A Review of Geospatial Content in IEEE Visualization Publications
Alexander Yoshizumi, Megan M. Coffer, Elyssa L. Collins, Mollie D. Gaines, Xiaojie Gao, Kate Jones, Ian R. McGregor, Katie A. McQuillan, Vinicius Perin, Laura M. Tomkins, Thom Worm, Laura Tateosian
AA Review of Geospatial Content in IEEE Visualization Publications
Alexander Yoshizumi * Megan M. Coffer † Elyssa L. Collins † Mollie D. Gaines † Xiaojie Gao † Kate Jones † Ian R. McGregor † Katie A. McQuillan † Vinicius Perin † Laura M. Tomkins † Thom Worm † Laura Tateosian * Center for Geospatial Analytics, North Carolina State UniversityFigure 1: Attributes of 94 IEEE VIS papers from years 2017-2019 found to have geospatial content. From top to bottom: data domain,geospatial nature of the paper (GEO), and VIS Conference paper type and track (TRK) are shown for each paper. Percentages onthe top band (lightest gray bars) correspond to data domain types. The GEO band marks papers as either containing both geospatialdata and a geospatial analysis (dark gray) or geospatial data only (light gray). The TRK band is colored by the VIS Conferencepaper types and tracks listed on Open Access VIS [15]. A BSTRACT
Geospatial analysis is crucial for addressing many of the world’smost pressing challenges. Given this, there is immense value inimproving and expanding the visualization techniques used to com-municate geospatial data. In this work, we explore this importantintersection – between geospatial analytics and visualization – byexamining a set of recent IEEE VIS Conference papers (a selec-tion from 2017-2019) to assess the inclusion of geospatial data andgeospatial analyses within these papers. After removing the paperswith no geospatial data, we organize the remaining literature intogeospatial data domain categories and provide insight into how thesecategories relate to VIS Conference paper types. We also contextu-alize our results by investigating the use of geospatial terms in IEEEVisualization publications over the last 30 years. Our work providesan understanding of the quantity and role of geospatial subject matterin recent IEEE VIS publications and supplies a foundation for futuremeta-analytical work around geospatial analytics and geovisualiza-tion that may shed light on opportunities for innovation.
Index Terms:
Human-centered computing—Visualization—Visualization application domains—Geographic visualization
NTRODUCTION
Geospatial factors play a key role in many of the world’s most press-ing challenges – pandemics, plant pest and pathogen spread, naturaldisasters, urban sprawl, pollution, human trafficking, food scarcity,and transportation (to name just a few). Additionally, as technolo- * e-mail: [email protected] —— [email protected] † These authors share second authorship and are listed alphabetically. gies such as GPS-equipped mobile devices, remote sensing satellites,and drones have proliferated, the centrality of georeferenced datahas only continued to grow. The complexity and volume of the dataand the importance of the issues at stake drive a need for innovativevisualization tools to support exploration and communication ofgeospatial information.As a focal event for the visualization community, the IEEE Vi-sualization (VIS) Conference profoundly influences the agenda forresearch in the visualization space. This influence includes iden-tifying new research directions, investigating novel analyses, andpresenting results that support a wide array of disciplines, geospatialanalytics included. Given this, understanding how geospatial subjectmatter is covered within the context of VIS conferences is important,as it can shed light on the intersection of geospatial analytics and vi-sualization, highlighting areas of interest and revealing opportunitiesfor innovation.In this work, we present our efforts to begin to unravel these con-nections. With a combination of word searches and close readings,we examine the role that geospatial subject matter plays in the IEEEvisualization research space, and we provide some fundamental con-text for how geospatial subject matter has been used within IEEEVIS Conference publications. Specifically, we offer the followingcontributions:• Descriptive analysis of the use of geospatial subject matter inIEEE VIS Conference papers from 2017 to 2019.• Temporal contextualization regarding how the role of geospa-tial subject matter varied across years.• Categorization of the 2017 to 2019 papers that leveragedgeospatial data by data domain.We anticipate this study will benefit both the visualization andgeospatial analytics communities by highlighting and clarifyingthe role of geospatial subject matter in recent VIS Conference publi-cations. a r X i v : . [ c s . D L ] S e p R ELATED W ORK
