Challenges and opportunities in mapping land use intensity globally

Challenges and opportunities in mapping land use intensity globally

Tobias Kuemmerle

, Karlheinz Erb

, Patrick Meyfroidt

, Daniel Mu¨ller

, Peter H Verburg

, Stephan Estel

, Helmut Haberl

, Patrick Hostert

,

Martin R Jepsen

, Thomas Kastner

, Christian Levers

, Marcus Lindner

, Christoph Plutzar

, Pieter Johannes Verkerk

, Emma H van der Zanden

and Anette Reenberg

Futureincreasesinland-basedproductionwillneedtofocus moreonsustainablyintensifyingexistingproductionsystems.

Unfortunately,ourunderstandingoftheglobalpatternsofland useintensityisweak,partlybecauselanduseintensityisa complex,multidimensionalterm,andpartlybecausewelack appropriatedatasetstoassesslanduseintensityacrossbroad geographicextents.Here,wereviewthestateoftheartregarding approachesformappinglanduseintensityandprovidea comprehensiveoverviewofavailableglobal-scaledatasetson landuseintensity.Wealsooutlinemajorchallengesand opportunities formapping land use intensity forcropland, grazing, andforestrysystems,andidentifykeyissuesforfutureresearch.

Addresses

1GeographyDepartment,Humboldt-UniversityBerlin,UnterdenLinden 6,10099Berlin,Germany

2EarthSystemAnalysis,PotsdamInstituteforClimateImpactResearch, 14412Potsdam,Germany

3InstituteofSocialEcologyVienna(SEC),Alpen-AdriaUniversita¨t Klagenfurt,Wien,Graz,1070Vienna,Austria

4GeorgesLemaitreEarthandClimateResearchCenter,EarthandLife Institute,F.R.S-FNRS&Universite´ CatholiquedeLouvain,1348Louvain- La-Neuve,Belgium

5LeibnizInstituteofAgriculturalDevelopmentinCentralandEastern Europe(IAMO),Theodor-Lieser-Str.2,06120Halle(Saale),Germany

6InstituteforEnvironmentalStudies,AmsterdamGlobalChange Institute,VUUniversity,Amsterdam,TheNetherlands

7DepartmentofGeographyandGeology,UniversityofCopenhagen, ØsterVoldgade10,DK-1350Copenhagen,Denmark

8EuropeanForestInstitute(EFI),SustainabilityandClimateChange Programme,Torikatu34,80100Joensuu,Finland

Correspondingauthor:Kuemmerle,Tobias(tobias.kuemmerle@geo.

hu-berlin.de)

CurrentOpinioninEnvironmentalSustainability2013,5:484–493 ThisreviewcomesfromathemedissueonHumansettlementsand industrialsystems

EditedbyPeterHVerburg,OleMertz,Karl-HeinzErbandGiovana Espindola

Availableonline29thJune2013

1877-3435#2013TheAuthors.Publishedby ElsevierB.V.

http://dx.doi.org/10.1016/j.cosust.2013.06.002

Introduction

Unless fundamental changes in consumption occur, land-based production of food, feed, fiber, and bioe- nergywillhavetoincreasesubstantiallytomeethuman- ity’s surging demands [1,2]. As land resources are becoming scarcer [3]much ofthis risein production must come from sustainably intensifying existing pro- duction systems [4]. Yet, land use science has so far mainly focusedon broad land coverconversions while thespatialpatternsintheintensityofcropland,grazing, and forestry systems remain highly unclear for most worldregions.

Thelackofdatasetstoadequatelyassesslanduseinten- sity and changes therein is particularly apparentat the globalscale,whereexistingdataonlanduseintensityare either coarse in scale (e.g. national-scale statistics) or connectedtoconsiderableuncertainties[5,6],orboth.

Existingdatagapstranslateintolargeuncertaintieswhen assessingtheworld’spotentialforincreasingland-based production,forminimizing theenvironmentaltrade-offs ofland use,or for assessingtheoutcomesofalternative landuse pathways suchas expansionversusintensifica- tion.Moreover,datagaps areparticularly indeveloping countries,which sometimes lackconsistent datacollec- tion and sharing frameworks, yet where land systems changeisextensive.

Threereasonsexplainthescarcityofglobal-scalelanduse intensitydatasets. First,landuse intensityisacomplex andmultidimensional phenomenon. Landuse intensity can refer to the land area farmed, the frequency of cultivation [7], the amount of capital-related inputs (e.g. fertilizer [8], irrigation [9], technology [10], or mechanization [11]), the crop yields from a particular area[12,13],ortheshare ofecosystem productivitythat isappropriated byhumans [14].Second, indicator defi- nitions may vary between disciplines or countries.

