Project Goals

The National UFO Reporting Center (http://www.nuforc.org) provides a site where people may report Unidentified Flying Objects (UFOs). This site collects information from observers about the location, time, shape and description of UFO sighting. These records may be accessed on the website but are currently displayed only as data tables. While this site allows visitors to view information by event date, state, shape and date posted, patterns between these parameters are not easily detected in a table format.

The goal of UFOvis is to provide a way to see possible patterns or correlations in the data spatially and temporally while providing other report details including duration and summary. For example, UFOvis could help determine multiple reportings of the same shape around the same location and the same time. It could also be useful in visualizing a possible path of navigation across a location if one shape type appears to spread out at regular time intervals. In addition, UFOvis could assist in finding cyclical patterns if certain shapes consistently appear at a similar time of year.

Related Work

Indian Emigration to California - 1960-2000:    by Vijay Viswanathan
http://www.sims.berkeley.edu/~vijay/InfoViz/project/GraphApplet.html


This visualization, by SIMS graduate Vijay Viswanathan, shows the Indian Emigration patterns to California between 1960 and 2000. This project is of interest to UFOVIS because of the relationships between the data elements: geographical location, time and quantity. The color range of the map, darker for more and lighter for less, allows an easy way to see which counties have a higher Indian population. The color range also makes it easier to see how the Indian population for the state of California changed over time. However, it is difficult to determine the changes for each individual county over time. Perhaps a future iteration could include a way to retain the history of the previous time selection while viewing the current year.

Election 2000 and 2004 Red and Blue States maps:
http://www.usatoday.com/news/politicselections/vote2004/countymap2000.htm


This visualization shows the United States Presidential election results for 2000 and 2004. Each state is either blue, indicating more Democratic votes than Republican votes or red, indicating more Republican votes than Democratic. This map is of interest to UFOVIS because it shows the relationship between: geographical location (state, county), color and voting results. Furthermore, this map offers different views of the data (by county, race, state Senate, state house, etc…) and options to “drill-down” for more detail. These features may be useful for UFOVIS to provide focus and context to our visualization. One drawback of this map, however, is that it does not show the relative population sizes of these states, which is an important factor for determining the popular vote. This map does not adequately show the relative number of votes between states and counties (unless you drill down) and cannot show whether the states were won with a slim margin. To better display a relationship to the number of votes, color gradations could be used, as in the map shown here: http://www.princeton.edu/~rvdb/JAVA/election2004/

TimeMap Korea: The University of Sydney's Archaeological Computing Laboratory and The National Institute of Korean History, Seoul
http://www.timemap.net/epublications/2002_animations/2002_shilla_animation.swf


This visualization uses animation and interactive features to show how the ancient political history of Korea changed over time. This visualization is particularly interesting to UFOVIS for its use of animation and color to show how different events are related to moments in time and location. Different colors are used to represent the tribes, and the animation shows how the boundaries of these tribes changes over the years in relation to particular political events which are shown in narrative on the sides. Red dots are used in the timeline at the bottom to show where there are clusters of political events happening around the same time periods, and to help situate the event at a particular point in time. The interaction provided by clicking on the timeline or on the tape-deck like console, lets you quickly move to particular moments in time. One critique of this visualization is that it can be a bit difficult to read the moving blocks of narrative, or to remember what happened before, without rewinding or pausing the animation. Perhaps this could be improved by allowing the user to decide when to move on to the next “scene” by having him/her click on a “next” button when they’ve finished reading the information presented on the screen.

UC Berkeley USGS Earthquake maps:
http://quake.wr.usgs.gov/recenteqs/Maps/120-36.htm


This map correlates the size of the square around a point to the magnitude of the earthquake in that area. This is of interest to UFOvis because we could use something similar, e.g. size of the shape icon, to indicate a concentration of a number of similar shapes sighted in that area or even during a particular time period. One drawback of using size to indicate the density of sightings is that some of the relative spacing between sightings may be lost. This may also end up occluding some data points, particularly in places with high numbers of sightings. One way to deal with this problem would be to use bigger icons to show high concentrations only on the zoom-out level to indicate possible areas of interest. That way, the actual frequencies and spatial distances can be shown in detail at the zoomed-in level.



