Knowledge with Global Change Science Ambiguous Geographies ...globe.umbc.edu/wp-content/uploads/2016/04/margulies_2016.pdf · Ambiguous Geographies: Connecting Case Study Knowledge

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=raag21

Download by: [University Of Maryland] Date: 11 April 2016, At: 13:25

Annals of the American Association of Geographers

ISSN: 2469-4452 (Print) 2469-4460 (Online) Journal homepage: http://www.tandfonline.com/loi/raag21

Ambiguous Geographies: Connecting Case StudyKnowledge with Global Change Science

Jared D. Margulies, Nicholas R. Maggliocca, Matthew D. Schmill & Erle C. Ellis

To cite this article: Jared D. Margulies, Nicholas R. Maggliocca, Matthew D. Schmill &Erle C. Ellis (2016): Ambiguous Geographies: Connecting Case Study Knowledge withGlobal Change Science, Annals of the American Association of Geographers, DOI:10.1080/24694452.2016.1142857

To link to this article: http://dx.doi.org/10.1080/24694452.2016.1142857

Published online: 06 Apr 2016.

Submit your article to this journal

Article views: 20

View related articles

View Crossmark data

http://www.tandfonline.com/action/journalInformation?journalCode=raag21http://www.tandfonline.com/loi/raag21http://www.tandfonline.com/action/showCitFormats?doi=10.1080/24694452.2016.1142857http://dx.doi.org/10.1080/24694452.2016.1142857http://www.tandfonline.com/action/authorSubmission?journalCode=raag21&page=instructionshttp://www.tandfonline.com/action/authorSubmission?journalCode=raag21&page=instructionshttp://www.tandfonline.com/doi/mlt/10.1080/24694452.2016.1142857http://www.tandfonline.com/doi/mlt/10.1080/24694452.2016.1142857http://crossmark.crossref.org/dialog/?doi=10.1080/24694452.2016.1142857&domain=pdf&date_stamp=2016-04-06http://crossmark.crossref.org/dialog/?doi=10.1080/24694452.2016.1142857&domain=pdf&date_stamp=2016-04-06

Ambiguous Geographies: Connecting Case StudyKnowledge with Global Change Science

Jared D. Margulies,* Nicholas R. Maggliocca,y Matthew D. Schmill,z and Erle C. Ellis*

*Department of Geography and Environmental Systems, University of Maryland, Baltimore CountyyNational Socio-Environmental Synthesis Center

zDepartment of Computer Science and Electrical Engineering, University of Maryland, Baltimore County

Case studies have long been a gold standard for investigating causal mechanisms in human–environment inter-actions. Yet it remains a challenge to generalize across case studies to produce knowledge at broader regionaland global scales even as the effort to do so, mostly using metastudy methods, has accelerated. One major obsta-cle is that the geographic context of case study knowledge is often presented in a vague and incomplete form,making it difficult to reuse and link with the regional and global contexts within which it was produced and istherefore most relevant. Here we assess the degree to which the quality of geographic description in publishedland change case studies limits their effective reuse in spatially explicit global and regional syntheses based on437 spatially bounded cases derived from 261 case studies used in published land change metastudies. Commonambiguities in published representations of case geographic contexts were identified and scored using threeindicators of geographic data quality for reuse in spatially explicit regional and global metastudy research. Sta-tistically significant differences in the quality of case geographic descriptions were evident among the six majordisciplinary categories examined, with the earth and planetary sciences evidencing greater clarity and confor-mance scores than other disciplines. The quality of case geography reporting showed no statistically significantimprovement over the past fifty years. By following a few simple and readily implemented guidelines, case geo-graphic context reporting could be radically improved, enabling more effective case study reuse in regional toglobal synthesis research, thereby yielding substantial benefits to both case study and synthesis researchers.Key Words: geographic representation, GIScience, metastudy, research synthesis, scale.

案例研究对于探讨人类—自然互动的因果机制而言, 长期作为黄金标准。但普遍化各个案例研究, 以在更广泛的区域及全球尺度中生产知识仍是个挑战, 尽管多半运用后设研究方法的努力已不断增加。其中一个主要的困难在于,案例研究知识的地理脉络,经常以模煳且不完整的形式呈现之, 使其难以被再利用,并难以连结至其被生产、因此最为相关的区域及全球脉络。我们在此根据已出版的土地变迁后设研究

所使用的二百六十一个案例研究中, 衍生而出的四百三十七个在空间上受限之案例, 评估在已出版的土地变迁案例研究中的地理描绘之质量, 限制它们在空间明确的全球及区域综合中有效再利用的程度。我们运用三项在空间明确的区域与全球后设研究中, 再利用的地理数据质量指标, 指认已出版的案例地理脉络再现中的普遍模煳性。在我们所检视的六大主要领域范畴中, 案例地理描绘质量中的显着统计差异相当明显, 其中地理与地球科学, 呈现出较其它领域更高的清晰度与一致性分数。案例地理学报告的质量显示, 过去五十年来在统计上并没有显着的进步。透过追踪数个简单且已实施的指导方针, 案例地理脉络报告可彻底改进, 并促成区域到全球综合研究中更有效的案例研究再利用, 因而同时对案例研究与综合研究者带来实质的益处。关键词：地理再现,地理信息科学,后设研究,研究综合,尺度。

Los estudios de caso han sido desde hace mucho tiempo el est�andar dorado para investigar los mecanismos cau-sales en las interacciones humano-ambientales. Sigue siendo un reto, sin embargo, generalizar de los estudios decaso para generar conocimiento a escalas m�as amplias regionales y globales, aun si el esfuerzo para lograrlo, prin-cipalmente usando m�etodos de metaestudio, ha sido incrementado. Un obst�aculo mayor es que el contexto geo-gr�afico del conocimiento por estudio de casos a menudo se presenta de forma vaga e incompleta, haciendodif�ıcil reusar y ligar con los contextos regionales y globales dentro de los cuales aquel fue producido, por lo quetiene mayor relevancia. En este art�ıculo evaluamos el grado con el que la calidad de la descripci�on geogr�aficaen estudios de casos publicados sobre cambios de la tierra restringe su reutilizaci�on efectiva en s�ınteis globales yregionales, espacialmente expl�ıcitas, basadas en 437 casos espacialmente demarcados, derivados de 261 estudiosde caso publicados en metaestudios sobre cambios de la tierra. Las ambiguedades comunes en representacionespublicadas de casos de contexto geogr�afico fueron identificadas y calificadas usando tres indicadores de calidadde los datos geogr�aficos para reutilizaci�on en investigaci�on de metaestudios regionales y globales espacialmenteexpl�ıcitos. Diferencias estad�ısticamente significativas en la calidad de descripciones geogr�aficas de caso fueronevidentes entre las seis mayores categor�ıas disciplinarias examinadas, con las ciencias de la tierra y las planetar-ias evidenciando mucha mayor claridad y marcas de conformidad que otras disciplinas. La calidad de los

Annals of the American Association of Geographers, 0(0) 2016, pp. 1–25� 2016 by American Association of GeographersInitial submission, July 2015; revised submission, October 2015; final acceptance, December 2015

Published by Taylor & Francis, LLC.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

informes sobre la geograf�ıa de casos no mostr�o una mejora estad�ısticamente significativa en los pasados cin-cuenta a~nos. Siguiendo unas pocas instrucciones simples y de f�acil implementaci�on el reporte del contexto geo-gr�afico del caso podr�ıa ser mejorado radicalmente, posibilitando un reuso del estudio de caso m�as efectivo en lainvestigaci�on de s�ıntesis de lo regional a lo global, generando de ese modo beneficios sustanciales para los inves-tigadores y para los estudios de casos y s�ıntesis. Palabras clave: representaci�on geogr�afica, ciencia SIG, metaestudio,s�ıntesis de investigaci�on, escala.

Synthesis research aimed at understanding thecauses and consequences of global social andenvironmental change is increasing rapidly,

supported by metastudy analysis of case studyresearch at local to regional scales (Turner et al.1990; Rindfuss et al. 2004; Rudel 2008; Cox 2015;Magliocca et al. 2015; van Vliet et al. 2016).Although case study research remains one of themost popular research methods for understandinghuman–environment interactions, translatingknowledge produced through local case studies intodata for broader-scale research synthesis efforts isconfronted by a variety of methodological challenges(Rindfuss et al. 2004; Keys and McConnell 2005;Turner, Lambin, and Reenberg 2007; Maglioccaet al. 2015). Here we assess the degree to which oneof these challenges, ambiguities in the geographicrepresentation of case study knowledge, might affectcase study reuse in global and regional synthesisresearch. We do so using a metastudy approach todescribe and evaluate the quality of geographic rep-resentations across a set of 437 cases extracted from261 case studies used in highly cited metastudies inthe field of land change science (Globe Cases Team2015).

The research presented here is motivated by twobasic research questions: (1) Do patterns in the qualityof geographic description exist across the case studyliterature of land change research and, if so, why? and(2) How might a more systematic approach to suchdescriptions facilitate more robust and precise reuse ofcase study knowledge in spatially explicit global andregional synthesis research? To examine these researchquestions, we applied a systematic quality coding pro-cedure to the 437 cases examined here to evaluate thequality of their geographic descriptions. Motivated byour research questions, we tested the following fourhypotheses:

1. Case quality scores vary across major academicdisciplines, with higher scores in the more geo-spatially oriented disciplines.

