16
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/311860912 Understanding Land System Change Through Scenario-Based Simulations: A Case Study from the Drylands in... Article in Environmental Management · December 2016 DOI: 10.1007/s00267-016-0802-3 CITATIONS 0 READS 45 4 authors, including: Some of the authors of this publication are also working on these related projects: Viable InTensification of Agricultural production through sustainable Landscape transition (VITAL) View project Investigating illegal wildlife trade: Innovative approaches to inform global conservation policy View project Zhifeng Liu Beijing Normal University 17 PUBLICATIONS 231 CITATIONS SEE PROFILE Peter H Verburg VU University Amsterdam 428 PUBLICATIONS 12,465 CITATIONS SEE PROFILE Chunyang He Beijing Normal University 65 PUBLICATIONS 1,172 CITATIONS SEE PROFILE All content following this page was uploaded by Zhifeng Liu on 26 December 2016. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document and are linked to publications on ResearchGate, letting you access and read them immediately.

Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/311860912

UnderstandingLandSystemChangeThroughScenario-BasedSimulations:ACaseStudyfromtheDrylandsin...

ArticleinEnvironmentalManagement·December2016

DOI:10.1007/s00267-016-0802-3

CITATIONS

0

READS

45

4authors,including:

Someoftheauthorsofthispublicationarealsoworkingontheserelatedprojects:

ViableInTensificationofAgriculturalproductionthroughsustainableLandscapetransition(VITAL)

Viewproject

Investigatingillegalwildlifetrade:InnovativeapproachestoinformglobalconservationpolicyView

project

ZhifengLiu

BeijingNormalUniversity

17PUBLICATIONS231CITATIONS

SEEPROFILE

PeterHVerburg

VUUniversityAmsterdam

428PUBLICATIONS12,465CITATIONS

SEEPROFILE

ChunyangHe

BeijingNormalUniversity

65PUBLICATIONS1,172CITATIONS

SEEPROFILE

AllcontentfollowingthispagewasuploadedbyZhifengLiuon26December2016.

Theuserhasrequestedenhancementofthedownloadedfile.Allin-textreferencesunderlinedinblueareaddedtotheoriginaldocumentandarelinkedtopublicationsonResearchGate,lettingyouaccessandreadthemimmediately.

Page 2: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

Environmental ManagementDOI 10.1007/s00267-016-0802-3

Understanding Land System Change Through Scenario-BasedSimulations: A Case Study from the Drylands in Northern China

Zhifeng Liu1,2 ● Peter H. Verburg2 ● Jianguo Wu1,3 ● Chunyang He1

Received: 27 June 2016 / Accepted: 6 December 2016© Springer Science+Business Media New York 2016

Abstract The drylands in northern China are expected toface dramatic land system change in the context of socio-economic development and environmental conservation.Recent studies have addressed changes of land cover withsocioeconomic development in the drylands in northernChina. However, the changes in land use intensity and thepotential role of environmental conservation measures haveyet to be adequately examined. Given the importance ofland management intensity to the ecological conditions andregional sustainability, our study projected land systemchange in Hohhot city in the drylands in northern Chinafrom 2013 to 2030. Here, land systems are defined ascombinations of land cover and land use intensity. Usingthe CLUMondo model, we simulated land system change inHohhot under three scenarios: a scenario following histor-ical trends, a scenario with strong socioeconomic and landuse planning, and a scenario focused on achieving envir-onmental conservation targets. Our results showed thatHohhot is likely to experience agricultural intensificationand urban growth under all three scenarios. The agriculturalintensity and the urban growth rate were much higher underthe historical trend scenario compared to those with more

planning interventions. The dynamics of grasslands dependstrongly on projections of livestock and other claims onland resources. In the historical trend scenario, intensivelygrazed grasslands increase whereas a large amount of thecurrent area of grasslands with livestock converts to forestunder the scenario with strong planning. Strong conversionfrom grasslands with livestock and extensive cropland tosemi-natural grasslands was estimated under the conserva-tion scenario. The findings provide an input into discussionsabout environmental management, planning and sustainableland system design for Hohhot.

Keywords Historical land use change ● Land use planning ●

Conservation ● Land use intensity ● Hohhot ● Land usemodel

Introduction

Drylands are areas characterized by a lack of water, whichconstrains primary production and nutrient cycling, i.e.,areas where the ratio of mean annual precipitation to meanannual potential evapotranspiration is less than 0.65 (MEA2005). The drylands in northern China (DNC) cover 41% ofChina’s land surface and support approximately 27% ofChina’s population. As a result, the DNC is one of the keyareas in relation to sustainable development due to its vastsize and multiple challenges of water scarcity, ecologicaldeficit, and poverty (Li et al. 2016; Wu et al. 2015; Yanget al. 2008). In the past few decades, the DNC has experi-enced rapid socioeconomic development. From 1990 to2010, the human development index of the DNC increasedfrom 0.6 to 0.8, a growth of 27.4% (Li et al. 2016). Suchrapid socioeconomic development has resulted in dramatic

* Zhifeng [email protected]

1 Center for Human-Environment System Sustainability (CHESS),State Key Laboratory of Earth Surface Processes and ResourceEcology, Beijing Normal University, Beijing 100875, People’sRepublic of China

2 Department of Earth Sciences, Environmental Geography group,VU University Amsterdam, De Boelelaan 1085, 1081 Amsterdam,HV, The Netherlands

3 School of Life Sciences and School of Sustainability, ArizonaState University, Tempe, AZ 85287, USA

Page 3: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

land system dynamics. Land system changes are definedhere as combinations of changes in land cover types andland use intensity (GLP 2005). For instance, urban areas inthe DNC increased from 8.1 thousand km2 in 1990–13.6thousand km2 in 2005, a growth of 68.6% (Liu et al. 2012).The land experiencing desertification in the DNC expandedfrom 2622 thousand km2 in 1994 to 2636 thousand km2 in2004 (Yang et al. 2008; Zhu 2006). In the context of dra-matic land system change, the DNC continue to face severalecological and environmental problems including biodi-versity loss, decline of net primary productivity (NPP), landdegradation and air pollution (Han et al. 2015; He et al.2014; Li et al. 2016; Tian and Qiao 2014).