Fish Eye Menus:
http://hcil.cs.umd.edu/trs/2000-12/2000-12.html


Fish Eye menus could be a useful feature for the UFOvis timeline because it could facilitate the browsing of a large number of date points, since it requires less clicking or scrolling. However, as this paper indicates, fish-eye menus are not as helpful for more “goal-directed tasks,” such as when looking for a specific date. In such cases, it might be more useful to have some sort of search feature.

top



Description of Visualization

Description of Data

The data used for this visualization is from the National UFO Reporting Center's website located at: http://www.nuforc.org/webreports.html. This site allows for people to report UFO sightings online by the event date, event time, duration, city, state, shape of craft and a summary description. The data set includes the United States and some other countries with sighting dates ranging from 1860 to 2005. For the purposes of this visualization, we chose to limit the scope to only California from January 2000 to October 2005.

Normalization and Assumptions

The current web form used to collect sighting information has limited control over user input. This has caused inconsistencies in the data since users have different concepts of time, duration and place. In order to provide a more accurate and consistent visualization we made several assumptions about the data set.

Location - For city there were entries that would describe a relative location instead of the proper city name. For example, "near I-5" or "approximately 15 miles south of Mendocino". If the relative location included a city name we chose to eliminate the descriptive text and use only the city name. If we could not tell where the city was we chose to eliminate that record from our visualization. There were also some sightings where "CA" was intended to be Canada, not California. Fortunately, there were familiar Canadian cities listed with those entries so we felt confident in eliminating those. The city name was not enough to determine the exact location of a sighting. For increased accuracy, it would be useful to have an address or even more precise the latitude and longitude of the sighting. However, this would be difficult information for the website to gather from people since people may not be at an exact address or have a GPS device to tell them the latitude and longitude. To deal with this discrepancy we decided to that sightings should be posted around the center of the city and spread out around the center to avoid occlusion.

Duration - There were several notions of duration ranging from "One Minute" to "25-30sec" to "Rock Climbing" or "Sunset". For the purposes of our visualization the default duration for a sighting is 1 hour since that is the smallest unit of time our timeline is going to allow. However, we kept the reported duration to be part of the sighting detail information so that this information was not completely lost.

Shapes - The data set includes 28 craft shapes. We assumed that several of these shapes could be similar especially when viewed at a distance. So we grouped similar shapes together to be represented by one icon in our visualization. Those groupings were:

Card sorting techniques with users would be useful to revise the vocabulary of the craft shapes.

Initial Design

Our visualization consists of these three main components: a checklist of shape categories, a map of California and a timeline.


Note: Mockup does not reflect actual data.

In our initial design we chose 18 distinct colors to represent the Shape of Craft categories. We considered using icons but decided these would be difficult to see on the map. A user may select any number of craft shapes via the checkboxes. A checked checkbox reveals that shape's color on both the map of California and the timeline. On the map, the points represent the location of the sighting and points on the timeline represent the time of the sighting. Through linking, a selected a point on the map causes the corresponding point on the timeline to appear. Conversely, selecting a point on the timeline causes that point on the map to appear. Our initial design also allows for brushing so that several points on the map may be selected and the corresponding points on the timeline are selected.Similarly, selecting several points on the timeline selects several points on the map. A user may zoom in on a section of the map where the same brushing and linking techniques hold true.

Upon selecting a point a details window shows on the map to describe textually the location, duration and summary information of the sighting. We included this feature since the duration and summary information could be of added interest to users.

Design Issue: Timeline One of our design issues was how to represent time. We wanted to create a timeline that would allow for multiple levels of granularity from year, to month, to day, to the hour. Inititally, we came up with 4 variations of this timeline: the toggle timeline, stacked timeline, popup timeline and a fisheye timeline.

Toggle timeline:
This timeline would default to the year setting allowing for a broad view of the dataset. Radio buttons would allow for navigation to another level of time. Then a user could use arrows to the left and right of the timeline to go back and forth in time. At each level of time the parent time would be indicated as text so that the user know what year, month or day they were on.


Stacked timeline:
This is a variation on the "toggle timeline". In the stacked timeline all the timelines for year, month, day and hour are stacked in one view. A year would be selected, then a month, day or hour. The points corresponding to shape would appear at the lowest level selected.


Pop-up timeline:
This is yet another iteration from the "stacked timeline". This allows for several levels of granularity to be seen at one time. A time range is selected and then the next level is shown with data points.