Finally, adequate approaches for measuring land use intensityandforintegratingvariousdatasourcesareoften missing(seeErbetal., thisissue).

Despite these issues, new opportunities are arisingto fill the existing data gaps and to derive new land use

Open access under CC BY-NC-SA license.

intensityindicators.Dataavailabilityisrapidlyimproving, and new algorithmsand computer processing capacities allowforbetteruseofthesedatasets.Here,ourgoalshere areto:

(a) Review approaches to measure and map land use intensityattheglobal scale,

(b) Provideanoverviewof spatiallyexplicitdatasetson land useintensity,and

(c) Outlineresearchgapsandopportunitiesformapping land useintensityglobally.

Measuring andmappinglanduse intensity

Conceptualframework

Ourconceptualframeworkoflanduseintensityfollows Erb et al. (this issue), which refers to land-based pro- ductioninabroadsense,includingagriculture,grazing, and forestry. In short,Erb et al.arguethat adequately addressinglanduseintensityand itsimpact onsociety andtheenvironmentrequiresconsideringthedifferent dimensionsoflanduseintensityinasystemicway.Land useactivitiestakeplaceinproductionsystems,whichare definedasintegratedsocio-ecologicalsystemswithboth biophysical (e.g. soils, climate, topography) and socio- economic properties (e.g. institutions, market integ- ration,population).Land-basedproductionthenencom- passes all activities that convert some combination of inputsintooutputs,dependentonthepropertiesofthe system(Figure1).Inputsintheclassicalsense referto the land area utilized, to capital (e.g. technology, mechanization, agrochemicals applied), andlabor (e.g.

theamountoflabor,knowledge)[15].Outputsreferto the production itself (e.g. harvests). Beyond outputs, land-based production impacts a range of ecosystem functions and services,as well asbiodiversity, human, social,andnaturalcapital,aswellaslandsystemresili- ence. These, usually unintended, impacts are here referred to as the outcomes of land-based production.

Measuringandmappingoutcomes,aswellasthetrade- offsbetweenproductionoutputandoutcomes(e.g.food versuscarbonstorageorbiodiversityloss),areattheheart of sustainability science, but beyond thescope ofthis manuscript.

Here, we focus on threetypes ofmetricsthat provide a quantitative, spatially explicit measure of land use intensityitselfandthusallowrankinglandusesystems orplaces accordingtotheirintensity(Figure 1):

(1) Inputmetricsmeasuretheintensityoflandusealong inputdimensions(e.g.fertilizer,croppingfrequency, rotation lengths).

(2) Output metrics relate outputs from the production systemto inputs(e.g.yields,capitalproductivity,or residue/felling ratiosinforestry).

(3) System metricsrelatetheinputsor outputsofland- basedproductiontosystemproperties(e.g.yieldgaps (actual versuspotential yield),human appropriation ofnetprimaryproduction(HANPP),orwoodfelling in relationto woodincrement).

Approachesformappinglanduseintensity

Approachesforderiving global-scalemetricsof landuse intensity at fine resolutions (i.e. 0.58 or finer) can be broadlygroupedintoapproachesbasedsolelyonremote sensingimageanalysis,andmethodsthatcombinesatel- lite observations with ground-based inventory data to derivegrid-levellanduse intensitymetrics(Table 1).

Satelliteremotesensing

Remotesensingisarguablythemostimportanttechnol- ogy available for mapping land use and land cover dynamicsacrossbroadgeographicextents.Imageaccess hassurgedoverthelastfewdecades,thespatial,spectral, and temporalresolution ofobservations haveincreased, anddataarchivescoverincreasinglylongertimeperiods, altogetherallowingformoredetailedassessmentsofland use changes than ever before. Strong advantages of remotesensingincludethesystematicacquisitionsetup, thespatially explicitnatureofmeasurements,and their consistencyacross politicalborders.Yet,land useinten- sitychangesareoftenrelatedtosubtlespectralchanges, andarethusnotoriouslyhardtoseparatefromtheback- groundvariabilityin thesystem(e.g.phenology,atmos- phericor topographiceffects). Apartfromafewnotable exceptions (see below), satellite-based methods do not generally provide direct measurements of land use intensity.

In terms of input metrics, remote sensing provides crucial information on the extent of land use, for example the global extent of agriculture (see [6]).