Inspired by claims that 80% of all human-generated data is geospa-tial [30], prior work has sought to identify metrics to quantify theamount of geospatial data in representative data collections. Hah-mann et al. proposed analyzing the network of links between data onthe Semantic Web to determine the degree of geospatial referencefor each node [14]. Later work from Hahmann et al. used networkanalysis on Wikipedia article links and cognitive analysis to iden-tify the articles as geospatial or non-geospatial [13]. Kienreich etal. proposed a geographic browsing system to anchor encyclope-dia articles based on geospatial references made within the articlecontent [24]. For this paper, we implemented our own cognitive anal-ysis to categorize papers containing geospatial versus non-geospatialcontent.As we considered approaches to assess the geospatial contentof IEEE VIS, we encountered a number of meta-studies focusedon popular visualization techniques, such as multiple-view layouts,trees, and glyphs [1] [20] [36] [12] [11]. Other meta-studies weremore general. van Ham used hierarchical clusters labeled with paperkeywords in order to examine discipline variability of visualizationstudies [37]. Isenberg et al. used author defined and expert chosenkeywords to create hierarchical clusters of IEEE publications to iden-tify trends and common themes in the visualization community [19].Isenberg et al. later expanded upon this by creating a dataset to betterunderstand trends of research in the visualization community [18].This work produced a VisPubData meta-collection which has contin-ued to be updated since its publication in 2017 and includes papertitles, abstracts, authors, DOI, and other metadata for each papersince 1990. In our work, we used elements of this dataset to situateour findings about recent papers within a longer-term context.
DENTIFYING G EOSPATIAL A NALYSIS
Given our core mission to identify geospatial papers within IEEEVIS Conference papers, defining the terms “geospatial” and “geospa-tial analysis” was a critical first step. The latter term, “geospatialanalysis,” was designated to mark the intrinsically geospatial qualityof a paper. We realized that evaluating papers effectively wouldrequire establishing working definitions for both terms. Significantthought was given to how we might best define these terms, both be-cause of their centrality to this paper but also for reproducibility andto ensure that all contributors used the same working definitions forthe analysis. Ultimately, the research team settled on the followingworking definitions:•
Geospatial : Involves georeferenced, GPS, or satellite data thatcaptures surface or atmospheric attributes of a planetary body.AND/OR, involves the use of a GIS or analytical techniquesspecifically associated with processing geographic data.•
Geospatial Analysis : A tool or analysis that was specificallydesigned for geospatial data or applications. AND/OR, ananalysis in which a geospatial component was fundamentalto understanding the results (i.e., if you removed it, the con-clusion(s) of the paper could be different). For our purposes,geospatial analysis was used as a short-hand to describe thefundamental geospatial quality of a paper.Using these working definitions, our intent was to take a broadapproach to what could be considered geospatial while still limitingthe assignment of the term geospatial to topics concerned withgeographic relationships.
As an initial foray into this meta-analytical work, we decided tofocus our attention on recent developments in the geovisualizationspace. As such, we cataloged IEEE VIS Conference publicationsfrom conference years 2017 to 2019 and collected metadata for eachpublication. Workshop papers were not included. For each paper, we collected the conference year, authors, title, VIS track, and sessiontitle from Open Access VIS [15], a collection of open access papersfrom VIS conferences (2017-present). Papers that were listed butnot linked within Open Access VIS were added manually to ourmetadata catalog. In total, we identified 585 papers for review.A detailed systematic review of each of the 585 papers was in-feasible for the scope of this work. As such, we needed to select agermane subset of the papers. To maximize the number of geospa-tial papers reviewed, we performed a word frequency analysis foreach paper, identified terms of interest, and then used those terms toassess likelihood that a given paper might be geospatial.
