O

Rs

Pa

b

5

a

ARRAA

KRELL

1

eteddtsoaWia

h1

Ecological Indicators 72 (2017) 399–410

Contents lists available at ScienceDirect

Ecological Indicators

j o ur na l ho me page: www.elsev ier .com/ locate /eco l ind

riginal Articles

egional ecosystem health response to rural land use change: A casetudy in Lijiang City, China

eng Jiana,∗, Liu Yanxua, Li Tianyia, Wu Jianshenga,b

Laboratory for Earth Surface Processes, Ministry of Education, College of Urban and Environmental Sciences, Peking University, Beijing 100871, ChinaKey Laboratory for Environmental and Urban Sciences, School of Urban Planning and Design, Shenzhen Graduate School, Peking University, Shenzhen18055, China

r t i c l e i n f o

rticle history:eceived 22 December 2015eceived in revised form 13 August 2016ccepted 16 August 2016vailable online 3 September 2016

eywords:egional ecosystem health assessmentcosystem servicesand use changeijiang City, China

a b s t r a c t

Quantitative analysis of the response of ecosystem health to rural land use change is required to compre-hend the human-nature coupling mechanism and to explore the process of global environmental change,which can interpret the ecological effects of regional land use and land cover change comprehensively.However, the existing regional ecosystem health assessment largely ignored either the internal connec-tion of ecosystem health to land use patterns or the internal representation of ecosystem services toecosystem health. Using Lijiang City of China as a study area, the average normalized difference vege-tation index (NDVI), landscape metrics, and ecosystem elasticity coefficient based on different land usetypes were used as quantitative indicators. Then the coefficient of spatial neighboring effect was intro-duced to characterize the adjacency effect on ecosystem services, and to generate the index of integratedecosystem health. The results showed the change of land use was close to 30% at county level from 1986to 2006, and forest land was the primary land use type. With respect to the declining physical health of

ecosystems in all the four counties, the integrated health experienced a slight increase in Lijiang County.The vast majority of towns’ ecosystem physical health and integrated health declined, while more than70% of towns did not change distinctly. Ecosystem physical health had distinct influence on the inte-grated ecosystem health, and ecosystem vitality was the main factor affecting the condition of physicalhealth. Emphasized in the interconnection of pattern and process, this study provided an ecosystemhealth approach to assessing the integrated ecological effects of regional land use change.

. Introduction

The upsurge of industrialization and urbanization withxpanded breadth and intensity of human activity has changedhe global ecosystem at unprecedented speed and scale, leading tocosystem degradation and significant threats to the survival andevelopment of human society. Facing the impacts of ecosystemegradation on human sustainable development, increasing atten-ions have been paid to how to monitor, evaluate, and regulate thetate and sources of risk, and the safety or sustainability degreef ecosystem health (Calow, 1992; Cairns et al., 1993; Dernbach

nd Mintz, 2011; Peng et al., 2007; Rapport, 1989; Suter, 1993;icklum and Davies, 1995). Ecosystem health refers to the abil-

ty to meet the reasonable requirements of human society and thebility to self-maintain and update the ecosystem (Lackey, 2001;

∗ Corresponding author.E-mail address: jianpeng@urban.pku.edu.cn (J. Peng).

ttp://dx.doi.org/10.1016/j.ecolind.2016.08.024470-160X/© 2016 Elsevier Ltd. All rights reserved.

© 2016 Elsevier Ltd. All rights reserved.

Rapport, 1998). Maintaining healthy ecosystems is fundamental toguarantee the achievement of sustainable socio-economic develop-ment, as natural ecosystems provide material basis and ecologicalservices for human survival (Rapport and Maffi, 2011). As the con-cept of health portrays essential features of a sustainable future(Patten and Costanza, 1997), when the health of natural ecosys-tem deteriorates severely from rapid socio-economic development,ecosystem health is considered as the purpose and basis of environ-mental management (Costanza, 1992, 2012). At the core positionof integrated ecosystem assessment, ecosystem health is regardedas one of the most important issues for ecosystem management inmacro ecological studies (Belaoussoff and Kevan, 2003; Burger andGochfeld, 2001; Pinto et al., 2013; Tiwari et al., 1998; Watson et al.,2005). Therefore, because ecosystem health represents integratedregional ecological quality from the two aspects of ecosystem struc-

ture and function, this concept is an important foundation forintegrated ecosystem assessment and management at macro scale.

Land use change has not only brought tremendous changes forthe structure of surface landscapes, but also affected materials

4 Indica

ccwcNct(eomebcYehdte

lllgcseuhtdamm

tnrLg2hitfiCus

2

2

so1jPMssht

00 J. Peng et al. / Ecological

irculation and energy flow of landscapes, resulting in signifi-ant change in regional climate, soil, biodiversity, hydrology andater resources, and profoundly impacting on regional ecologi-

al processes (Lavigne and Gunnell, 2006; Nainggolan et al., 2013;iedertscheider et al., 2012; Torres et al., 2014). Land use and landover change (LUCC) affects the ability of ecosystems to providehe services and biodiversity on which humans ultimately dependWan et al., 2015). A comparative LUCC study that compares differ-nt change processes could help to highlight the essential featuresf regional land use change and to provide an effective assess-ent of the ecological effects of regional land use change (Napton

t al., 2010; Turner, 2010). Currently, quantitative assessment hasecome mainstream in single factor analysis of LUCC’s ecologi-al effects (Imhoff et al., 1997; Weng, 2001; Weng and Lo, 2001;okohari et al., 2001). However, to reveal the integrated ecologicalffects of LUCC besides the single ecological processes, compre-ensive indicators are still in need (Shi et al., 2009), in spite of theifficulty to accurately measure the contribution of land use pat-erns to the variety of regional eco-environmental quality (Zhangt al., 2015).

Ecosystem services can directly link regional land use with eco-ogical quality; and the proportional relationship between differentand use types and economic value of ecosystem services calcu-ated by Costanza et al. (1997), has been widely applied to quantifylobal, national, or regional ecological quality corresponding to aertain land use structure. Thus, the quantification of ecosystemervices can help to make an integrated assessment of ecologicalffects with the change of quantitative structure of regional landse types (Li et al., 2007). The introduction of regional ecosystemealth indicators based on ecosystem services and land use pat-erns to characterize eco-environmental quality will establish airect bridge between LUCC and regional ecological state. However,lthough applying ecosystem health indicators to comprehensivelyeasure the influence of LUCC on regional ecological quality is aeaningful exploration, the detailed process is still obscure.Unlike the rapid urbanization with substantial transformation

o construction land in the east of China, the southwest of theation is mountainous with abundant biological resources, expe-iencing slow but noticeable land use change (Peng et al., 2008).ijiang City in the southwest of China has been identified as alobally significant region for its rich biodiversity (Xu and Wilkes,004), which is an ideal study area to perform a regional ecosystemealth assessment. Thus, besides the theoretical object of complet-

ng a quantitative assessment of the LUCC’s ecological effects inhe view of ecosystem health, two detailed research objects were:rstly, to clarify the process of regional land use change in Lijiangity; and secondly, to integrate landscape composition and config-ration metrics, as well as the neighborhood effects of ecosystemervices in the paradigm of ecosystem health assessment.

. Study area and data source

.1. Study area

Lijiang City is located in the middle-upper stream of the Jin-ha River in the northwest of Yunnan Province, covering an areaf 21,219 km2. The longitude of Lijiang City ranges from 99◦23′ to01◦31′E, with the latitude from 25◦59′ to 27◦51′N. Located in the

oint position of Qinghai–Tibet Plateau and the Yunnan–Guizhoulateau, and crossing the two geomorphic units of Hengduan

ountains Canyon and Western Yunnan Plateau, Lijiang City has

ubstantially more mountains than plains, with over 2000 m aboveea level accounts for 92.3% of the total area (Fig. 1). Lijiang Cityas the multi-year average precipitation of about 1000 mm, andhe annual mean temperature of 12.6–19.9 ◦C. Although the heat

tors 72 (2017) 399–410

is insufficient, tourism resources, biodiversity and water resourceshighlight the huge potential for regional development. In the viewof land use structure, forest land, grassland, and dryland accountfor a considerable proportion of the land use, while the lowestproportion is construction land (Peng et al., 2008).