Furthermore,satellite imagetimeseries allowinsome casesfordeterminingcroppingcycles(e.g.[16,17]),the extent offallow land [18], or the frequency of fallow periods[19].Mappinggrazingpressureandforestman- agementeffectsacrossbroadgeographicextentsremains achallenge,althoughsomepromisingapplicationsexist [20,21].Advanceshavealsobeenmaderegardingmap- pingindividualcroptypes[22–24]orfordistinguishing irrigated fromrainfedagriculture(e.g.[25,26]).Finally, remote sensing can provide some information on the spatialconfigurationoflanduse,suchasfieldsize([27]

see Figure S1), which can be important for mapping capitalandlaborintensity(e.g.largefieldsasindicatorof agri-businessfarming).

Remotesensing canalso helpto derive outputmetrics.

Examples include yield estimates [28,29] or timber volumes extracted [30], although global applications of thiskindarestilllacking.Likewise,satellitescanassistin

generatinglanduseintensitysystemmetrics,forexample bymeasuringreferencestatesandchangesinecosystem properties such as net primary production [31], carbon stocks [30] or forest tree species composition [32,33], althoughmanyoftheseapproachesarestillexperimental

and cannot readily be applied across broad geographic extents.Despitethesepromisingdevelopmentsthough, algorithmsfor mappingoutputs and system metricsare notmatureenoughtoberoutinelyappliedtomaplarger regions.

Figure1

System/Output metrics (e.g., HANPP, yield gaps)

Input metrics (e.g., land/labor

ratio, N application)

System properties (e.g., NPP, land tenure)

Production system Output metrics (e.g., efficiency analysis, yields)

Inputs (e.g., land

capital, labor)

Outputs (e.g., food or

timber harvested)

System metrics (e.g., yield gaps,

HANPP) Current Opinion in Environmental Sustainability

Schematicoverviewoflanduseintensitymetrics.Metrics(orangeboxes)arequantitative,spatiallyexplicitmeasuresoflanduseintensityderivedby relatingdifferentdimensionstoeachother.Inputmetricsmeasuretheintensityoflandusealongdifferentinputdimensions(e.g.fertilizer/land,labor/

land).Outputmetricsrelateoutputsfromtheproductionsystemtoinputs(e.g.yields,residue/fellingratiosinforestry).Systemmetricsrelatetheinputs oroutputsofland-basedproductiontosystemproperties(e.g.actual/potentialyieldratios(i.e.yieldgaps),woodfellingtowoodincrementratios).

Table1

Datasourcescharacterizinglanduseintensityacrossbroadspatialextents

Datasource Description Extent Unitofobservation Examples

Satelliteimagery Measurementsofthespectral propertiesoflandsurfaces

Variable(localto regionaltoglobal coverage,dependingon thesensorsystem)

Pixel Landcover(e.g.cropping

area),landcoverchange (e.g.loggedarea),vegetation indices,NPP,albedo, surfacetemperature International

statistics

Reconcilednationalstatistics fromvarioussources

National(globalcoverage) Nations(sometimes subnational)

FAO(e.g.labor,capital, pesticideuse,agricultural production,landusearea, forestryuse),FAOForest ResourceAssessments Census(total

population)

Agricultureorforestry statistics(usually basedonquestionnaires)

National/subnational Administrativeunits Populationandhousing census,taxreports Survey(sample

ofpopulation)

Agricultureorforestry statistics(usually basedonquestionnaires orinterviewsofa stratifiedsample ofthepopulation)

National/subnational Individual,household,plot LUCASdatabase;living standardsurveys,national forestinventories

Cadastredata Landproperty boundariesandass ociatedinformation

Individualproperties Propertyboundaries Landtenure,national landregisters

Combiningsatellitedataandground-basedinventory data

A range of land use intensity metricsprimarily rely on ground-based inventory data, often combined with remote sensing information. Three approaches are frequently usedtotranslateinventorydataintospatially explicitlanduse intensitymetrics.

First, interpolation techniques can derive maps from point-based measurements such as national forest

inventories[34],theLandUse/CoverAreaFrameSurvey (LUCAS)of theEuropeanUnion[35],ornationalfarm- levelsurveys. Awidearrayof deterministicand geosta- tistical interpolation techniques can be used to derive grid-based land use intensity metrics from such point datasets (for a review see [36]).The potentialof these techniques is illustrated by advances in mapping tree species [37,38] or age-class distributions [39] based on forestryinventorydata,ormapsoffieldsizeandagricul- tural landscape patterns derived fromthe LUCAS data

Figure2

0

No agriculture Field size (ha)

Major lakes

> 10 ha

0.5 ha

250 500 km

Current Opinion in Environmental Sustainability

MapofcroplandfieldsizesforEuropederivedfrominterpolatingground-basedsurveydatafromtheLandUse/CoverAreaFrameSurvey(LUCAS)of theEuropeanUnionusinganordinaryKrigingapproach.

