1 Performance of versus Topographic Representations for Different s Debra MacIvor Savage, Eric N. Wiebe an Hugh A. Devine North Carolina State University Raleigh, North Carolina In this stuy, a performance comparison is mae between an topographic representations for solving ifferent tasks. The tasks involve answering questions that either i or i not require elevation information an were either focuse or integrative. Integrative questions require unerstaning the relationship of three separate locations. representations (i.e., contour maps) only showe a clear avantage for focuse, non-elevation questions with representations not showing a clear avantage for any other task conition. There were interactions between task type, imensionality an elevation. In aition, the integrative, elevation questions were clearly more ifficult regarless of imension. In aition, participants visualization ability, as measure by a paper-foling test, correlate with their task performance. Further work is propose to look at how experience interacts with these tasks an topographic representations. INTRODUCTION There are many types of problems that can be most effectively solve by visualizing geographic (spatial) relationships. Such applications as zoning an lan use planning, natural resource planning, geographic eucation, park an recreation planning an maintenance, utility infrastructure maintenance an repair, mining, transportation planning an elivery route planning all require visualizing the relationship between terrain an objects locate in geographic space. Typically, planar topographic contour maps are use in these applications, but woul the integrate isplay of all three imensions in a perspective view provie an avantage for spatial visualization over the traitional topographic map? Some evience for map superiority for both focuse attention an integrate tasks has been foun (Hollans, Pierce, & Magee, 1995). In fact, perspective isplays can introuce problems ue to foreshortening of the epth (Y) of the map when it is tilte at an oblique angle (Smallman, St. John, & Cowen, 2002). Other stuies have foun little or no avantage for over isplays when scales an tick marks were use to facilitate ata extraction (Bennett, Payne, Calcaterra, & Nittoli, 2000; Meserth & Hollans, 1999). (Haskell & Wickens, 1993) foun that for integrate spatial tasks, a perspective (oblique) isplay may be preferable to a isplay, an for tasks requiring focuse attention, separable or separate isplays may be better. Topographic Maps Topographic contour maps are (usually) static isplays or harcopy maps use as geographic visualization tools. They inclue rectangular gris of latitue an longitue, as well as contour lines. Contour lines are lines rawn through points of equal elevation at regular intervals such as 10 meters of elevation change. Contour maps may be shae into ifferent elevation zones (Muehrcke & Muehrcke, 1998). The topographic contour map can be consiere a separable isplay, which shoul facilitate focuse attention tasks (Garner, 1976). The elevation values are represente as elevation isolines, while the latitues an longitues are represente as a regular gri. Aitionally, the elevation changes are enhance visually by grauate shaing; low areas are ark while higher elevations are light. Finally some of the contour lines have numeric elevation labels. Although rich in etaile geographic information, contour maps have been foun ifficult to rea by inexperience users (Gilhooly, Woo, Kinnear, & Green, 1988). In aition, research has shown that can be superior to topographic maps representations for precise relative positioning tasks, but not for shape unerstaning. (St. John, Cowen, Smallman, & Oonk, 2001; St. John & Smallman, 2000). The purpose of this research was to investigate whether the aition of shape (but no other cues such as shaows) to the traitional topographic maps is enough to gain the avantages of for shape unerstaning, while retaining the relative positioning avantage of the map. For this purpose, all of the traitional characteristics of the topographic maps were retaine: equally space latitue an longitue lines with numeric labels, elevation measure in isolines with numeric labels, an grauate shaing from ark (low elevations) to light (high elevations). In aition, the representations consiste of the maps rape over the terrain. s were of two types: s, which are similar to the relative position tasks in St. John (2001), experiments 5 an 6; an Integrate s, similar to the shape unerstaning tasks use in St. John (2001), experiment 4 (See Table 1). Table 1. Mapping of an Integrate tasks to Relative Position (RP) an Shape Unerstaning (SU) s (St. John et al., 2001). Non- A-high-B (RP) A-to-B (RP) Integrate A-see-B (SU)