There are four counties and one district in Lijiang City now,i.e. Gucheng District, Yulong County, Yongsheng County, Huap-ing County, and Ninglang County. However, before 2002, GuchengDistrict and Yulong County were combined as one Lijiang County.As a result, the four counties of Lijiang, Yongsheng, Huaping, andNinglang were selected for comparative study. Generally speaking,the socioeconomic development was obviously not in equilibriumin the four counties of Lijiang City. The most densely populatedwas Yongsheng County, with the least for Ninglang County. Thegross domestic product of Lijiang County accounted for the highestproportion of the whole city (35–50%), while the economic aggre-gate of Ninglang County was minimal (10–16%). Lijiang County wasalso the heart of the city’s tourism development, while tourismaccounts for a very small proportion of the city’s total tourism rev-enue in Yongsheng County and Huaping County. Overall, the levelsof socioeconomic development of the four counties were, LijiangCounty, Huaping County, Yongsheng County, and Ninglang Countyin descending order. The similar natural environmental basis anddifferent stages of socioeconomic development which triggers landuse change had make the study area one of the ideal targets for acomparative LUCC study.

2.2. Data source and processing

The study used the Landsat TM remote sensing image data.The detailed stripe number/date were as follows: 131041/1986-11-02, 131042/1986-11-02, 132041/1986-11-25, 131041/2006-01-25,131042/2006-01-25, and 132041/2006-02-01. Related fieldworkdata in 2000, 2001, 2002, and 2005 were adopted as auxiliary forinterpretation. After image geometry correction, this study usedsupervised classification and visual correction methods to obtainland use maps in different periods with the support of the remotesensing image processing software ERDAS 8.4 and geographic infor-mation system ARC/INFO 8.3. There were 8 kinds of land useclassified, i.e. paddy field, dryland, forest land, grassland, bare land,water body, glacier and snow, and construction land. The over-all classification accuracy in 1986 and 2006 was 83.2% and 85.9%,respectively; and the Kappa coefficient was 0.77 and 0.83, respec-tively. Then FRAGSTATS 3.3 was used to calculate the landscapemetrics and the transformation matrix of landscape componentsduring the study period. Thereafter, the study analyzed the char-acteristics of land use change. Meanwhile, according to the bandratio derived from band 3 and band 4, comparable normalized dif-ference vegetation index (NDVI) values of each pixel during thestudy period were also obtained.

3. Methods

3.1. Land use change measurement

In this study, the transformation matrix method was used toquantify transition probabilities of different land use types, andthen, a comparative analysis was undertaken. That is, for any two ofthe land use maps Ai*j and Bi*j, using the formula (1) of map algebramethod, we can obtain land use change map Ci*j from periods A toB, which reflects the change in the land use types and their spatial

distribution.

Ci∗j = Ai∗j × 10 + Bi∗j(applicable for land use types ≤ 10) (1)

Landscape ecology, which emphasizes on the interactionbetween spatial patterns and ecological process, provides effi-

J. Peng et al. / Ecological Indicators 72 (2017) 399–410 401

n of L

ceticltoimc

bvarnapolltbecmug[

3

evlE(mrpf

Fig. 1. The locatio

ient concepts and methods for ecosystem health assessment (Suot al., 2008; Turner et al., 1989, 2001). Measures of landscape pat-erns generally include three aspects: morphological features ofndividual units of the landscape, spatial configuration features ofomponents of the landscape, and diversity features of the overallandscape. In this study, to measure and analyze land use change ofhe study area, we selected landscape diversity index to measureverall diversity features; aggregation index and fragmentationndex to quantify spatial configuration features; and area-weighted

ean patch fractal dimension (AWMPFD) to indicate morphologi-al features (Peng et al., 2010).

In detail, landscape diversity reflects differences of the num-er and area percentage of each landscape component. A higheralue means more kinds of landscape components or more bal-nced area proportion of each component. Landscape aggregationeflects the agglomeration degree of different landscape compo-ents. A higher value means the landscape is constituted by a fewgglomerated large patches, with a smaller value for scattered smallatches. Landscape fragmentation reflects the interference degreef anthropogenic activities on the landscape to a certain extent. Theandscape fractal dimension reflects the complexity of patches; theower the value is, the more regular shaped patches there are, andhe greater are the patches affected by natural or human distur-ances. When the AWMPFD takes the average fractal dimension ofach patch, it is based on weighted area, and mainly reflects thehanging shape trends of large patches; thus, fractal dimension isore capable of reflecting change in landscape patterns. The val-

es of landscape diversity index and fragmentation index are bothreater than 0; the value of aggregation index is in the range of0,1], with [1,2] for AWMPFD.

.2. Indicators for ecosystem health assessment

Assessing integrated ecosystem health that couples ecological,conomic, and human health together, needs to integrate humanalues and ecological processes. Promoting health to the regionalevel means the integration of natural, social, and health sciences.cosystem health was usually measured through three aspects

Costanza, 1992; Rapport, 1989; Rapport et al., 1998): vigor, which

eant activity, metabolism, or primary productivity; organization,epresenting diversity and connectivity between ecosystem com-onents; and resilience, which referred to the capability to reboundrom perturbations. Additionally, health is not the antonym of dis-

ijiang City, China.

abilities, but an expression of carrying capacity (Rapport, 1992).Lack of ecosystem services and management would result in thedecline of ecological carrying capacity, thereby reducing healthlevel. Therefore, an integrated healthy ecosystem should high-light the contribution of ecosystem services, and could benefit theecosystem through homeostasis, the absence of disease, diversityor complexity, stability or resilience, vigor or scope, and a balancebetween ecosystem components (Gao et al., 2013).

3.2.1. Ecosystem vigorNatural ecosystem vigor is generally characterized through

vegetation productivity indicators. As vegetation production issignificantly positive correlated to the normalized difference vege-tation index (NDVI) (Phillips et al., 2008), NDVI was usually chosenas an indicator to measure the vigor of natural ecosystems. Basedon the band ratio method, NDVI is defined as the ratio of the dif-ference between the near-infrared waveband (NIR) and the visiblered band (R), which can be expressed as NDVI = (NIR-R)/(NIR + R).For Landsat TM images, R and NIR are the third and fourth bands,respectively.

3.2.2. Ecosystem organizationRegional ecosystem organization can be quantitatively eval-

uated from three aspects: landscape heterogeneity, landscapeconnectivity, and the connectivity of patches with importantecological functions, which includes such landscape metrics aslandscape diversity, landscape fractal dimension, landscape frag-mentation, landscape aggregation, forest land fragmentation, andpatch cohesion of forest land. The diversity of the landscape,which expresses the richness and area evenness of landscape patchtypes, can be used as a core index to determine landscape hetero-geneity. Because of the dominant area proportion of forest land,the fragmentation of forest land is the core index to determinepatch connectivity of important ecological functions. Thus this twoindexes were set as the highest weight. Because fragmentation isregarded as a core index to determine connectivity, the weight oflandscape fragmentation should be higher. Therefore, the index

weights in the formula were set as follows:

O = 0.35 × LH + 0.3 × LC + 0.35 × CC

= 0.25 × DI + 0.1 × FD + 0.2 × FN1 + 0.1 × RC + 0.25

×FN2 + 0.1 × COHESION

(2)

4 Indica

wsilsf

3

nrtrtntwptetdnrhrac

E

rre7

fi1lssowwcbatistec

3

tettwtCtw

02 J. Peng et al. / Ecological

here O is the index of regional ecosystem organization, LH is land-cape heterogeneity index, LC is landscape connectivity index, CCs patch connectivity index of important ecological functions, DI isandscape diversity, FD is landscape fractal dimension, FN1 is land-cape fragmentation, RC is landscape aggregation, FN2 is forest landragmentation, and COHESION is patch cohesion of forest land.

.2.3. Ecosystem elasticityA healthy ecosystem that is resistant and resilient to exter-

al disturbances can keep species composition and productivityelatively stable throughout a long time (López et al., 2013). Elas-icity, one of the important standards to assess ecosystem health,efers to the ecosystem’s ability of structure and behavioral pat-ern to rebound to the initial stage in the presence of human oratural disturbances. That is to say, elasticity can be characterizedhrough the resistance and resilience to external disturbances. Theeights of resistance and resilience were determined based on therobability of external disturbances exceeding or not exceedinghe self-adjustment ability of regional natural ecosystems. If thexternal disturbances have a high damage to natural ecosystems,he resilience should be emphasized; otherwise if the externalisturbances have not exceeded the self-adjustment ability ofatural ecosystems, the resistance is more important. Becauseegional development in Lijiang City is not fast with relatively lowuman disturbances, the weight of resistance should be higher thanesilience. Thus, the weight of resilience and resistance were sets 0.3 and 0.7, respectively. Then the regional ecosystem elasticityould be calculated as follows:

= 0.3 × Resil + 0.7 × Resist (3)

where E is the index of regional ecosystem elasticity, Resilepresents ecosystem resilience, and Resist refers to ecosystemesistance. That is to say, for each land use type in Lijiang City, thelasticity coefficient is equal to 30% of resilience coefficient plus0% of resistance coefficient (Table 1).