Source:LUCASprimarydata2009,http://epp.eurostat.ec.europa.eu/portal/page/portal/lucas/data/lucas_primary_data_2009.

(Figure 2). Despite this potential though, no global ground-observationdatasetwhich wouldallowforinter- polatingbroad-scaleland useintensitymetricscurrently existstoourknowledge.

Asecondgroupofmethodsdisaggregatesspatiallyaggre- gated data. Most datasetscontaining land use intensity information are available in aggregated form, because microdata (e.g. farm-level surveys, tax records) cannot be made available widely due to privacy issues, are difficult to interpret (e.g. individual forest inventory plots), are often not geocoded, or were only gathered foradministrativeunits(e.g.timberharvests). Disaggre- gationtechniques commonlycombineharmonized land usestatisticswithhigh-resolutionlandcoverinformation, forexampletomapforestgrowingstock[40]ortheextent of differentcrops[13,41].Morecomplex disaggregation techniques use a wider range of ancillary data, for example crop-type maps to produce global N and P fertilizer application maps [8], or census data and land cover mapsto produce rainfedand irrigated agriculture maps [9,42,43], which can then be used to map the croppingintensityofagriculture[44].Similarly,livestock patternsinEurope[45]andglobally[46]weremappedby disaggregatinglivestockstatistics.

A third group of land use intensity metrics combines measurements, either from satellites or on the ground, withmodeloutputs.Thisisparticularlyimportantregard- ing systemmetrics, which typically rely on areference valuethatisless straightforwardtomeasurethaninputs andoutputsalone.Forexample,globalyieldgaps[47,48] weremapped byfirstderiving efficiency frontiersusing econometric modeling, and then calculating the gap between actual and potential yields at the grid level [10].AnotherexampleisHANPP[14],whichisdefined as the difference between the fraction of actual NPP remaininginanecosystemafterharvest(e.g.determined with methods that combine ground-based and remote sensing data) and potential productivity from dynamic globalvegetationmodels.

Availableglobal metricsand datagaps

A review of available global-scale, gridded land use intensitymetricsreveals thata numberof suchmetrics arealreadyavailable,butlargedatagapsremaininterms ofdimensionsand sectorscovered(TableS1).

Croplandsystems

Thedatasituationisarguablybestin termsofcropland intensitymetrics.Arangeofmapsdepictglobalcropland area, crop distribution, cropping frequency, and the extent of irrigated as well as fallow cropland. Several maps also depict the amount of organic and mineral fertilizerapplied. Fewerdatasetscapture outputor sys- temmetrics,themostnotableoftheseareacomprehen- sive yield dataset and several global yield gap maps.

Moreover, time series for some metrics (e.g. cropland area) exist (Table S1). Although a comparatively high numberofcroplandintensitydatasetsexist,itshouldbe notedthoughthatthesedatasetsareoftenconnectedto considerableuncertainty.

Datagapsexistparticularlywithregardtocapital-related inputs(e.g.spatially explicitdatasetsonmechanization, pesticide application, or investment in agriculture) and laborinputs(e.g.thenumber,share,andskill-levelofthe agriculturalworkforce). We also currently lack detailed information on the extent and pattern of agroforestry, crop rotations, shifting cultivationsystems, and organic versus conventional cropping. Finally, the quality of many of the existing cropland intensity metrics could beimprovedfurther.

Grazingsystems

Global data on grazing systems are particularly scarce (Table S1). While indicatorsof livestock densities and major livestock products (meat, milk, eggs) exist, con- siderablegapsrelatetotheextentofgrazinglandandthe amount and types of biomass grazed. Likewise, infor- mationonother inputindicators ismissing,particularly regarding the spatial pattern of feed and forage pro- ductionand consumption,fertilizer applied to pastures, grasslanddrainage,andthepatternsoflaborandcapital inputsconnectedto livestocksystems.

Forestrysystems

Very few global forestry intensity metrics exist (Table S1). A number of datasets provide information on the currentextent,biomass,andgrowingstockofforests,and theareaofforestmanagement canbeapproximated via theexclusion of wilderness areas [49].However, major gapsincludeabetterunderstandingofthecharacteristics of forests (e.g. tree species composition, age-class or diameter-classdistributions,increment),thespatialpat- terns and types of forest management (e.g. close-to- nature versus monoculture, rare versus frequent man- agement),andtheinputs(e.g. fertilizer,labor, mechan- ization)andoutputs(e.g.timbervolumesextracted,non- timbergoods)of forestry.

