A global assessment of market accessibility and market ...

IOP PUBLISHING ENVIRONMENTAL RESEARCH LETTERS

Environ Res Lett 6 (2011) 034019 (12pp) doi1010881748-932663034019

A global assessment of market accessibilityand market influence for globalenvironmental change studies

Peter H Verburg1 Erle C Ellis2 and Aurelien Letourneau3

1 Institute for Environmental Studies Amsterdam Global Change Institute VU UniversityAmsterdam De Boelelaan 1087 1081 HV Amsterdam The Netherlands2 Department of Geography amp Environmental Systems University of Maryland BaltimoreCounty Baltimore MD 21250 USA3 UMR 5175 Centre drsquoEcologie Fonctionnelle amp Evolutive Centre National de la RechercheScientifique 1919 Route de Mende 34293 Montpellier cedex 5 France

E-mail PeterVerburgivmvunl

Received 21 April 2011Accepted for publication 2 August 2011Published 19 August 2011Online at stacksioporgERL6034019

AbstractMarkets influence the global patterns of urbanization deforestation agriculture and other landuse systems Yet market influence is rarely incorporated into spatially explicit global studies ofenvironmental change largely because consistent global data are lacking below the nationallevel Here we present the first high spatial resolution gridded data depicting market influenceglobally The data jointly represent variations in both market strength and accessibility based onthree market influence indices derived from an index of accessibility to market locations andnational level gross domestic product (purchasing power parity) These indices show strongcorrespondence with human population density while also revealing several distinct and usefulrelationships with other global environmental patterns As market influence grows the need forhigh resolution global data on market influence and its dynamics will become increasinglyimportant to understanding and forecasting global environmental change

Keywords accessibility land use environmental change markets economy global GDPworld

1 Introduction

Human interactions with local environments have far reachingconsequences for the functioning of the Earth systemincluding global climate biodiversity and biogeochemistry(Rockstrom et al 2009) Humanndashenvironment interactionsvary greatly in type intensity and duration across Earthrsquos landsurface depending on a wide variety of dynamic influencesat global regional and local scales including climate landsuitability for use governance and economic systems andtheir history of interaction at specific locations (Ellis andRamankutty 2008 Gallup et al 1999 Liverman and Cuesta2008 Rindfuss et al 2008) For this reason high spatialresolution data on both human and environmental systems

are needed to understand and assess the local causes andconsequences of global environmental change processes drivenby human interactions with land

High spatial resolution global data for land cover soilsand other biophysical variables are now widely availablefrom remote sensing and the coordinated efforts of globalobservation networks (Achard et al 2007 Batjes 2009 Heroldet al 2008 Hijmans et al 2005 Sanchez et al 2009 Schneideret al 2009) High spatial resolution data on human systemsare much less available (Hibbard et al 2010 Verburg et al2011) While global agencies including the FAO and theWorld Bank produce annual harmonized inventories of manyhuman variables at national level these data are not generallyavailable at sub-national scales As a result global level

1748-932611034019+12$3300 copy 2011 IOP Publishing Ltd Printed in the UK1

Environ Res Lett 6 (2011) 034019 P H Verburg et al

analysis of environmental change tends to be biased towardbiophysical processes or restricted to national levels (Rudel2009) Exceptions include high resolution gridded data forhuman population density (Dobson et al 2000 ORNL 2008)the use of land for urban settlements crops pastures andlivestock created by disaggregating national and sub-nationaldatasets using spatial models (Achard et al 2007 Heroldet al 2008 Klein Goldewijk et al 2011 Kruska et al 2003Ramankutty and Foley 1998 Schneider et al 2009) and evenfor crop management (Monfreda et al 2008 Sacks et al 2010)

For socioeconomic variables globally standardized datahave been limited to population density and gross domesticproduct (GDP) related measures Most socioeconomic datasetsare the result of spatial downscaling of (sub)national statisticswith the help of topographic data (Baer 2009) Gallup et al(1999) produced the first global gridded map of lsquoGDP densityrsquoby multiplying national level GDP by human populationdensity Poverty data have been downscaled by nighttimelights observed from remote sensing (Elvidge et al 2009) Inthis study correlations between national level poverty statisticsand average national level nighttime light intensity were usedto assign poverty values to individual pixels Doll et al(2006) similarly use nighttime light intensity to map regionaleconomic activity Influential maps of economic activity havebeen prepared using national and sub-national statistics andglobal gridded population data (Nordhaus 2006 Nordhaus andChen 2009) These studies show that variation within nationsof such socioeconomic parameters is often larger than variationof average values between nations If only average valuesfrom national level datasets are used there is a large risk formisinterpretations due to the many non-linear relations withinhumanndashenvironment interactions and the notion of ecologicalfallacy (Easterling 1997)

Markets have become one of the most important factorsdriving human activities and their interactions with the globalenvironment Markets link local activities to larger regionsand global processes through trade For farmers marketsare a means to sell their products to consumers while at thesame time markets provide access to inputs such as fertilizerand pesticides to increase production (Keys and McConnell2005) Markets also provide a strong incentive for investmentsand production choices (Chomitz and Gray 1996 Walker2004) In one of the early theories on the geography of landuse Von Thunen describes land use choices as a result ofmarket prices and transport costs to the market The locationof markets is therefore since long considered an importantdeterminant of land change especially in a strongly globalizingworld Peet (1969) describes the role of markets within thecolonial system indicating that markets influence productiondecisions over large distances Since colonial times globaltrade has increased many fold and improved accessibility hasmade global market conditions even more important (Britz andHertel 2011 Lambin and Meyfroidt 2011 Meijl et al 2006)Market access is also listed as one of the main determinantsof deforestation in a meta-analysis of case studies aroundthe world (Geist and Lambin 2002 Rudel 2005) and is usedin many regional studies of land change as an importantdeterminant (Nelson et al 2004 Pfaff 1999 Verburg et al2004)

Transportation infrastructure largely determines the accesspeople have to markets and this market access tendsto drive further infrastructure expansion (Hansen 1959)Infrastructure construction and operation is by itself aprimary driver of environmental change (Doyle and Havlick2009) Infrastructure expansion and associated environmentalchanges are driven by economic demand for the services thatinfrastructure provides combined with the political will andability to facilitate the implementation of the infrastructureconstruction and operational programs As a result theprocesses of increasing market influence and improving marketaccess tend to be interwoven