2 The four hypotheses investigate in this stuy are: 1. For focuse questions of a single value of Latitue or Longitue at a single location, we expecte the flat map an the map woul be equally effective, ue to the gri marks available in both maps to ai iniviual value extraction. These tasks were similar to the St John (2001) Experiment 6 tasks with gri an topo, but only inclue tasks which i not require altitue jugments. The gri marks support the separability of the spatial imensions of latitue an longitue, an shoul work equally well for an. labele, an then rape over a terrain image erive from elevation (Z) values at regular intervals. Each image was tilte to an angle which woul allow the correct answer to be perceive. Five tasks of each type were create: 1), elevation (5 questions) 2), no elevation (5 questions) 3) Integrate, elevation (5 questions) 4) Integrate, no elevation (5 questions) 2. For focuse questions of a single value of at a single location, we preicte that the map woul be more effective, ue to the aitional cognitive processing effort require to extract the thir imension of elevation from a topographic map. This task is similar to the St. John (2001) Experiment 6 Gri an Topo conition, elevation (altitue) only. Although St. John (2001) foun the topo map better for these tasks, our representation has more exaggerate terrain, an so shoul facilitate relative elevation ecisions. 3. For tasks requiring relative jugment of istance between several points in a plane (latitue/longitue), we preicte that the map woul be more effective than the map. These tasks were similar to the St. John (2001) Experiment 6 latitue an longitue questions, except that our tasks involve choosing the nearest or farthest of several points from a single target point. This mae these questions more integrate, requiring relative jugment of istances in latitue AND longitue among several points. Question: How much farther East is the green point than the blue point? A) 1 B) 2 C) 3 D) 5 E) 6 4. For integrate tasks to etermine terrain shape, we preicte that the map woul be most effective, ue to the aitional information processing require in orer to project a map into a mental image in orer to solve the problem. s were similar to St. John (2001) Experiment 4, but a variety of realistic problem scenarios were presente in aition to A-see-B type tasks. EXPERIMENT Metho Participants. This stuy inclue 68 college freshman an sophomores, 34 were ranomly assigne to maps, an 34 receive the maps. Design. The esign was 2x2x2, with 2 types of isplays ( an ) between subjects, an 2 major types of tasks (focuse an integrate) being within subjects (see Figure 1 for a focuse task an Figure 2 for an integrate task). Each task type was further ivie into those which require knowlege of elevation to perform the task, an those which i not. There were five questions in each cell, for a total of twenty questions per participant. All questions were isplaye at 800X600 on a 19 color isplay. Stimuli. Maps were create from USGS Digital Maps (DEMs) of Western North Carolina counties. Contour lines were create from elevation ata, shae, Figure 1. Example of a focuse, no-elevation task. The upper version is the bottom is the same map in. Proceure. Each participant was ranomly assigne to the or conition. All tasks of each type were presente to participants in ranomize orer. Response time was measure as was correctness of response. Participants were instructe to answer as quickly an accurately as possible. The answers were multiple choice. Participants were given a survey regaring their previous experience with GIS or other visualization training, or geography (or hyrology, rafting, scientific visualization). Aitionally, participants were teste for spatial ability using the Paper Foling Test (Ekstrom, French, Harman, & Dermen, 1976). Paper foling ability was
3 use as a covariate in this stuy in orer to control for spatial visualization ability. Table 2. Mean Accuracy Scores for / Pairs Integrate M=.83, SD=.24 M=.45, SD=.18 Non- M=.80, SD=.25 M=.71, SD=.20 Well water pollution was foun at the location labele Contamination. Question: Given that water an water pollutants usually flow ownhill, which point is the most likely source of the pollution (contamination)? A) B) C) D) E) Average scores Nonelevation Integrate Non- Integrate types Figure 3. Accuracy Scores type/ by Dimension. Figure 2. Example of an integrate, elevation-require task. The upper version is an the bottom is the same map in. RESULTS Accuracy Accuracy results were analyze using a three-way ANCOVA with repeate measures in two factors ( type an ). The main effect of the covariate (paper fol score) was significant, F(1,63) = 6.45, p =.014. The main effect of Dimension was also significant, F(1,63) = 4.47, p =.038 (Figure 3). Aitionally, the type x interaction was significant, F(1,63) = 5.08, p =.028 (Table 2). The type x x Dimension interaction was not significant. For the focuse, non-elevation task, post-hoc analysis showe a significant ifference between the an groups in their mean accuracy scores (Mean =.87, Mean =.73, p =.023.) Figure 3 also shows significantly lower scores for the integrate elevation task in both an conitions (p <.001) than for the other tasktype/elevation cells. Time Time results were analyze using a three-way ANCOVA with repeate measures in two factors (type an ). The main effect of the covariate was significant, F(1,63) = 4.69, p =.034. The main effect of Dimension was not significant in this analysis, F(1,63) =.22, p =.64 (Figure 2). The main effect for was significant, F(1,63) = 11.35, p = Aitionally, the type x Dimension interaction was significant, F(1,63) = 5.05, p =.028 (Table 3). The type x x Dimension interaction was not significant. For focuse, non-elevation tasks, post-hoc analysis showe a significant ifference between the an groups in their mean times (Mean =13.6, Mean =16.6), p =.012. Table 3. Mean Times for ype/ Pairs Integrate M=19.7 SD=4.6 M=27.2, SD=6.4 M=21.3, SD=6.0 M=25.2, SD=5.7