Discussion

Challengesformappinglanduseintensityglobally Despite considerable recent progress, the mapping of global land use intensity continues to face major chal- lenges.First,fine-scalelanduseintensitydatawithglobal coverage remains scarce, particularly regarding grazing andforestrysystems.Statisticaldataarefrequently only availableat thenationalscale, systematicground-based data collection covers only a few regions, and remote sensing struggles to capture the often subtle spectral effects of land use intensity changes. Data gaps are unfortunately largest in developing countries, many of whichexperiencerapidlandusechange,andarethought

Table2

Researchprioritiesforglobal-scale,spatiallyexplicitdatasetsneededtoimproveandextendtheexistingsetoflanduseintensitymetrics

Datasetormetricneeded Potentialmappingapproach

Croplandsystems

Improvedmapsofcroplandextent,especially foruncertainregions

(e.g.SubSaharanAfrica)andcroppingsystems(e.g.

shiftingcultivation),aswellascroplandabandonment

Satelliteremotesensing,atmultiplescales(includingimagesfineenoughtocapture landusepatterns),potentiallyincombinationwithcensusdataorlocalsamplingsurveys

Improvedmapsofcroppingcycles,incl.fallowcycles Analysesofsatelliteimagetimeseries,combinedwithcropcalendars[61,62]and agriculturalcensusdata[44]and/orcrowd-sourcedinformationonfarmingpractices Laborintensityormechanization Disaggregationofstatisticaldata(e.g.agriculturallaborforce)withancillarydata(e.g.

remoteness,populationdensity,landusesystems).Harmonizedcollectionofstatistical data,preferablyatsubnationalscale,needed.Newremotesensingdatasets(e.g.field size)couldimproveestimations

Pesticideuse Disaggregationapproachsimilartothoseusedtogenerateglobalfertilizerapplication maps.Structuredcollectionandaccesstodataonpesticideuse(e.g.viafarmsurveys) andsales(e.g.subnationalstatistics)needed

Capitalinvestmentandcapitalproductivity Structureddatacollectionneeded.Capitalproductivitycouldbemappedbyrelating investmentstorevenues(e.g.usingyieldmapsandpriceestimates)

Organicfarmingextent Disaggregation/downscalingofnationalorsubnationaldataonorganicfarmingextent.

Closelinkstoseveraloftheabovemetrics(e.g.pesticideuse) Grazingsystems

Shareoffeed/foragefromnaturalvegetation (versuscroplandandpermanentpastures)

Collectionandhomogenizationoffeed/foragedataatthesubnationalscaleneeded, potentiallyincombinationwithcrowd-sourcedinformationongrazingpractices.Such informationcouldbeusedtogetherwithcroplandextentandlivestockdensitymaps[63]

Extentofgrazingandtypesofvegetationthat isgrazed(e.g.grasslands,forests)

Improvedvegetationmapsfromremotesensingincombinationwithdisaggregated livestockstatisticsandinformationongrazingpractices(seeabove)

Foragequality Remotesensing(vegetationstructure,productivity)possiblyincombinationwith ecosystemmodels,andcrowd-sourcedinformationonlivestocksystems Improvedmapsoftheshareofanimalsinfeedlots

versusgrazing/free-ranginganimals

Collectionandhomogenizationofsuchdataatthesubnationalscaleneeded.

Disaggregation/downscalingcouldbesubstantiallyimprovedbyimplementing informationongrazingsystems(typeofvegetationgrazed,foragequality) Improvedestimatesoffertilizer(mineraland

manure)usedingrazingsystemsand manuretransferredtocropland

Disaggregation/downscalingcouldbesubstantiallyimprovedbyimplementing informationongrazingsystems(typeofvegetationgrazed,feedfromnaturalvegetation versusfarmland,foragequality)

Watermanagementongrazingland Informationongrazingextentincombinationwithinformationonirrigationequipment, climatedataandsatelliteremotesensing

Labororcapitalinputstograzingsystems Collectionandhomogenizationofdataonlabor(e.g.#personsengagedwithgrazing/

livestockhusbandry)andcapital-relatedinputs(e.g.fences,fertilizer,vaccination)of grazingsystemsneeded.Disaggregation/downscalingwouldbepossibleusing indicatorsonlivestockdistributionandgrazingpractices

Forestrysystems

Forestmanagementtypes(e.g.agroforestryversus plantationsversusmanagednaturalforest versusunmanagedforests)

Collectionandhomogenizationofsubnationaldataontheextentofplantationsneeded.