For global-scale studies high spatial resolution data onthe influence of markets is lacking This letter developsand demonstrates the first high spatial resolution globaldatasets of market influence indices including their strengthsand weaknesses as predictors of the global spatial patternsof land use and land cover human populations biomesanthromes (anthropogenic biomes globally significantecological patterns created by sustained interactions betweenhumans and ecosystems) plant species richness and netprimary production

2 Data and methods

21 Development of market influence indices

To derive a globally consistent indicator of market influencethe concept of market influence was matched with availableindependent global datasets It is assumed that marketinfluence at a specific location is determined as a function ofaccessibility to markets and the importance of these marketsWhile the importance of individual markets is difficult tomeasure and consistent data at the level of individual marketsare lacking national level GDP is a general indicator of marketimportance across individual nations as it measures a nationrsquosoverall economic output in terms of the market value of allfinal goods and services made within the borders of a nation ina year

Accessibility to markets may be measured in manydifferent ways and a wide literature of accessibility measuresis available (Geurs and van Wee 2004 Kwan et al 2003 Leiand Church 2010) Accessibility is loosely defined by Ingram(1971) as the inherent characteristic (or advantage) of a placewith respect to overcoming some form of spatially operatingsource of friction (for example time or distance) Differentauthors have discussed various measures of accessibilityvarying from simple line distance between two locationsto measures that account for the infrastructure network andtravel costs Distance measures (also called connectivitymeasures) are the simplest class of location-based accessibilitymeasures We have chosen to base the accessibility not solelyon distance but also account for the infrastructure and a numberof terrain characteristics that impede access to the marketsTherefore our measure is based on travel time rather thanon the absolute distance Many studies have modified suchsimple location-based accessibility measures by accountingfor some aspects related to behavior and perception In

2


Table 1 List of global datasets used for calculating the market influence index

Variable YearSpatialcharacteristics Source

Road network rivers 1979ndash1999 Vector map 11 M National Geospatial Intelligence Agency (NGA)VMAP0

Slope mdash Slope derived fromresampled altitudedata so the slopeonly captures theoverall topographyand is no measureof the real slope

Based on SRTM elevation data (Farr et al 2007)resampled to 1 km

Wetlands Approx1990ndash2000

30 s resolution mapcontaining differentwetland types

(Lehner and Doll 2004)

Cities gt750 000 2003 Point data Selected from UNEP major urban agglomerationdatabase (wwwgeodatagridunepch)

Cities gt50 000 Approx2000

Point data Database compiled by the Joint Research Centre of theEuropean Commission (Nelson 2008) based on theGPW database CIESIN Columbia University and theWorld Bank database of air pollution in World cities

Maritime ports 2005 Point data harborswith size lsquolargersquo areselecteda

Global Maritime Ports Database produced by GeneralDynamics Advanced Information Systems

Population density 2000 30 s resolution grid GPWv3 database Center for International EarthScience Information Network (CIESIN) ColumbiaUniversity and Centro Internacional de AgriculturaTropical (CIAT) 2005 Gridded Population of theWorld Version 3 (GPWv3) Palisades NYSocioeconomic Data and Applications Center(SEDAC) Columbia University Available athttpsedacciesincolumbiaedugpw

Gross domesticproduct (on apurchasing powerparity basis)

2010 in caseof missingdata earlieryears

National level CIA World factbook(wwwciagovlibrarypublicationsthe-world-factbook)

a The size classification in the database is based on a combination of attributes including area facilities and wharfspace Ports classified as large must at least be able to accommodate vessels over 500 feet

so-called potential accessibility measures the influence of alocation is diminishing by travel time (Geurs and van Wee2004 Guy 1983 Ingram 1971) By integrating measures ofeconomic strength and accessibility we have created a simpleand straightforward indicator for market influence

22 Data

A variety of publicly available global datasets were used incalculating market influence indices (table 1) For a numberof variables different alternative datasets were available anda selection was made based on global consistency and fitwith the specific application International data on roadsare extremely patchy and inconsistent with frequent gapsand many large changes in time that are often quicklyreversed (Canning 1998) Most available datasets arebased on sources of individual nations (Nelson et al 2006)Different nations define roads differently and the definitionof a road often changes within nations over time Ruralroads above a certain quality threshold are often centrallycontrolled while urban roads are controlled by municipalauthorities leading to an underreporting of urban and low-

quality rural roads controlled by the central authority Wehave used the VMAP0 database of infrastructure given itspublic availability and global consistency as compared tomore recent databases Though more recent databasescontain much more detail in road patterns for a numberof nations differences in detail between countries is alsomuch greater derailing the construction of globally consistentaccessibility indices (Nelson et al 2006) Rivers werealso taken from VMAP0 which is generally considered toprovide the most comprehensive and consistent global rivernetwork data currently available It is based on the USDMA (now NGA) Operational Navigation Charts Althoughmore detailed satellite-based river network data are available(Lehner 2005) these include many smaller streams that arenot navigable and therefore not adding to accessibility Grossdomestic product (GDP) values on a purchasing power parity(PPP) basis were obtained from the CIA Factbook and a linkwith a map of national territories was made to allow spatialrepresentation GDP measured in PPP was chosen over GDPmeasured at market exchange rates to standardize internationalcomparisons

3


Figure 1 Overview of the different steps involved in calculating the market influence indices

23 Calculation of market access and market influenceindices

Calculation of market influence indices consisted of twodistinct steps (figure 1) first an index of access to national andinternational markets was calculated and second two indicesof market influence one by combining national GDP datadirectly with the access index and the other by downscalingnational GDP using a measure of economic density Allcalculations are made at a spatial resolution of 1 km2 in EckertIV Equal Area Projection using a geographical informationsystem (GIS)

231 Market access index The calculation of market accessis based on a set of destinations that people travel to anda measure of the costs of traveling either in distance timeor monetary costs For this study we have used two groupsof locations as destinations The first group represents largedomestic and international markets Consistent data are notavailable at global scale representing the locations of marketsTherefore as a proxy for these locations cities or urbanagglomerations with more than 750000 inhabitants were usedCitiesagglomerations of such a size are in any case locationsof important domestic markets while they often have airportsimportant for the importexport of the country In addition tothese large maritime ports are included as important locationsrepresenting the influence of international markets The secondgroup represents locations that are important destinations asregional and domestic markets All towns and cities with apopulation of more than 50000 inhabitants were selected