To generate word frequencies for each paper, we performed somepreprocessing on the 585 documents in our corpus (PDF to textconversion, lowercasing and removing punctuation, numbers, URLs,stop words, short words, and excessively long words). We thencreated an n x m word frequency matrix, where n is the number ofunique words from the entire corpus and m = To assess the geospatial qualities of the 220 papers selected, wedeveloped questions to evaluate elements that we might expect froma geospatial paper. The questions we used and their associatedcolumn headings in our dataset (in brackets) are listed below:1. Is the title geospatial? [Geospatial Title]2. How many geospatial figures are included? [Geospatial FigureCount]3. How many total figures are included? [Total Figure Count]4. Does the paper use geospatial data? [Geospatial Data]5. Was the tool and/or analysis a geospatial analysis? [GeospatialAnalysis]While the last two questions are self-evidently central to ourresearch, the first three questions served as a means of exploringif the title and percent of geospatial figures could be used as aneffective filter for geospatial content.To answer these five questions, we performed a blind, double-entry review. In a first round, each team member answered questions1-5 for a distinct set of 20 papers. The second round repeated thisprocedure with new paper-reviewer pairings, and reviewers were notprovided access to round one decisions. A final round of reviews wasperformed to resolve disagreements across rounds one and two. Thisround only re-examined conflicting responses. Responses that didnot exhibit disagreement were assumed correct and excluded fromhis final review. Previous decisions and comments were provided toinform final review decisions. Final reviews were conducted in pairsso that the final answer would represent a joint conclusion betweentwo reviewers.
After completion of this initial meta-analysis, papers identified ascontaining either geospatial data (Q4) or a geospatial analysis (Q5)were categorized by geospatial data usage to examine which geospa-tial domains had received attention in recent VIS Conference publi-cations. To this end, two reviewers examined each paper to identifywhere geospatial data was used. Specifically, they generated key-words and short descriptions of data used within the paper and thensynthesized these observations to identify data domains and developan initial categorization schema to describe data use in each paper.A separate review team then coalesced this information and madefinal decisions on category selection.
To provide context for our core work, we also wanted to look forartifacts of geospatial data and geovisualization in IEEE VIS overtime since its inception in 1990. The VisPubData meta-collectionby Isenberg et al. [18] currently includes metadata for 1990-2018,plus tentative metadata for 2019. Assuming the contents of the titles,abstracts, and author keyword lists are a reasonable indicator ofgeospatial content, we decided to leverage this meta-collection forfurther contextualization, adding any missing titles and abstracts,and using the first paragraph of the introduction as a proxy for theabstract when papers themselves did not provide an abstract. Then,we devised a short list of 66 geospatial key terms based on the ex-pertise of the collaborators. This list was more conservative thanthe one used to filter papers for inspection in Section 3.2 becausethis was to be a one-pass, automated process not followed by a closereading. Overloaded terms – for example: “map” – were excludedfrom this list, and terms unlikely to be used in non-geospatial con-texts – such as “choropleth” – were included to reduce inclusion offalse positives. After prepossessing (lower-casing, tokenizing, andstemming), the title, abstract, and author keyword list of each paper(1990-2019) were searched for occurrences of each term.
ESULTS
Within the top 220 papers that we investigated, 94 contained geospa-tial data, and, of those 94 papers, 64 constituted a geospatial analysis(full meta-data results are available online). Across all 220 papers,the average percentage of geospatial figures per paper was 17.65%.Of those papers that constituted a geospatial analysis, the averagepercentage of geospatial figures per paper was 47.40%. Only twopapers had geospatial content in every figure [5] [39]. For papers notclassified as geospatial analysis, the average percentage of geospatialfigures per paper was only 5.44%.By year, the 2017 VIS Conference contained 15 papers that qual-ified as geospatial analyses, the 2018 VIS Conference contained26 papers that qualified as geospatial analyses, and the 2019 VISConference contained 23 papers that qualified as geospatial analyses(Fig. 2). Because the amount of papers captured in the top 220varied across years (2017 comprised 22.73% of reviewed papers,2018 comprised 39.55%, and 2019 comprised 37.73%), Fig. 2 wasorganized to display percentages as a proportion of all papers in agiven year (including those not reviewed). This decision was alsoinformed by our confidence that we likely captured a large majorityof the geospatial-related papers across all years.In reviewing our search term frequency filtering, we observedthat 33 of the 64 papers that were determined to include geospatialanalysis contained 65 or fewer search term hits (Fig. 3). The other 31 https://go.ncsu.edu/ieee geovisualization review Figure 2: Percentages of all VIS papers each year that used geospatialdata (purple) or qualified as geospatial analysis (blue). papers contained between 66 and 523 search terms hits. The paperwith the largest amount of search term hits – 523 – contained 271occurrences of the word “cartograms” across its 14-page text [32].