2. Case quality scores differ by geographic entitytype, with higher scores among entity types with

clearer and more replicable boundaries (e.g.,administrative units or watersheds compared tovillages or pastures).

3. Case quality scores vary by land use type, withhigher scores among more intensively managedland use types (e.g., dense settlements comparedto rangelands).

4. Case quality scores improve over time based onpublication date, with more recent studies pro-ducing higher quality scores.

Informed by our results and the experiential knowl-edge acquired through the process of case scoring,we also present readily implemented guidelines fordescribing the geographic context of case studies toimprove their effective reuse in regional and globalresearch synthesis.

Representing Case Study Space

Our primary research questions are motivated by adesire to better understand how the quality of geo-graphic descriptions might affect research synthesisefforts based on the reuse of empirical knowledgereported in published case studies. The process ofdefining the geographic context within which casestudy knowledge has been gained in terms of an areaof Earth’s land surface sets the terms by which thisknowledge can be interpreted and used by others(Keys and McConnell 2005; Downey 2006; Kwan2012; Karl et al. 2013). Defining the unit of analysisof a case study, or “bounding of the case,” is consid-ered an essential step in the development of a casestudy protocol (Yin 2013, 33). Most recently in rela-tion to case study synthesis research, Cox (2014)raised the distinction between case studies (a unit ofobservation) and cases (a unit of analysis). A casestudy typically takes the form of a published paperor report and might include one or more cases that aresearcher conducting synthesis research can bothextract data from and apply coding procedures to.The boundaries of a case might be spatial, temporal,or present in the form of another concrete

2 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

delineation between who or what is being analyzedin a case analysis and who or what is excluded (Yin2013). Yet to date, guides on case study design andreporting have paid insufficient attention to charac-terizing the appropriate geographic descriptors forcases that are spatially bounded in both the casestudy literature and across the empirical environ-mental social sciences (Ragin and Becker 1992;Flyvbjerg 2006; Yin 2013; Cox 2014, 2015).

Although the past two decades have seen a flour-ishing body of research problematizing and theorizingon scale and spatial representation, particularly withinhuman geography (for a review of some key works,see Marston 2000; Brenner 2001; Marston, Jones, andWoodward 2005; Sayre 2005; Miller 2007; Moore2008; among others), for researchers investigatinghuman–environment interactions with cumulativeglobal consequences, such as the loss of carbon orbiodiversity in response to land change, there remainsthe practical problem of adequately identifying astudy’s geographic extent on the Earth’s surface sothat its spatially explicit regional and global contextscan be assessed and integrated into synthesis research(Turner et al. 1990; Karl et al. 2013; Magliocca et al.2015). The field of land change science in particular,with its focus on patterns and processes of land useand modification of land systems, has long sought todraw generalizable patterns and trends of human–environment relations out of locally conducted casestudies (Turner, Hanham, and Portararo 1977; Rind-fuss et al. 2004; Turner, Lambin, and Reenberg 2007;Rudel 2008; Magliocca et al. 2015; Verburg et al.2015; van Vliet et al. 2016). It is therefore necessaryto distinguish and describe those aspects of caseknowledge that have localizable spatial contexts sothey can be used in generating spatially explicitregional and global knowledge of land change pro-cesses. Although there are important ethical consider-ations researchers must consider when choosing howto describe the geographic context of a case, thereare simple and basic improvements most researcherscan and should employ in describing the geographiccontext of case research.

Geographic Context in Synthesis Research

Accurate geographic descriptions of the boundariesof case knowledge are especially important in meta-study synthesis research on environmental change.Metastudies of case studies are increasingly used to

make general inferences on land change patterns andprocesses at global and regional scales using empiricaldata drawn from case studies conducted at more local-ized spatial scales (Lambin and Geist 2006; Rudel2008; Verburg, Neumann, and Nol 2011; Cox 2015;Magliocca et al. 2015; van Vliet et al. 2016). Landchange scientists are interested in a diversity of factorsshaping land systems, including demographic, eco-nomic, cultural, institutional, technological, and eco-logical mechanisms, and their interactions at multiplespatial and temporal scales (Lambin and Geist 2006).The influence of many of these factors on land systemdynamics has been found to be scale dependent andnonstationary over space (e.g., population density andmarket access [Verburg, Ellis, and Letourneau 2011];agricultural intensity [Laney 2002]). Spatially explicitand accurate reporting of a case’s geographic extent istherefore especially important for metastudy researchin which studies across multiple sites and geographiclocations are compared and integrated (Karl et al.2013; Magliocca et al. 2015).

Despite an acceleration of synthesis research inland change science using local case knowledge(Magliocca et al. 2015), the challenges to syntheticknowledge creation across different scales of observa-tion and analysis are exposed in the persistent diffi-culties in “scaling up” case study research to gainbroader insight on patterns of environmental change(Sayre 2005). Although there is a long history ofcomparative case study research in the social sciences(e.g., Murdock and White 1969) and there have beenrecent advances in case study synthesis methods suchas the social–ecological systems meta-analysis data-base (e.g., SESMAD; Cox 2014), the difficulties ofengaging in research to make broader observations onland change through synthesis research remain. Oneof the greatest barriers to such synthesis efforts is thecomparability of individual cases and the relativefacility for other researchers to extract data from pub-lished studies for secondary analysis (Magliocca et al.2015). Nevertheless, metastudies of case studyresearch conducted at local to regional spatial scalesremain an important and growing research strategyfor generating regional and global understanding ofcoupled human and environmental systems, as itis otherwise difficult to observe the coupling of socialand environmental patterns and processes byother methods, despite the promise of remote sensingand volunteered geographic information (Rindfusset al. 2004; Goodchild and Li 2012; Magliocca et al.2015).

Connecting Case Study Knowledge with Global Change Science 3

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Exploring Ambiguous Geographies

This article is based on geographic descriptionsprovided in a set of 437 cases compiled, coded, andmapped as part of the GLOBE project (Ellis 2012)by a team of trained students at the University ofMaryland, Baltimore County. Common ambiguitiesin the reporting of case geographic contexts areidentified and scored relative to the degree towhich the quality of their geographic reporting ena-bles their reuse for spatially explicit regional andglobal metastudy synthesis. Variation in the qualityof case geographic representation is assessed as afunction of discipline, time, geographic entity type,and land use system, demonstrating a remarkablyconsistent lack of clarity in these descriptions across

most disciplines that has changed little over thepast fifty years.

In the process of mapping these cases, the diversityand commonality of ambiguous geographic descrip-tions was made clear, as illustrated in Figure 1,demonstrating the importance of precise in-text andgeospatial representation of case geographic context,especially when findings on multiple cases are pre-sented within the same publication. The causes of thiswidespread and continuing ambiguity are evaluatedand discussed together with readily implemented strat-egies for improving the communication of the spatialcontexts of case study research in an effort to advancespatially explicit regional to global metastudy synthesisresearch within land change science and broader spa-tial sciences communities.

Figure 1. Example of geographic ambiguities emerging through translating local case study geographies for use in metastudies. In this exam-ple, a fictitious case study of five villages is translated in four different ways based on a map and in-text description of the study sites. The sub-sequent depictions (displayed on the right) were produced by three different undergraduate students at the University of Maryland,Baltimore County, when provided the initial fictitious description (left). Both the illustrative map and in-text description represent commonforms of representing case geographies based on our review of 437 cases analyzed in this article. (Color figure available online.)

4 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Method

Case Study Acquisition

A total of 444 cases were identified for research byreproducing the case study collections used in eightpublished metastudies chosen for their subject breadthacross land change science, ranging from biofuel pro-duction, deforestation, and agricultural abandonmentin the tropics to cropland change and risk manage-ment in pastoral systems (Table 1). Cases wereselected from published metastudies as these wereassumed to represent cases especially suitable for meta-study synthesis. The original source of each case study(journal articles, book chapters, books) was acquiredin hard copy or electronically. Cases were excludedfrom analysis when no original source could be located(one case), the original source was located but therewas insufficient geographic information included inthe source to map the case location (two cases), andtheir geographic extent exceeded 5 million km2 (theapproximate size of the Amazon rainforest), a limitimposed to exclude large regional studies (four cases),producing a total collection of 437 cases. Many indi-vidual case study sources reported on multiple cases, inwhich data were presented for more than one geo-graphic extent. For instance, an urban land changestudy might produce multiple unique cases based onseparate cities for which data were reported. Individualcases were identified within sources to correspond withthe same number of cases utilized in the original meta-study they were used in, based on analysis of sourcetext, figures, and tables.

Case Preparation Procedure

Cases were prepared for analysis using proceduresfor spatially explicit case study entry into the onlinecase database of the GLOBE project, as described later(Global Collaboration Engine; Ellis 2012; Schmillet al. 2014; Young and Lutters 2015). Full biblio-graphic information on the published study fromwhich each case was derived was first entered intoGLOBE, followed by a map of the geographic extentof the case and an automated scoring of case geographydata quality pedigree (Table 2), as detailed in the fol-lowing section and in greater detail in Figure A1 inthe Appendix. Cases were entered into GLOBEbetween March 2012 and March 2014 by a trainedteam of nine undergraduate and graduate studentsfrom the Department of Geography and Environmen-tal Systems at University of Maryland, BaltimoreCounty. All of the students had at least an introduc-tory course in geography and geographical techniquesat the time of coding cases. Additionally, seven of thestudents had taken at least two geographic informationsystems (GIS) courses (many of whom were workingtoward certification) and thus understood the require-ments of georeferencing the geographic extents ofcases contained within a case study.