Travel time accounting for infrastructure and some aspectsof terrain was used to measure the accessibility to the selecteddestinations Although costs of travel means of transport andthe quality of infrastructure have a high variation a globallyuniform approach was used to represent differences in traveltime to each of the two groups of destinations Table 2 providesan overview of the assumed velocities to calculate the totaltravel time Assumed velocities are within the range of 75ndash100 of the most common speed limits for the different road

Table 2 Speed assumed for different types of infrastructure andterrain to calculate the travel time to the nearest market

Infrastructureterrain type Speed (km hminus1)

Highways 100Primarysecondary roads 65Tertiary roads 40Railways 70Large rivers 10Canals 5Seas lakes and reservoirs 2Off-road

Flat or gentle slopes small rivers 5Moderate slopes 3Steep slopes 1Marshes Swamps Bogs Peatlandsa 3

International borders Speed is dividedby 10 over adistance of 1 km

a Only wetland classes covering more than 50 of thedesignated area in the database are considered

types while the speed on large rivers is accounting for waitingtimes at sluices etc It is assumed that it is also possible totravel outside the infrastructure network represented in the dataas many smaller roads are available Off-road speed is howeverassumed to be relatively low to account for deviations fromstraight-line connections especially in mountain and wetlandlandscapes that pose barriers and often have a lower densityof smaller roads Because the calculated travel times areconverted to an index it is the relative speed across differenttypes of infrastructure and terrain that is important ratherthan the absolute values Air transportation is ignored in theanalysis as it mainly links urban areas that are designatedas destinations in the analysis and consequently have a highaccessibility

The calculated travel times to the two different types ofdestination are integrated into one index accounting for travelbehavior As distance or travel time increases people are lesslikely to travel to certain locations and curvilinear functionsof distance are more suitable than linear relationships Ingram

4


(1971) and Guy (1983) compare different functional forms andconclude that a Gaussian curve is the most applicable for thequantitative measurement of accessibility A Gaussian functionhas a slow rate of decline in the region close to the origin thusallowing for the zone where the frictional effects of distance onaccessibility is low following

ai j = Sj exp

(minus d2

i j

v

)(1)

where ai j is the relative accessibility of point i to destinationj and di j is the distance between points i and j For eachlocation i ai j is calculated for the closest destination j forrespectively the nationalinternational market locations and theregional markets The importance (size or frequency of visit)of destination j is Sj and v is a constant specific to the studyIn our study we have assigned Sj a value of 1 for the nationaland international market locations (including ports) and a valueof 05 for the regional market locations This arbitrary choiceof values allows us to distinguish the influence of the differenttypes of market Within equation (1) v is a constant Accordingto Ingram (1971) this constant may be set at the averagesquared distance between all points considered However norational is provided Guy (1983) indicates that a value for v

may be chosen such that the steepest part of the graph (that isthe point of inflexion) is at a predetermined distance from theorigin In this case it can be shown that

v = 2d2lowast (2)

where dlowast is the distance from i at which accessibility is deemedto decline at the most rapid rate Since we assume thatthe influence of large market locations is stronger than theinfluence of regional markets we have calculated the constantfor the two groups of destinations with values for the inflectionpoint of respectively 2 h for large markets and 45 min forsmaller regional markets In the final accessibility index wehave used the maximum value ai j for the nationalinternationalmarkets and the regional markets Figure 2 provides anillustration of the decline in accessibility index with the traveltime from a large market

232 Market influence indices Two different indices ofmarket influence were developed (figure 1) The simplestmarket influence index characterizes market influence bysimply multiplying the accessibility index by national level percapita GDP values independent of population density dataThis creates an index of market influence expressed in $ percapita (market access is dimensionless) essentially treatingmarket influence as the multiplicative effect of local marketaccess and national market importance (GDPcapita measuredin PPP) A second index market density incorporatespopulation density data in an effort to allocate marketimportance (GDPcapita) across space before combining itwith market access thereby describing market influence as adensity expressed in $ kmminus2 This is accomplished by dividingthe local market accessibility index by the national average ofthe market access index and then multiplying this by the PPPand population density as illustrated in figure 1 The globaldata for both indices are available for download at wwwivmvunlmarketinfluence

Figure 2 Illustration of the decrease in market access index withtravel time from the location of a large city At 150 and 250 min fromthe large city two regional markets are located

24 Analysis of market influence in relation to other globalpatterns

To investigate the utility of our new global market indicestheir relationships with existing global data for human andenvironmental variables were assessed Global land cover datawere obtained from the GlobCover v22 dataset (httpionia1esrinesaint) and simplified into ten land cover classes bymerging forest and scrubland types Spatial data at 5 arc minresolution were obtained for human population density in year2000 (Klein Goldewijk et al 2010 based on Landscan (Dobsonet al 2000)) anthromes (Ellis et al 2010) potential vegetationbiomes (Ramankutty and Foley 1999) potential net primaryproductivity (NPP Haberl et al 2007) and potential plantspecies richness in regional landscapes (based on Kreft andJetz 2007) For analysis population density NPP and plantspecies richness were stratified into 11 classes (including alsquozero classrsquo) covering their full range of variation Global datawere processed at 5 arc minute resolution using zonal statisticsin a GIS to create a database allowing calculation of land area-weighted statistics for market indices and population density inrelation to other global patterns

3 Results

31 Spatial patterns in market influence indices

Figure 3 illustrates global patterns in market influencedescribed by each of our three indices As expected marketinfluence is strongest near large cities in all indices decliningto zero in regions without human populations Also asexpected the market access index is saturated near 10 in largecities (figure 3(a)) declining to moderate values in denselypopulated regions which have an abundance of smaller citiesand towns as intercity distances are so small that the indexnever declines below 02 Examples of such regions are WestAfrica Central America and parts of South America India andChina

The market influence index (figure 3(b)) differs fromthe market access index most clearly in areas with similar

5


Figure 3 Global overview of the market access index (A) the market influence index (B) and the market influence density index (C)

population densities but different levels of market strength(GDP per capita) These differences originate in the differenteconomic conditions of these regions For example India hashigh scores in the market access index similar to Europeyet the market influence index is much lower than Europe asa result of the regionrsquos relatively low GDP per capita Theeffects of differences in GDP are even clearer if we lookat the maps in more detail as in figure 4 which zooms into a part of the South-East Asian region Market accessis strong in Peninsular Malaysia around Kuala Lumpur andalso around the city of Medan in Indonesia The majordifferences between these two regions only become apparentwhen market influence is expressed in per capita units usingthe market influence index with the lower GDP of Indonesia

clearly creating different spatial patterns of market influencethan in Malaysia areas that are otherwise similar in terms ofpopulation densities and environmental conditions Only whenmarket influence is adjusted for human population densityas it is in the market density index do the most denselypopulated developing regions of the world tend to becomemore prominent as in China (figure 3(c)) and the island of Javain Indonesia (figure 4)