Figure 3: Comparing search term counts to our results. The greenline shows paper counts for the full 585 papers, binned by searchterm counts. The bars show our results for the top 220 papers (or-ange, purple, and blue for non-geospatial, geospatial data only, andgeospatial analysis). The main chart has breaks near each end of thehorizontal axis. The inset shows the full graph.
As indicated by the supporting analysis, the occurrence of paperscontaining at least one geospatial keyword appeared to increasethrough time (Fig. 4). When aggregated by decade, this pattern moreclearly constituted an upward trend (geospatial metadata: 6.5% inthe 1990-1999 papers, 9.5% in the 2000-2009 papers, and 11.9% in2010-2019 papers). VisPubData contained metadata only for SciVis,InfoVis, and VAST papers; however, the OAVis collection alsoincluded journal and short papers, complicating direct comparisonbetween these two for years 2017-2019. VisPubData contained 215fewer papers for these years. The supporting analysis found keyterms in the metadata of 11% of the 2017-2019 papers. Consideringour close-reading decisions to hold geospatial analysis as groundtruth, these 2017-2019 supporting analysis results contained 13 false-negatives, 15 false-positives, 20 true-positives, and 86 true-negatives,plus 236 negative results for papers not chosen for close reading.The classification of geospatial-data-using papers by domainsyielded 14 categories (Fig. 1): (1) Multi-Domain, (2) AtmosphericScience, (3) Movement, (4) Cartography, (5) Social Media, (6)Urban Planning, (7) Marine Science, (8) Demography, (9) PlanetaryScience, (10) Geoscience, (11) Economics, (12) Education, (13)Text, and (14) Art.Multi-Domain papers (23%) applied their work to geospatial datafrom two or more domains (e.g., Li et al. [27] visualized both air pol-lutants and socioeconomic data). Atmospheric Science (22%) paperswere related to climate [21], air pollution [9], and meteorology [23].Movement papers (18%) included applications using trajectory [2]and origin-destination data [3]. The Cartography category (10%)apers applied tools/analysis to map features [33] [34]. That is,the map features were themselves considered to be the data. Therewere also several Social Media (9%) [26] [28] and Urban Planningapplications (5%) [31] [22]. Marine Science (4%) chiefly capturedflow visualization papers [10]. Smaller categories were folded intothe “Other” class in Fig. 1: two papers using demographic data[25] [38] and one each of the Planetary Science [6], Geoscience [17],Economics [4], Education [16], Text [7], and Art [8] domains. HE R OLE OF G EOSPATIAL S UBJECT M ATTER
Assuming our filtering captured the majority of geospatial-relatedpapers in the dataset, it appeared that the presence of geospatialanalysis and the use of geospatial data in IEEE VIS Conferencepublications remained relatively stable across 2017, 2018, and 2019(Fig. 2). As we might expect, fundamentally geospatial papers werefound to contain a notably higher percentage of geospatial figuresthan non-geospatial papers on average.Unpacking the role of geospatial subject matter in our studyproved fairly difficult. Even with concise working definitions, manydecisions still required nuance, and we found that reviewers couldoften reasonably arrive at divergent conclusions. Additionally, wefound that adhering strongly to our working definitions sometimesmeant that a paper could use only geospatial data examples but stillnot be marked as a geospatial analysis. For example, in a paperfrom Lui et al. [29], the authors demonstrated their uncertaintyvisualization technique strictly on cyclone data, but, because theirtechniques were not explicitly for geospatial data, the paper wasultimately not counted as a geospatial analysis. Disagreementscaused by these types of issues contributed to the need for the thirdround of reviews.
The supporting analysis – spanning years 1990 to 2019 – indicatedan increasing trend across decades with a fair amount of variabilitybetween years within each decade. That said, our ability to drawdefinitive conclusions from our supporting work was limited as thepattern in the data could represent (1) a change in the amount ofgeospatial papers, (2) a change in the popularity and usage of searchterms we selected, or (3) both. The core analysis was potentiallymore robust to this because search terms were selected through aposteriori selection.