Case geographic extents were mapped based on theclearest geographic description of the spatial extent ofeach case for which data were utilized in the originalciting metastudy, based on thorough study of the text,tables, and figures within each original source. Thefirst step in mapping case geographic extents was toidentify the geographic entity (e.g., forest, watershed,

Table 1. List of eight metastudies from the field of land change science and topics of extracted case studies

Meta-study TopicNo. of cases (coefficient of

variation D 0.83)Turner, Lambin, and Reenberg (1977) Relationships between population density and agricultural intensity 28Keys and McConnell (2005) Agricultural intensification in the global tropics 93Kauffman, Hughes, and Heider (2009) Rates of deforestation and resulting carbon emissions as well as

land-use changes including agricultural abandonment in theneotropics

19

Achten and Verchot (2011) Implications of land-use change emission on the climate-changemitigation potential of different biofuel production systems

16

Moritz et al. (2011) Social risk-management strategy variations within pastoral systemsin the neotropics

22

Eclesia et al. (2012) Replacement of native vegetation by pastures and tree plantations 54Van Vliet et al. (2012) Trends, drivers, and impacts of changes in swidden cultivation in

tropical forest-agriculture frontiers156

Van Vliet, Reenberg, and Rasmussen(2013)

Cropland change as well as their driving forces and perceivedimpacts within the Sahel region of Africa

49


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

village; Table A1 in the Appendix) and the reportedarea (km2) of the extent for which case data were pre-sented as the basis for determining the optimal type ofgeographic representation (points, lines, polygons; rel-ative spatial scale of each geographic entity). The geo-graphic entity of each case was then mapped in theGLOBE online database either by scanning, register-ing, and digitizing published maps in a GIS (shapefilesuploaded into GLOBE), identifying known places anddigitizing these in a GIS or directly in GLOBE usingonline vector mapping tools or by selecting existingpublished kml or shapefiles of known places (GlobalAdministrative Areas 2012; International Union forConservation of Nature and United Nations Environ-ment Programme-World Conservation ManagementCenter 2015). Geographic coordinates and point

geometries were used if no more complete geographicinformation were available in the source. The finalsource data, data quality scores (additional informa-tion later), and geographic representation (onlinemap) were then validated by the mapping team leaderbefore the case was committed to the database. Thefull collection of 437 cases used in this study are sharedonline with the public in the GLOBE system for inter-active geovisualization, analysis, and downloading(Globe Cases Team 2015).

Case Geography Data Quality Scoring

To test for systematic biases in case geographicrepresentation across academic disciplines, geographicentity types, land systems, and time, a data quality

Table 2. Case quality scoring rubric for describing data quality of cases based on how well the geographic entity for whichcase study knowledge is reported (the source data) is described as a spatial unit of Earth’s land surface (case geometry)

Score ProvenanceClarity (case contributor is the

author/site expert)Clarity (case contributor is not

the author/site expert) Conformance

4 Geometry created byauthor/site expert

Geographic entity conformsperfectly with the dataprovided in the source

Geographic entity and geometryfully and professionallydescribed in original source orcorrespond precisely toentities for which precisegeographic data are available

Geometry is entered by uploadingan SHP file or an existinggeometry is selected, the area ofthe geometry entered intoGLOBE agrees with thatreported in the geographicdescription, and a polygon orprecise point geometry is used torepresent the site

3 Geometry not entered byauthor/site expert, andpolygon or precise pointgeometry is used torepresent the site

N/A Geographic entity and geometryare clear in original source,but mapping of the sitegeometry requires someinterpretation before it can bemapped

Geometry is entered using the mapdraw function, the area of thegeometry entered into GLOBEagrees with that reported in thegeographic description, and adetailed polygon or precise pointgeometry is used to represent thesite

2 Geometry entered bytrained GLOBE teammember, approximatepoint geometry is usedto represent the site

Geographic entity conformsroughly to the data providedin the source

Geographic entity describedroughly in original source

The area of the geometry enteredagrees with that reported in thegeographic description, but theClarity Score is less than orequal to 2

1 Geometry entered by acontributor withoutdirect site knowledge,approximate pointgeometry is used torepresent the site

Geographic entity does notclearly conform to the dataprovided in the source

Geographic entity not clearlydescribed in original source

The area of the geometry entereddoes not agree with that reportedin the geographic description;that is, the spatial scales do notmatch

0 Source of the casegeometry is unknown

Data provided in the source donot clearly conform togeographic entities that canbe described here

Geographic entity descriptionmissing or completelyambiguous

Geometry type is unknown or nodata were entered

Note: See Appendix for more detailed information on case quality scoring algorithm.

6 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

pedigree system was used to score the quality of theconformance, provenance, and clarity of geographicrepresentation for each case, using the data qualitypedigree rubric specified in Table 2 and the algorithmimplemented in GLOBE as detailed in Figure A11

(Funtowicz and Ravetz 1990; Costanza, Funtowicz,and Ravetz 1992). Conformance scores were automati-cally computed by the GLOBE system and used to ratespatial agreement between the source reported area ofthe case and the geographic area of the case as com-puted from the mapped geographic entity, as well asthe appropriateness of the geography type (point, poly-gon, line) for the reported geographic entity. Prove-nance scores rated the relative expertise of the casecontributor (study author, expert on site, GIS expert,nonexpert, etc.) and were automatically assigned bythe GLOBE system based on the case contributor’sindication of whether or not they were an author ofthe case source. This was not a useful metric in thisstudy, however, as all cases were contributed by theGLOBE Cases team and thus granted the same score.Clarity scores rated how clearly the geographic entitywas described in the source such that the highestscores required precise geographic descriptions ineither detailed maps, GIS files, or precise coordinates.

Unlike conformance and provenance scores, clarityscores were determined by the GLOBE Cases team.Clarity scores were vetted through an iterative consen-sus-based process. Students were provided with a datapedigree rubric (Table 2) developed by the GLOBEteam. Explanations of the process through which eachstudent arrived at a given clarity score were recordedand provided as Contributor’s Notes (which are view-able to the public online) for every case. Weekly teammeetings were held to review each coded case and theContributor’s Notes that each student provided. Eachcase was presented to the rest of the team and the scoringlogic critiqued.When disagreements about the case scor-ing emerged, the group vetted alternative scoring ration-ales and settled on a final scoring by consensus. Finalcommitment of each case into GLOBE was then con-ducted by one of two team leaders (article coauthors).Thus, quality assurance and score validation were per-formed in an iterative and participatory manner, whichultimately resulted in 100 percent concordance amongstudent scorers, eliminating the need for intercoder reli-ability metrics. The iterative group process was the mostappropriate approach due to the inherently subjectivenature of study site representation, and it also helped torefine the data pedigree and ensure scoring decisionsthat accounted for a diversity of perspectives.

Disciplinary Coding

To test the hypothesis that case quality scores varyamong academic disciplines, cases were coded basedon the major disciplinary and subdisciplinary affilia-tion of the journals in which the studies were pub-lished following the coding protocol of Maglioccaet al. (2015). Cases not obtained from peer-reviewedjournals (books, theses, reports, etc.) were coded basedon title publication for major disciplinary type only. Astandard set of disciplines and subdisciplines wastaken from www.journalseek.net and cross-referencedwith the journal subject area database found atwww.scimagojr.com when multiple journals were clas-sified by multiple disciplines. Only journals explicitlycategorized as multidisciplinary or interdisciplinary(e.g., Science, Nature, Human Ecology, etc.) arereported here as multidisciplinary.

Statistical Analysis

Statistical analysis was conducted using SPSS ver-sion 22 (IBM, Armonk, NY, USA). The original clar-ity and conformance score range from 1 to 4 (low tohigh) was collapsed into a dichotomized low–highscoring rubric owing to the low frequency of 1 and 4clarity scores (N D 43) and 1 and 3 conformancescores (N D 90). Scores of 1 and 2 were reclassified as0 (low), and scores of 3 and 4 were reclassified as 1(high). The decision to collapse the scoring categorieswas made to maximize the sample size of categoriescompared in subsequent analyses to test Hypotheses 1through 4. Statistical comparisons among dichoto-mized clarity and conformance scores across disciplin-ary categories, geographic entity, time periods, andland use types used the Kruskal–Wallis H test (one-way analysis of variance on ranks; Kruskal and Wallis1952). The Kruskal–Wallis H test was selected as themost appropriate nonparametric method to comparedistributions of scores across independent samplesowing to the test’s statistical power when comparingmore than two samples with small sample sizes in mul-tiple pairwise comparisons (Kruskal and Wallis 1952).