Disaggregation/downscalingcouldbeimprovedbyremotesensinginformation(forest types,foreststructure)andancillarydata(e.g.wildernessdatasets)

Improvedforesttypemaps Newremotesensingdata(e.g.high-resolution,multi/hyperspectralsensorssuchasthe upcomingSentinel-2sensor)orjointuseofdata(Lidar,radar,andopticaldata)may allowformovingbeyondbroadforesttypes(currentlybroadleaved,mixedandneedle- leavedforests)

Forestharvestingrates Disaggregationofforestharvestingstatisticsusingforestareamaps,forest managementtypes,andmarketaccessibilityproxies(e.g.traveldistancetomarkets, infrastructurenetwork,terrainruggedness)

Improvestandingvolume/biomassmaps Remotesensing,forexampleviacombininginformationonforesttypesandforest structure[64,65,66]

Incrementmapandshareofharvestinincrement Dynamicglobalvegetationmodelsincombinationwithimprovedforest,extent,forest type,andforestharvestingmaps

Age-classdistributionsandmanagement frequencymaps(e.g.rareversusfrequent)

Collectionandhomogenizationofnational/subnationaldataonforestageand managementcyclesneeded.Informationcouldcomepartlyfromremotesensing(e.g.

logginghistories),surveydata,orcrowd-sourcing Forestryinputs(e.g.fertilizer,labor,

mechanization,drainage)

Collectionandhomogenizationofnational/subnationaldataondifferentinputsis needed.Suchdatacouldbedisaggregatedusingmapsofforestryextentand/orforest managementtypes(seeabove)

to harbormajor potentials for furtherintensifying land- based production. Second, existing datasets are often inconsistent in time (e.g. due to changes in survey methods or data processing), space (e.g. political boundarychanges),ormaplegends,requiringsubstantial homogenization efforts. Third, uncertainty of existing land use intensity metrics is often high (e.g. due to positionalinaccuracy,unreliableinputdata,orprocessing algorithmslimitations), as highlighted by thelarge dis- crepancies of alternative global cropland extent maps [50] or fertilizer application maps [8,51], and remain largelyunquantified, because formal validationis often lacking.Where alternativemapsexist,uncertaintiescan bereducedbycombingseveralmapsintoa‘hybrid’map [52,53]. Errors mayalso varysubstantially in space and there is a risk of error propagation in more complex datasets(e.g.landcovermapsareneededtoderivecrop distributions,whichareneededtodisaggregatefertilizer statistics). Fourth, interpolation or disaggregation often reliesoncovariates(e.g.locationfactors)whichmayresult in endogeneity problems in subsequent analyses. For example, the FAO’s Gridded Livestock of the World [46]usesremotesensedvegetationmeasurestodistribute livestock,andthuscannotbeusedtoanalyzetheeffects of livestock density on vegetation. This endogeneityis sometimes difficult to trace, emphasizing the need to clearlydocumenthow datasets wereconstructed.Fifth, globaldatasetsaretypically coarse,resultinginsubstan- tialbiasinareaestimates[54]orwhendownscalingdata [55].Finally,substantialconceptualchallengesremainin ordertoframelanduseintensityglobally(seeErbetal., thisissue).

Opportunitiesforanimprovedmappingoflanduse intensity

Progressindataaccessand algorithmdevelopmentpro- vide opportunities for developing new and improved global land use intensity metrics. Advances in remote sensingarerapid,withagrowingnumberofsensorsand increasingaccesstoimagearchives(e.g.theUSGSLand- sat archive), as well as new algorithms able to handle complex datastructures (e.g. machine learning, geosta- tistical,ordataminingtools).Weseethreemainavenues for an improved mapping of land use intensity: First, longer and more consistent image time series which capture phenology may help to map cropping cycles andtoreconstructlandusehistories.Second,multi-scale applications(e.g.jointuse LandsatandMODISorSen- tinel1/2/3images)seempromisingregardingovercoming resolution-dependent limitations. Finally, merging data from different sensor systems (e.g. optical, radar, or LIDAR)mayprovidenewinsightsintolanduseintensity.

Despite these opportunities, however, the integration of remote sensing and ground-based data will remain crucial. Although statistically rigorous, ground-based surveys are increasingly implemented, for example

sampling-based national forest inventories [34] or the LUCAS survey in theEuropean Union[35], therepre- vails ahugelack of high-quality, ground-based dataon landmanagement,especiallyforthoseregionswhereland use changesrapidly.Newmeans for ground-baseddata collectionareemergingthough,forexamplecrowd-sour- cingcouldbecomeanimportantsourceofgeocodedland usedata[56]andcanhelptovalidategloballandusemaps [50]. Finally, new technologies for field or plot-based monitoring are also becoming available and affordable (e.g.wirelesscommunicationandsolar-poweredsensors) [5].