In the number determination of resilience and resistance coef-cients, a decimal score was used to assign the coefficients as [0.1,] with the interval of 0.1. The self-adjustment ability of the 8

and use types had been ranked. The glacier and snow which isensitive to climate change should have the lowest ability. In con-ideration of ecosystem integrity, natural ecosystems may recoverr resist more easily when experiencing external disturbances,hile human-dominated ecosystems may have a lower ability andill thus suffer greater destruction (Peng et al., 2015). Therefore,

onstruction land and dryland were determined with lower num-ers, while higher numbers for grassland and water body. Besides,ccording to the different implication of resilience and resistance,he numbers were not in consistence. The ecosystem in bare lands difficult to resist external disturbances, whereas the simple orometimes inferior ecosystem is easy to recover; and the ecosys-em in forest land with high ecosystem integrity is robust to resistxternal disturbances, whereas it needs more time to recover to theomplicate ecosystem structure and function.

.2.4. Ecosystem servicesNatural restriction and social preference are necessary condi-

ions to maintain ecosystem health (Rapport, 1995). Therefore,cosystem services determine the regional ecosystem health, andheir quantity and sustainability are key standards to assess ecosys-em health. Ecosystem services coefficients for each land use type

ere determined referring to the ratio among the average ecosys-

em services’ values of different ecosystem types calculated byostanza et al. (1997). According to the practical situation of ecosys-em services in the study area, the ecosystem services coefficientsere assigned as [0,1] taking forest land as a standard, with dryland

tors 72 (2017) 399–410

lower than the paddy field (Peng et al., 2015). As a result, the coef-ficients of ecosystem services were set as follows: forest land 1.0,paddy field 0.325, dryland 0.295, grassland 0.426, bare land 0.035,water body 0.932, glacier and snow 0.015, and construction land0.015.

However, the ecosystem services coefficients only consideredfunctional differences and the area ratio of different ecosystems,and did not involve the spatial neighboring interaction effect ofvarious kinds of ecosystems. In a view of ecosystem services flow,regional ecosystem services can be delivered to the spatial neigh-boring area (Bagstad et al., 2013). For example, the forest ecosystemis beneficial for the services provision of adjacent ecosystems, whilethe construction land has an adverse effect. To interpret the eco-logical connectivity between different land use types, an affinitymatrix has been proposed with assigned values in 10 divisions from0 to 1 in the Barcelona Metropolitan Area (Marulli and Mallarach,2005). However, because some ecological interactions among var-ious land use types would be negative, the 10 divisions from −5 to5 is regarded to be more reasonable. The human-dominated landuse types usually deliver a negative effect on the spatial neighbor-ing area, while the adjacency of natural ecosystems may benefiteach other. In particular, the ecosystem in forest land and grass-land usually deliver more ecosystem services to the surroundingsthan paddy field and dryland because of their multifunctionality(Palacios-Agundez et al., 2015; Santika et al., 2015). Therefore, thecoefficients of spatial neighboring effect on ecosystem services ofdifferent land use types were assigned with a range of [−5%, 5%](Table 2). The ecosystem services of image pixels in the study areawere determined by their land use types and land use type of theirfour adjoining points, with regional ecosystem services quantifiedby the average value of all the pixels, which could be expressed asfollows:

ESp = ESc × (100 + SNc)/100 (4)

where ESp is ecosystem services of the pixel, ESc is the coefficientof ecosystem services for specific land use type, and SNc is thecoefficient of spatial neighboring effect.

3.3. Ecosystem health assessment

As mentioned above, the three aspects of vigor, organization,and elasticity can provide comprehensive characterization of anecosystem’s ability to maintain its structure and function steadily,which can be defined as ecosystem physical health. In accordancewith the definition of the ecosystem health index proposed byCostanza (1992), and considering the comparability of the physicalhealth index and ecosystem services index, this study modified theproduct of vigor, organization, and elasticity, to the cube root of theproduct (Lu and Li, 2003; Suo et al., 2008):

PH = 3√

V × O × E (5)

where PH is the index of regional ecosystem physical health, V isregional ecosystem vigor, O is regional ecosystem organization, andE is the elasticity of regional ecosystems.

The integrated health level of regional ecosystems depends onthe ecosystem’s own health status. Comparing with the physicalhealth focusing on the status of ecosystem itself, the judgementcriteria of regional integrated ecosystem health are whether thesupply of ecosystem services can meet human demands. Ecosys-tems with sufficient biodiversity to maintain multifunctionality can

provide humans with a broad range of health through sustainableprovision of ecosystem services (Speldewinde et al., 2015). Even ifecosystem’s own physical health is good enough, if it cannot pro-vide humans with sufficient ecosystem services, it is still unhealthyfor humans. Therefore, regional integrated ecosystem health equals

J. Peng et al. / Ecological Indicators 72 (2017) 399–410 403

Table 1Ecological elasticity coefficient of land use types in Lijiang City.

Land use type Paddy field Dryland Forest land Grassland Bare land Water body Glacier and snow Construction land

Resilience coefficient 0.3 0.4 0.5 0.8 1.0 0.7 0.1 0.2Resistance coefficient 0.6 0.5 1.0 0.7 0.2 0.8 0.1 0.3Elasticity coefficient 0.51 0.47 0.85 0.73 0.44 0.77 0.10 0.27

Table 2Spatial neighboring effect coefficient on ecosystem services of land use types in Lijiang City.

Adjacent pixel type Target pixel type

Paddy field Dryland Forest land Grassland Bare land Water body Glacier and snow Construction land

Paddy field +4 +2 −1 −1 −1 −5 −1 +1Dryland +1 +2 −1 −1 −1 −4 −1 +1Forest land +5 +4 +5 +5 +4 +5 +3 +4Grassland +2 +2 +3 +2 +3 +3 +2 +2Bare land −3 −3 −1 −3 +1 −4 −1 −1Water body +4 +2 +5 +4 +4 +5 +1 +4

* adjac

ts

H

wE

mwmmpitt

D

D

4

4

aalaatsmwtidfaCt

Glacier and snow −5 −5 −4 −1Construction land +2 +2 −2 −3

Numerical value shows the variation percentage of target pixel type resulted from

o the square root of the product of physical health and ecosystemervices:

=√

PH × ESV (6)

here H is the index of regional integrated ecosystem health, andSV is regional ecosystem services.

In the standardization of original indexes mentioned above,aximum difference normalization method with the range of [0,1]as used. For the temporal change assessment, a distinct judge-ent of data sequence was carried out using a standard scoreethod. It was assumed that data sequence of change value of

hysical health index or integrated health index for all the assess-ng towns was Xi (i = 1,2,3, . . . n) with n for the number of assessingowns, the arithmetic mean of the data sequence was X̄ , and then,he distinct change of a township can be judged as follows:

istinctly positive change :Xi − X̄√(Xi − X̄

)2⁄n

> 1 (7)

istinctly negative change :Xi − X̄√(Xi − X̄

)2⁄n

< −1 (8)

. Results

.1. Land use change

Affected by the similar natural environment, the land use char-cteristics of the four counties were relatively similar. Forest landccounted for the absolute proportion of more than 60%, with grass-and and dryland each for a considerable proportion of about 10%,nd paddy fields for about 4%. Both the proportions of bare landnd water body were low of about 1%, with the lowest propor-ion of less than 1% for construction land. The similarity of land usetructure indicated that the differences of socioeconomic develop-ent in the four counties of Lijiang City were not large enough,hich had not resulted in the discriminating land use transforma-

ion. This might because each county’s economy was too low tonduce significant changes of land use. With regard to the structural

ifferences of land use for the four counties within the study period,orest land accounted for a little higher proportion in Lijiang Countynd Huaping County, dryland in Yongsheng County and Ninglangounty accounted for a slightly higher proportion than the otherwo counties, paddy field in Yongsheng County accounted for a

−1 −2 +1 −1−2 −5 −3 +1

ent pixel type.

slightly higher proportion with Ninglang County for a little lowerproportion, and water body in Yongsheng County accounted for ahigher proportion than the other counties.