To solve these problems, a series of ecological con-servation and restoration projects were initiated around2000, including the Beijing–Tianjin Wind and Sand SourceControl Project, the Natural Forest Protection Project, andthe Grain to Green Project, which marked a turning point ofland system change in the DNC (Cao et al. 2011; Wanget al. 2007; Wu et al. 2015). As a result, grasslandsincreased from 86.7 million to 88 million ha, and forestcover increased from 17.6 to 20% in the period 2001–2010in Inner Mongolia in the DNC while land desertification hasbeen partially curbed (Wu et al. 2015). To plan for futuredevelopments and achieve landscape sustainability underfurther changes in socioeconomic conditions, an under-standing of the recent spatial dynamics and an explorationof possible future trajectories of land system change in theDNC is required (Berling-Wolff and Wu 2004; Jeneretteand Wu 2001; Wu 2013; Wu et al. 2014, 2015).

Recently, the simulation of land use change has become afrequently used approach for understanding and predictingland system changes in the DNC (Chen et al. 2008; Huanget al. 2014; Xu et al. 2009). For example, Chen et al. (2008)and Xu et al. (2009) simulated the land system change in theagro-pastoral transitional zone in the DNC based on theconversion of land use and its effects model. Huang et al.(2014) simulated the land system change in the DNC from2005 to 2030 based on the improved land use systemdynamics model. Deng et al. (2013) simulated the landsystem change in Qinghai Province in the DNC from 2010to 2050 using the land use change dynamics model. Thesestudies have provided significant insight into the drivers,processes, and consequences of changes in land cover typesunder socioeconomic development in the DNC. However,because all of these studies have only accounted for landcover changes, the roles of land management and intensifi-cation of land use have been disregarded, even though landuse intensity has seen major changes in the region and hasstrong implications for ecological and environmental sus-tainability. The main reasons for these deficiencies are: (1)most of these models can only simulate the changes in landcover types; and (2) the models were mainly driven by a

demand for urban and agricultural areas, ignoring otherdemands on land resources such as the claims for forest andnatural land to benefit environmental conservation (He et al.2005; van Asselen and Verburg 2013; Verburg andOvermars 2009; Verburg et al. 2006).

Similar constraints hold for most of the available landuse models, in which the units of simulation are often pixelswith a specific land use. For example, this holds for cellularautomata models, which are often used in urban simulationbut occasionally in rural areas as well (e.g., Jenerette andWu 2001; Wu 2002; Kamusoko et al. 2009). While agent-based models are another type of land use model, none-theless they represent the actual decision making of theactors of land change in the model structure, the outcomesof which are mostly still defined in terms of land coverchanges (Matthews et al. 2007; Li and Liu 2008; Parker2008). Developed by van Asselen and Verburg (2013), theCLUMondo model provides an innovative approach tousing a land system rather than a land cover representation.The CLUMondo model can not only allocate land systemchanges as determined by the regional demands of landcover types but also simulate land system changes inresponse to demands for various ecosystem services, whichcan be provided by multiple land systems at the same time(van Asselen and Verburg 2013; Ornetsmüller et al. 2016).The CLUMondo model allocates demands endogenously toeither changes in land cover type or land use intensity,depending on the location conditions, the land availabilityand the competitive advantage of the different competingland systems.

Our objective was to project the spatiotemporal patternsof land system change under socioeconomic developmentand environmental conservation in Hohhot city, a typicalcity in the DNC. To achieve this purpose, we first para-meterized the CLUMondo model based on a land systemmap of data from 2000, and then validated it using a landsystem map of 2013 data. Next, we used the parameterizedCLUMondo model to simulate land system changes from2013 to 2030 under three scenarios that represent varioussocioeconomic pathways and environmental conservationtargets. Finally, we assessed the differences in land systemchange under the scenarios and the sensitivity of theCLUMondo model toward some of our assumptions.

Study Area and Data

Study Area

The entire administrative area of Hohhot (110°46′–112°10′E, 40°51′–41°8′N) was used as the study area, with a totalland area of 17.2 thousand km2 (Statistical Bureau inHohhot 2013) (Fig. 1). The elevation in Hohhot declines

Environmental Management

Page 4: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

from the higher northwest (i.e., the Daqing Mountain) andsoutheast (i.e., the Manhan Mountain) to the lower centralarea (i.e., the Tumote Plain), with an average elevation of1050 m (Fig. 1). Hohhot is in the middle temperate zone,with a semi-arid, continental and monsoon climate (Zhanget al. 2013). Across the region, the mean annual temperatureranges from 3.5 to 8 °C, and the annual precipitation variesfrom approximately 337–418 mm, with 81% of the annualprecipitation falling between June and September when thetemperature is also high (Statistical Bureau in Hohhot2013). The city consists of the city proper of Hohhot,Wuchuan county, Tumotezuo banner, Tuoketuo county,Helingeer county, and Qingshuihe county, where the totalpopulation was approximately 3 million in 2012 (StatisticalBureau in Hohhot 2013) (Fig. 1).

Data Sources

The land use/cover data for 2000 and 2013, with a resolutionof 30m, were obtained from the national land use/coverdataset (NLCD) of China in the Data Sharing Infrastructure ofthe Earth System Science at the Chinese Academy of Science(http://www.geodata.cn/Portal/index.jsp, accessed August 30,2015). These NLCD datasets were produced through thevisual interpretation of Landsat Thematic Mapper images,with an accuracy of land use/cover classification greater than90% and six land use/cover classes (cropland, forest, grass-land, water, built-up area, and bare) (Liu et al. 2010).

Moderate-resolution imaging spectroradiometer 16-daycomposite images of normalized difference vegetation index(NDVI) for 2000 and 2013, with a resolution of 250m, wereobtained from the National Aeronautics and Space Admin-istration (http://ladsweb.nascom.nasa.gov, accessed August30, 2015). These data were radiometrically calibrated, pre-cisely georeferenced, and corrected for atmospheric effectsbefore distribution, and each image was a composite of themaximum NDVI values observed over 16 days.

County-level socioeconomic data, including crop pro-duction and livestock number, for 2000 and 2013 werecollected from the Hohhot Economic Statistical Yearbook(Statistical Bureau in Hohhot, 2013). The digital elevationmodel data with a resolution of 30 m was obtained from theInternational Scientific Data Service Platform (http://datamirror.csdb.cn/dem/files/ys.jsp, accessed August 30,2015). The meteorological data of temperature, precipita-tion, and solar radiation from 2000 to 2013 were obtainedfrom the Chinese Meteorological Information Center (http://data.cma.gov.cn, accessed August 30, 2015). The soilcharacteristic data were obtained from the HarmonizedWorld Soil Database (version 1.1) (FAO/IIASA/ISRIC/ISS-CAS/JRC, 2009). A number of geographic ancillary datawere collected from the National Administration of Sur-veying, Mapping, and Geoinformation of China, including

administrative boundaries, rivers, railways, highways,national roads, provincial roads, and city centers. All datawere georegistered to the same coordinate system andresampled to a spatial resolution of 300 m.

