Dynamic visualization variables in animation to support monitoring of spatial phenomena
Connie A. Blok
Our world is dynamic. Changes occur constantly in almost all phenomena at or near the earth surface. Understanding these dynamics to discover spatio-temporal patterns, relationships and trends is an important step towards the solution of many global, regional and local problems (e.g. global warming, desertification, pollution, endangered food security). Data that may reveal the dynamics are abundantly available nowadays, even to such an extent that these data are not even fully exploited yet, particularly remotely sensed data. These data are often studied using computational methods, but the human ability to quickly see shapes, patterns, relationships, trends and movement is very powerful. If qualitative, visual methods and techniques to explore and analyze the data can be integrated with computationally based functions, preferably in one environment, this would yield a rich range of tools to support problem-solving decisions.
The research focused on animation to support monitoring of spatial phenomena. Experts involved in monitoring want to keep track of the dynamics exhibited by specific phenomena. Animated representations are dynamic by nature and they enable users to quickly observe real world changes, even small ones. But animations have limitations as well; they may overwhelm the observer with sequences of rapid, volatile changes. Is it possible to extract relevant information, or to acquire knowledge from an animation? Evidence in the geosciences is mixed, and very little is known about how an animated representation is actually used. Slocum et al. (2000) indicate that more research is required, particularly on animation design and use. My research focused on both aspects, but design was limited to the variables of the temporal dimension of animated representations (the dynamic visualization variables): moment of display, order, duration and frequency. One of my main research objectives was to develop methods to incorporate the variables in animation design in such a way that information relevant for monitoring could be discovered or acquired. My other main objective was to gain knowledge on the strategies and reasoning applied by experts in monitoring during the use of these variables in an animated exploration environment. Following on from this, I was hoping that the results would shed light on methods and tools to use animations effectively, and that this would lead to design recommendations and a theoretical framework for the application of the dynamic visualization variables.
The methodological approach consisted of four phases. The first three phases are described below. The last phase, formulation of the results, is incorporated in the section ‘Main results’.
User task analysis. Overall monitoring goals, objectives and questions of geoscientists involved in monitoring have been identified based on literature review and interviews with domain experts. As a case study, satellite data containing a commonly used vegetation index (NDVI) were used. Aspects of change that are important for monitoring were also identified in this phase.
Creation of an environment in which answers to monitoring questions can be sought by visually exploring animations. Potential problems related to a reliance on visual input for changes (obtained from dynamic representations) have been investigated. Dangers such as ‘change blindness’ and ‘inattentional blindness’ (caused by problems to focus attention on simultaneously occurring changes) are described. The conclusion is that although there are limitations in the human capability to see change, there are possibilities for partly overcoming the problems as well, particularly if users can interact with the animated representation. In order to enable users to observe the earlier identified aspects of change, important for monitoring, I investigated which characteristics of change can be visually perceived in animated representations. This led to a framework of concepts, describing the characteristics in general terms. Following on from this, it is supposed that seeing change in an animated representation is able to trigger domain knowledge that is relevant in the search for answers to monitoring questions. The main cognitive tasks that are involved in visual exploration of patterns on maps (identification and comparison) were distinguished. Next, the dynamic visualization variables were investigated in depth. Four (already mentioned) variables have been distinguished, as well as the relationships between them. Further investigation focused on ways to use these variables to depict aspects of spatial data and to control the animated representation by interactions. The assumption was made that the dynamic visualization variables generate certain effects. If you know which effects a user wants – and preferably also for which question or task – then it might be possible to provide tools or develop methods that generate those effects. A prototype animation environment, aNimVis, was produced in which ideas about applications of the dynamic visualization variables have been incorporated. The data set used in the prototype consisted of pre-processed ten-day synthesis images (SPOT 4 VEGETATION) of a part of Iran.
Empirical testing. The first version of the prototype was evaluated in a session with domain experts (a focus group). Results led to some adjustments. This was followed by a detailed evaluation by domain experts of the use of the adjusted prototype. Data for the evaluation were gathered during individual sessions in which the participants were thinking aloud while they familiarized with the prototype and while they were performing a task. Verbal and action protocols could be generated from the integrated video recordings made during the sessions. In addition, data were gathered with post-test interviews and a questionnaire. The main results are summarized below.
The ‘problem-solving behaviours’ of potential users, who attempted to answer some typical monitoring questions, have been analysed in various ways. An analysis of the problem-solving phases revealed that ‘selection of time’ is most frequently used. This is followed by ‘identification and comparison’, then ‘comparison’. Users were able to extract relevant information from the animation. Three main animation use strategies could be identified. Some users mostly want to reduce the amount of data, and then focus on subsets that are relevant for further exploration. Other users mostly want to play the animation almost continuously, taking time to observe patterns, without much further interaction, because that often distracts. Between these two groups, there are users who are mostly playing, or playing and stepping, but they also frequently interact with the data using various controls. Tool use paths or trajectories reveal similarities and deviations in tool use among participants. Basic tools – such as media player controls (like play, stop, forward, backward) and toggles with base map layers – were used by all participants. The same applies for temporal selection tools (see also the problem solving phases) and for the tuning mode. For tuning, this is unexpected, since users have to play or step through two simultaneously displayed animations: a perceptually and conceptually difficult task. Actual tool use by the participants was compared with a model that predicts tool use. Predictions were based on the effects that are generated by the dynamic visualization variables. Implicit effects (automatically occurring when an animation is played) and special effects (intentionally caused by interaction) were distinguished. Use of the implicit effects ‘dynamic behaviour’ and ‘rate of change’ is always high: the majority uses (at least once for each question in the task given) the play mode of the animation. Analysis of the use of special effects revealed that ‘visual isolation’ and ‘review’ are the most important effects that users want to generate. This can be explained by the overwhelming character of an animation. If comparisons in time have to be made, ‘synchronization’ of two animations (in the tuning mode) is also highly important. One of the earlier proposed theoretical effects, ‘swapping’, can be dropped from the list of special effects. Finally, the feedback gathered from the users during the evaluation sessions revealed problems with the application, good and bad aspects, overall (high) usability ratings and desired refinements and extensions. The most commonly expressed wish was to extend functionalities in the tuning mode. All participants would like to use aNimVis as a complementary environment to visually explore and analyze data if it was integrated into a GIS or image processing environment.
A start has been made with the establishment of a theoretical framework for application of the dynamic visualization variables. Taking the effects that users want to generate when they perform tasks with animated representations into account seems a more fruitful approach than looking at measurement levels of data. The latter was introduced by Bertin for the graphic variables, but cannot be automatically transferred to the dynamic visualization variables. In my research I attempted to link desired effects to (monitoring) tasks. Some links could be established, but more tasks and different questions need to be taken into account to establish and extend the relationships further. Given the different strategies of animation use applied in the context of this research, it seems that those users who want to visually isolate part of the complex content can best be served with additional tools, beyond the ones examined here. Tools that enable a reduction of the complex content are desirable anyway, to avoid problems such as change blindness and inattentional blindness. Such tools will enable effective use of animations. Tool design ‘guidelines’ can perhaps also be derived from the extensive user feedback. The directions for tuning, in particular, are clear.