32 Global relationships between market influence andhuman and environmental patterns

Figure 5 illustrates global relationships between our threemarket indices and major global patterns in human population

6


Figure 4 Detailed views of the accessibility and market influence indices for part of the South-East Asian region

density land cover anthromes biomes potential NPP andpotential plant species richness (an indicator of biodiversity)Global relationships with human population density are alsodepicted at the top of the figure illustrating the strength ofthis most basic measure of human influence and allowingthe relative utility and additional strengths of different marketindicators to be observed Moreover the means medians andinter-quartile ranges in these charts make clear that the globalvariables used to assess relationships among variables arehighly non-linear skewed and full of inherent variation oftenbecause large numbers of low and zero values are combinedwith small numbers of extremely high values in some caseseven causing mean values to exceed the inter-quartile range

The strongest relationships evident in the charts of figure 5depict the strong positive correspondence between marketinfluence and human population density Urban areas inparticular tend to score lsquooff the chartsrsquo on all three indicesof market influence an obvious result given that the locationof cities was used to derive these indices The verystrong relationship of all three market influence indices withpopulation density outside urban areas is also unsurprisinggiven that maps of market access and population density areboth derived from similar input maps including not onlysettlement maps but also maps of transportation and land use

Therefore the distributions of other indicators with respectto market influence are more interesting especially those thatdeviate from patterns indicated by population density by itself

The clear global relationships between market influenceand agricultural land cover in figure 5 appear to correspond tothe classical lsquoVon Thunenrsquo patterns in which (intensive) arableagriculture predominates in areas of highest market influencefollowed by mosaic landscapes (cropnatural being mostlycrops naturalcrops vice versa) grazing land and savannahsand natural land cover types All of the market influenceindices show these patterns as does human population density

Relationships between anthrome classes and marketinfluence indices also resemble those with population densitywith notable exceptions Villages tended toward higherpopulation densities than dense settlements yet their marketaccess was similar and their market influence tended to besignificantly lower in terms of market density This makessense in that villages generally persist only in developingnations with long histories of subsistence economies whiledense settlements which have low levels of agriculturalland use mostly occur in wealthier nations relying solelyon commercial agricultural systems Comparisons amongcroplands rangelands and semi-natural anthromes withsimilar population densities are also interesting (lsquoResidentialrsquo

7


Figure 5 Global patterns in population density market access and market influence indices in relation to population density land cover(sorted by access value) anthromes (Ellis et al 2010) biomes (Ramankutty and Foley 1999) potential net primary productivity (NPP Haberlet al 2007) and potential plant species richness (Kreft and Jetz 2007) Colored bars are area-weighted means diamonds are medians errorbars depict inter-quartile range

anthromes have 10ndash100 lsquoPopulatedrsquo 1ndash10 and lsquoRemotersquoanthromes lt1 persons kmminus2) As predicted by Von Thunenmarket access and influence are greater in croplands than inrangelands and semi-natural lands Further market influenceis on average higher in semi-natural woodlands than inrangelands indicating perhaps the abandonment of agricultureor conservation of nonagricultural areas in more marketinfluenced areas

Relationships between markets and biomes NPP andplant species richness are weaker than those with anthromesand anthropogenic land cover The strongest relationshipobserved was between market influence and biomes witha significantly higher market influence by all indices in thetemperate woodlands confirming the prevalence of denseand wealthy populations across the temperate woodlandsInterestingly market density and market access appearedto be more strongly associated with temperate woodlandsthan population density a fairly exceptional result across allmeasures in figure 5 Population density appeared to haveslightly stronger relationships with NPP and plant species

richness than did market influence indices with very low levelsof NPP and plant species richness strongly associated with lowpopulations market access and influence and all of these haveon average higher at intermediate middle levels of NPP andplant species richness though inherent variation in the valuesis greater than the apparent trends

4 Discussion and conclusions

41 Evaluation of results

The global market influence indices presented in this letteroffer an important addition to the data available for analysisand modeling of global environmental change and land useExisting global models such as IMAGE (Eickhout et al 2007van Vuuren et al 2010) which is used in a wide range ofglobal environmental assessments account for accessibilityeffects using only a simple distance to city measure Themarket influence data presented here can be used directlyin IMAGE and other such models Compared to earlier

8


published global accessibility datasets (Nelson 2008) this newdataset distinguishes different types of destinations includingimportant maritime ports and accounts for the strength ofmarkets allowing derivation of multiple distinct and usefulindicators of the global patterns of market influence Theinclusion of ports in this study is especially important asthese can drive environmental changes such as deforestationand plantation development wherever these are constructed tosupport commodity export markets such as bananas in centralAmerica (Kok and Veldkamp 2001)

The market influence indices presented here do have somesimilarities with earlier efforts to downscale gross domesticproduct The simplest method for determining the spatialspread of GDP is to assume that GDP is equally distributedamong the inhabitants of a nation or region so that the spreadand influence of GDP is directly related to the distributionof population (Metzger et al 2010 Nordhaus 2006) TheG-Econ database uses sub-national GDP values as a basisfor such downscaling thereby capturing to some extent thedifferences in GDPcapita within countries (Nordhaus andChen 2009) The representation in this database does howevernot explicitly incorporate income differences between urbanand rural regions and it also creates data that are very tightlyand artifactually correlated with population density Grubleret al (2007) have downscaled national level GDP values usingincome distribution statistics by assuming that the richest 20of a countryrsquos population reside in urban areas The remaining80 of the population and their income share is assumed to bedistributed according to the remaining ruralndashurban populationdistribution A third alternative to account for sub-nationaldifferences in per capita GDP is the use of nighttime lightsas a measure of economic activity (Doll et al 2006 Suttonand Costanza 2002) While the indices presented hereespecially market density in some ways resemble existingGDP downscaling efforts they differ in one very importantway all three of the new indicators couple market strength(GDP) with variations in market access across Earthrsquos landusing detailed infrastructure datasets enabling much higherspatial resolutions than earlier studies