Methods

We first performed a land system classification to producethe land system maps by integrating the land cover and landuse management intensity for 2000 and 2013 in Hohhot.Second, we parameterized the CLUMondo model usingempirical relations between biophysical and socioeconomicexplanatory variables and the spatial distribution of landsystems in 2000. A run of the model across the period2000–2013 was used to validate the model. Three scenarioswere defined to simulate land system change from 2013 to2030. The first scenario was based on the historical trend ofland system change from 2000 to 2013 and assumed thatthese trends would continue into the future. The secondscenario was based on a review of documents depicting thesocioeconomic and land use planning envisaged for theregion, while the third scenario represents the conditions thatwere assumed to best correspond with an implementation ofplanning and policy to meet the biodiversity conservationtargets set in policy documents for the region. Figure 2aprovides an overview of the steps of our methodology.

Land System Classification

We classified the variation of land cover and land useintensity into ten land system types starting from the land

Fig. 1 The location of the study area

Environmental Management

Page 5: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

use/cover maps for Hohhot that were available for 2000 and2013 (Table 1). First, croplands under three levels of agri-cultural intensity (i.e., intensive, medium intensive, andextensive) were identified using NPP, which is a widely usedindicator for characterizing land use intensity (Kuemmerleet al. 2013) (Table 1). Following Kuemmerle et al. (2013),we assumed that croplands with a higher NPP have a higherlevel of agricultural intensity. Based on this assumption, thecroplands were classified using the following formula:

ClassCropi;t ¼

0 NPPi;t>Tint

1 NPPi;t � Tint&NPPi;t>Tmed

2 NPPi;t � Tmed

:

8>>>>>><>>>>>>:

ð1Þ

where ClassCropi;t is the land system type of pixel i incropland in year t, while the values of 0, 1, and 2 denote theland system types of intensive croplands, medium intensivecroplands, and extensive croplands, respectively. NPPi,t isthe NPP of pixel i in year t. Tint and Tmed are thresholds ofNPP. In Hohhot, NPPi,t was calculated using NDVI andphotosynthetically active radiation (PAR), which wasincluded as a fixed fraction (0.47) of solar radiation in2000 and 2013, based on the formula developed byBuyantuyev and Wu (2009):

NPP ¼Z

NDVI ´PAR ð2Þ

Tint and Tmed were determined by the natural breaksclassification based on the NPP of cropland in 2013. Spe-cifically, Tint and Tmed were, respectively, equal to 877 and676 NDVI*PAR.

Fig. 2 Flowchart. a A flowchartof the key steps in the simulationof land system change. b Mainconcept and workflow of theCLUMondo model

Environmental Management

Page 6: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

Second, the grasslands and built-up areas were classifiedaccording to their location. The grasslands were classifiedinto semi-natural grasslands and grasslands with livestockbased on the policy that grasslands in protected areas cannotbe grazed (Standing Committee of the People’s Congress inInner Mongolia, 2013). Additionally, built-up areas wereclassified into urban built-up areas and villages based on theapproach developed by Angel et al. (2011). The basic ideais to automatically group built-up areas into several clustersbased on the distance of influence of each built-up area, andthen to classify clusters containing city centers as the urbanbuilt-up area (Angel et al. 2011). In Hohhot, the sixadministrative centers were used as city centers to identifyurban built-up areas (Fig. 1), while the other built-up areaswere identified as villages in 2000 and 2013.

3.2. Parameterization of the CLUMondo Model

The basic principle of the CLUMondo model is to determinethe spatial allocation of land systems through ensuring thatthe allocated land systems fulfill the regional demands forecosystem goods and services in consideration of the typicalsupply of these goods and services by the different landsystems (van Asselen and Verburg 2013) (Fig. 2b). Duringan iterative procedure, the model allocates changes in landsystems to fulfill the regional demands by considering thedifferences in location suitabilities for different land systems,the comparative advantage of the different land systems insupplying the goods and services, and the restrictions andconstraints of land system conversion at a specific locationas set by policies, natural conditions or the land use history.

The demands considered in our study included cropproduction, livestock, urban land, forest areas, and naturalareas (including forest and semi-natural grassland). To meetthese demands, the land systems are allocated according totheir capacity to supply these demands as well as by the

local biophysical and socioeconomic location conditions.Specifically, the model allocates the land system (LS) withthe highest transition potential (Ptran) for grid cell (i) at time(t). The transition potential was calculated using the localsuitability (Ploc), the conversion resistance (Pres) and thecompetitive advantage of a land system (Pcomp):

Ptrant;i;LS ¼ Ploct;i;LS þ PresLS þ Pcompt;LS ð3Þ

According to the location factors selected by Huang et al.(2014), we used 17 biophysical and socioeconomic explana-tory variables (S), including climate factors, topographic fac-tors, soil characteristic factors, and locational factors (Table 2),to calculate local suitability based on the following formula:

Log Ploct;i;LS= 1� Ploct;i;LS� �� � ¼ β0 þ

X

j

βj;LS � Sj;t;i;

ð4Þwhere j denotes the number of variables; the weight βj,LS forthe variable Sj and the constant β0 are determined by esti-mating a logistic model based on the spatial distribution ofland systems in the land system map for 2000.

We estimated the conversion resistance factor for eachland system based on the costs of conversion and infor-mation on local policies on land use (Office of Land andResources in Inner Mongolia 2010) (Table 3). The resis-tance factors of urban built-up areas and villages with highcost for conversion were set at 1.0, while the resistancefactors of medium intensive croplands, extensive croplands,and bare land with low cost for conversion were set at 0.8.The resistance factors of other land systems were set at 0.9.

Table 2 Biophysical and socioeconomic explanatory variables

Category Factors

Climate factors Mean annual temperature

Annual precipitation

Climate zone

Topographic factors Elevation

Slope

Aspect

Soil characteristic factors Soil organic carbon

Percentage of clay

Percentage of sand

Percentage of silt

Locational factors Distance to city center

Distance to county center

Distance to railway

Distance to highway

Distance to national road

Distance to provincial road

Distance to river

Table 1 The land system classes and their corresponding land use/cover classes

Name of class in land use/covermap

Name of class in land systemmap

Cropland Intensive cropland

Medium intensive cropland

Extensive cropland

Forest Forest

Grassland Semi-natural grassland

Grassland with livestock

Water Water

Built-up area Urban built-up area

Villages

Bare Bare

Environmental Management

Page 7: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

To link land systems to their supply of ecosystem goodsand services, we developed a land system lookuptable to indicate the relative order of the land systems in theircontribution to fulfilling a specific demand as well as theircapacity to supply this demand. During the iteration, thisallows the model to check whether allocated land systemsfulfill the demands for goods and services. The competitiveadvantage of land systems, at a specific location, witha higher hierarchical level is increased in the case of under-supply while the competitive advantage is decreased in thecase of oversupply (Table 3). The crop production providedby cropland systems and the livestock number carried bygrassland with livestock systems was calculated based onstatistical data, while the NPP was used as ancillary data formeasuring differences in crop production among the croplandsystems under the three levels of agricultural intensity(Table 3).