Across all tests, statistical distributions of clarityand conformance scores differed across independentvariable groups as assessed by visual inspection of box-plots Pairwise comparisons among categorical groupsused Dunn’s (1964) procedure with a Bonferroni cor-rection for multiple comparisons as a post hoc analysis;adjusted p values are presented throughout the resultssection and in the figures and tables. It is important to


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

http://www.journalseek.nethttp://www.scimagojr.comhttp://www.scimagojr.comhttp://www.scimagojr.comhttp://www.scimagojr.comhttp://www.scimagojr.com

note that when unadjusted p values are corrected formultiple comparisons they can obtain a value of 1.0after adjustment if the unadjusted p value multipliedby the number of categories being compared exceeds1.0. Asymptomatic test statistical significance levelsare reported as the value of the chi-square statisticrather than the Kruskal–Wallis H statistic, but theyare the same value using this statistical test (Kruskaland Wallis 1952).

Dichotomous clarity and conformance scores before(N D 228) and after the year 2005 (N D 209; the yearGoogle Earth was introduced, a popular, free, and

relatively precise online mapping tool) were comparedusing the Mann–Whitney U test, which is the equiva-lent nonparametric statistical test to the Kruskal–Wallis test for when there are only two groups beingcompared (Wilcoxon–Mann–Whitney test; Mann andWhitney 1947). This statistical analysis was conductedto test the hypothesis that there would be statisticallysignificantly higher quality scores after the introduc-tion of Google Earth (studies after 2005) given itsability to offer researchers lacking more advanced geo-spatial skills a simple and relatively precise tool fordescribing the geographic context of case studies.

Figure 2. Concept diagram for determining whether a case meets criteria for spatially explicit sharing of case study knowledge. The conceptdiagram was developed through an iterative and reflexive research process following the compilation, synthesis, and reproduction of 437cases as well as their geographic descriptions and spatial extents.

8 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Results

Through the iterative process of coding and map-ping 437 cases, general patterns of ambiguity in casestudy geographic descriptions were identified, reveal-ing that basic guidelines for these descriptions mighthelp to overcome barriers to case study knowledge

reuse in spatially explicit synthesis research. Statisticalresults are then presented to test our four main hypoth-eses, that case quality scores would vary across majoracademic disciplines, by geographic entity type, byland use type, and over time based on publication date(and, relatedly, that scores would be higher after theavailability of Google Earth in 2005).

Figure 3. Illustrations of several of the most common forms of ambiguous geographies encountered during the process of reproducing 437case geographies. The reproduced geographic descriptions (four map descriptions, two in-text descriptions) display common ambiguities asdescribed in detail Table 3.2 The illustrations highlight how case geographic descriptions that might appear adequate to authors andreviewers often lack sufficiently detailed information to reproduce and reuse these in spatially explicit metastudy research. (A) A commongeographic description of remote sensing studies in which the border of the case is also the border of the figure (boundary representation).(B) A common representation of village studies in which the village or villages are only depicted with point locations at the country scale(point vs. nonpoint geographies, scale of representation), and only coarse geographic coordinates of study locations are provided (coordi-nates). (C) An example of a common representation of villages where only a coarse study area boundary is provided without the precise loca-tion of study villages (area value, scale of representation, local landmarks). (D) A local case description lacking sufficient geographic contextor description for reproducing a study area (coordinates, scales of representation, local landmarks, boundary representation). (E) and (F) Twocommon forms of in-text descriptions of case geographic areas that are insufficient for precise georeferencing of case geographic areas withoutadditional maps and geographic information (in-text descriptors, ephemeral or colloquial descriptors).


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Table 3. Common sources of geographic ambiguities in case studies and suggested improvements for the spatially explicitsharing of case study knowledge

Typology of ambiguity Specific form Description Limitation Suggested improvement

Descriptive In-text description Only in-text descriptionof study area providedfor spatially explicit(e.g., nonpoint)geographic areas

Limits ability of otherresearchers to georeference aspatially explicit study area

In-text study areadescriptors should beaccompanied by a mapor set of maps

Ephemeral orcolloquialdescriptors

In-text description ofstudy area onlyreports colloquial orephemeral study areanames

Study area might be inaccuratelymapped due to confusion overlocation (e.g., a colloquialname might be very commonand a study might be mappedto the wrong location)

Additional (e.g., formaladministrative names)should also be reportedalongside colloquial orephemeral study areanames

Geographic Area value No area value of studyprovided for aspatially explicit casegeography

Area values allow otherresearchers to check theaccuracy of their owngeoreferencing of a study andimprove accuracy ofgeographic reporting

Report study area values forspatially explicit casegeographies

Point versusnonpointgeographies

Studies include a point-based geographywhen they shouldinclude a line orpolygon geographyfor a study occurringover a spatiallyexplicit area

Point geographies do notaccurately describegeographic areas except forvery small study sites.Reporting point geographiesinstead of nonpointgeographies limitsreplicability and reduces theaccuracy of a case geography

Unless a study area is verysmall (typically <1 km2), a nonpointgeography is most likelya more accuraterepresentation of a studyarea

Georeferencing Coordinates Only rough estimates oflatitude andlongitude coordinatesfor a study areprovided

Providing one set of coordinates(latitude, longitude) for alarge study area limits theability of other researchers toaccurately locate orgeoreference a study area

The most specificcoordinates possibleshould be providedrather than one set ofcoordinates intended torepresent a large area

Local landmarks Local landmarks are notprovided asgeographic context instudy area maps

Local landmarks (e.g., rivers,administrative boundaries,etc.) improve the ability ofother researchers toaccurately georeference astudy area

Include local landmarks onstudy area mapswhenever possible toincrease the accuracy ofgeoreferencing

Scale ofrepresentation

Only including onescale of visualrepresentation of astudy geography isprovided

Often sources provide either alocalized geometry or aregional one when bothwould be better for accurategeoreferencing

Include both a local studygeographic extent aswell as map with greatergeographic extentwhenever possible andappropriate

Boundaryrepresentations

The border of the figureis also the study siteboundary

When the study site boundary isused as the outermost borderin a study area’s map, otherresearchers have littleperipheral information to usefor georeferencing the study(common in remote sensingstudies)

Place study area withinbroader geographicextent when visuallydescribing the area ofinterest

10 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

In the process of mapping the geographic contextsof 437 cases, a systematic understanding was developedof the most common ambiguities in case study geo-graphic descriptions that have the potential to hinderaccurate and precise reproduction and reuse of casestudies in spatially explicit regional and globalresearch synthesis efforts. This process also enabled usto understand what geographic information is mostuseful for authors to share in case studies to reduceimprecision and error when individual cases are reusedin synthesis research. The information presented inFigures 2 and 3 and Table 3 was developed through aniterative and consensus-based research process involv-ing both the study authors and the team of graduateand undergraduate students involved in the mappingand coding of cases examined in this study.

In Figure 2, we present a practical rubric for decidingwhat elements of a spatially bounded case can andshould be shared for reuse in spatially explicit regionaland global knowledge generation. To overcome thechallenges of vague or ambiguous presentations of casegeographies, Figure 2 also provides three basic require-ments for researchers determining whether a specificcase meets the essential criteria for sharing a spatiallyexplicit case geography, and Table 3 describes simpleimprovements that can be made to case geographicdescriptions by case creators. Illustrative visual examplesof cases exhibiting many of these forms of ambiguous

geographic representation described in Table 3 are pre-sented in Figure 3 through six different geographicdepictions, with ambiguity types corresponding to thoselisted in Table 3 indicated in parentheses in the figurelegend. These results are intended to assist case studyresearchers in both avoiding the presentation of ambig-uous or imprecise geographic information with casestudies (Table 3 and Figure 3), as well as basic guide-lines for determining whether and what geographicinformation should be presented in spatially explicitcase study research publications (Figure 2).

Quality Scores by Discipline

The distribution of 437 cases across major andminor disciplines is shown in Figure 4. Dichotomizedclarity scores were statistically significantly differentacross disciplines (p < 0.0005, Kruskal–Wallis H test).Dichotomized conformance scores were also statisti-cally significantly different across disciplinary catego-ries (p < .0005, Kruskal–Wallis H test). Earth andplanetary sciences mean rank dichotomized clarity andconformance scores were statistically significantlyhigher than all other major disciplinary groups (p <0.05, Kruskal–Wallis H test; Table 4). Mean clarityand conformance values with confidence intervals bydiscipline are displayed in Figure 5. Based on these

Figure 4. Number (%) and distribution of 437 cases extracted from eight land change science metastudies coded by major and minor disci-plinary categories. (Color figure available online.)


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Figure 5. Mean conformance and clarity scores by major discipline type with standard error bars (confidence interval D 95 percent) for 437cases from eight land change science metastudies. Geography is displayed on the right side of the graph for comparative purposes but thosecases are included under the social sciences category for all statistics presented in the article and were not tested as a statistically indepen-dent sample.