There was no significant change in the land use structure in thefour counties during 1986–2006 (Fig. 2), but the area of each landuse type presented different changes in each county. In details,the more prominent change was that glacier and snow had dis-appeared in Yongsheng County and Huaping County, and therehad been a general increase of construction land in each county.In Lijiang County, forest land increased and grassland decreasedsignificantly; Yongsheng’s forest land and grassland decreased sig-nificantly while dryland and bare land increased; and forest landdecreased while dryland and grassland increased slightly in Huap-ing County and Ninglang County. In general, for the transformationof each county’s land use quantitative structure, the biggest changeoccurred in Lijiang County, followed by Yongsheng County, andthe minimum change occurred in Huaping County and NinglangCounty, which were associated closely with total economy of eachcounty.

There were large differences in the extent of landscape diver-sity and its variation in the study area from 1986 to 2006 (Table 3).Yongsheng’s landscape diversity index was significantly higherthan those of the other three counties, indicating it had a morebalanced area proportion of land use types. The landscape diversityindex of Lijiang County had a larger decline in the study period, indi-cating a significant increase in the differences of area proportion ofland use types. This was mainly caused by the increase of forestland, which accounted for the absolute area ratio of the county.On the contrary, Yongsheng’s landscape diversity index increasedsignificantly while the indexes for both Ninglang County and Huap-ing County increased slightly. In particular, as glacier and snowdisappeared completely at the end of the study period, the num-ber of land use types in Yongsheng County and Huaping Countydecreased, indicating the land use structure gradually balanced.The proportional decline of forest land was the direct reason forthe increase of landscape diversity index in the three counties.

The indexes of landscape aggregation of the four counties wereat medium levels (Table 3), showing the existence of a few largepatches of forest land, paddy fields, and glacier and snow. The land-scape aggregation index increased in Lijiang County and decreased

in the other three counties, showing that the influencing degreeof a few large patches increased over time in Liang County, buttended to weaken in the other three counties. The overall land-scape fragmentation of the four counties was relatively low in the

404 J. Peng et al. / Ecological Indicators 72 (2017) 399–410

Fig. 2. Land use patterns in Lijiang City during 1986–2006.

Table 3Landscape metrics change in each county of Lijiang City from 1986 to 2006.

Landscape diversity Landscape aggregation Landscape fragmentation

1986 2006 1986 2006 1986 2006

0000

sfdtitita

lowttwsamIsgdrsla

4

tv

Lijiang County 0.9778 0.8720

Yongsheng County 1.0805 1.1706

Ninglang County 0.9391 0.9648

Huaping County 0.8574 0.9139

tudy period. However, the changing trend was dissimilar in theour counties; the landscape fragmentation index of Lijiang Countyeclined slightly, reflecting that the orderliness of human activi-ies in Lijiang County was enhanced. The landscape fragmentationndex tended to increase in the other three counties, among whichhe maximal occurred in Yongsheng County, showing that thempact of anthropogenic disturbance intensity to landscape evolu-ion increased significantly in the counties of Yongsheng, Ninglang,nd Huaping.

As shown in Table 4, there was similar AWMPFD between eachand use type in the four counties on the whole, with the AWMPFDf forest land highest in each county, showing that its patch shapeas relatively complex with low regularity, and forest land was

he least one of land use types affected by the anthropogenic dis-urbances. The AWMPFD of dryland and grassland ranked second,hile that of bare land, paddy field, water body, and glacier and

now were relatively low. Construction land was the greatest oneffected by anthropogenic activity, which had the simplest andost regular geometrical patch shape and the minimum AWMPFD.

t should also be noted that spatial distribution of glacier andnow was strictly restricted by climatic factors, being distributedenerally in high mountains and extending from the tops of hillsown to nearby snow lines. Thus, glacier and snow patches hadelatively simple and single geometrical shape, being mostly arc-haped, and its landscape fractal dimension was correspondinglyower, although glacier and snow was influenced by the minimalnthropogenic disturbance in all land use types.

.2. Change of ecosystem health indicators

Classified with natural breaks method, the change of ecosys-em vigor level was most obvious (Fig. 3). In 1986, the highestigor was in Jin’an Town in Lijiang County (1.00), with the low-

.6697 0.7100 0.0664 0.0566

.6378 0.5712 0.0685 0.0904

.6643 0.6476 0.0858 0.0931

.7011 0.6482 0.0703 0.0895

est in Dayan Town in Lijiang County (0.23). In 2006, the highestwas in Shitou Town in Lijiang County (0.71), and the lowest wasremained in Dayan Town in Lijiang County (0.00). The vigor ofDayan Town remained at the minimum with substantial decline,which was mainly due to the town being the seat of local govern-ment at both county and municipal levels, with the largest urbanbuilt-up area among all the towns. Urban built-up areas in Dayantown had expanded quickly during the study period, leading toa relatively low ratio of natural vegetation area which was occu-pied by construction land continually. In contrast, the high vigortowns were far from the economic centers of the four counties,with traffic isolation and sparse population of townships. During1986–2006, due to the enhancement of human activities, someforest land and grassland were transformed to arable land and con-struction land. Thus the vast majority of towns showed significantlylower vigor, indicating decreased regional ecosystem vitality. Thelargest decline occurred in Xinyingpan Town of Ninglang County,and the smallest decline occurred in Chenghai Town of YongshengCounty. During 1986–2006, the relative differences of vigor at townscale showed a slight decreasing trend, with the difference betweenthe maximum and minimum vigor values decreased from 0.77 to0.71.

The ecosystem organization level changed mainly in the middleand south part of the study area (Fig. 3). In 1986, Judian Town ofLijiang County had the highest ecosystem organization (0.63), andPaomaping Town of Ninglang County had the lowest (0.19). In 2006,Shitou Town of Lijiang County had the highest (1.00), while the low-est was remained Paomaping Town (0.00). During 1986–2006, the

ecosystem organization showed an increase in Lijiang County, withdownward trends in the other counties. Among them, the largestincrease occurred in Shitou Town of Lijiang County, reaching 0.64,with Songping Town of Yongsheng County for the largest declineof −0.48. During 1986–2006, the relative differences of ecosystem

J. Peng et al. / Ecological Indicators 72 (2017) 399–410 405

Table 4AWMPFD Change of land use types in Lijiang City from 1986 to 2006.

Paddy field Dryland Forest land Grassland Bare land Water body Glacier and snow Construction land

Lijiang County 1986 1.1438 1.2109 1.4224 1.2688 1.1786 1.2223 1.2841 1.07852006 1.1667 1.2442 1.3903 1.2391 1.1928 1.2013 1.1848 1.1380

Yongsheng County 1986 1.1452 1.2762 1.4151 1.2182 1.1542 1.1106 1.1494 1.08172006 1.1591 1.2689 1.3879 1.2236 1.1849 1.2456 / 1.1086

Ninglang County 1986 1.1522 1.2224 1.4346 1.2273 1.1835 1.1143 1.1500 1.09762006 1.1703 1.2635 1.4289 1.2595 1.1615 1.1329 1.1397 1.0587

Huaping County 1986 1.1867 1.2106 1.3938 1.2083 1.1779 1.1805 1.1396 1.05762006 1.1747 1.2303 1.4002 1.2173 1.1779 1.1664 / 1.0981

s for e

oti

fethfrmsttttwettTaDftcpf

d

Fig. 3. Change of ecosystem health indicator

rganization at town scale increased slightly in the four counties;he difference between the maximum and minimum of the indexncreased from 0.44 to 1.00.

The ecosystem elasticity level changed inconsistent in the dif-erent parts of the study area (Fig. 3). In 1986, the highest ecosystemlasticity was in Songping Town of Yongsheng County (0.97) andhe lowest was in Dayan Town of Lijiang County (0.43). In 2006, theighest was in Liming Town of Lijiang County (1.00) with the lowest

or Dayan Town too (0.00). The ecosystem elasticity of Dayan Townemained at the minimum with substantial decline, which wasainly because the town had the highest area proportion of con-

truction land with sustained urban growth. As shown in Table 1,he elasticity coefficient of construction land was just higher thanhat of glacier and snow in all land use types, and was far lowerhan the others. Furthermore, in the towns with the higher ecosys-em elasticity, the intensity of human activities was low, togetherith the high area proportion of forest land. During 1986–2006, the

cosystem elasticity of the towns in Lijiang County showed a risingrend, in contrast with the downward trend of ecosystem elas-icity for most towns in Ninglang County and Yongsheng County.he largest increase occurred in Longpan Town of Lijiang County,nd the largest decline occurred in Dayan Town of Lijiang County.uring 1986–2006, ecosystem elasticity showed not far relative dif-

erences. The difference between the maximum and minimum ofhe index increased from 0.54 to 1.00. In contrast with the elasticity

oefficients of all land use types, it could be inferred that the arearoportion of forest land and construction land was an importantactor affecting the ecosystem elasticity of towns.