Additionally, a conversion matrix was used to indicatewhich land system conversions were possible. According tothe local land use policy (Office of Land and Resources inInner Mongolia 2010), conversions from intensive crop-lands, forest, semi-natural grasslands, water, and urbanbuilt-up areas were not allowed. Similar to Verburg andOvermars (2009), we assumed that conversions fromextensive croplands to medium intensive croplands andfrom medium intensive croplands to intensive croplandsrequired at least 5 years, thus avoiding multiple swaps inland system within a short period.

Scenario Setting

To capture land system change under different levels ofsocioeconomic development and environmental conserva-tion, three scenarios were developed according to the his-torical trend of land system change from 2000 to 2013, theland use planning from 2013 to 2030, and the biodiversityconservation targets, respectively (Table 4). The three sce-narios differ not only in the absolute changes in demands forgoods and services from land systems but also in terms ofthe demands considered in the scenario. These differencesreflect the importance given to different land system func-tions under the different scenarios. Specifically, in the his-torical trend scenario (TREND), three demands for cropproduction, livestock, and urban land were included. Thesedemands were calculated for the period from 2013 to 2030using the annual change rates in supply from 2000 to 2013in Hohhot (Table 4).

In the scenario based on socioeconomic developmentand strong land use planning (PLANNED), demands forcrop production, livestock, urban land, and forest develop-ment were considered. In this case, the demands werederived from the projected annual change rates from 2010to 2020 as stated in the planning documents concerningland use for Hohhot as well as the national planning onmedium- and long-term food security (National Develop-ment and Reform Commission of China, 2008; Office ofLand and Resources in Inner Mongolia 2010) (Table 4).

Table 3 The resistance factors and lookup values for each land system

Land system Resistancefactor

Cropproduction(tons/pixel)

Livestock(heads/pixel)

Urban land(km2/pixel)

Forest(km2/pixel)

Natural land(km2/pixel)

Intensive cropland 0.9 3 22.8 0 0 0 0 0 0 0 0

Medium intensive cropland 0.8 2 18.1 0 0 0 0 0 0 0 0

Extensive cropland 0.8 1 13.5 0 0 0 0 0 0 0 0

Forest 0.9 0 0 0 0 0 0 1 0.09 1 0.09

Semi-natural grassland 0.9 0 0 0 0 0 0 0 0 1 0.09

Grassland with livestock 0.9 0 0 1 53.5 0 0 0 0 0 0

Water 0.9 −1 0 −1 0 −1 0 −1 0 −1 0

Urban built−up area 1.0 −1 0 −1 0 1 0.09 −1 0 −1 0

Villages 1.0 −1 0 −1 0 0 0 −1 0 −1 0

Bare 0.8 0 0 0 0 0 0 0 0 0 0

For each demand type, the first column presents a ranking of the land systems*, the second column presents the average value of the ecosystemservice supplied by the land system in 2013**

*The land system with the higher value (≥0) has the higher order for fulfilling the corresponding ecosystem service. The land system is excludedfrom consideration for fulfilling the corresponding ecosystem service by coding this system ‘−1’**The average value of the ecosystem service supplied by the land system in 2013 was calculated on the basis of the crop production and thelivestock number from statistical data and the total area of corresponding land system. In particular, the NPP was used as ancillary data forcalculating the crop production supplied by different land systems. For example, Crop production of intensive cropland= Total crop production *(Total NPP of intensive cropland/Total NPP of cropland)

Environmental Management

Page 8: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

In the scenario with strong conservation targets (CON-SERVATION), the same demands for crop production,livestock, and urban land were used as in the scenario withstrong land use planning (PLANNED) to meet the baselineof socioeconomic development in Hohhot. However, thedemand for forest development was replaced by a demandfor protected areas to meet the national conservation areatarget. An estimate of the regional implementation of thistarget was made based on the existing protected area and theprotected area target at the ecoregional scale derived fromButchart et al. (2015). Natural land systems, includingforest and semi-natural grasslands, were assumed to con-tribute to the protected area (Table 4).

Results

Validation of the Land System Classification and theCLUMondo model

We used census data to validate the land system classifi-cation concerning the identified intensity levels of croplandin Hohhot in 2000 and 2013. Intensity levels were based onthe NPP data, and the census data provided an independent(but low spatial resolution) alternative. A correlation ana-lysis was performed between the crop production per hec-tare from statistical data in each county and the percentageof croplands at different intensity levels from our results(Fig. 3). The assessment shows that the percentage ofintensive croplands had a positive correlation (R= 0.83)with crop production per hectare at a significance level of0.01 (Fig. 3a), whereas the percentage of extensive crop-lands had a significantly negative correlation with cropproduction per hectare (R= −0.64, P< 0.01) (Fig. 3b). Thesignificant correlations suggested that the differences in

intensity of cropland management were well captured by theNPP proxy in the land system classification.

To validate the CLUMondo model, we simulated theland system changes in Hohhot from 2000 to 2013 based ondemands of the crop production and the livestock number

Table 4 A description of thescenarios

Demands Target in 2030 Annual growth rate from 2013 to 2030

Trend Planned Conservation Trend (%) Planned (%) Conservation (%)

Crop production (milliontons)

2.1 1.6 1.6 3.0 1.0* 1.0*

Livestock (million heads) 4.7 3.7 3.7 2.5 0.9* 0.9*

Urban land (km2) 275.0 249.5 249.5 2.0 1.4** 1.4**

Forest (km2) 3458.2 2.2**

Natural land (includingforest and semi-naturalgrassland) (km2)

4828.5 2.4

*Calculated according to the national planning on medium- and long-term food security (2008–2020)published by the National Development and Reform Commission of China (2008)

**Calculated according to the planning on land use in Inner Mongolia (2006–2020) published by the Officeof Land and Resources in Inner Mongolia (2010)