Table 4. Matrix showing results with adjusted p values with a Bonferroni correction for multiple comparisons for major disci-plinary categories (N D 437) for dichotomous clarity (top) and dichotmous conformance (bottom) scores

Clarity Multidisciplinary EconomicsEnvironmental

sciencesBiologicalsciences

Socialsciences

Earth andplanetary sciences

Multidisciplinary 1.0 1.0 0.266 0.061 0.0001Economics 1.0 1.0 1.0 0.026Environmental sciences 1.0 1.0 0.0001Biological sciences 1.0 0.018Social sciences 0.01Earth and planetary sciences

Conformance Multidisciplinary EconomicsEnvironmental

sciencesBiologicalsciences

Socialsciences

Earth andplanetary sciences

Multidisciplinary 1.0 1.0 0.228 0.024 0.0001Economics 1.0 1.0 1.0 0.02Environmental sciences 1.0 0.616 0.0001Biological sciences 1.0 0.005Social sciences 0.004Earth and planetary sciences

Note: Statistically significant different pairwise comparisons are shown in bold (p < 0.05, Kruskal–Wallis H test).

12 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

results, we were able to accept the hypothesis thatthere are disciplinary differences in the quality of geo-graphic reporting of case studies, with geospatial disci-plines (earth and planetary sciences) evidencinghigher quality scores than other disciplines.

Quality Scores by Geographic Entity Type

Statistically significant differences in clarity scoreswere observed across the eleven most common geo-graphic entities in the collection (N D 381; sixteen

Figure 6. Mean clarity (top) and conformance scores (bottom) by most common geographic entity types with confidence interval error bars(confidence interval D 95 percent) for 381 cases from eight land change science metastudies. Bars ordered from lowest to highest meanscores.


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

entities with fewer than eleven cases were omittedfrom analysis; p < 0.0005, Kruskal–Wallis H test).3

There were also statistically significant differences inconformance scores across the eleven most commongeographic entities in the collection (p < 0.0005,Kruskal–Wallis H test). Entity types watershed andcounty had the highest mean clarity and conformancescores (Figure 6). Statistically significant differences

in mean rank dichotomized clarity and conformancescores between entity types are indicated in Table 5(p < 0.05, Kruskal–Wallis H test). Mean and meanrank clarity and conformance scores by geographicentity are presented in Table 6.

To expand the sample size of categories by entitytypes and look for further patterns in the data set, geo-graphic entities were recategorized by a broader

Table 5. Matrix showing results with adjusted p values with a Bonferroni correction for multiple comparisons for eleven geo-graphic entity types (N D 381) for dichotomous clarity (top) and dichotomous conformance (bottom) scores

Clarity Farm TownStudyarea Forest Pasture Province Village

Remotesensing image Region County Watershed

Farm 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.007 0.019Town 1.0 1.0 1.0 1.0 1.0 1.0 0.525 0.054 0.1Study area 1.0 1.0 1.0 1.0 1.0 0.074 0.013 0.047Forest 1.0 1.0 1.0 1.0 1.0 0.110 0.217Pasture 1.0 1.0 1.0 0.672 0.071 0.165Province 1.0 1.0 1.0 0.252 0.448Village 1.0 1.0 0.267 0.563Remote sensing image 1.0 1.0 1.0Region 1.0 1.0County 1.0Watershed

Conformance Farm Town Study area Forest Pasture Province Village Remote sensing image Region County Watershed

Farm 1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.005 0.014Town 1.0 1.0 1.0 1.0 1.0 1.0 1.0 0.042 0.079Study area 1.0 1.0 1.0 1.0 1.0 1.0 0.009 0.036Forest 1.0 1.0 1.0 1.0 1.0 0.036 0.086Pasture 1.0 1.0 1.0 1.0 0.055 0.013Province 1.0 1.0 1.0 0.206 0.376Village 1.0 1.0 0.219 0.476Remote sensing image 1.0 0.485 0.798Region 1.0 1.0County 1.0Watershed

Note: Statistically significant different pairwise comparisons are shown in bold (p < 0.05, Kruskal–Wallis H test).

Table 6. Mean and mean rank clarity and conformance scores across the eleven most frequent geographic entity types (N D381), sorted high to low by mean rank clarity score (Kruskal–Wallis H test)

Geographic entity Mean clarity score Mean rank clarity score Mean conformance score Mean rank conformance score

Watershed 2.7 267.9 3.4 271.9County 2.6 261.5 3.3 265.5Region 2.5 223.7 2.9 199.3Remote sensing image 2.3 198.0 2.7 202.0Village 2.2 193.9 2.5 197.9Province 2.3 184.9 2.5 188.9Pasture 2.1 175.7 1.8 179.7Forest 2.1 175.3 2.2 172.5Study area 2.0 167.5 2.3 171.5Town 2.2 158.3 2.0 162.3Farm 1.9 145.1 1.7 149.1

14 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

typology into political, observational, and land units;no statistically significant differences in clarity or con-formance scores among these categories were observed(Table A.2, p > 0.05, Kruskal–Wallis H test).

Quality Scores by Anthrome

The anthrome level classification of 437 case loca-tions was determined (Ellis et al. 2010; Schmill et al.2014). Cases spanned all six anthrome levels—wild-lands (n D 13), seminatural (n D 184), rangelands (nD 110), croplands (n D 76), villages (n D 39), anddense settlements (n D 15; Ellis et al. 2010)—but nostatistically significant differences were observedamong their dichotomous clarity or conformancescores (p > 0.05, Kruskal–Wallis H test). We weretherefore unable to accept the hypothesis that moreintensively managed land use types (e.g., dense settle-ments, villages) would have statistically significantlyhigher quality scores than less intensively managedland use types (e.g., wildlands or rangelands).

Quality Scores by Publication Date

We failed to accept the hypothesis that clarity andconformance scores would improve over time. Clarity

and conformance scores showed no general temporaltrend but did show statistically significant differencesbased on the publication date of cases when testedacross seventeen temporally binned groups using anequal percentile binning strategy as shown in Figure 7,but we found no interpretable trend in the results overtime (5.56 percent of total cases per bin; p < 0.0005,Kruskal–Wallis H test). Number of bins was selectedbased on an iterative visual binning of the data acrosstime to ensure a sufficient number of temporal cut-points to capture changes in geographic quality report-ing over time alongside the rapid acceleration ofgeospatial tools beginning in the 1990s. When testedfor a change in clarity and conformance scores beforeand after the introduction of Google Earth in 2005, nostatistically significant differences in dichotomizedscores were observed between cases published beforeversus after 2005 (p > 0.5, Mann–Whitney U test).

Discussion

For case study researchers who define spatiallyexplicit units of knowledge sharing in their publishedwork, the basic requirements outlined in Figure 2 arestraightforward and relatively easy to meet with tech-niques commonly available to all. It is therefore all themore striking that these simple methods for geographic

Figure 7. Ninety-five percent confidence intervals of mean clarity and conformance scores for 437 cases across seventeen equal percentilebins (5.56 percent of cases per bin). Mean interpolation lines across bins are presented as a visual aid.


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

data sharing are not consistently applied in the pub-lished case study literature. A frequent example is theuse of point locations, rather than polygons, to describegeographic entities that cover significant areas of theEarth’s surface. In sixty-seven cases, geographicdescriptions did not allow the geographic context of acase to be reproduced in greater detail than as a point(area covered D 0 km2) despite the presentation ofcase knowledge representing a geographic entity suchas a city or forest that quite likely covered areas of atleast a square kilometer or greater. Except for caseswith very small geographic extents, such as studies ofindividual fields or ecological observational plots, stud-ies with spatial units of knowledge generation coveringgeographic extents of one hectare and greater shouldutilize polygon representations, not points. Althoughit is understandable that case study researchers mightsometimes feel that coupling their case study knowl-edge sharing within spatially explicit areas of theEarth’s surface will inadequately or incompletelydescribe the geographic contexts of their work, for themany studies meeting the criteria in Figure 2, the shar-ing of precise geographic contexts together with caseknowledge would greatly improve ongoing spatiallyexplicit regional and global synthesis efforts across theland change and environmental social sciences.

Spatial Social Sciences Need to Do Better Geography

The results of this study indicate that some disci-plines are more inclined to publish more precise geo-graphic descriptions than others, with cases publishedin journals categorized within earth and planetary sci-ences producing clearer and more easily reproduciblespatially explicit case geographic descriptions thanthose published in other journal disciplinary categories(Table 4). Likely, this finding is explained by the com-mon use of GIS and other geospatial tools in this disci-plinary category (satellite imagery, remote sensingscenes, etc.) and a general familiarity with producingand using spatially explicit knowledge and data atregional to global spatial scales. Surprisingly, casespublished in journals categorized within geography(presented within the broader category of social scien-ces; see Figure 4) tended toward lower clarity and con-formance scores than earth and planetary sciences, butthe differences between scores for geography cases as asubdiscipline (n D 37) and earth and planetary scien-ces were not statistically significantly different whencompared as independent categories in a separate sta-tistical test (p > 0.05, Kruskal–Wallis H test).

The reasons why the clarity of geographic descrip-tions published in an explicitly spatial discipline mightbe lower than those of other disciplines cannot bedecided from the data presented here owing to a rela-tively small sample size and the absence of more detailedfactors in this study. The interdisciplinary nature ofgeography and its diversity of methodologicalapproaches is one possibility (Kwan 2004), along withthe possibility of a bias toward the study of types or scalesof geographic entities, land systems, or geographicextents that are more difficult to spatially delineate com-pared with those commonly used in other disciplines.The median reported geographic extent of cases in geog-raphy (19.5 km2) was much smaller than those of theearth and planetary sciences (1,250 km2), and themajority of cases in geography represented knowledgefrom sites scaled from 1 ha to 100 km2 (56 percent).Yet the complete set of studies conducted at this scale(N D 118) had modestly higher conformance scoresthan those at larger scales (>100 to 1,000 km2). It ispossible that further studies specifically examiningthese relationships within the discipline of geographymight reveal intradisciplinary biases in geographicextents or entities leading to lower clarity and confor-mance scores.