Papers that used geospatial data in our results spanned a wide gamutof data domains and were present in all of the VIS tracks (Fig. 1TRK band). The SciVis, InfoVis, and VAST tracks each made upapproximately one quarter of the 220 papers isolated for review inthis paper. They were represented at roughly the same levels withinthe 94 papers that were found to use geospatial data (16 SciVis, 28
Figure 4: Percentage of papers per given year from 1990 to 2019 withone or more geospatial key terms in the title, abstract, and/or authorkeyword list, arranged by conference track.
InfoVis, and 25 VAST papers). For years, community members haveraised questions around the SciVis and InfoVis separation [35]. Thisrepresentation of geospatial applications across SciVis, InfoVis, andVAST tracks seemed to support the blurring of these divisions.The TRK band in Fig. 1 indicates that some tracks were morefrequently associated with certain topics. These associations par-tially, though not entirely, aligned with our expectations. SciVishas traditionally been thought to focus primarily on physical data(volume, flow, geospatial forms, etc.). SciVis is associated withAtmospheric Science (longest red band), but we might expect evenmore of the Atmospheric Science applications (typically physical)to be presented in SciVis. Though InfoVis has been thought to focuson abstract, nonphysical data, InfoVis is associated with the Multi-Domain category (long green band). This may be due, in part, tothe abundance of geospatial data, such as demographics, that canprovide convenient applications to demonstrate or test the versatilityof a new technique on familiar data. In fact, close to half of Info-Vis papers in the Multi-Domain category used geospatial data butwere not presenting intrinsically geospatial analyses (Fig. 1 GEOband). InfoVis is also associated with Cartography (another longgreen band), perhaps because maps are an abstract representation ofphysical data. VAST, which focuses on interactive analytics tools,is associated with Movement (long orange band) perhaps becauseinteractive displays are often a good solution when both geographicspace and time need to be expressed.
ONCLUSIONS
In our investigation of the role of geospatial subject matter in recentIEEE VIS publications, we found that 94 of the 220 papers wereviewed made use of geospatial data, and, of those papers, 64constituted fundamentally geospatial analyses. We organized the 94papers containing geospatial content into categories based on thedomains of geospatial data used within the papers and contextualizedhow those groupings related to VIS Conference paper types andtracks. An inventory of geospatial term usage in 30 years of IEEEVisualization metadata indicated an increasing trend across decades.Some questions that we would have liked to address remain unan-swered. For example, though we applied close-reading to morethan a third of papers from 2017-2019, a more comprehensive re-view could provide more information about the context of the useof geospatial subject matter and possible research gaps that exist(potential opportunities). To that end, we believe future analyseswould benefit from more sophisticated text analysis methods, suchas text classification, for filtering papers of interest and identifyingtopic groupings.Reflecting on the research we did review, we here consider the at-tractiveness of geospatial data and tools for communicating research.We note that while geospatial analysis papers often presented vi-sualization tools designed specifically for geospatial applications,a number of authors developed tools for non-geospatial purposesand included applications of those tools to geospatial tasks. Theseauthors may be motivated not only by the abundance of multivariategeospatial data, but also by the accessibility of geospatial data for abroad audience, as geospatial data and tools are widely used acrossscientific domains and are also commonly consumed by the generalpublic.Understanding of the role of geospatial subject matter in recentIEEE VIS provided by our work can be used for several purposes.Though this paper focuses on observations of interest to the visual-ization community, the results also provide a roadmap for geospatialaudiences interested in recent developments in IEEE VIS. The re-sults can also be useful for teaching the topic of geovisualization.Furthermore, our work may serve as a foundation for future meta-analytical investigations across other bodies of literature for a viewof how geospatial content is leveraged and thought about in thebroader scientific community.
EFERENCES [1] H. M. Al-maneea and J. C. Roberts. Towards quantifying multipleview layouts in visualisation as seen from research publications. In , pp. 121–121. IEEE, 2019.[2] G. Andrienko, N. Andrienko, G. Fuchs, and J. M. C. Garcia. Clusteringtrajectories by relevant parts for air traffic analysis.
IEEE Transactionson Visualization and Computer Graphics , 24(1):802–811, January2018. doi: 10.1109/TVCG.2017.2744322[3] G. Andrienko, N. Andrienko, G. Fuchs, and J. Wood. Revealingpatterns and trends of mass mobility through spatial and temporalabstraction of origin-destination movement data.