Fig. 3 Validation of the land system classification. a Correlationbetween crop production per hectare and percentage of intensivecropland by county. b Correlation between crop production per hectareand percentage of extensive cropland by county

Environmental Management

Page 9: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

derived from the statistical data and the urban land areachanges derived from the land system maps. We comparedthe simulated land system map with the actual land systemmap in Hohhot in 2013. Following van Vliet et al. (2011)and Pontius Jr and Millones (2011), six accuracy assess-ment indices of overall accuracy (OA), quantity disagree-ment (QD), allocation disagreement (AD), Kappasimulation (KSimulation), Kappa transition (KTransition), andKappa transition location (KTransloc) were selected for vali-dation. Kappa is an indicator frequently used for measuringthe accuracy of a simulated land system, the details ofKappa and the other indicators used in our study can befound in van Vliet et al. (2011). The results showed that theland system changes were simulated with relatively highaccuracy with an OA of 76.8, QD of 1.5, AD of 21.7%,KSimulation of 0.13, KTransition of 0.70, and KTransloc of 0.19compared with the performance of similar regional-scaleland use model applications (van Vliet et al. 2011).

Land System Change from 2000 to 2013

Hohhot experienced agricultural intensification in terms ofcroplands as well as a growth in grasslands with livestockand urban built-up areas from 2000 to 2013 (Fig. 4). Thearea of intensive croplands increased from 516.9 to 2065.8km2, a growth of approximately three times (Fig. 4a). Thearea of medium intensive croplands increased from 1623.7to 3205.3 km2, a growth of 97.4% (Fig. 4a). The increases ofintensive and medium intensive croplands mainly occurredin the south and northeast parts of Hohhot (Fig. 4b). Addi-tionally, the area of grasslands with livestock increased from5370.3 to 5474.2 km2, a growth of almost 2% (Fig. 4a). Thegrowth in grasslands with livestock was found in thenorthwest and south parts of Hohhot (Fig. 4b). In this period,the urban built-up areas increased by 37% (Fig. 4).

Land System Changes from 2013 to 2030 UnderDifferent Scenarios

The degree of agricultural intensification varied sig-nificantly under the different scenarios (Fig. 5; Table 5).Specifically, a large agricultural intensification was pro-jected with a 140% increase of intensive croplands, a 50%decrease of medium intensive croplands and a 90% decreaseof extensive cropland in Hohhot from 2013 to 2030 underthe TREND scenario. A middle level of agricultural inten-sification was projected under the CONSERVATION sce-nario with a near 50% increase of intensive croplands, a12% decrease of medium intensive croplands, and a 66%decrease of extensive croplands. Although the demand foragricultural commodities is the same as in the CON-SERVATION scenario, only a small intensification wasforeseen under the PLANNED scenario with a growth in

intensive and medium intensive croplands of approximately8 and a 33% decline of extensive croplands. Such agri-cultural intensification was mainly estimated in the north-west and south parts of Hohhot (Fig. 5c).

The forest, the semi-natural grasslands, and the grass-lands with livestock were projected to have diverse devel-opments under the different scenarios (Fig. 5; Table 5).Under the TREND scenario, the grasslands with livestockwere projected to increase by 1%, without changes in theforest and semi-natural grassland areas. Under the PLAN-NED scenario, the increased demand for forest developmentled to a conversion of grasslands with livestock to forestmainly in the north part of Hohhot (Fig. 5c). Under theCONSERVATION scenario, the semi-natural grassland wasprojected to increase 1.4 times, mainly through the con-version of grasslands with livestock and extensive crop-lands in the north part of Hohhot (Fig. 5c).

Discussion and Conclusions

Validation and Sensitivity Analysis of the CLUMondoModel

To evaluate the validity of a land system model, a com-parison between the accuracy of the land system model andthe accuracy of a no-change model is often used (van Vlietet al. 2011). Accordingly, we used the actual land system in2000 as the result of the no-change model. Then, we cal-culated the accuracy of the no-change model by directlycomparing the actual land system in 2000 with that in 2013(Fig. 6a, c and Table 6). The accuracy of this very basic no-change model was compared, through a series of metrics,with the accuracy of the simulated land system distributionfor 2013. We found that the OA, KSimulation, and KTransition

of the simulated land system based on the CLUMondomodel were, for our simulations, all significantly higherthan the same metrics for the no-change model, whereas QDwas much lower (Table 6; Fig. 6a–c). All metrics thusdenoted that the CLUMondo model provides an effectiveapproach for understanding land system change in Hohhot.

Some variables of the CLUMondo model are estimatedbased on expert knowledge and, thus, subject to uncertainty.Therefore, we tested the extent to which the CLUMondooutputs for this region are sensitive to the conversionresistance parameters and the restrictions set in the con-version matrix. If the conversion resistance (Pres) was notincluded in the allocation procedure, the model wouldallocate the land system more strongly toward the suitabilitymaps derived from the location factors, resulting in the landsystem shown in Fig. 6d, which differs from the actual landsystem observed in 2013 with a much lower OA (65.0%)and KSimulation (0.04) (Table 6). Furthermore, if all of the

Environmental Management

Page 10: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

conversions between land systems were allowed, i.e.,without conversion restrictions, the CLUMondo model

would result in the land systems indicated in Fig. 6e, with alow OA (63.7%) and KSimulation (0.02) as well (Table 6).

Fig. 4 Land system changefrom 2000 to 2013 in Hohhot. aQuantitative change from 2000to 2013. b Spatial pattern of landsystems in 2000 and 2013. cChange in spatial pattern of landsystems from 2000 to 2013

Environmental Management

Page 11: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

Fig. 5 Land system change from 2013 to 2030 in Hohhot under different scenarios. a Quantitative change. b Land system maps in 2030. c Changein spatial pattern

Environmental Management

Page 12: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

In principle, these parameters of conversion resistanceand restrictions on conversions could be set through modelcalibration if temporally consistent series of land systemsare available. In this study, only the observed land systemmaps for 2000 and 2013 were available, and the 2013 mapwas needed for validation. Therefore, these parameters wereset according to expert knowledge and follow the localpolicy on land use (Office of Land and Resources in InnerMongolia 2010). The results of the sensitivity analysis showthat the accuracy of the simulations may be improved ifthese parameters can be calibrated based on a time-series ofthe observed data.