Fuzzy Boundaries Produce Fuzzy Data

The hypothesis that quality scores would differ bygeographic entity type is supported by the results pre-sented in Table 5 (p < 0.0005). As frequently mappedunits, it is intuitive that watershed (reproducible basedon terrain data maps in a GIS) and county (an easilyreproducible administrative unit) would receive higherclarity and conformance scores compared to moreambiguous geography types such as farm, town, orstudy area that have less explicit spatial delineationsand are more difficult to map and reproduce from pub-lished studies (Tables 5 and 6). To further investigatethis hypothesis, a post hoc analysis combining entitytypes into broader categories (political, observational,and land units) was conducted but did not reveal sig-nificant differences or further explain differences inscores across entity types (p > 0.05, Table A2). Quali-tatively, there were no apparent patterns between geo-graphic entities with higher quality scores anddisciplines with higher scores, but the limited numberof cases across entity types by disciplines preventedquantitative comparison (Table A3).

The results of the statistical tests do raise the issue ofhow one should best represent geographic entities with

16 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

fuzzy boundaries or with multiple ways of demarcatingboundaries. For example, villages represent a particu-larly fuzzy form of geographic entity (e.g., the boundaryof a village could be based on an administrative bound-ary, informal local knowledge, or parcel sizes; Figure 1),and we recommend that researchers be explicit indescribing how such boundaries are defined. We arenot advocating, however, for a one-size-fits-allapproach to how the boundaries of such an entityshould be defined; such decisions need to be made byindividual researchers informed by the context of thestudy. Instead, individual cases should sufficientlydescribe how a boundary was selected, and present suffi-cient information to improve the clarity and reproduc-ibility of the geographic extent of the case (Figure 2).

Geographic Description Has Not Improved OverTime

We were surprised by the finding that clarity andconformance scores did not improve over time (Fig-ure 7). The dramatic growth in availability of geospa-tial tools, including Global Positioning System (GPS),GIS, and especially free and open-source mapping pro-grams such as QGIS and Google Earth, was expected tocause long-term increases in case geographic qualityscores over the time frame of this study (1936–2012).The absence of any statistically significant upwardtrend in the quality of case geographic representationwas therefore both unexpected and striking (Figure 7).What is clear is that the remarkable advances in geo-spatial tool availability of recent decades have, inthemselves, had little effect on the quality of geo-graphic representation in published case study research.This statistical finding mirrors the subjective experi-ence of the team in mapping the 437 cases employed inthese analyses and helped drive us to elaborate thesewidespread long-term practices of ambiguous geo-graphic description in Table 3 and Figure 3.

A Persistent Problem: Ambiguous SpatialityChallenges Synthesis Research

There are many different reasons why studies operat-ing within a spatial context might be difficult or evenimpossible to describe within Cartesian space, justifi-ably leading to ambiguous geographic descriptions(Figure 2). In studies emphasizing interactive pro-cesses, spatial fluidity, and the interconnectivity ofsites, these spatially delimited approaches to geographic

representation might be impossible to reconcile withsome research agendas and might even be seen as pro-moting notions of hierarchical scale that certain studiesseek to deconstruct or critique. Nevertheless, for manyresearchers, including critical scholars and humangeographers, the boundaries of political administrativeunits, biophysical areas, or artificial study plots mightalso be essential to a study’s design or even the object ofstudy itself. Accurately and precisely mapping theseboundaries and sharing this information with othershas the potential to enable broader and more generalanalyses aimed at understanding how global processesand flows are acted out on and across social sites glob-ally and within multiple geographic contexts.

It is relevant to note how other spatially orienteddisciplines have also grappled with questions of scalingbetween local and global research in efforts to producegeneralizable theories on environmental change(Rindfuss et al. 2004, 2007; Lambin and Geist 2006;Verburg, Neumann, and Nol 2011; Verburg et al.2012). Although physical geography and land changescience might engage less critically in their conceptu-alizations of scale and space as analytical tools (Moore2008), there is nevertheless a robust literature outsidethe remit of human geography asking related questionsabout spatial representation and linkages betweenfine-grained studies of relatively small geographicextents and global patterns and processes (Jelinski andWu 1996; Geist and Lambin 2002; Kwan 2004; Lam-bin and Geist 2006; Goodchild, Yuan, and Cova 2007;Turner, Lambin, and Reenberg 2007; Goodchild 2004;Karl et al. 2013). In the GISciences, theoretical andtechnological research has advanced methodologiesfor selecting and demarcating the appropriate spatio-temporal contexts exerting influence on study subjects(Kwan 2012, 2013). Kwan (2000, 2012, 2013) andGoodchild (2004, 2012) described how theGISciences and new spatial technologies such as GPStracking can help reconcile issues related to the modi-fiable areal unit problem (MAUP; Openshaw 1984)and the more recently described uncertain geographiccontext problem (UGCoP) to improve the selectionof appropriate spatiotemporal contexts and zones ofanalysis used in social science studies. These advancesin describing and conceiving of temporal units of caseanalysis present additional challenges in how caseresearchers make clear the boundaries of a case bothspatially and temporally. By highlighting the persistentproblem of ambiguous geographic description in thereporting and sharing of spatially explicit case studyknowledge, our work aims to complement rather than


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

conflict with efforts to advance these important theo-retical and methodological engagements with scaleand spatial representation.

Improving the Representation and Sharing ofSpatially Explicit Knowledge

Despite the finding that earth and planetary scien-ces studies appear to represent case geographies in amore spatially explicit and clear manner comparedwith other major disciplines, our results have notrevealed any specific causal relationships that mightexplain differences in the relative quality of geo-graphic descriptions across land change science meta-studies. Still, by metastudy and exploration of casestudy geographic reporting, it has become absolutelyclear that there is a basic need to overcome disciplin-ary cultural tolerances to ambiguous geographic repre-sentation in spatial research. As has been previouslydemonstrated for ecological studies, even the inclusionof accurate geographic coordinates representing astudy area’s centroid as a scale-neutral point are oftenlacking from published studies, a relatively poor formof geographic representation for spatially boundedcases covering an area of the Earth’s surface greaterthan one hectare (Karl et al. 2013). The results pre-sented here reinforce the notion that there is a needfor greater development of common language andguidelines for describing the geographic context ofspatially explicit case research. We believe that theguidelines presented in this article begin to addressthis particular barrier to knowledge synthesis.

In addition to the recommendations outlined inFigure 2 and Table 3, there are other practical oppor-tunities for improving the replicability of spatiallyexplicit knowledge and how it is shared across a diver-sity of spatially oriented scholarship. First, we believethat it is essential that more scholarly journalsand their publishers enable—or, better, require—researchers to share and make available for freedownloadable spatial files (shapefiles or kml) of thegeographic extent of studies. Although an increasingnumber of journals and publishers offer this option,many, including top-tier geography journals such asthe Annals of the American Association of Geographersand The Professional Geographer do not explicitly do so.This will enable synthesis researchers to understandthe geographic extent across which the findings of astudy are valid and avoid producing errors in attempt-ing to reproduce case geographies themselves. In themeantime, we encourage researchers to make such files

available and downloadable through their own per-sonal or institutional Web sites.

Second, recently developed tools such as GLOBE(globe.umbc.edu) and JournalMap (www.journalmap.org) are important new platforms in which researcherscan share, compare, and download the geographiclocation or extents of case studies and conduct analy-ses connecting local case study research with globaldata sets (Ellis 2012; Karl et al. 2013). Such effortsrepresent an important development for spatially ori-ented disciplines to understand the global and regionalcontexts of local case study research in a spatiallyexplicit manner. We hope that more researchers willconsider using such platforms to share their researchin a spatially explicit manner that preserves the geo-graphic fidelity of their work. Third, we note thatopen data sharing has been shown to provide signifi-cant benefits to the authors of published studies, byincreasing the reuse and citation of published work, afundamental reason why individual case studyresearchers should embrace the processes of open shar-ing of their published work in the most data-rich for-mats available (Piwowar and Vision 2013).

Conclusions

The divide between local and global knowledge gen-eration in the social and environmental sciences islikely to persist. This study, however, identifies onesource of this division and helps to bridge this divide byenhancing the spatially explicit reuse of knowledgegenerated at more local geographic extents in globaland regional scale synthetic research. Although ouranalysis draws on a limited set of cases used in eightland change metastudies, its results are more broadlyrelevant to all who produce case studies in local geo-graphic contexts and to those who use them to synthe-size broader scale insights. Although critiques of scalespecificity are merited, there is a clear lack of significantimprovement in case geographic descriptions overtime, despite advances in widely available tools to sup-port this. We suggest that the prevalence of ambiguousgeographic representations observed over time has littleto no relation to the scale-theoretical concerns of casestudy researchers but rather has resulted from the toler-ance of ambiguous geographic descriptions in the publi-cations of some disciplines, geography among them,even when the geographic contexts of case knowledgeare explicit in principle. We hope that in highlightingpractical strategies for clear and concise case geographiccontext reporting, this work will help to improve efforts

18 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

http://www.journalmap.orghttp://www.journalmap.org

to connect fine-grained and coarser-grained researchagendas and toward an overall improvement in howsocial and environmental scientists report on and usethe geographic contexts of their research.