42 Relating markets to land use and anthropogenic globalchange

Associations between market influence indices and land coverdemonstrate that the most intensive modifications of naturalland cover urbanization land clearing and cultivation arefound in areas with the highest market influence In thatsense market influence is a proxy for the intensity of humanalteration of natural environments and shows similar patternsas the human footprint calculated by Sanderson et al (2002)However such an interpretation should be made with extremecare large areas in South Asia and Africa are denselypopulated with intensive agricultural systems focused mainlyon subsistence In spite of the large impact of these systems onthe environment the influence of markets is (still) relativelysmallmdashand this can be differentiated by comparing marketdensity with the market influence index with lower values ofthe latter highlighting subsistence regions

Besides altering land cover patterns market access isalso believed to determine the intensity of land managementpractices including use of irrigation fertilizers and otheragrichemicals (Lambin et al 2001) Easy access to marketsis likely to raise land prices and favor intensive cultivationpractices and access to inputs such as fertilizers and otherchemicals (Keys and McConnell 2005 Verburg et al 20002004) In a global-scale analysis of the spatial distribution ofgrain yields Neumann et al (2010) used the market influenceindex presented in this letter as one of the factors explainingthe gap between actual yield and the highest attainable yield(as defined by a frontier function) For all three crops analyzed(rice wheat maize) the market influence index yielded asignificant relation with the efficiency in production ie uponhigher values of the market influence index yields tended to becloser to the maximum attainable yield The authors used boththe market influence and a standard accessibility index Bothindicators were significant in the estimated regression modelsclearly indicating the additional value of the market influenceindex to the more traditional accessibility indicators

There is a close relationship between population densityand all market influence indices (figure 5) Is this just anartifact of the similar inputs used to map these variables oris there a good theoretical reason for this strong relationshipTheory would indicate the latter Markets emerge in space asa function of population density with low densities capableof feeding themselves without markets and having less laborand consumptive demand to offer the marketplace whilehigher densities would increase the need for support for andadvantages of the marketplace However at the same timethe results indicate that not all regions with high populationdensities evolve into strong markets Examples of regionswith high population densities and low values of the marketinfluence index are Nigeria Ethiopia and the rural areas ofSouthern Asia (eg large parts of Bangladesh) Many of suchlandscapes with high population densities rely on subsistencefarming and have poorly developed markets Moreoverintegration of the rural hinterland into the market economystrongly depends on infrastructural conditions It is especiallythese aspects that are captured in our new indices together withpatterns depending more on market forces than on populationpatterns especially for the market access and influence indiceswhich were derived independent of population data

Accessibility and economic development are also relatedto other geographic factors such as terrain and climateThe provision of infrastructure is significantly correlatedwith geography particularly for poorer countries probablybecause the costs and benefits of infrastructure vary withgeography This implies that the impact of infrastructureon economic growth may depend on geography and thatgeographical considerations should be taken into accountwhen analyzing these effects (Canning and Pedroni 2008Krugman 1999) Although such relations have been exploredby several authors (Nordhaus 2006) it is obvious that largeregional deviations occur Therefore the inclusion of dataon the spatial distribution of socioeconomic conditions inglobal environmental change studies will remain essentialThe high spatial resolution indicator datasets presented in this

9


letter can potentially make an important contribution to globalchange assessments and models by addressing one of the mostimportant dimensions of humanndashenvironment processes themarketplace

43 Limitations

As in any global analysis data available for use in our analysisare subject to the inconsistencies and other limitations ofinternational socioeconomic data collections (Verburg et al2011) Publicly available road data in most regions areoutdated and major extensions have been made to the roadsystem in recent years perhaps most notably in China Alsothe level of detail of road maps is inconsistent between nationsleading to further bias

Our use of an arbitrary cut-off for city size does notnecessarily reflect the relative strength and global importanceof their markets Some cities with less than 750 000 inhabitantsare important international markets and are disregarded in thisanalysis Also the selection of ports does not fully account forthe type of commodities shipped in the port Finally the weightgiven to respectively large and smaller market locations wasarbitrarily chosen similar to the distance decay coefficient ofthe Gaussian function of the accessibility index Although allchoices were discussed at various workshops and accountedfor literature and observations it is clear that for differentregions such values might best be chosen differently In thisstudy it was decided to derive a consistent global measureA sensitivity analysis on these underlying assumptions revealsthat although locally the patterns may change the overall globalpattern in the market influence index remains the same Whensub-national data on GDP become more easily available fora larger range of countries (preferably all of them) it willbecome possible to specify the strength of regional markets inmore detail based on sub-national GDP levels (Nordhaus andChen 2009)

In a world where large amounts of money are spentto monitor global biophysical variables from space it isremarkable that there is no international effort or institutionin place to annually obtain and share globally consistent highspatial resolution socioeconomic data such as sub-nationalGDP demographics and road data Global environmentalchanges are increasingly driven by the dynamics and spatialpatterning of market forces Given appropriate data thesimple and straightforward indicators presented here wouldmake it possible to assess changes in market influence onlocal environments as rapidly as infrastructure and GDP datacould become available Under a more advanced regime ofglobal socioeconomic data gathering and distribution thesemaps could be used to monitor changes in the global patternsof market influence on global environmental change

44 Applications

The three different market indices developed in this letterrepresent three different aspects of market influence on landuse decisions Choice of an optimal market index or indicesfor a specific application will therefore depend on the purposeof the analysis The market access index is the simplest

measure indicating only the degree to which market accesstime and the relative scale of markets (large cities versustowns) produces the global patterns of market influenceindependent of total population size or wealth This indexis therefore the best measure for applications in which thefixed infrastructure of the marketplace is the main interest(density of market locations and transportation infrastructure)The market influence index expands on this infrastructuraleffect by also incorporating the effects of regional and nationalvariations in economic power expressed by GDP making thisindex the most useful for investigating the more dynamic anddevelopment-related effects of the marketplace but withoutadjusting for variations in population size The marketdensity index is the most comprehensive measure taking intoaccount variations in market infrastructure national economicpower and population size potentially offering the most usefulindicator of the overall influence of the marketplace on landuse decisions However in studies that intend to considervariations in population density as an independent variablethe Market density index should be avoided because it alreadyincludes population density making Market influence indexthe better choice