Additionally, our analysis used a land system classifi-cation instead of the regularly used land cover classifica-tions as a basis of the simulations. The advantage of thisapproach is that it can account for both land cover and landuse intensity changes. Recently, some studies have definedland systems that are much wider than the combination ofland cover and land use intensity (Verburg et al. 2013). Forexample, Verburg et al. (2013) defined land systems as “theterrestrial component of the Earth system”, which “encom-pass all processes and activities related to the human use ofland, including socioeconomic, technological and organi-zational investments and arrangements, as well as the

Table 5 Change in land system from 2013 to 2030 in Hohhot under the different scenarios

Land system Area in 2013(km2)

Area in 2030 (km2) Change rate from 2013 to 2030

Trend Planned Conservation Trend (%) Planned (%) Conservation (%)

Intensive cropland 2065.8 4886.5 2221.1 3040.4 136.5 7.5 47.2

Medium intensive cropland 3205.3 1594.1 3493.5 2809.1 −50.3 9.0 −12.4Extensive cropland 1339.7 130.1 896.0 454.7 −90.3 −33.1 −66.1Forest 2498.8 2498.8 3373.6 2498.8 0.0 35.0 0.0

Semi-natural grassland 957.2 957.2 957.2 2288.1 0.0 0.0 139.1

Grassland with livestock 5474.2 5530.4 4552.9 4479.4 1.0 −16.8 −18.2Water 306.7 306.7 306.7 306.7 0.0 0.0 0.0

Urban built-up area 203.1 275.0 249.6 249.7 35.4 22.9 22.9

Villages 783.5 783.5 783.5 783.5 0.0 0.0 0.0

Bare 323.2 195.0 323.2 247.1 −39.7 0.0 −23.6

Fig. 6 Comparison of the actual land system and the simulated land system in 2013 in Hohhot

Environmental Management

Page 13: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

benefits gained from land and the unintended social andecological outcomes of societal activities”. Following thesedefinitions, others have used significantly more complexland system classifications that also capture some of thetemporal and spatial configuration effects of the land system(Muller et al. 2014; Ornetsmüller et al. 2016). For our studyarea, the classification used only captures the main elementsof variation in land systems due to the data limitations.Therefore, our simulations may be further improved byperforming a more comprehensive land system classifica-tion with an integration of multiple data (e.g., socio-economic statistic data, field survey data, and remotesensing data).

Land Cover Change and Land use Intensification UnderSocioeconomic Development and EnvironmentalConservation in Hohhot

The scenarios indicate that the region is likely to face bothland cover change and land use intensification in the contextof socioeconomic development and environmental con-servation from 2013 to 2030. In particular, urban built-upareas were estimated to increase rapidly under all threescenarios in the context of large scale of urbanization in theDNC (Li et al. 2016). All scenarios indicated that agri-cultural land cover change is rather limited. Most of theincreasing demand for agricultural goods must be suppliedby agricultural intensity, although there is a large range ofdifferences in intensification among the scenarios. Underthe TREND scenario, the annual growth rate of demand forcrop production was approximately three times that of theother scenarios (Table 4), thus the degree of agriculturalintensification was projected to be much higher (Table 5;Fig. 5). In addition, under the CONSERVATION scenario,a number of extensive croplands was projected to convert tosemi-natural grasslands to achieve the environmental con-servation target (Table 4, Table 5; Fig. 5). Due to thiscompetition for space, the agricultural intensity under theCONSERVATION scenario was simulated to be greaterthan that under the PLANNED scenario to compensate forthe cropland loss, although the demands for crop productionwere the same under the two scenarios (Table 5; Fig. 5).Furthermore, the grasslands with livestock, forest and semi-natural grasslands face different trajectories of change asresult of different demands under the three scenarios, thecompetition with other land systems and the local suit-ability. Under the TREND scenario, the demand for live-stock was significantly higher, and no environmental targetwas set. Thus grasslands with livestock were projected toincrease while the areas covered by forest and semi-naturalgrasslands were projected to remain the same (Table 4,Table 5; Fig. 5). The PLANNED scenario shows the con-sequences and trade-offs of an environmental target forT

able

6Accuracyassessmentof

thesimulated

land

system

in20

13

Accuracyassessmentindex

The

CLUMon

domod

elThe

no-chang

emod

el*

The

CLUMon

domod

elwith

outconv

ersion

resistance**

The

CLUMon

domod

elwith

outconv

ersion

restriction*

**

Overallaccuracy

(OA)

76.8%

68.5%

65.0%

63.7%

Quantity

disagreement(Q

D)

1.5%

19.5%

17.3%

16.2%

Allo

catio

ndisagreement(A

D)

21.7%

12.0%

17.7%

20.1%

Kappa

simulation(K

Sim

ulation)

0.13

0.00

0.04

0.02

Kappa

transitio

n(K

Transition)

0.70

0.00

0.34

0.12

Kappa

transitio

nlocatio

n(K

Transloc)

0.19

0.12

0.17

*The

land

system

map

from

2000

was

used

**The

conv

ersion

resistance

ofeach

land

system

was

setas

0

***A

llconv

ersion

sam

ongland

system

swereallowed

Environmental Management

Page 14: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

increasing forest as was set according to land use planningin Hohhot (Office of Land and Resources in Inner Mongolia2010). Such a target would potentially result in a largeextent of conversion from grasslands with livestock to forest(Table 5; Fig. 5). Under the CONSERVATION scenario, adifferent trajectory is observed as targets were set forincreasing natural land, including forest and semi-naturalgrasslands (Table 4). Because the suitability of semi-naturalgrasslands is much higher than that of forest in most regionsof Hohhot due to the semi-arid climate (Cao et al. 2011;Wang et al. 2007), such a target mainly resulted in theconversion from grasslands with livestock to semi-naturalgrasslands (Table 5; Fig. 5). However, intensification onremaining lands indicates the tradeoff of such a ‘landsparing’ policy.

Our findings are consistent with earlier studies withrespect to the projected land cover changes. For example,based on the SD model, Hasbagen et al. (2008) found thatthe grasslands would decrease whereas the forest wouldincrease in Hohhot from 2010 to 2020, showing a consistenttrend with our results from the PLANNED scenario.

Implications for Environmental Management andSustainable Land System Design

Our results regarding simulated land system change hadsome potential options to inform environmental manage-ment and sustainable land system design in Hohhot. Thedifferent scenarios visualize spatiotemporal patterns of landsystem change under different levels of socioeconomicdevelopment and environmental conservation. The resultsprovide the necessary input to assess the potential impactsof such alternative developments on ecosystems andenvironments. For example, as a tradeoff of development,regional urban growth would cause natural habitat loss andfragmentation (Liu et al. 2016a, b), biodiversity decline (Heet al. 2014), decreases in carbon storage (He et al. 2016),and pollution of water, soil and air (Chen 2007; Han et al.2015). The strong agricultural intensification projected inthe scenario would potentially lead to water scarcity and soilacidification (Guo et al. 2010; Li et al. 2016). At the sametime, increases of natural land, e.g., forest and semi-naturalgrasslands, could have positive effects on the environment(Cao et al. 2011; Wang et al. 2007; Wang et al. 2010). Suchassessments are essential to foresee the possible negativeimpacts of development or environmental targets and topromote effective environmental management and landsystem design in Hohhot (Turner et al. 2013; Verburg et al.2013).

