, O. (1995). From balance of nature to hierarchical patch dynamics: a

paradigm shift in ecology. Quarterly Review of Biology, 70: 439–466.Xiao, D.-N., Zhao, Y., Sun, Z. & Zhang, H. (1990). Study on the variation of landscape pattern

in the west suburbs of Shenyang. Chinese Journal of Applied Ecology, 1: 75–84 (in Chinese).Xiao, H.-L. (1999). Compiling the Land Use Map of the Heihe River Basin. Lanzhou Institute of

Desert Research, Chinese Academy of Sciences (in Chinese).