The ecosystem services level increased in the west part andecreased in the east part of the study area (Fig. 3). In 1986, the

ach town in Lijiang City from 1986 to 2006.

highest ecosystem services was in Songping Town of YongshengCounty (0.96), and the lowest was in Dayan Town of Lijiang County(0.24). In 2006, the highest was in Liming Town of Lijiang County(1.00), and the lowest was still in Dayan Town (0.00). The ecosystemservices of Dayan Town remained at the minimum with substan-tial decline, which was also mainly due to the area proportion ofconstruction land being the highest with continue increase, consid-ering that ecosystem services coefficient of construction land wasthe lowest in all land use types. In contrast, there was the highestarea ratio of forest land in the towns with the highest ecosystemservices. During 1986–2006, the ecosystem services of the townsin Lijiang County showed a rising trend. On the contrary, therewere decline trends of ecosystem services for most of the townsof the other three counties. The largest increase occurred in Long-pan Town of Lijiang County, reaching 0.20, and the largest declineoccurred in Dayan Town of Lijiang County, reaching −0.24. During1986–2006, ecosystem services showed moderately relative differ-ences at town scale; and the difference between the maximum andminimum of the index increased from 0.72 to 1.00. In contrast withthe ecosystem services coefficients of all land use types, it couldbe inferred that the area proportion of forest land and constructionland was also an important factor affecting the ecosystem servicesof towns.

4.3. Ecosystem health change

The change of ecosystem physical health was more obvious inthe west part than in the east part of the study area (Fig. 4). In1986, the highest town of the ecosystem physical health indexwas Songping (0.99) of Yongsheng County, and the lowest was

406 J. Peng et al. / Ecological Indicators 72 (2017) 399–410

for e

Deatmittosfhtv

otcppetLCaMLL

ohCHpiwipwis

csT

Fig. 4. Change of ecosystem physical health

ayan (0.31) of Lijiang County. In 2006, the highest town of thecosystem physical health index was Shitou (1.00) of Lijiang Countynd Dayan remained the lowest town (0.00). During 1986–2006,he ecosystem physical health of towns in Lijiang County showed

ainly a descending trend, but there were also nine towns increas-ng slightly. The ecosystem physical health index of all the towns inhe other three counties showed a downward trend. Among them,he growth of Shitou Town in Lijiang County reached a maximumf 0.25, and the largest decline occurred in Songping Town of Yong-heng County, reaching −0.46. In the study period, each town of theour counties scored a little differently in the ecosystem physicalealth but their differences tended to increase slightly. The rela-ive difference at town scale between the maximum and minimumalues increased from 0.68 to 1.00.

On account of township differences, the average change ratef the ecosystem vigor was significantly higher than the ecosys-em organization and the elasticity, and therefore, ecosystem vigorould be regarded as the major influencing factor for ecosystemhysical health in the four counties. The change of ecosystemhysical health index showed that for the vast majority of towns,cosystem physical health did not change distinctly. Only eightowns decreased distinctly, i.e. Dayan Town and Jinjiang Town ofijiang County, Xingquan Town and Yongxing Town of Huapingounty, Guanghua Town and Songping Town of Yongsheng County,nd Chanzhanhe Town and Jinmian Town of Ningliang County.eanwhile, the ecosystem physical health in such nine towns of

ijiang County increased distinctly as Shigu, Daju, Jiuhe, Liming,ongpan, Ludian, Shitou, Tacheng, and Tai’an.

Among the four counties, Huaping County had the highest indexf integrated ecosystem health in 1986. The integrated ecosystemealth of the other three counties were similar, with Yongshengounty for the lowest. In 2006, the highest was Lijiang County, withuaping County and Ninglang County taking jointly the secondlace, and Yongsheng County was the lowest. During 1986–2006,

ntegrated ecosystem health of Lijiang County increased slightly,hile those of the other three counties decreased. Comparing the

ntegrated ecosystem health, ecosystem services, and ecosystemhysical health, it could be found that integrated ecosystem healthas influenced by its two components in roughly the same amount

n each county, and the influence of the ecosystem services waslightly higher than that of ecosystem physical health.

The distinctness judgement of integrated ecosystem healthhange at town scale showed that the vast majority of town-hips’ integrated ecosystem health did not change distinctly (Fig. 5).here were also nine towns of Lijiang County showing distinct

ach town in Lijiang City from 1986 to 2006.

increase, i.e. Shigu Town, Daju Town, Jiuhe Town, Liming Town,Longpan Town, Ludian Town, Shitou Town, Tacheng Town, andTai’an Town. On the contrary, only eight towns experienced dis-tinct decrease, i.e. Dayan Town, Jin’an Town and Jinjiang Town ofLijiang County, Yongxing Town of Huaping County, Guanghua Townand Songping Town of Yongsheng County, and Chanzhanhe Townand Jinmian Town of Ningliang County. Comparing the distinctchanges of ecosystem physical health and integrated ecosystemhealth, it could be found that the towns of distinct change of inte-grated ecosystem health were highly similar but not completely inconsistence with that of ecosystem physical health.

In 1986, the highest town of integrated ecosystem health amongthe four counties was Songping (1.00) of Yongsheng County, andthe lowest was Dayan (0.27) of Lijiang County. In 2006, the highesttown was Shitou (1.00) of Lijiang County with Dayan (0.00) stillfor the lowest. From 1986–2006, the integrated ecosystem healthof towns in Lijiang County mainly showed a rising tendency, andonly eight towns decreased slightly. The vast majority of towns inthe other three counties showed decreasing trends, and only fourtowns increased slightly, i.e. Yongbei Town, Chenghai Town, andQina Town of Yongsheng County, and Yongning Town of NingliangCounty. Among all the towns of the study area, Jiuhe Town of LijiangCounty had the largest increasing rate of 23.07%, with Dayan Townof Lijiang County for the largest decreasing rate of −100%. Duringthe study period, the relative differences of integrated ecosystemhealth at town scale increased obviously. The relative differencebetween the maximum and minimum values increased from 0.73to 1.00. Comparing the spatiotemporal differences of ecosystemphysical health and integrated ecosystem health at town scale, thesynchronicity of the two indicators could be found, which referredto a distinct influence of ecosystem physical health on integratedecosystem health, although the township differences of ecosystemservices were larger than that of ecosystem physical health.

5. Discussion

5.1. Indicators and models for ecosystem health assessment

Ecosystem health is a system diagnostic for ecosystem statecharacteristics. Health is a relative concept based on subjective

human expectations, while there is no such thing as absolute healthstandard. Thus, regional ecosystem health assessment should focuson exploring temporal dynamics and spatial differences of regionalecosystem health, rather than artificial judgement about whetherin a specific region the natural ecosystems or ecosystems’ mosaic

J. Peng et al. / Ecological Indicators 72 (2017) 399–410 407

h for e

araa

freb(amaeLatfctqe“o

ruaqp2

Fig. 5. Change of integrated ecosystem healt

re healthy at a particular time point. Therefore, it should be moreeasonable and practical to highlight whether regional ecosystemsre healthier or weaker than to judge whether regional ecosystemsre healthy or not.

A new idea from this study is to assess regional ecosystem healthrom the perspective of land use change. In a view of regional envi-onmental change, land use change can effectively integrate naturalcological processes with socioeconomic development, and hasecome a current research frontier of global environmental changePeng et al., 2008). LUCC has provided an important integratedpproach for geography to analyzing the human-nature couplingechanism, exploring the process of global environmental change,

nd promoting regional sustainable development. In particular, thecological effects have been regarded as one of the key issues ofUCC studies. Focusing on global scale together with practicingt regional scale, is an analytical paradigm of geography to studyraditional human–land relationship. As an important basis andundamental prerequisite of ecological security, ecosystem healthan represent the integrated functional status of natural ecosys-ems, and thus can be used to quantify regional eco-environmentaluality and associated ecological effects of LUCC. Therefore, thecosystem health assessment focusing on the “change” instead ofthreshold” will contribute to highlight the essential characteristicsf regional LUCC.