In addition, we found obvious differences of change inforest and semi-natural grasslands under the two environ-mental conservation strategies. According to local planningon land use (Office of Land and Resources in Inner

Mongolia 2010), the forest was projected to increase rapidly(Table 5; Fig. 5). However, land use planning for increasingforest with inadequate consideration of local suitabilitycould result in negative impacts on the environment inHohhot (Cao et al. 2011; Wang et al. 2007). In Hohhot,where the mean annual precipitation is lower than 400 mm,increased forest coverage would consume a large amount ofthe groundwater (Cao et al. 2011). Furthermore, the vege-tation coverage and NPP of the new forest would be con-strained, which would lead to reduced positive effects onwater and soil retention and ecological restoration (Caoet al. 2011; Wang et al. 2007). In contrast, our analysisshowed that in most places in the Hohhot, semi-naturalgrassland has much better suitability (Fig. 5).

Acknowledgements We would like to thank Dr. Jasper van Vliet,David A. Eitelberg, and Dr. Žiga Malek from VU UniversityAmsterdam for their helpful suggestions on the article. This work hasbeen supported in part by the National Natural Science Foundation ofChina (Grant No. 41501195), the National Basic Research Program ofChina (Grant Nos. 2014CB954303 & 2014CB954302) and the fund-ing from European Research Council under the European Union’sSeventh Framework Program ERC Grant Agreement nr. 311819(GLOLAND). It was also supported by the Youth Scholars Program ofBeijing Normal University (Grant No. 2014NT02) and the State KeyLaboratory of Earth Surface Processes and Resource Ecology (GrantNo. 2015-RC-01).

Compliance with Ethical Standards

Conflict of Interest The authors declare that they have no com-peting interests.

References

Angel S, Parent J, Civco DL, Blei A, Potere D (2011) The dimensionsof global urban expansion: estimates and projections of allcountries, 2000–2050. Prog Plan 75:53–107

Berling-Wolff S, Wu JG (2004) Modeling urban landscape dynamics:a case study in Phoenix, USA. Urban Ecosyst 7:215–240

Butchart SH et al. (2015) Shortfalls and solutions for meeting nationaland global conservation area targets. Conserv Lett 8:329–337

Buyantuyev A, Wu JG (2009) Urbanization alters spatiotemporalpatterns of ecosystem primary production: a case study of thePhoenix metropolitan region, USA. J Arid Environ 73:512–520.doi:10.1016/j.jaridenv.2008.12.015

Cao SX, Chen L, Shankman D, Wang CM, Wang XB, Zhang H(2011) Excessive reliance on afforestation in China’s arid andsemi-arid regions: Lessons in ecological restoration. Earth SciRev 104:240–245. doi:10.1016/j.earscirev.2010.11.002

Chen J (2007) Rapid urbanization in China: a real challenge to soilprotection and food security. Catena 69:1–15

Chen YH, Li XB, Su W, Li Y (2008) Simulating the optimal land-usepattern in the farming-pastoral transitional zone of NorthernChina. Comput Environ Urban 32:407–414. doi:10.1016/j.compenvurbsys.2008.01.001

Deng XZ, Huang JK, Lin YZ, Shi QL (2013) Interactions betweenclimate, socioeconomics, and land dynamics in Qinghai Province,China: a LUCD model-based numerical experiment. AdvMeteorol 10.1155/2013/297926

Environmental Management

Page 15: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

FAO/IIASA/ISRIC/ISS-CAS/JRC (2009) Harmonized world soildatabase (version 1.1). FAO, IIASA, Rome, Laxenburg

Global Land Project (GLP) (2005) Science plan and implementationstrategy. IGBP Secretariat, Stockholm

Guo JH et al. (2010) Significant acidification in major Chinese crop-lands. Science 327:1008–1010

Han LJ, Zhou WQ, Li WF (2015) City as a major source area of fineparticulate (PM 2.5) in China. Environ Pollut 206:183–187

Hasbagen BaoY, Li B (2008) The systematical study of the landcarrying capacity of Hohhot. J Arid Land Resour Environ22:26–32. in Chinese

He CY, Liu ZF, Tian J, Ma Q (2014) Urban expansion dynamics andnatural habitat loss in China: a multi-scale landscape perspective.Glob Change Biol 20:2886–2902. doi:10.1111/gcb.12553

He CY et al. (2005) Developing land use scenario dynamics model bythe integration of system dynamics model and cellular automatamodel. Sci China Ser D 48:1979–1989

He CY, Zhang D, Huang QX, Zhao YY (2016) Assessing the potentialimpacts of urban expansion on regional carbon storage by linkingthe LUSD-urban and InVEST models. Environ Model Softw75:44–58

Huang QX, He CY, Liu ZF, Shi PJ (2014) Modeling the impacts ofdrying trend scenarios on land systems in northern Chinausing an integrated SD and CA model. Sci China Earth Sci57:839–854

Jenerette GD, Wu JG (2001) Analysis and simulation of land-usechange in the central Arizona–Phoenix region, USA. Landsc Ecol16:611–626

Kamusoko C, Aniya M, Adi B, Manjoro M (2009) Rural sustainabilityunder threat in Zimbabwe—simulation of future land use/coverchanges in the Bindura district based on the Markov-cellularautomata model. Appl Geogr 29:435–447. doi:10.1016/j.apgeog.2008.10.002

Kuemmerle T et al. (2013) Challenges and opportunities in mappingland use intensity globally. Curr Opin Environ Sustain5:484–493. doi:10.1016/j.cosust.2013.06.002

Li JW, Liu ZF, He CY, Tu W, Sun ZX (2016) Are the drylands innorthern China sustainable? A perspective from ecological foot-print dynamics from 1990 to 2010. Sci Total Environ553:223–231. doi:10.1016/j.scitotenv.2016.02.088

Li X, Liu XP (2008) Embedding sustainable development strategies inagent-based models for use as a planning tool International. JGeogr Inf Sci 22:21–45. doi:10.1080/13658810701228686

Liu J, Zhang Q, Hu Y (2012) Regional differences of China’s urbanexpansion from late 20th to early 21st century based on remotesensing information. Chin Geogr Sci 22:1–14