Our results also highlight the large differences betweenurban and rural regions worldwide Many global databaseson social and economic characteristics disregard urbanruraldifferences by presenting national level GDP as a fixedinfluence across nations The new indicators we presentoffer the first spatially explicit global approximation of thedistribution of economic assets and influence within and acrossnations The large spatial variations observable in these mapsare in themselves drivers of a wide assortment of global changeprocesses (Grau and Aide 2008 Mcdonald et al 2009 Thurowand Kilman 2009) As a result these new datasets and indicesare the first step toward providing a high spatial resolutionglobal overview of regional and local disparities in economicdevelopment

Sanderson et al (2002) mapped a human impact indicator(human footprint) and Ellis and Ramankutty (2008) mappedthe global patterns sustained direct human interaction withterrestrial ecosystems (anthromes) using empirical analysesof existing global data These global assessments of humaninfluence on local environments are mere descriptions unableto fully explain or predict the causes of these global patterns inthe environmental changes driven by human activity Humaninteractions with local environments depend heavily on thestate of local economic development and the level of localinteraction with domestic and global markets (Meyfroidt andLambin 2009 Rudel et al 2005) It is therefore essential thatmarket influence be incorporated in future efforts to map andclassify anthromes and other more dynamic and robust globalrepresentations of human interactions with the environmentthat drive global environmental change

Acknowledgments

We acknowledge ESA and the ESA GlobCover Project led byMEDIAS FrancePOSTEL for using the GlobCover data Thisletter is based on work funded by the Netherlands Organization

10


for Scientific Research (NWO) The work presented inthis letter contributes to the Global Land Project (wwwgloballandprojectorg)

References

Achard F DeFries R Eva H Hansen M Mayaux P and Stibig H J2007 Pan-tropical monitoring of deforestation Environ ResLett 2 045022

Baer P 2009 Equity in climate-economy scenarios the importance ofsubnational income distribution Environ Res Lett 4 015007

Batjes N H 2009 Harmonized soil profile data for applications atglobal and continental scales updates to the WISE databaseSoil Use Manag 25 124ndash7

Britz W and Hertel T W 2011 Impacts of EU biofuels directives onglobal markets and EU environmental quality an integrated PEglobal CGE analysis Agric Ecosyst Environ 142 102ndash9

Canning D 1998 A database of world stocks of infrastructure1950ndash95 World Bank Econ Rev 12 529ndash47

Canning D and Pedroni P 2008 Infrastructure long-run economicgrowth and causality tests for cointegrated panels TheManchester School 76 504ndash27

Chomitz K M and Gray D A 1996 Roads land use anddeforestation a spatial model applied to belize World BankEcon Rev 10 487ndash512

Dobson J E Bright E A Coleman P R Durfee R C and Worley BA 2000 LandScan a global population database for estimatingpopulations at risk Photogramm Eng Remote Sens 66 849ndash57

Doll C N H Muller J P and Morley J G 2006 Mapping regionaleconomic activity from night-time light satellite imagery EcolEcon 57 75ndash92

Doyle M W and Havlick D G 2009 Infrastructure and theenvironment Annu Rev Environ Resources 34 349ndash73

Easterling W E 1997 Why regional studies are needed in thedevelopment of full-scale integrated assessment modelling ofglobal change processes Global Environ Change A 7 337ndash56

Eickhout B Van Meijl H Tabeau A and van Rheenen T 2007Economic and ecological consequences of four European landuse scenarios Land Use Policy 24 562ndash75

Ellis E C Klein Goldewijk K Siebert S Lightman D andRamankutty N 2010 Anthropogenic transformation of thebiomes 1700 to 2000 Global Ecol Biogeogr 19 589ndash606

Ellis E C and Ramankutty N 2008 Putting people in the mapanthropogenic biomes of the world Front Ecol Environ6 439ndash47

Elvidge C D Sutton P C Ghosh T Tuttle B T Baugh K EBhaduri B and Bright E 2009 A global poverty map derivedfrom satellite data Comput Geosci 35 1652ndash60

Farr T G et al 2007 The shuttle radar topography mission RevGeophys 45 RG2004

Gallup J L Sachs J D and Mellinger A D 1999 Geography andeconomic development Int Reg Sci Rev 22 179ndash232

Geist H J and Lambin E F 2002 Proximate causes and underlyingdriving forces of tropical deforestation Bioscience 52 143ndash50

Geurs K T and van Wee B 2004 Accessibility evaluation of land-useand transport strategies review and research directionsJ Transp Geogr 12 127ndash40

Grau R and Aide M 2008 Globalization and land-use transitions inLatin America Ecol Soc 13 16 wwwecologyandsocietyorgvol13iss2art16

Grubler A OrsquoNeill B Riahi K Chirkov V Goujon A Kolp PPrommer I Scherbov S and Slentoe E 2007 Regional nationaland spatially explicit scenarios of demographic and economicchange based on SRES Technol Forecast Soc Change74 980ndash1029

Guy C 1983 The assessment of access to local shopping EnvironPlan 10 219ndash38

Haberl H Erb K H Krausmann F Gaube V Bondeau A Plutzar CGingrich S Lucht W and Fischer-Kowalski M 2007

Quantifying and mapping the human appropriation of netprimary production in earthrsquos terrestrial ecosystems Proc NatlAcad Sci 104 12942ndash7

Hansen W G 1959 How accessibility shapes land use J Am InstPlan 25 73ndash6

Herold M Mayaux P Woodcock C E Baccini A andSchmullius C 2008 Some challenges in global land covermapping an assessment of agreement and accuracy in existing1 km datasets Remote Sens Environ 112 2538ndash56

Hibbard K Janetos A van Vuuren D P Pongratz J Rose S KBetts R Herold M and Feddema J J 2010 Research priorities inland use and land-cover change for the Earth system andintegrated assessment modelling Int J Climatol 30 2118ndash28

Hijmans R J Cameron S Para J L Jonges P G and Jarvis A 2005Very high resolution interpolated climate surfaces for globalland areas Int J Climatol 25 1965ndash78

Ingram D R 1971 The concept of accessibility a search for anoperational form Reg Stud 5 101ndash7

Keys E and McConnell W J 2005 Global change and theintensification of agriculture in the tropics Global EnvironChange A 15 320ndash37

Klein Goldewijk K Beusen A and Janssen P 2010 Long-termdynamic modeling of global population and built-up area in aspatially explicit way HYDE 31 The Holocene 20 565ndash73