Based on the classical framework of vigor–organization–esilience, integrated indicators and a fuzzy comprehensive eval-

ation model are often used to conduct regional health statessessment (Su et al., 2010; Yu et al., 2013). Although some newuantifications have been introduced such as the models of setair analysis and emergy indicators (Su et al., 2009; Su and Fath,012), further understanding of classical framework is still needed

ach town in Lijiang City from 1986 to 2006.

to interpret the inner relationship among the dimensions of ecosys-tem health at regional scale. In this study, representative indicatorsof vigor–organization–elasticity incorporated with the index ofecosystem services revised through a spatial adjacency matrix,helped to form a new framework of vigor-organization-elasticity-services for regional ecosystem health assessment, with moreeffective spatial explicit quantification.

The quantification of each indicator is not arbitrary. To measurethe vigor indicator, primary productivity is the most commonlyused index (Costanza and Michael, 1999; Boesch, 2000), whichcan be estimated by NDVI (Li et al., 2014). At ecosystem or land-scape scale, the organization is defined as the aggregation of thestructural and functional aspects of landscape (Müller, 2005), andthis attribute of landscape organization with structural and func-tional interactions can be simplified by landscape metrics in thediscipline of landscape ecology (Turner, 2005; Turner et al., 1989,2001). Land use is an important aspect in the concept of ecosystemresilience (Colding, 2007; Foster et al., 2003), and the separationof resilience and resistance can make the assessment of ecosys-tem health more reasonable (Whitford et al., 1999). In terms of theecological connectivity index built on land use types (Marulli andMallarach, 2005), the ecosystem services flow can be determinedin a landscape connectivity view, which considers that the ecosys-tems or landscape components can be benefited from the spatialadjacent ecosystems or landscape components (Peng et al., 2015).Although the coefficients and weights were set to be fit for the land-

scape characteristics in Lijiang City, the establishment of indicatorsis reasonably conveying the initial object to integrate the land useindices for ecosystem health assessment. Therefore, the model ofintegrated ecosystem health assessment with the practical quan-tifications of the four aspects of indicators is the main contribution

4 Indica

ooo

evtohdohwicitTmfiisotwsamrifpes

5

aohlhses(sh

dstemti1hopccccn

08 J. Peng et al. / Ecological

f this study to land use change research, which can be applied tother cities or countries after an adjustment of parameter settingf the coefficients and weights.

Moreover, the weight determination for the parameters ofcosystem organization, ecosystem elasticity and ecosystem ser-ices should be highlighted. The number of weight can affecthe results, which is consequential and the exact differencef even 0.1 precision cannot be ignored. However, ecosystemealth has a definitiveness framework with its calculation in non-efinitiveness, neither with weight determination. The importancef each index could make sense, whereas the importance degreeas indeterminacies. The indeterminacies cannot be removed in theeighting solution as this kind of weight scoring for an integrated

ndicator is hard to be validated. Therefore, the weights in this arti-le is perhaps of individual uniqueness, and should be revised whenntroducing the framework to other regions in the world. Facinghese limits, a label of “wicked problem” should be introduced.his paradigm highlights that numerous problems in planning,anagement, and policy-making are by “nature wicked” with

ve properties: irreversible consequentiality, non-definitivenessn problem solution, indeterminacy in problem formulation, non-olubility, and individual uniqueness (Xiang, 2013). For the weightsf parameters in assessing ecosystem organization, ecosystem elas-icity and ecosystem services, these five properties all appeared,hich meant that the weight determination with accurate preci-

ion might be inexistence. Nevertheless, besides being awarenessnd take acceptance to “wicked problem”, researchers can alsoake adaptation to it (Xiang, 2013). In the article, the logistic

anking orders of parameter importance have been described, andmportance of the weights has reasonable interpretation. There-ore, although the weight determination is not perfect, the datarocessing is far from quantified in arbitrary. The researchers’ fullxperience and perception of this local study area could makeense.

.2. Measurement goals and future research directions

As an overall guideline, ecosystem health goes beyond the char-cterization of ecological pressure levels, as the latter is concernednly with its own specific ecological issues, whereas ecosystemealth considers a wide range of socioeconomic, public health,

egal, and policy fields (Rapport and Singh, 2006). Consideringuman center doctrine related aspects of ecosystem propertiesuch as beliefs, morals, values, and ethics, the methodology ofcosystem assessment for integrated environmental managementhould never be unique in the multiple connotation of healthKapustka and Landis, 1998). Therefore, the determination of mea-urement goals should be the prerequisite work in ecosystemealth assessment.

The measurement goals of ecosystem health can be roughlyivided into two parts: one is to describe the integrity of ecosystemtructure and function under the pressure of human activity; andhe other is to understand the influence on human health from thevolution of ecosystem structure and function. After a bright start,any ecosystem health studies tended to support the human cen-

er doctrine, which set human health state as the measurement goaln assessing specific ecosystem structure and function (Boetzkes,998; Light, 1998). However, the research issue derived from thisuman center doctrine is rooted in public health rather than ecol-gy. In the main focus of ecological influences, this public healtherspective cannot clarify the ecological importance and natural

onstraints, especially when countering with the question of whatauses environmental problems, namely ecological response. Inomparison, the ecosystem health indicator rooted in ecology isapable of characterizing the overall ecological function state ofatural ecosystems, and is the core content and fundamental way

tors 72 (2017) 399–410

to assess ecosystem quality comprehensively (Shen et al., 2016;Tecchio et al., 2015). Therefore, promoting ecological connotationsof ecosystem health in theory and methodology dimensions shouldbe a rewarding task in the disciplinary affiliation of ecology.

Without doubt, no model can depict all goals, but the practi-cal evaluation of ecosystem health is necessary to pay attentionsto the key indicator dimensions (Rapport, 1995). As a complexconcept, research on ecosystem health is not limited to understand-ing from naturalistic perspectives, but also reducing the negativeimpact of population growth and consumption on ecosystemsunder the background of inevitable ecosystem change. Ecosystemhealth should be realized based on biological protection, togetherwith restricting human activities, and emphasizing environmentalsustainability (Belaoussoff and Kevan, 1998; Kevan et al., 1997a,1997b). Precautionary measures that can ensure ecological sus-tainability would be helpful to guarantee ecosystem services fromlong-term perspectives, which are beneficial to health (Hales et al.,2004). Therefore, beyond the traditional assessment framework ofvigor, organization, and resilience which indicated a naturalisticecosystem quality, the core of health assessment should be high-lighted both the biophysical processes and the ecological servicesfor humanity.

Currently, global ecological problems, such as forest degrada-tion, watershed pollution, and fish reduction are mainly resolvedby the help of specific disciplines. From a multi-disciplinaryintegration perspective of integrated ecosystem health, the eco-logical problems of environmental degradation, species protection,and human existence can be answered in a comprehensive way(Bebianno et al., 2015). The land use changes of particular land-scape is such a media that endow the ecosystem health study anintegrated view (Walker et al., 2001). However, insufficient atten-tions are paid to conditions, processes, and causal relationships ofecosystem change resulting from land use change. The influenc-ing mechanisms of land use change on ecosystem health is stilldifficult to identify. Although this study aimed to investigate thehealth change of natural ecosystems, it highlighted that the inter-action of human socioeconomic systems and natural ecosystemsremained to be analyzed more deeply. Important directions forfuture research included the integration of socioeconomic indi-cators, the identification of driving factors that affected healthand corresponding mechanisms, and possible regulatory measures.Moreover, the links between socioeconomic drivers, urbanizationpatterns, and their influences on the ecosystem functions and thusecosystem health had to be deeply investigated.

6. Conclusion

Assessing regional ecosystem health is fundamental for explor-ing the ecological effects generated by global environmentalchange, and is important for analyzing comprehensively thehuman-nature coupling mechanism. However, existing studiesseldom considered the impact of land use pattern change andecosystem services on regional ecosystem health. Quantitativeresearch on the ecosystem health response to regional land usechange could contribute to comprehensively judging the macro-ecological effects of land use change, which is an importantintegrated approach of LUCC study.