4 Time in Secons nonelevation s Integrate nonelevation Integrate elevation Figure 4. Reaction time type/ by Dimension. DISCUSSION This analysis reveale a significant avantage in both accuracy an time for simple tasks that o not require elevation ata. This implies that for, the map representation use in this stuy may interfere with task performance, which is consistent with the research of others (Haskell & Wickens, 1993; St. John et al., 2001). One might assume that for focuse tasks requiring elevation (similar to St. John et al A- high-b tasks), scores woul be better an times woul be reuce for those participants using the maps, because the thir imension of elevation is ae to the maps as a reunant cue. However, we foun that no significant ifference in accuracy an time between an for this task type, in contrast to St. John et al (who foun a ifference in accuracy but not time). Our result is puzzling, but coul be ue to the simplicity of the tasks: The elevation lines were clearly marke, there were reunant shaing cues inicating high an low elevations, an the tasks were to compare elevations at two points or extract the elevation value at a single point. We preicte that for integrate tasks not requiring elevation ata, scores woul be better an times woul be less for those using maps, ue to the misunerstaning of the effects of foreshortening in maps (Smallman et al., 2002). We foun that there was no significant ifference in accuracy an time between an for this task type. Finally, we preicte that for integrating tasks requiring elevation ata (shape unerstaning tasks), there woul be an avantage to using maps (Haskell & Wickens, 1993; St.John, Cowen, Smallman, & Oonk, 2001). However, we foun that there was no significant ifference between an in either accuracy or time for this task type. It is worth noting that spatial visualization ability, as measure by the Paper Foling Test seeme to be a strong inicator as to how iniviuals woul perform on these types of tasks. In aition, shape unerstaning tasks showe themselves to be the most ifficult whether they were paire with or maps. On the other han, focuse, nonelevation questions were the easiest, with an avantage going to the maps, as preicte by St. John et al., CONCLUSIONS s which were essentially an require simple focuse ata extraction were best supporte by maps, even when tic marks an mesh were inclue on the maps. There was no apparent avantage of maps for those tasks requiring elevation information, nor was there a isavantage for integrate tasks which i not require elevation information. Although more research nees to be one, there is little support in these results using for using this style of topographic maps in problem solving an ata extraction tasks. Future research is planne to inclue analysis of experience level with topographic maps as a separate factor, as well as inclusion of aitional questions of each type. References Bennett, K. B., Payne, M., Calcaterra, J., & Nittoli, B. (2000). An empirical comparison of alternative methoologies for the evaluation of configural isplays. Human Factors, 42(2), Garner, W. R. (1976). Interaction of Stimulus Dimensions in Concept an Choice Processes. Cognitive Psychology, 8(1), Gilhooly, K. J., Woo, M., Kinnear, P. R., & Green, C. (1988). Skill in Map Reaing an Memory for Maps. Quarterly Journal of Experimental Psychology Section a-human Experimental Psychology, 40(1), Haskell, I. D., & Wickens, C. D. (1993). Two- an threeimensional isplays for aviation: A theoretical an empirical comparison. International-Journal-of- Aviation-Psychology, 3(2), Hollans, J. G., Pierce, B. J., & Magee, L. E. (1995). Displaying quantitative information in two an three imensions. Paper presente at the Human Factors an Ergonomics Society 39th Annual Meeting, San Diego, CA. Meserth, T. A., & Hollans, J. G. (1999). Comparing an Displays for Tren Estimation: The Effects of Display Augmentation. In Proceeings of the Human Factors an Ergonomics Society 43r Annual Meeting, Houston, Texas, September 27-October 1, (Vol. 2, pp ). Santa Monica, California: The Human Factors an Ergonomics Society.
5 Muehrcke, P. C., & Muehrcke, J. O. (1998). Map Use: Reaing, Analysis an Interpretation (Fourth e.). Maison, WI: JP Publications. Smallman, H. S., St. John, M., & Cowen, M. B. (2002). Use an Misuse of Linear Perspective in the Perceptual Reconstruction of 3-D Perspective View Displays. In Proceeings of the 46th Annual Meeting of the Human Factors an Ergonomics Society (pp ). Santa Monica, California: Human Factors an Ergonomics Society. St. John, M., Cowen, M. B., Smallman, H. S., & Oonk, H. M. (2001). The Use of an Displays for Shape- Unerstaning versus Relative-Position. Human Factors, 43(1), St. John, M., & Smallman, H. S. (2000). Designing for the : Sometimes 2-D Is Just Plane Better. In Ergonomics for the New Millennium. Proceeings of the XIVth Triennial Congress of the International Ergonomics Association an 44th Annual Meeting of the Human Factors an Ergonomics Society, San Diego, California, USA, July 29-August 4, 2000 (Vol. 3, pp ). Santa Monica, California: Human Factors an Ergonomics Society. St.John, M., Cowen, M. B., Smallman, H. S., & Oonk, H. M. (2001). The Use of an Displays for Shape- Unerstaning versus Relative-Position. Human Factors, 43(1),