Acknowledgments

The authors thank three anonymous reviewers andMei-Po Kwan for their thoughtful comments and con-structive suggestions on earlier revisions of the article.This research would have been impossible without theassistance and hard work of the GLOBE Cases Team atUniversity of Maryland, Baltimore County (UMBC),including Gailynn Milligan, Joseph Milligan, LaureenEchiverri, Brandon Cottom, Michael Glassman, Mat-thew Gregory, Marissa Lenoce, and Anna Johnson.Lindsey Gordon and Christopher Zink of the CasesTeam deserve particular mention for their long-termdedication to the project and insight on forms of geo-graphic ambiguity in the studies reviewed here. Finally,we thank David Lansing at UMBC for his thoughts ongeographical scale that were helpful during the earlydevelopment of the article.

Funding

This material is based on work supported by theU.S. National Science Foundation under grant NSF#1125210 and cosponsored by the Global Land Project(www.globallandproject.org) and the InternationalNetwork of Research on Coupled Human and NaturalSystems (www.chans-net.org). Any opinions, findings,and conclusions or recommendations expressed in thismaterial are those of the authors and do not necessar-ily reflect the views of the National ScienceFoundation.

Notes1. Additional case scoring documentation is available at

http://globe.umbc.edu/documentation-overview/cases-documentation/.

2. Maps and descriptions are reproductions of actual geo-graphic descriptions encountered during research. Toretain author and publication confidentiality, placenames, land use classification types, coordinates, andlocations on continent-scale maps (7b, 7c) wereremoved and replaced with generic placeholder text.All figures presented here demonstrate common formsof case geographic descriptions encountered during thereview and reproduction of 437 cases. The descriptionsselected and presented here were chosen for their cleardepiction of these issues, not because they represented

especially poor case geographic descriptions. Biblio-graphic information for figure sources is not included toprotect the identities of the authors but is available onrequest from the first author.

3. Geoentity analysis excludes fifty-six studies from lesscommon entity types: basin (n D 2), catchment (n D5), city (n D 2), country (n D 4), district (n D 9), island(n D 3), municipality (n D 4), parcel (n D 1), park (n D2), plot (n D 3), protected area (n D 5), quadrat (n D2), river (n D 1), state (n D 3), and unknown (n D 10)geographic entities.

References

Achten, W., and L. Verchot. 2011. Implications of biodie-sel-induced land-use changes for CO2 emissions: Casestudies in tropical America, Africa, and SoutheastAsia. Ecology and Society 16 (4): 14.

Brenner, N. 2001. The limits to scale? Methodologicalreflections on scalar structuration. Progress in HumanGeography 25 (4): 591–614.

Costanza, R., S. O. Funtowicz, and J. R. Ravetz. 1992. Assessingand communicating data quality in policy-relevantresearch.EnvironmentalManagement 16 (1): 121–31.

Cox, M. 2014. Understanding large social-ecological sys-tems: Introducing the SESMAD project. InternationalJournal of the Commons 8 (2): 265–76.

———. 2015. A basic guide for empirical environmentalsocial science. Ecology and Society 20 (1): 63.

Downey, L. 2006. Using geographic information systems toreconceptualize spatial relationships and ecological con-text.AJS: American Journal of Sociology 112 (2): 567–612.

Dunn, O. J. 1964. Multiple comparisons using rank sums.Technometrics 6 (3): 241–52.

Eclesia, R., E. Jobbagy, R. Jackson, F. Biganzoli, and G.Pineiro. 2012. Shifts in soil organic carbon for planta-tion and pasture establishment in native forests andgrasslands of South America. Global Change Biology 18(10): 3237–51.

Ellis, E. C. 2012. The GLOBE Project: Accelerating globalsynthesis of local studies in land change science. News-letter of the Global Land Project 8:5–6.

Ellis, E. C., K. Klein Goldewijk, S. Siebert, D. Lightman,and N. Ramankutty. 2010. Anthropogenic transforma-tion of the biomes, 1700 to 2000. Global Ecology andBiogeography 19 (5): 589–606.

Flyvbjerg, B. 2006. Five misunderstandings about case-studyresearch. Qualitative Inquiry 12 (2): 219–45.

Funtowicz, S. O., and J. R. Ravetz. 1990.Uncertainty and qual-ity in science for policy. (Vol. 15). New York: Springer.

Geist, H. J., and E. F. Lambin. 2002. Proximate causes andunderlying driving forces of tropical deforestation. Bio-Science 52 (2): 143.

Global Administrative Areas. 2012. GADM database ofGlobal Administrative Areas, version 2.0. http://www.gadm.org (last accessed 2 January 2016).

Globe Cases Team. 2015. Margulies case collection.GLOBE Collection of Georeferenced Case Studies. http://globe.umbc.edu/app/#/manager/collections/view/990(last accessed 5 December 2015).

Goodchild, M. F. 2004. GIScience, geography, form, andprocess. Annals of the Association of American Geogra-phers 94 (4): 709–14.


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

http://www.globallandproject.orghttp://www.chans-net.orghttp://www.chans-net.orghttp://www.chans-net.orghttp://www.chans-net.orghttp://www.chans-net.orghttp://globe.umbc.edu/documentation-overview/cases-documentation/http://globe.umbc.edu/documentation-overview/cases-documentation/http://www.gadm.orghttp://www.gadm.orghttp://globe.umbc.edu/app/#/manager/collections/view/990http://globe.umbc.edu/app/#/manager/collections/view/990

——. 2012. The future of Digital Earth. Annals of GIS 18(2): 93–98.

Goodchild, M., R. Haining, and S. Wise. 1992. IntegratingGIS and spatial data analysis: Problems and possibili-ties. International Journal of Geographical InformationSystems 6 (5): 407–23.

Goodchild, M. F., and L. Li. 2012. Assuring the quality ofvolunteered geographic information. Spatial Statistics1:110–20.

Goodchild, M. F., M. Yuan, and T. J. Cova. 2007. Towardsa general theory of geographic representation in GIS.International Journal of Geographical Information Science21 (3): 239–60.

IBM Corp. Released 2013. IBM SPSS Statistics for Macin-tosh, Version 22.0. Armonk, NY: IBM Corp.

International Union for Conservation of Nature andUnited Nations Environment Programme-World Con-servation Management Center. 2015. The world data-base on protected areas (WDPA). Cambridge, UK:United Nations Environment Programme-World Con-servation Management Center. http://www.protectedplanet.net (last accessed 2 January 2016).

Jelinski, D. E., and J. Wu. 1996. The modifiable areal unitproblem and implications for landscape ecology. Land-scape Ecology 11 (3): 129–40.

Karl, J. W., J. E. Herrick, R. S. Unnasch, J. K. Gillan, E.C. Ellis, W. G. Lutters, and L. J. Martin. 2013. Dis-covering ecologically-relevant knowledge from pub-lished studies through geo-semantic searching.BioScience 63:674–82.

Kauffman, J., F. Hughes, and C. Heider. 2009. Carbon pooland biomass dynamics associated with deforestation,land use, and agricultural abandonment in the neotrop-ics. Ecological Applications 19 (5): 1211–22.

Keys, E., and W. J. McConnell. 2005. Global change andthe intensification of agriculture in the tropics. GlobalEnvironmental Change 15 (4): 320–37.

Kruskal, W. H., and W. A. Wallis. 1952. Use of ranks inone-criterion variance analysis. Journal of the AmericanStatistical Association, 47 (260): 583–621.

Kwan, M. P. 2000. Analysis of human spatial behavior in aGIS environment: Recent developments and futureprospects. Journal of Geographical Systems 2 (1): 85–90.

———. 2004. Beyond difference: From canonical geogra-phy to hybrid geographies. Annals of the Association ofAmerican Geographers 94 (4): 756–63.

———. 2012. The uncertain geographic context problem.Annals of the Association of American Geographers 102(5): 958–68.

———. 2013. Beyond space (as we knew it): Toward tem-porally integrated geographies of segregation, health,and accessibility. Annals of the Association of AmericanGeographers 103 (5): 1078–86.

Lambin, E. F., and H. J. Geist. 2006. Land-use and land-coverchange: Local processes and global impacts. New York:Springer.

Laney, R. M. 2002. Disaggregating induced intensificationfor land-change analysis: A case study from Madagas-car. Annals of the Association of American Geographers92 (4): 702–26.

Magliocca, N. R., T. K. Rudel, P. H. Verburg, W. J. McCon-nell, O. Mertz, K. Gerstner, A. Heinimann, and E. C.

Ellis. 2015. Synthesis in land change science: Methodo-logical patterns, challenges, and guidelines. RegionalEnvironmental Change 15 (2): 211–26.

Mann, H. B., and D. R. Whitney. 1947. On a test ofwhether one of two-random variables is stochasticallylarger than the other. The Annals of Mathematical Statis-tics 18 (1): 50–60.