This study analyzed ecosystem health response to land usechange in the four counties of Lijiang City in China during1986–2006. The extent of land use change was the highest in

Ninglang County of 30.22% changed, with the lowest in LijiangCounty for 24.37% changed. Landscape diversity in YongshengCounty was the highest, and landscape diversity and landscapefragmentation only decreased in Lijiang County. The integratedecosystem health was slightly better than ecosystem physical

Indica

hstitobheawcaraw

A

S

R

B

B

B

B

B

B

B

C

C

C

C

C

C

C

D

F

G

H

I

K

K

K

J. Peng et al. / Ecological

ealth in the four counties. Moreover, ecosystem physical healthhowed a downward trend in the four counties, with a similarendency for the integrated ecosystem health except the slightlyncrease in Lijiang County. Ecosystem vigor was the dominant fac-or that resulted in the spatial difference and temporal changef ecosystem physical health at town scale. The synchronicityetween ecosystem physical health and integrated ecosystemealth in spatiotemporal differences resulted in a distinct influ-nce of ecosystem physical health on integrated ecosystem health,lthough the relative differences of regional ecosystem servicesere larger than that of ecosystem physical health. It is the main

ontribution of this study to conduct regional ecosystem healthssessment in the perspective of LUCC, which will help to the rep-esentation of health indicators in the methodology, and provide

quantitative approach for assessing ecological effects associatedith regional land use change.

cknowledgment

This research was financially supported by the National Naturalcience Foundation of China (No. 41322004).

eferences

agstad, K.J., Johnson, G.W., Voigt, B., Villa, F., 2013. Spatial dynamics of ecosystemservice flows: A comprehensive approach to quantifying actual services.Ecosyst. Serv. 4, 117–125.

ebianno, M.J., Pereira, C.G., Rey, F., Cravo, A., Duarte, D., D’Errico, G., Regoli, F.,2015. Integrated approach to assess ecosystem health in harbor areas. Sci.Total Environ. 514, 92–107.

elaoussoff, S., Kevan, P.G., 1998. Toward an ecological approach for theassessment of ecosystem health. Ecosyst. Health 4 (1), 4–8.

elaoussoff, S., Kevan, P.G., 2003. Are there ecological foundations for ecosystemhealth? Environmentalist 23 (3), 255–263.

oesch, D.F., 2000. Measuring the health of the Chesapeake Bay: towardintegration and prediction. Environ. Res. 82 (2), 134–142.

oetzkes, E., 1998. Gender, risk and scientific proceduralism. Ecosyst. Health 4 (3),162–169.

urger, J., Gochfeld, M., 2001. On developing bioindicators for human andecological health. Environ. Monit. Assess. 66 (1), 23–46.

airns, J., Mccormick, P.V., Niederlehner, B.R., 1993. A proposed framework fordeveloping indicators of ecosystem health. Hydrobiologia 263 (1), 1–44.

alow, P., 1992. Can ecosystems be healthy? Critical considerations of concepts. J.Aquat. Ecosyst. Health 1 (1), 1–15.

olding, J., 2007. ‘Ecological land-use complementation’ for building resilience inurban ecosystems. Landsc. Urban Plan. 81 (1), 46–55.

ostanza, R., Michael, M., 1999. What is a healthy ecosystem? Aquat. Ecol. 33 (1),105–115.

ostanza, R., d’Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,Naeem, S., ONeill, R.V., Paruelo, J., 1997. The value of the world’s ecosystemservices and natural capital. Nature 387 (6630), 253–260.

ostanza, R., 1992. Toward an operational definition of ecosystem health. In:Costanza, R., Norton, B.G., Haskell, B.D. (Eds.), Ecosystem Health: New Goals forEnvironmental Management. Island Press, Washington, DC, pp. 239–256.

ostanza, R., 2012. Ecosystem health and ecological engineering. Ecol. Eng. 45 (8),24–29.

ernbach, J.C., Mintz, J.A., 2011. Environmental laws and sustainability: anintroduction. Sustainability 3 (3), 531–540.

oster, D., Swanson, F., Aber, J., Burke, I., Brokaw, N., Tilman, D., Knapp, A., 2003.The importance of land-use legacies to ecology and conservation. Bioscience53 (1), 77–88.

ao, Y., Yu, G.R., He, N.P., 2013. Equilibration of the terrestrial water, nitrogen, andcarbon cycles: advocating a health threshold for carbon storage. Ecol. Eng. 57(4), 366–374.

ales, S., Butler, C., Woodward, A., Corvalan, C., 2004. Health aspects of themillennium ecosystem assessment. EcoHealth 1 (2), 124–128.

mhoff, M.L., Lawrence, W.T., Elvidge, C.D., Paul, T., Levine, E., Privalsky, M.V.,Brown, V., 1997. Using nighttime DMSP/OLS images of city lights to estimatethe impact of urban land use on soil resources in the US. Remote Sens. Environ.59 (1), 105–117.

apustka, L.A., Landis, W.G., 1998. Ecology: the science versus the myth. Hum.

Ecol. Risk Assess. 4 (4), 829–838.

evan, P.G., Greco, C.F., Belaoussoff, S., 1997a. Log-normality of biodiversity andabundance in diagnosis and measuring of ecosystem health: pesticide stresson pollinators on blueberry heaths. J. Appl. Ecol. 34 (5), 1122–1136.

evan, P.G., Thomas, V.G., Belaoussoff, S., 1997b. AgrECOLture: defining theecology in agriculture. J. Sustain. Agric. 9 (2), 109–129.

tors 72 (2017) 399–410 409

López, D.R., Brizuela, M.A., Willems, P., Aguiar, M.R., Siffredi, G., Bran, D., 2013.Linking ecosystem resistance, resilience, and stability in steppes of NorthPatagonia. Ecol. Indic. 24 (1), 1–11.

Lackey, R.T., 2001. Values, policy, and ecosystem health. Bioscience 51 (6),437–443.

Lavigne, F., Gunnell, Y., 2006. Land cover change and abrupt environmentalimpacts on Javan volcanoes, Indonesia: a long-term perspective on recentevents. Reg. Environ. Change 6 (1-2), 86–100.

Li, R.Q., Dong, M., Cui, J.Y., Zhang, L.L., Cui, Q.G., He, W.M., 2007. Quantification ofthe impact of land use changes on ecosystem services: a case study in Pingbiancounty, China. Environ. Monit. Assess. 128 (1-3), 503–510.

Li, Z.Q., Xu, D.D., Guo, X.L., 2014. Remote sensing of ecosystem health:opportunities, challenges, and future perspectives. Sensors 14, 21117–21139.

Light, A., 1998. Environmental ethics and environmental risk assessment:expanding the scope of ecosystem health. Ecosyst. Health 4 (3), 147–151.

Lu, F., Li, Z.Z., 2003. A model of ecosystem health and its application. Ecol. Model.170 (1), 55–59.

Müller, F., 2005. Indicating ecosystem and landscape organization. Ecol. Indic. 5(4), 280–294.

Marulli, J., Mallarach, J.M., 2005. A GIS methodology for assessing ecologicalconnectivity: application to the Barcelona Metropolitan Area. Landsc. UrbanPlan. 71 (2), 243–262.

Nainggolan, D., Termansen, M., Reed, M.S., Cebollero, E.D., Hubacek, K., 2013.Farmer typology, future scenarios and the implications for ecosystem serviceprovision: a case study from south-eastern Spain. Reg. Environ. Change 13 (3),601–614.

Napton, D.E., Auch, R.F., Headley, R., Taylor, J.L., 2010. Land changes and theirdriving forces in the Southeastern United States. Reg. Environ. Change 10 (1),37–53.

Niedertscheider, M., Gingrich, S., Erb, K.H., 2012. Changes in land use in SouthAfrica between 1961 and 2006: an integrated socio-ecological analysis basedon the human appropriation of net primary production framework. Reg.Environ. Change 12 (4), 715–727.

Palacios-Agundez, I., Onaindia, M., Barraqueta, P., Madariaga, I., 2015. Provisioningecosystem services supply and demand: the role of landscape management toreinforce supply and promote synergies with other ecosystem services. LandUse Policy 47, 145–155.

Patten, B.C., Costanza, R., 1997. Logical interrelations between four sustainabilityparameters: stability, continuation, longevity, and health. Ecosyst. Health 3 (3),136–142.

Peng, J., Wang, Y.L., Wu, J.S., Zhang, Y.Q., 2007. Evaluation for regional ecosystemhealth: methodology and research progress. Acta Ecol. Sin. 27 (11), 4877–4885.