Thewayforward

A few general recommendations for assessing land use intensity patterns at the global scale emerge from our review. First, considerable progress can be made with alreadyexistingdata,forexamplebycombiningmultiple datasets from different sources and across scales. Data access isa key challengein this context, and efforts to developplatforms and protocols to compile, share, and distribute land use datasets, such as the GEOSHARE initiative (www.geoshareproject.org), are urgently needed.Second,furtherstandardizationandharmoniza- tionofexistinglandusedatasetsaswellasdatacollection protocols areneeded, similar to efforts focusing onland cover(e.g.[57]).Asground-baseddataareessentialformost land use intensity metrics, implementing new sampling schemesandstandardizingexistingnationalschemesare crucial.Third,thereisanurgentneedtovalidateexisting globaldatasets and todocumentuncertainty, biases,and potentialerrorpropagation(i.e.uncertaintyofinputdata- sets).Itiscrucial,foreachdataset,totransparentlydocu- mentthecovariates usedandassumptions made,so that subsequentuserscanavoidendogeneityproblems.Fourth, betterintegrationofobservationaldata(from satellitesor theground)intoprocess-basedmodelsisneededtoadvance themappingofsystemmetrics.Finally,timeseriesformost land use variables do currently not exist, but would be importanttoassesspastchangesinlanduseintensityandits environmentaloutcomes.

Futureresearchshouldfocusonimprovingexistingland useintensitymetricsandonfillingdatagaps(Table 2), prioritizing those sectors and indicators where data deficienciesarelargest.Regardingcroplanduseintensity, the already relatively rich set of metrics needs further improvements (e.g. global cropland extent, cropping cycles,fertilizeruse),andcouldbeextended(e.g.pesticide use). Uncertainty is generally larger regarding grazing systems,andbetterinformationongrazingextent,especi- ally on the distribution of grazing among the different vegetationtypes, and feedproduction and consumption is urgentlyneeded. Datagaps appear biggest regarding globalforestryintensity,forwhichmajoradvancescouldbe madefrommapsof broadtypesofforestrysystems(e.g.

plantations,agroforestry,managedandunmanagednatural

forests,aswellasforestharvesting).Inadditiontoquan- titativemetrics,afruitfulfieldforfutureresearchwouldbe to advance the global mapping of land use systems [58,59].

Abettercharacterizationofthespatialpatternsofglobal land use intensity is crucial to monitor the various environmental andsocietalimpacts oflanduse,andto understand the drivers ofchanging landuse intensity.

Giventhemultidimensionalnatureoflanduseintensity, afocusonmultiplemetricswithinasystemsperspective isneeded.Themetricswediscussedhereprovideeither aquantitativemeasureofoneaspectoflanduseintensity (i.e. input and output metrics), or of the aggregated effects oflanduseintensity(i.e.systemmetrics).Both types of metrics complement each other, as single metrics are relatively easy to compute and interpret, butdonotprovideacoherentpictureofintensification, whereas system metrics, by aggregating multiple pro- cesses, hamper the understanding of the relations be- tween different system components [60]. Ample opportunities exist toadvanceboth typesofmetricsin parallel to arrive at a second generation of land use intensity metrics. Such metrics wouldbe a major step towardconfrontingthesustainabilitychallengeinglobal landuse,butdeveloping,harmonizing,maintaining,and sharingthesedatasetsrelated willrequiresubstantially investments from scientists and funding organizations alike.

Acknowledgements

ThisresearchwaspartlyfundedbytheEuropeanCommission(Integrated ProjectVOLANTEFP7-ENV-2010-265104)andtheGlobalLandProject (www.globallandproject.org)andwearegratefulforthissupport.TKu acknowledgessupportbytheEinsteinFoundation,Berlin(Germany).KHE acknowledgesfundingfromERCStartingGrant263522LUISE.HH,KHE andTKaacknowledgefundingfromtheAustrianScienceFunds(FWF), projectP20812-G11.WethankL.Kehoeforhelpfulcommentsonearlier manuscriptversions,andtwoanonymousreviewersforconstructiveand veryhelpfulremarks.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/

j.cosust.2013.06.002.

References and recommendedreading

Papersofparticularinterest,publishedwithintheperiodofreview, havebeenhighlightedas:

ofspecialinterest ofoutstandinginterest

1. GodfrayHCJ,BeddingtonJR,CruteIR,HaddadL,LawrenceD, MuirJF,PrettyJ,RobinsonS,ThomasSM,ToulminC:Food security:thechallengeoffeeding9billionpeople.Science 2010,327:812-818.

2.