Marston, S. A. 2000. The social construction of scale. Prog-ress in Human Geography 24 (2): 219–42.

Marston, S. A., J. P. Jones, and K. Woodward. 2005. Humangeography without scale. Transactions of the Institute ofBritish Geographers 30 (4): 416–32.

Miller, H. 2007. Place-based versus people-based geographicinformation science. Geography Compass 1 (3): 503–35.

Moore, A. 2008. Rethinking scale as a geographical cate-gory: From analysis to practice. Progress in Human Geog-raphy 32 (2): 203–25.

Moritz, M., J. Giblin, M. Ciccone, A. Davis, J. Fuhrman, M.Kimiaie, S. Madzsar, K. Olson, and M. Senn. 2011.Social risk management strategies in pastoral systems:A qualitative comparative analysis. Cross-CulturalResearch 45 (3): 286–317.

Murdock, G. P., and D. R. White. 1969. Standard cross-cul-tural sample. Ethnology 8:329–69.

Openshaw, S. 1984. The modifiable areal unit problem. Nor-wich, UK: Geo Books.

Piwowar, H. A., and T. J. Vision. 2013. Data reuse and theopen data citation advantage. PeerJ 1:e175.

Ragin, C. C., and H. S. Becker. (Eds.) (1992). What is acase? Exploring the foundations of social inquiry. Cam-bridge, UK: Cambridge University Press.

Rindfuss, R. R., B. Entwisle, S. J. Walsh, C. F. Mena, C. M.Erlien, and C. J. Gray. 2007. Frontier land use change:Synthesis, challenges, and next steps. Annals of theAssociation of American Geographers 97 (4): 739–54.

Rindfuss, R. R., S. J. Walsh, B. L. Turner, J. Fox, and V.Mishra. 2004. Developing a science of land change:Challenges and methodological issues. Proceedings of theNational Academy of Sciences of the United States ofAmerica 101 (39): 13976–81.

Rudel, T. K. 2008. Meta-analyses of case studies: A methodfor studying regional and global environmental change.Global Environmental Change 18 (1): 1825.

Sayre, N. F. 2005. Ecological and geographical scale: Paral-lels and potential for integration. Progress in HumanGeography 29 (3): 276–90.

Schmill, M. D., L. M. Gordon, N. R. Magliocca, E. C. Ellis,and T. Oates. 2014. GLOBE: Analytics for assessingglobal representativeness. In 2014 Fifth InternationalConference on Computing for Geospatial Research andApplication, 25–32. Washington, DC: IEEE.

Turner, B. L., II, R. O. Hanham, and A. V. Portararo. 1977.Population pressure and agricultural intensity. Annalsof the Association of American Geographers 67 (3): 384–96.

Turner, B. L., II, R. E. Kasperson, W. B. Meyer, K. M. Dow, D.Golding, J. X. Kasperson, R. C. Mitchell, and S. J. Ratick.1990. Two types of global environmental change: Defini-tional and spatial-scale issues in their human dimensions.Global Environmental Change 1:14–22.

Turner, B. L., II, E. F. Lambin, and A. Reenberg. 2007. Theemergence of land change science for global

20 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

http://www.protectedplanet.nethttp://www.protectedplanet.net

environmental change and sustainability. Proceedings ofthe National Academy of Sciences 104 (52): 20666–71.

van Vliet, J., N. R. Magliocca, B. Buchner, E. Cook, J. M.Rey Benayas, E. C. Ellis, A. Heinimann, et al. 2016.Meta-studies in land use science: Current coverage andprospects. Ambio 45 (1): 15–28.

van Vliet, N., O. Mertz, A. Heinimann, T. Langanke, C.Adams, P. Messerli, S. Leisz, et al. 2012. Trends,drivers and impacts of changes in swidden cultiva-tion in tropical forest-agriculture frontiers: A globalassessment. Global Environmental Change 22 (2):418–29.

van Vliet, J., N. R. Magliocca, B. B€uchner, E. Cook, J. M.Rey Benayas, E. C. Ellis, A. Heinimann, et al. 2016.Meta-studies in land use science: Current coverage andprospects. Ambio 45 (1): 15–28.

van Vliet, N., A. Reenberg, and L. V. Rasmussen. 2013. Sci-entific documentation of crop land changes in theSahel: A half empty box of knowledge to support poli-cy?. Journal of Arid Environments 95:1–13.

Verburg, P. H., N. Crossman, E. C. Ellis, A. Heinimann, P.Hostert, O. Mertz, H. Nagendra,et al. Forthcoming.Land system science and sustainable development ofthe earth system: A global land project perspective.Anthropocene. Advance online publication.

Verburg, P. H., E. C. Ellis, and A. Letourneau. 2011. Aglobal assessment of market accessibility and marketinfluence for global environmental change studies.Environmental Research Letters 6 (3): 034019.

Verburg, P. H., K. Neumann, and L. Nol. 2011. Challengesin using land use and land cover data for global changestudies. Global Change Biology 17 (2): 974–89.

Verburg, P. H., S. van Asselen, E. H. van der Zanden, andE. Stehfest. 2012. The representation of landscapes inglobal scale assessments of environmental change.Landscape Ecology 28 (6): 1067–80.

Yin, R. K. 2013. Case study research: Design and methods. 5thed. Thousand Oaks, CA: Sage.

Young, A. L., and W. G. Lutters. 2015. (Re)defining landchange science through synthetic research practices. InProceedings of the 18th ACM Conference on ComputerSupported Cooperative Work & Social Computing, ed. D.Cosley, A. Forte, L. Ciolfi, and D. McDonald, 431–42.Vancouver, BC, Canada: ACM.

JARED D. MARGULIES is a PhD Candidate in theDepartment of Geography and Environmental Systems atthe University of Maryland, Baltimore County, Baltimore,MD 21250. E-mail: [email protected]. His researchinterests include human–animal relations, political ecolo-gies of conservation, and spatiotemporal contexts in geo-graphical research.

NICHOLAS R. MAGLIOCCA is an Assistant ResearchProfessor at the National Socio-Environmental SynthesisCenter, University of Maryland, Annapolis, MD 21401. E-mail: [email protected]. His research interests includemeta-analysis methodology and agent-based modeling ofcoupled human–natural systems, particularly in the contextof the causes and consequences of land-use change.

MATTHEW D. SCHMILL is an Assistant Research Scien-tist in the Computer Science and Electrical EngineeringDepartment at the University of Maryland, BaltimoreCounty, Baltimore, MD 21250. E-mail: [email protected] is the system architect for the GLOBE Project (www.globe.umbc.edu). He currently studies scientific collabora-tion, geocomputation, data mining, and machine learning.

ERLE C. ELLIS is Professor of Geography & EnvironmentalSystems at the University of Maryland, Baltimore County,and Visiting Professor of Landscape Architecture at theGraduate School of Design, Harvard University. E-mail:[email protected]. His research investigates human–environ-ment interactions at local, regional, and global scales.


Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

http://www.globe.umbc.eduhttp://www.globe.umbc.edu

Appendix

An analysis reclassifying all of the GLOBE geo-graphic entities types into political units, observa-tional units, and land units revealed no statisticallysignificant differences in dichotomous (high–low) clar-ity and conformance scores based on a Kruskal–WallisH test (Figure A2). The binning structure is describedin Table A2. A Kruskal–Wallis H test was conductedto determine if there were differences in dichotomousclarity and conformance scores between observationalunit (n D 72), land unit (n D 112), and political unit(n D 243) geographic entity categories. Values aremean ranks unless otherwise stated. Distributions ofunit scores were not similar for all groups, as assessedby visual inspection of a boxplot. Unit scores increasedfrom observational units (200.41), to land units(205.28), to political units (222.05) based on clarityscores and from observational units (204.41), to landunits (207.38), to political units (219.90) based onconformance scores, but the differences were not sta-tistically significant for clarity, x2(2) D 3.914, p D.141, or conformance, x2(2) D 2.165, p D .339.

Figure A1. Conceptual flowchart and algorithm visualization forhow GLOBE case quality scores are generated based on a pedigreescoring rubric (outlined in Table 2). (Color figure available online.)

22 Margulies et al.

Dow

nloa

ded

by [

Uni

vers

ity O

f M

aryl

and]

at 1

3:25

11

Apr

il 20

16

Table A1. Geographic entity types with definitions and examples as employed in the coding and case creation procedure for437 cases

Name Definition Examples

Archaeological site An archaeological site with area less than 100 ha Archaeological research sites < 100 ha (larger, seearchaeological complex)

Archaeological complex An area of archaeological observation with areagreater than 100 ha

An archaeological site group, or large urban complexor cluster

Built structure Human built structures, including buildings, airports,dams, hospitals, etc. Note: Linear structuresincluding irrigation canals have their own geoentity

School, hospital, power station, airport

Catchment An area of land where surface water converges to asingle point at a lower elevation, usually the exit ofthe basin, where the waters join another water body

Map or geometry of drainage basin, catchment areaprovided

City A relatively lar

Knowledge with Global Change Science Ambiguous Geographies ...globe.umbc.edu/wp-content/uploads/2016/04/margulies_2016.pdf · Ambiguous Geographies: Connecting Case Study Knowledge

Documents