A Study on 3D Surface Graph Representations

Abstract: Surface graphs have been used in many application domains to represent three-dimensional (3D) data. Another approach to representing 3D data is making projections onto two-dimensional (2D) graphs. This approach will result in multiple displays, which is time-consuming in switching between different screens for a different perspective. In this work, we study the performance of 3D version of popular 2D visualization techniques for time series: horizon graph, small multiple, and simple line graph. We explore discrimination tasks with respect to each visualization technique that requires simultaneous representations. We demonstrate our study by visualizing saturated thickness of the Ogallala aquifer - the Southern High Plains Aquifer of Texas in multiple years. For the evaluation, we design comparison and discrimination tasks and automatically record result performed by a group of students at a university. Our results show that 3D small multiples perform well with stable accuracy over numbers of occurrences. On the other hand, shared-space visualization within a single 3D coordinate system is more efficient with small number of simultaneous graphs. 3D horizon graph loses its competence in the 3D coordinate system with the lowest accuracy comparing to other techniques. Our demonstration of 3D spatial-temporal is also presented on the Southern High Plains Aquifer of Texas from 2010 to 2016.