Peng, J., Wu, J.S., Yin, H., Li, Z.G., Chang, Q., Mu, T.L., 2008. Rural land use changeduring 1986–2002 in Lijiang, China, based on remote sensing and GIS data.Sensors 8 (12), 8201–8223.

Peng, J., Wang, Y.L., Zhang, Y., Wu, J.S., Li, W.F., Li, Y., 2010. Evaluating theeffectiveness of landscape metrics in quantifying spatial patterns. Ecol. Indic.10 (2), 217–223.

Peng, J., Liu, Y.X., Wu, J.S., Lv, H.L., Hu, X.X., 2015. Linking ecosystem services andlandscape patterns to assess urban ecosystem health: a case study in ShenzhenCity, China. Landsc. Urban Plan. 143, 56–68.

Phillips, L.B., Hansen, A.J., Flather, C.H., 2008. Evaluating the species energyrelationship with the newest measures of ecosystem energy: NDVI versusMODIS primary production. Remote Sens. Environ. 112 (9), 3538–3549.

Pinto, U., Maheshwari, B.L., Ollerton, R.L., 2013. Analysis of long-term water qualityfor effective river health monitoring in peri-urban landscapes—a case study ofthe Hawkesbury–Nepean river system in NSW, Australia. Environ. Monit.Assess. 185 (6), 4551–4569.

Rapport, D.J., Maffi, L., 2011. Eco-cultural health, global health, and sustainability.Ecol. Res. 26 (6), 1039–1049.

Rapport, D.J., Singh, A., 2006. An ecohealth-based framework for state ofenvironment reporting. Ecol. Indic. 6 (2), 409–428.

Rapport, D.J., Costanza, R., McMichael, A.J., 1998. Assessing ecosystem health.Trends Ecol. Evol. 13 (10), 397–402.

Rapport, D.J., 1989. What constitutes ecosystem health? Perspect. Biol. Med. 33 (1),120–132.

Rapport, D.J., 1992. Defining the practice of clinical ecology. In: Costanza, R.,Norton, G., Haskell, B. (Eds.), Ecosystem Health: New Goals for EnvironmentalManagement. Island Press, Washington, DC, pp. 144–156.

Rapport, D.J., 1995. Ecosystem health: more than a metaphor? Environ. Values 4(4), 287–309.

Rapport, D.J., 1998. Defining ecosystem health. In: Rapport, D.J., Gaudet, C.L.,Costanza, R., Epstein, P.R., Levins, R. (Eds.), Ecosystem Health. BlackwellScience Inc, Malden, Massachusetts, pp. 18–33.

Santika, T., Meijaard, E., Wilson, K.A., 2015. Designing multifunctional landscapesfor forest conservation. Environ. Res. Lett. 10 (11), 114012.

Shen, C.C., Shi, H.H., Zheng, W., Ding, D.W., 2016. Spatial heterogeneity ofecosystem health and its sensitivity to pressure in the waters of nearshorearchipelago. Ecol. Indic. 61 (2), 822–832.

Shi, Z.H., Chen, L.D., Hao, J.P., Wang, T.W., Cai, C.F., 2009. The effects of land use

change on environmental quality in the red soil hilly region, China: a casestudy in Xianning County. Environ. Monit. Assess. 150 (1-4), 295–306.

Speldewinde, P.C., Slaney, D., Weinstein, P., 2015. Is restoring an ecosystem goodfor your health? Sci. Total Environ. 502, 276–279.

Su, M.R., Fath, B.D., 2012. Spatial distribution of urban ecosystem health inGuangzhou, China. Ecol. Indic. 15 (1), 122–130.

4 Indica

S

S

S

S

T

T

T

T

T

T

T

W

10 J. Peng et al. / Ecological

u, M.R., Yang, Z.F., Chen, B., 2009. Set pair analysis for urban ecosystem healthassessment. Commun. Nonlinear Sci. Numer. Simul. 14 (4), 1773–1780.

u, M.R., Fath, B.D., Yang, Z.F., 2010. Urban ecosystem health assessment: a review.Sci. Total Environ. 408 (12), 2425–2434.

uo, A.N., Xiong, Y.C., Wang, T.M., Yue, D.X., Ge, J.P., 2008. Ecosystem healthassessment of the Jinghe River watershed on the Huangtu Plateau. EcoHealth 5(2), 127–136.

uter II, G.W., 1993. A critique of ecosystem health concepts and indexes. Environ.Toxicol. Chem. 12 (9), 1533–1539.

ecchio, S., Chaalali, A., Raoux, A., Rius, A.T., Lequesne, J., Girardin, V., Lassalle, G.,Cachera, M., Riou, P., Lobry, J., Dauvin, J.C., Niquil, N., 2015. Evaluatingecosystem-level anthropogenic impacts in a stressed transitionalenvironment: the case of the Seine estuary. Ecol. Indic. 61 (2), 833–845.

iwari, B.K., Barik, S.K., Tripathi, R.S., 1998. Biodiversity value status and strategiesfor conservation of Sacred Groves of Meghalaya, India. Ecosyst. Health 4 (1),20–32.

orres, R., Gasparri, N.I., Blendinger, P.G., Grau, H.R., 2014. Land-use and land-covereffects on regional biodiversity distribution in a subtropical dry forest: ahierarchical integrative multi-taxa study. Reg. Environ. Change 14 (4),1549–1561.

urner, M.G., Gardner, R.H., Dale, V.H., O’Neill, R.V., 1989. Predicting the spread ofdisturbance across heterogenous landscape. Oikos 55 (1), 121–129.

urner, M.G., Gardner, R.H., O’Neill, R.V., 2001. Landscape Ecology in Theory andPractice. Springer-Verlag, New York.

urner, M.G., 2005. Landscape ecology in North America: past, present, and future.Ecology 86 (8), 1967–1974.

urner II, B.L., 2010. Land change in the southern Yucatan: case studies in landchange science. Reg. Environ. Change 10, 169–174.

alker, J., Thompson, C.H., Reddell, P., Rapport, D.J., 2001. The importance oflandscape age in influencing landscape health. Ecosyst. Health 7 (1), 7–14.

tors 72 (2017) 399–410

Wan, L.H., Zhang, Y.W., Zhang, X.Y., Qi, S.Q., Na, X.D., 2015. Comparison of landuse/land cover change and landscape patterns in Honghe National NatureReserve and the surrounding Jiansanjiang Region, China. Ecol. Indic. 51,205–214.

Watson, R.T., Jonahan, P., Gubler, D.J., Parson, E.A., Vincent, J.H., 2005.Environmental health implications of global climate change. J. Environ. Monit.7 (9), 834–843.

Weng, Q.H., Lo, C.P., 2001. Spatial analysis of urban growth impacts on vegetativegreenness with Landsat TM data. Geocarto Int. 16 (4), 19–28.

Weng, Q., 2001. A remote sensing-GIS evaluation of urban expansion and itsimpact on surface temperature in the Zhujiang Delta, China. Int. J. RemoteSens. 22 (10), 1999–2014.

Whitford, W.G., Rapport, D.J., deSoyza, A.G., 1999. Using resistance and resiliencemeasurements for ‘fitness’ tests in ecosystem health. J. Environ. Manag. 57 (1),21–29.

Wicklum, D., Davies, R.W., 1995. Ecosystem health and integrity. Can. J. Bot. 73 (7),997–1000.

Xiang, W.N., 2013. Working with wicked problems in socio-ecological systems:awareness, acceptance, and adaptation. Landsc. Urban Plan. 110, 1–4.

Xu, J.C., Wilkes, A., 2004. Biodiversity impact analysis in Northwest Yunnan,Southwest China. Biodivers. Conserv. 13 (5), 959–983.

Yokohari, M., Brown, R.D., Kato, Y., Yamamoto, S., 2001. The cooling effect of paddyfields on summertime air temperature in residential Tokyo, Japan. Landsc.Urban Plan. 53 (1–4), 17–27.

Yu, G.M., Yu, Q.W., Hu, L.M., Zhang, S., Fu, T.T., Zhou, X., He, X.L., Liu, Y.A., Wang, S.,Jia, H.H., 2013. Ecosystem health assessment based on analysis of a land usedatabase. Appl. Geogr. 44, 154–164.

Zhang, K., Li, Y.P., Huang, G.H., You, L., Jin, S.W., 2015. Modeling for regionalecosystem sustainable development under uncertainty—a case study ofDongying, China. Sci. Total Environ. 533, 462–475.