2-D vs 2½-D Terrain Visualization Study


Do we make choose better routes through terrain when we have a overhead, 2-dimensional view with a symbolic representation of the elevations, or when we have a 2½-D view (often called “3D”) that we can rotate, pan and see the actual height of the terrain?

I conducted a study in the spring of 2009 to examine whether a user could plot a shorter route through mountainous terrain using an overhead, 2-D view or a “3D” 2½-D view. Although 2½-D tools, such as Google Earth, have become extremely popular over the past few years, there has been little research on how well users can perform with these types of 2½D tools. If we understand which visualization provides the most efficient way for humans to interpret and plan routes through difficult terrain, we can develop better mapping and path-planning tools.

2-D, topographic visualization using Google Maps

2½-D Visualization using Google Earth

Conducting a literature review, previous research has clearly shown that 2-D views are better than fixed 2½-D views. This is an intuitive finding – a fixed 2½-D view of terrain will obscure some areas and misrepresent others to the user. However, no research had been conducted on comparing a 2-D view to a 2½-D view where the user could pan, rotate and zoom the views as needed.

Our hypothesis was that with more advanced interface tools, novice users would perform better (shorter routes, fewer waypoints and constructed faster) with a 2½-D tool than a 2-D tool. This follows the intuition that novice users can intuitively understand a 2½-D map better than the symbolic, topographic, representation of terrain on a overhead 2-D map.

To test this hypothesis, I constructed a Ruby on Rails application which performed a partially-crossed, randomized 2×2 factor study. Participants were randomly distributed into an even set of testing “blocks.” There were four blocks, comprised of two different factors: 2 types of terrain and the 2 types of map visualizations This allowed us to easily run many participants through the experiment with a minimal time investment on our part.

Consistency between the visualizations was important, as to eliminate any variability in performance due to different interfaces, terrain data, etc. Therefore, I designed two interfaces, built on top of Google Maps and Google Earth, which presented the exact same interface for both types of visualizations.

With 42 participants recruited from the MIT community, we found that users performed better overall using the 2-D tool than the 2½-D tool. In many cases though, we observed interaction effects between the tools, indicating no solid statistical difference could be inferred. This result may be due to users not fully using the 2½-D tool’s panning and rotating features, in which case the users were essentially using a fixed 2½-D visualization, which would be consisten with previous research. Currently, a second study is in progress which will specifically examine participants use of panning and rotating with a 2½-D tool.