TilmanD,BalzerC,HillJ,BefortBL:Globalfooddemandandthe sustainableintensificationofagriculture.ProcNatlAcadSciU SA2011,108:20260-20264.

Highlightsthestrongrelationshipbetweeneconomicdevelopmentand dietchange,andtheroleofsustainableintensificationtomeetfuturefood demands.

3.

LambinEF,MeyfroidtP:Globallandusechange,economic globalization,andtheloominglandscarcity.ProcNatlAcadSci USA2011,108:3465-3472.

Discussesglobalizationeffectsonlandssystemsandapotentialscarcity ofadditionalavailablecropland.

4. FoleyJA,RamankuttyN,BraumanKA,CassidyES,GerberJS, JohnstonM,MuellerND,O/’ConnellC,RayDK,WestPCetal.:

Solutionsforacultivatedplanet.Nature2011, 478:337-342.

5.

ZaksDPM,KucharikCJ:Dataandmonitoringneedsforamore ecologicalagriculture.EnvironResLett2011:6.

Reviews thedata and technology needed for establishing an agro- ecologicalinsitumonitoringnetwork.

6.

VerburgPH,NeumannK,NolL:Challengesinusinglanduseand landcoverdataforglobalchangestudies.GlobalChangeBiol 2011,17:974-989.

Reviews availability,uncertainty, andconsistencyofavailable broad- scalelandcoverdatasets.

7. BoserupE:TheConditionsofAgriculturalGrowth.TheEconomics ofAgrarianChangeUnderPopulationPressure. London:George Allen&UnwinLtd.;1965,.

8. PotterP,RamankuttyN,BennettEM,DonnerSD:Characterizing thespatialpatternsofglobalfertilizerapplicationandmanure production.EarthInteractions2010,14:1-22.

9. PortmannFT,SiebertS,Do¨llP:MIRCA2000—globalmonthly irrigatedandrainfedcropareasaroundtheyear2000:anew high-resolutiondatasetforagriculturalandhydrological modeling.GlobalBiogeochemCycles2010,24:GB1011.

10. DietrichJP,SchmitzC,Mu¨llerC,FaderM,Lotze-CampenH, PoppA:Measuringagriculturalland-useintensity—aglobal analysisusingamodel-assistedapproach.EcolModel2012, 232:109-118.

11. TurnerBL,DoolittleWE:Theconceptandmeasureof agriculturalintensity.ProfGeogr1978,30:297-301.

12. NettingRM:Smallholders,Householders:FarmFamiliesandthe EcologyofIntensive,SustainableAgriculture. Stanford:Stanford UniversityPress;1993,.

13. MonfredaC,RamankuttyN,FoleyJA:Farmingtheplanet:2.

Geographicdistributionofcropareas,yields,physiological types,andnetprimaryproductionintheyear2000.Global BiogeochemCycles2008,22:GB1022.

14. HaberlH,ErbKH,KrausmannF,GaubeV,BondeauA,PlutzarC, GingrichS,LuchtW,Fischer-KowalskiM:Quantifyingand mappingthehumanappropriationofnetprimaryproductionin earth’sterrestrialecosystems.ProcNatlAcadSciUSA2007, 104:12942-12947.

15. BinswangerHP,RosenzweigMR:Behaviouralandmaterial determinantsofproductionrelationsinagriculture.JDevStud 1986,22:503-539.

16. RedoDJ,MillingtonAC:Ahybridapproachtomappingland-use modificationandland-covertransitionfromMODIStime- seriesdata:acasestudyfromtheBolivianseasonaltropics.

RemoteSensEnviron2011,115:353-372.

17. CanisiusF,TurralH,MoldenD:Fourieranalysisofhistorical NOAAtimeseriesdatatoestimatebimodalagriculture.IntJ RemoteSens2007,28:5503-5522.

18.

AlcantaraC,RadeloffVC,PrishchepovAV,KuemmerleT:

Mappingabandonedagriculturewithmulti-temporalMODIS satellitedata.RemoteSensEnviron2012,124:334-347.

Applies MODIStimeseriestomap post-Sovietlandabandonmentin Russia.

19. TottrupC,RasmussenMS,EklundhL,Jo¨nssonP:Mapping fractionalforestcoveracrossthehighlandsofmainland SoutheastAsiausingMODISdataandregressiontree modelling.IntJRemoteSens2007,28:23-46.

20. Ro¨derA,UdelhovenT,HillJ,delBarrioG,TsiourlisG:Trend analysisofLandsat-TMand-ETM+imagerytomonitorgrazing impactinarangelandecosysteminNorthernGreece.Remote SensEnviron2008,112:2863-2875.