Klein Goldewijk K Beusen A van Drecht G and de Vos M 2011 TheHYDE 31 spatially explicit database of human-induced globalland-use change over the past 12 000 years Global EcolBiogeogr 20 73ndash86

Kok K and Veldkamp A 2001 Evaluating impact of spatial scales onland use pattern analysis in Central America Agric EcosystEnviron 85 205ndash21

Kreft H and Jetz W 2007 Global patterns and determinants ofvascular plant diversity Proc Natl Acad Sci 104 5925ndash30

Krugman P 1999 The role of geography in development Int Reg SciRev 22 142ndash61

Kruska R L Reid R S Thornton P K Henninger N andKristjanson P M 2003 Mapping livestock-oriented agriculturalproduction systems for the developing world Agric Syst77 39ndash63

Kwan M-P Murray A T OrsquoKelly M E and Tiefelsdorf M 2003Recent advance in accessibility research representationmethodology and applications J Geogr Syst 5 129ndash38

Lambin E F and Meyfroidt P 2011 Global land use change economicglobalization and the looming land scarcity Proc Natl AcadSci 108 3465ndash72

Lambin E F et al 2001 The causes of land-use and land-cover changemoving beyond the myths Global Environ Change A 4 261ndash9

Lehner B 2005 HydroSHEDS Technical Documentation Version 10(Washington DC World Wildlife Fund US)

Lehner B and Doll P 2004 Development and validation of a globaldatabase of lakes reservoirs and wetlands J Hydrol 296 1ndash22

Lei T L and Church R L 2010 Mapping transit-based accessintegrating GIS routes and schedules Int J Geogr Inf Sci24 283ndash304

Liverman D M and Cuesta R M R 2008 Human interactions with theEarth system people and pixels revisited Earth Surf ProcessLandf 33 1458ndash71

Mcdonald R I Forman R T T Kareiva P Neugarten R Salzer D andFisher J 2009 Urban effects distance and protected areas in anurbanizing world Landsc Urban Plan 93 63ndash75

Meijl H v van Rheenen T Tabeau A and Eickhout B 2006 Theimpact of different policy environments on agricultural land usein Europe Agric Ecosyst Environ 114 21ndash38

Metzger M J Bunce R G H van Eupen M and Mirtl M 2010 Anassessment of long term ecosystem research activities acrossEuropean socio-ecological gradients J Environ Manag91 1357ndash65

Meyfroidt P and Lambin E F 2009 Forest transition in Vietnam anddisplacement of deforestation abroad Proc Natl Acad Sci106 16139ndash44

11


Monfreda C Ramankutty N and Foley J A 2008 Farming the planet2 Geographic distribution of crop areas yields physiologicaltypes and net primary production in the year 2000 GlobalBiogeochem Cycles 22 GB1022

Nelson A 2008 Travel Time to Major Cities A Global Map ofAccessibility (Luxembourg Office for Official Publications ofthe European Communities)

Nelson A de Sherbinin A and Pozzi F 2006 Towards development ofa high quality public domain global roads database Data Sci J5 223ndash65

Nelson G De Pinto A Harris V and Stone S 2004 Land use and roadimprovements a spatial perspective Int Reg Sci Rev27 297ndash325

Neumann K Verburg P H Stehfest E and Mnller C 2010 The yieldgap of global grain production a spatial analysis Agric Syst103 316ndash26

Nordhaus W D 2006 Geography and macroeconomics new data andnew findings Proc Natl Acad Sci USA 103 3510ndash7

Nordhaus W D and Chen X 2009 Geography graphics andeconomics B E J Econ Anal Policy 9 1doi1022021935-16822072

Peet J R 1969 The spatial expansion of commercial agriculture in thenineteenth century a von Thunen interpretation Econ Geogr45 283ndash301

Pfaff A S P 1999 What drives deforestation in the Brazilian AmazonEvidence from satellite and socioeconomic data J EnvironEcon Manag 37 26ndash43

Ramankutty N and Foley J A 1998 Characterizing patterns of globalland use an analysis of global croplands data GlobalBiogeochem Cycles 12 667ndash85

Ramankutty N and Foley J A 1999 Estimating historical changes inglobal land cover croplands from 1700 to 1992 GlobalBiogeochem Cycles 13 997ndash1027

Rindfuss R R et al 2008 Land use change complexity andcomparisons J Land Use Sci 3 1ndash10

Rockstrom J et al 2009 A safe operating space for humanity Nature461 472ndash5

Rudel T K 2005 Tropical Forests Regional Paths of Destruction andRegeneration in the Late Twentieth Century (New YorkColumbia University Press)

Rudel T K 2009 Tree farms driving forces and regional patterns inthe global expansion of forest plantations Land Use Policy26 545ndash50

Rudel T K Coomes O T Moran E Achard F Angelsen A Xu J andLambin E 2005 Forest transitions towards a globalunderstanding of land use change Global Environ Change A15 23ndash31

Sacks W J Deryng D Foley J and Ramankutty N 2010 Crop plantingdates an analysis of global patterns Global Ecol Biogeogr 19607ndash20

Sanchez P A et al 2009 Digital soil map of the world Science325 680ndash1

Sanderson E W Jaiteh M Levy M A Redford K H Wannebo A Vand Woolmer G 2002 The human footprint and the last of thewild Bioscience 52 891ndash904

Schneider A Friedl M A and Potere D 2009 A new map of globalurban extent from MODIS satellite data Environ Res Lett4 044003

Sutton P C and Costanza R 2002 Global estimates of market andnon-market values derived from nighttime satellite imageryland cover and ecosystem service valuation Ecol Econ41 509ndash27

Thurow R and Kilman S 2009 Enough Why the Worldrsquos PoorestStarve in an Age of Plenty (New York Public Affairs)

van Vuuren D P Stehfest E den Elzen M G J van Vliet J andIsaac M 2010 Exploring IMAGE model scenarios that keepgreenhouse gas radiative forcing below 3 Wm2 in 2100 EnergyEcon 32 1105ndash20

Verburg P H Chen Y Q and Veldkamp A 2000 Spatial explorationsof land-use change and grain production in China AgricEcosyst Environ 82 333ndash54

Verburg P H Neumann K and Nol L 2011 Challenges in using landuse and land cover data for global change studies GlobalChange Biol 17 974ndash89

Verburg P H Overmars K P and Witte N 2004 Accessibility and landuse patterns at the forest fringe in the Northeastern part of thePhilippines Geograph J 170 238ndash55