IEEE Transactionson Visualization and Computer Graphics , 23(9):2120–2136, September2017. doi: 10.1109/TVCG.2017.2666146[4] A. Arleo, C. Tsigkanos, C. Jia, R. A. Leite, I. Murturi, M. Klaffenbck,S. Dustdar, M. Wimmer, S. Miksch, and J. Sorger. Sabrina: Modelingand visualization of financial data over time with incremental domainknowledge. In , pp. 51–55,2019.[5] A. Arleo, C. Tsigkanos, C. Jia, R. A. Leite, I. Murturi, M. Klaffenb¨ock,S. Dustdar, M. Wimmer, S. Miksch, and J. Sorger. Sabrina: Modelingand visualization of financial data over time with incremental domainknowledge. In , pp. 51–55.IEEE, 2019.[6] K. Bladin, E. Axelsson, E. Broberg, C. Emmart, P. Ljung, A. Bock,and A. Ynnerman. Globe browsing: Contextualized spatio-temporalplanetary surface visualization.
IEEE Transactions on Visualizationand Computer Graphics , 24(1):802–811, January 2018. doi: 10.1109/TVCG.2017.2743958[7] H. Chung, S. P. Dasari, S. Nandhakumar, and C. Andrews. Cricto:Supporting sensemaking through crowdsourced information schema-tization. In , pp. 139–150, 2017.[8] L. Crissaff, L. Wood Ruby, S. Deutch, R. L. DuBois, J. Fekete, J. Freire,and C. Silva. Aries: Enabling visual exploration and organization ofart image collections.
IEEE Computer Graphics and Applications ,38(1):91–108, 2018.[9] Z. Deng, D. Weng, J. Chen, R. Liu, Z. Wang, J. Bao, Y. Zheng, andY. Wu. Airvis: Visual analytics of air pollution propagation.
IEEETransactions on Visualization and Computer Graphics , 26(1):800–810,2020.[10] A. Friederici, H. T. M. Kele, I. Hoteit, T. Weinkauf, H. Theisel, andM. Hadwiger. A lagrangian method for extracting eddy boundaries inthe red sea and the gulf of aden. In . IEEE, October 2018.[11] J. Fuchs, P. Isenberg, A. Bezerianos, and D. Keim. A systematicreview of experimental studies on data glyphs.
IEEE transactions onvisualization and computer graphics , 23(7):1863–1879, 2016.[12] M. Graham and J. Kennedy. A survey of multiple tree visualisation.
Information Visualization , 9(4):235–252, 2010.[13] S. Hahmann and D. Burghardt. How much information is geospa-tially referenced? networks and cognition.
International Journal ofGeographical Information Science , 27(6):1171–1189, 2013.[14] S. Hahmann, D. Burghardt, and B. Weber. 80% of all informationis geospatially referenced??? towards a research framework: Usingthe semantic web for (in) validating this famous geo assertion. In
Proceedings of the 14th AGILE Conference on Geographic InformationScience , 2011.[15] S. Haroz. Open practices in visualization research: Opinion paper. In , pp. 46–52. IEEE, 2018.[16] H. He, O. Zheng, and B. Dong. Vusphere: Visual analysis of videoutilization in online distance education. In , pp. 25–35, 2018.[17] D. A. Hyde, T. R. Hall, and J. Caers. Vrge: An immersive visualizationapplication for the geosciences. In , pp. 1–5. IEEE, 2018.[18] P. Isenberg, F. Heimerl, S. Koch, T. Isenberg, P. Xu, C. Stolper, M. Sedl-mair, J. Chen, T. M¨oller, and J. Stasko. vispubdata.org: A metadatacollection about IEEE visualization (VIS) publications.
IEEE Transac- tions on Visualization and Computer Graphics , 23(9):2199–2206, Sept.2017. doi: 10.1109/TVCG.2016.2615308[19] P. Isenberg, T. Isenberg, M. Sedlmair, J. Chen, and T. M¨oller. Toward adeeper understanding of visualization through keyword analysis. arXivpreprint arXiv:1408.3297 , 2014.[20] S. J¨urgensmann and H.-J. Schulz. Poster: A visual survey of treevisualization. In
Proceedings of IEEE Information Visualization , vol. 5,p. 7. IEEE Press, 2010.[21] C. P. Kappe, M. Bttinger, and H. Leitte. Exploring variability withinensembles of decadal climate predictions.