Liu JY et al. (2010) Spatial patterns and driving forces of land usechange in China during the early 21st century. J Geogr Sci20:483–494

Liu ZF, He CY, Wu JG (2016a) General spatiotemporal patterns ofurbanization: an examination of 16 World cities. Sustainability8:41

Liu ZF, He CY, Wu JG (2016b) The relationship between habitat lossand fragmentation during urbanization: an empirical evaluationfrom 16 world cities. PLoS One 11:e0154613. doi:10.1371/journal.pone.0154613

Matthews RB, Gilbert NG, Roach A, Polhill JG, Gotts NM (2007)Agent-based land-use models: a review of applications. LandscEcol 22:1447–1459. doi:10.1007/s10980-007-9135-1

Millennium Ecosystem Assessment (MEA) (2005) Ecosystems andhuman well-being: current state and trends. Island Press,Washington

Muller D, Sun ZL, Vongvisouk T, Pflugmacher D, Xu JC, Mertz O(2014) Regime shifts limit the predictability of land-systemchange. Glob Environ Change 28:75–83. doi:10.1016/j.gloenvcha.2014.06.003

National Development and Reform Commission of China (2008) Thenational planning on medium- and long-term food security(2008–2020). http://www.gov.cn/jrzg/2008-11/13/content_1148414.htm

Office of Land and Resources in Inner Mongolia (2010) The planningon land use in Inner Mongolia (2006–2020). http://www.mlr.gov.cn/tdsc/tdgh/201006/t20100623_152592.htm

Ornetsmüller C, Verburg PH, Heinimann A (2016) Scenarios of landsystem change in the Lao PDR: transitions in response to alter-native demands on goods and services provided by the land. ApplGeogr 75:1–11

Parker DC, Hessl A, Davis SC (2008) Complexity, land-use modeling,and the human dimension: fundamental challenges for mappingunknown outcome spaces. Geoforum 39:789–804. doi:10.1016/j.geoforum.2007.05.005

Pontius Jr RG, Millones M (2011) Death to Kappa: birth of quantitydisagreement and allocation disagreement for accuracy assess-ment. Int J Remote Sens 32:4407–4429

Standing Committee of the People’s Congress in Inner Mongolia (2013)Regulations of Inner Mongolia autonomous region on Daqingmountain national nature reserve. http://dqs.nmglyt.gov.cn/

Statistical Bureau in Hohhot (2013) Hohhot economic statisticalyearbook. China Statistics Press, Beijing

Tian GJ, Qiao Z (2014) Assessing the impact of the urbanizationprocess on net primary productivity in China in1989–2000. Environ Pollut 184:320–326. doi:10.1016/j.envpol.2013.09.012

Turner IIBL, Janetos AC, Verbug PH, Murray AT (2013) Land systemarchitecture: using land systems to adapt and mitigate globalenvironmental change. Pacific Northwest National Laboratory(PNNL), Richland, WA

van Asselen S, Verburg PH (2013) Land cover change or land-useintensification: simulating land system change with a global-scaleland change model. Glob Change Biol 19:3648–3667. doi:10.1111/gcb.12331

van Vliet J, Bregt AK, Hagen-Zanker A (2011) Revisiting Kappa toaccount for change in the accuracy assessment of land-use changemodels. Ecol Model 222:1367–1375. doi:10.1016/j.ecolmodel.2011.01.017

Verburg PH, Erb K-H, Mertz O, Espindola G (2013) Land systemscience: between global challenges and local realities. Curr OpinEnviron Sustain 5:433–437

Verburg PH, Overmars KP (2009) Combining top–down andbottom–up dynamics in land use modeling: exploring the futureof abandoned farmlands in Europe with the Dyna-CLUEmodel. Landsc Ecol 24:1167–1181. doi:10.1007/s10980-009-9355-7

Verburg PH, Schulp CJE, Witte N, Veldkamp A (2006) Downscalingof land use change scenarios to assess the dynamics of Europeanlandscapes. Agric Ecosyst Environ 114:39–56. doi:10.1016/j.agee.2005.11.024

Wang XH, Lu CH, Fang JF, Shen YC (2007) Implications fordevelopment of grain-for-green policy based on cropland suit-ability evaluation in desertification-affected north China.Land Use Policy 24:417–424. doi:10.1016/j.landusepol.2006.05.005

Wang XM, Zhang CX, Hasi E, Dong ZB (2010) Has the Three NorthsForest Shelterbelt Program solved the desertification and duststorm problems in arid and semiarid China? J Arid Environ74:13–22. doi:10.1016/j.jaridenv.2009.08.001

Wu FL (2002) Calibration of stochastic cellular automata: the appli-cation to rural-urban land conversions. Int J Geogr Inf Sci16:795–818. doi:10.1080/13658810210157769

Wu JG (2013) Landscape sustainability science: ecosystem servicesand human well-being in changing landscapes. Landsc Ecol28:999–1023

Environmental Management

Page 16: Understanding Land System Change Through Scenario-Based ...leml.asu.edu/jingle/Wu-Publications-PDFs/2017/Liu_etal-2016... · Environmental Management DOI 10.1007/s00267-016-0802-3

Wu JG, He CY, Zhang QY, Yu DY, Huang GL, Huang QX (2014)Integrative modeling and strategic planning for regional sustain-ability under climate change. Adv Earth Sci 29:1315–1324. inChinese

Wu JG, Zhang Q, Li A, Liang CZ (2015) Historical landscapedynamics of Inner Mongolia: patterns, drivers, and impacts.Landsc Ecol 30:1579–1598. doi:10.1007/s10980-015-0209-1

Xu X, Gao Q, Liu YH, Wang JA, Zhang Y (2009) Coupling a land usemodel and an ecosystem model for a crop-pasture zone.Ecol Model 220:2503–2511. doi:10.1016/j.ecolmodel.2009.04.043

Yang XH, Ci LJ, Zhang XS (2008) Dryland characteristics and itsoptimized eco-productive paradigms for sustainable developmentin China. Nat Resour Forum 32:215–227. doi:10.1111/j.1477-8947.2008.00201.x

Zhang F et al. (2013) Spatial and seasonal variations of pesticidecontamination in agricultural soils and crops sample from anintensive horticulture area of Hohhot, North-West China. EnvironMonit Assess 185:6893–6908. doi:10.1007/s10661-013-3073-y

Zhu L (2006) Dynamics of desertification and sandification in China.China Agricultural Press, Beijing, in Chinese

Environmental Management

View publication statsView publication stats