Walker R 2004 Theorizing land-cover and land-use change the caseof tropical deforestation Int Reg Sci Rev 27 247ndash70

12

1 Introduction

2 Data and methods


22 Data

23 Calculation of market access and market influence indices

24 Analysis of market influence in relation to other global patterns

3 Results


32 Global relationships between market influence and human and environmental patterns



42 Relating markets to land use and anthropogenic global change

43 Limitations

44 Applications

Acknowledgments

References


analysis of environmental change tends to be biased towardbiophysical processes or restricted to national levels (Rudel2009) Exceptions include high resolution gridded data forhuman population density (Dobson et al 2000 ORNL 2008)the use of land for urban settlements crops pastures andlivestock created by disaggregating national and sub-nationaldatasets using spatial models (Achard et al 2007 Heroldet al 2008 Klein Goldewijk et al 2011 Kruska et al 2003Ramankutty and Foley 1998 Schneider et al 2009) and evenfor crop management (Monfreda et al 2008 Sacks et al 2010)

For socioeconomic variables globally standardized datahave been limited to population density and gross domesticproduct (GDP) related measures Most socioeconomic datasetsare the result of spatial downscaling of (sub)national statisticswith the help of topographic data (Baer 2009) Gallup et al(1999) produced the first global gridded map of lsquoGDP densityrsquoby multiplying national level GDP by human populationdensity Poverty data have been downscaled by nighttimelights observed from remote sensing (Elvidge et al 2009) Inthis study correlations between national level poverty statisticsand average national level nighttime light intensity were usedto assign poverty values to individual pixels Doll et al(2006) similarly use nighttime light intensity to map regionaleconomic activity Influential maps of economic activity havebeen prepared using national and sub-national statistics andglobal gridded population data (Nordhaus 2006 Nordhaus andChen 2009) These studies show that variation within nationsof such socioeconomic parameters is often larger than variationof average values between nations If only average valuesfrom national level datasets are used there is a large risk formisinterpretations due to the many non-linear relations withinhumanndashenvironment interactions and the notion of ecologicalfallacy (Easterling 1997)

Markets have become one of the most important factorsdriving human activities and their interactions with the globalenvironment Markets link local activities to larger regionsand global processes through trade For farmers marketsare a means to sell their products to consumers while at thesame time markets provide access to inputs such as fertilizerand pesticides to increase production (Keys and McConnell2005) Markets also provide a strong incentive for investmentsand production choices (Chomitz and Gray 1996 Walker2004) In one of the early theories on the geography of landuse Von Thunen describes land use choices as a result ofmarket prices and transport costs to the market The locationof markets is therefore since long considered an importantdeterminant of land change especially in a strongly globalizingworld Peet (1969) describes the role of markets within thecolonial system indicating that markets influence productiondecisions over large distances Since colonial times globaltrade has increased many fold and improved accessibility hasmade global market conditions even more important (Britz andHertel 2011 Lambin and Meyfroidt 2011 Meijl et al 2006)Market access is also listed as one of the main determinantsof deforestation in a meta-analysis of case studies aroundthe world (Geist and Lambin 2002 Rudel 2005) and is usedin many regional studies of land change as an importantdeterminant (Nelson et al 2004 Pfaff 1999 Verburg et al2004)

Transportation infrastructure largely determines the accesspeople have to markets and this market access tendsto drive further infrastructure expansion (Hansen 1959)Infrastructure construction and operation is by itself aprimary driver of environmental change (Doyle and Havlick2009) Infrastructure expansion and associated environmentalchanges are driven by economic demand for the services thatinfrastructure provides combined with the political will andability to facilitate the implementation of the infrastructureconstruction and operational programs As a result theprocesses of increasing market influence and improving marketaccess tend to be interwoven

For global-scale studies high spatial resolution data onthe influence of markets is lacking This letter developsand demonstrates the first high spatial resolution globaldatasets of market influence indices including their strengthsand weaknesses as predictors of the global spatial patternsof land use and land cover human populations biomesanthromes (anthropogenic biomes globally significantecological patterns created by sustained interactions betweenhumans and ecosystems) plant species richness and netprimary production

2 Data and methods


To derive a globally consistent indicator of market influencethe concept of market influence was matched with availableindependent global datasets It is assumed that marketinfluence at a specific location is determined as a function ofaccessibility to markets and the importance of these marketsWhile the importance of individual markets is difficult tomeasure and consistent data at the level of individual marketsare lacking national level GDP is a general indicator of marketimportance across individual nations as it measures a nationrsquosoverall economic output in terms of the market value of allfinal goods and services made within the borders of a nation ina year

Accessibility to markets may be measured in manydifferent ways and a wide literature of accessibility measuresis available (Geurs and van Wee 2004 Kwan et al 2003 Leiand Church 2010) Accessibility is loosely defined by Ingram(1971) as the inherent characteristic (or advantage) of a placewith respect to overcoming some form of spatially operatingsource of friction (for example time or distance) Differentauthors have discussed various measures of accessibilityvarying from simple line distance between two locationsto measures that account for the infrastructure network andtravel costs Distance measures (also called connectivitymeasures) are the simplest class of location-based accessibilitymeasures We have chosen to base the accessibility not solelyon distance but also account for the infrastructure and a numberof terrain characteristics that impede access to the marketsTherefore our measure is based on travel time rather thanon the absolute distance Many studies have modified suchsimple location-based accessibility measures by accountingfor some aspects related to behavior and perception In

2





















22 Data



3















4



ai j = Sj exp

(minus d2

i j

v

)(1)



v = 2d2lowast (2)






3 Results




5







6








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References





















22 Data



3















4



ai j = Sj exp

(minus d2

i j

v

)(1)



v = 2d2lowast (2)






3 Results




5







6








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References















4



ai j = Sj exp

(minus d2

i j

v

)(1)



v = 2d2lowast (2)






3 Results




5







6








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References



ai j = Sj exp

(minus d2

i j

v

)(1)



v = 2d2lowast (2)






3 Results




5







6








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References







6








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References








7









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References









8









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References









9



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References



43 Limitations




44 Applications





Acknowledgments


10



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References



References
















































11





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References





























12

1 Introduction

2 Data and methods


22 Data



3 Results






43 Limitations

44 Applications

Acknowledgments

References

A global assessment of market accessibility and market ...

Documents

Transcript of A global assessment of market accessibility and market ...