IEEE Transactions on Visu-alization and Computer Graphics , 25(3):1499–1512, 2019.[22] T. Kauer, S. Joglekar, M. Redi, L. M. Aiello, and D. Quercia. Mappingand visualizing deep-learning urban beautification.
IEEE computergraphics and applications , 38(5):70–83, 2018.[23] M. Kern, T. Hewson, A. Schtler, R. Westermann, and M. Rautenhaus.Interactive 3d visual analysis of atmospheric fronts.
IEEE Transactionson Visualization and Computer Graphics , 25(1):1080–1090, 2019.[24] W. Kienreich, M. Granitzer, and M. Lux. Geospatial anchoring of en-cyclopedia articles. In
Tenth International Conference on InformationVisualisation (IV’06) , pp. 211–215. IEEE, 2006.[25] M. Le Goc, C. Perin, S. Follmer, J. Fekete, and P. Dragicevic. Dynamiccomposite data physicalization using wheeled micro-robots.
IEEETransactions on Visualization and Computer Graphics , 25(1):737–747,2019.[26] J. Li, S. Chen, W. Chen, G. Andrienko, and N. Andrienko. Semantics-space-time cube: A conceptual framework for systematic analysisof texts in space and time.
IEEE Transactions on Visualization andComputer Graphics , 26(4):1789–1806, 2020.[27] J. Li, S. Chen, K. Zhang, G. Andrienko, and N. Andrienko. Cope:Interactive exploration of co-occurrence patterns in spatial time se-ries.
IEEE Transactions on Visualization and Computer Graphics ,25(8):2554–2567, 2019.[28] C. Liu, C. Wu, H. Shao, and X. Yuan. Smartcube: An adaptive datamanagement architecture for the real-time visualization of spatiotem-poral datasets.
IEEE Transactions on Visualization and ComputerGraphics , 26(1):790–799, 2020.[29] L. Liu, L. Padilla, S. H. Creem-Regehr, and D. H. House. Visualizinguncertain tropical cyclone predictions using representative samplesfrom ensembles of forecast tracks.
IEEE Transactions on Visualizationand Computer Graphics , 25(1):882–891, 2019.[30] A. M. MacEachren and M.-J. Kraak. Research challenges in geovisual-ization.
Cartography and geographic information science , 28(1):3–12,2001.[31] F. Miranda, H. Doraiswamy, M. Lage, L. Wilson, M. Hsieh, and C. T.Silva. Shadow accrual maps: Efficient accumulation of city-scaleshadows over time.
IEEE transactions on visualization and computergraphics , 25(3):1559–1574, 2018.[32] S. Nusrat, M. J. Alam, and S. Kobourov. Evaluating cartogram effec-tiveness.
IEEE Transactions on Visualization and Computer Graphics ,24(2):1077–1090, 2018.[33] S. Nusrat, M. J. Alam, and S. Kobourov. Evaluating cartogram effec-tiveness.
IEEE Transactions on Visualization and Computer Graphics ,24(2):1077–1090, 2018.[34] M. Reckziegel and S. Jnicke. Time varying predominance tag maps.In , pp. 231–235, 2019.[35] T.-M. Rhyne. Does the difference between information and scientific vi-sualization really matter?
IEEE Computer Graphics and Applications ,23(3):6–8, 2003.[36] H.-J. Schulz. Treevis. net: A tree visualization reference.
IEEEComputer Graphics and Applications , 31(6):11–15, 2011.[37] F. van Ham. Case study: Visualizing visualization. In
IEEE Symposiumon Information Visualization , pp. r5–r5. IEEE, 2004.[38] L. Wilkinson. Visualizing big data outliers through distributed aggre-gation.
IEEE Transactions on Visualization and Computer Graphics ,24(1):256–266, 2018.[39] Z. Zhou, L. Meng, C. Tang, Y. Zhao, Z. Guo, M. Hu, and W. Chen.Visual abstraction of large scale geospatial origin-destination move-ment data.