Fisheye timeline:
The fisheye timeline would allow for the four levels of time to appear on one timeline. A user could drag the fisheye to the data point of interest and see the hour the sighting was reported. At the same time they could get a high level view of the data points.

Design Revisions

Based on peer review and usability testing we have made revisions to our initial design.

Shapes: Originally we had 18 distinct colors to represent the different craft shape categories. We carefully selected these colors so that they were the same intensity so that each shape was of equal importance. A peer review session suggested that we combine color and icon shapes at the zoomed level as a semantic zoom. Participants of our usability study thought it would be helpful to see the color and icon shape at the zoomed out level as well.

Another recommendation was to show all of the shapes upon opening the visualization. This would allow for a user to see patterns more easily and could serve as a serendipitous jumping off point to explore further.

Timeline: After peer review we narrowed our timeline options down to the pop-up and the fisheye. During the usability study we asked participants to comment on both timelines. Each timeline offered pros and cons for selecting a date range or a particular date (see usability results). Our last participant led us to the current, simplified timeline design.

This timeline shows the entire time frame, from 2000 to 2005. The arrows have been removed because our dataset does not use the data outside of this range. All the data points are then plotted on the timeline. When the user rolls over a data point more detailed time information, "January 19, 2002 at 1:00pm" appears above the timeline. By clicking on a point the date information stays above the timeline while the user explores the rest of the timeline. The user can then select multiple data points with detailed time information by holding ctrl and clicking. A user could also select multiple points by clicking and dragging over several points. These selection processes would still link with points on the map. Finding a specific date then becomes an exploration of data points relative to the date you are looking for. This should suffice since finding a date is not the goal of this visualization. The goal is finding a date with a craft shape, times without data points become uninteresting unless to see patterns of time where crafts are not reported.

This image shows a user rolling over a circle shape and getting the date information:



After the user selects a data point, the time stays visable while he may continue to browse other points on the timeline:



click to view html mockup of design revisions      Note: Mockup does not reflect actual data.

How This Visualization Provides Insight on the Problem

These features aim to assist a user of the visualization in seeing patterns across location and time. Points that are the same shape (color) occurring close in location and time may be the same craft reported by multiple people. For example, there may be two triangle shapes near Sacramento that correspond to two point on the timeline that occur at the same time. Additionally, cyclical patterns may be determined if a shape appears in the same location at a given time interval (year, month, day or hour). For instance, a flare shape may be reported outside of Bakersfield consistently every June.

top

Usability Study

General Demographic Information:
To test the usability of UFOvis, we tested a paper prototype of UFOvis with three participants. All three were college-educated: one holds a Ph.D., another is in graduate school getting her Master’s. The participants range in age from 25-37 and are in the following occupations: scientist, graduate student and assistant director of student services.

We screened our testing participants to make sure that they believed in the possibility of UFO's to some extent. This was to ensure that they would be more inclined to look for patterns within the data using UFOvis, rather than trying to prove the opposite. All participants acknowledged the possible existence of UFO’s, but also qualified their responses by saying that UFO’s were more likely of human origin rather than extra-terrestrial.

Methodology:
The participants were given three different information-gathering tasks to complete using the paper prototype interface. One team member acted as “the computer,” reacting to the participant’s actions by displaying different interfaces and features accordingly. The other team member assigned the tasks, interviewed the participant and took notes of the observations.

The first task asked the participant to find out which UFO-shapes were more prevalent in the San Francisco area, in order to find out how the participant would interact with the zoom-in and icon shape features.

The second and third tasks asked the participant to verify reports of sightings of flares, disk and circular shaped UFO’s in and around Bakersfield. The participant was encouraged to find out more about these sighting details and timing to see if they could find any relevant correlations in the data.

After completing these tasks, the participant was shown the fish-eye and pop-up timelines to compare the functionality of both designs. We asked the participant about their opinions on how easy it was to do the following: discern the different shape icons on the map and in the key, select a date range or a specific date on the two different timelines, and find patterns and correlations within the map and timeline data.

Participant Responses:
To accomplish the tasks, all three participants clicked on specified city on the zoomed-out map (see Figure 1.) to get a zoomed-in map with a more detailed local view. Next, they selected “check all” to see all of the data points on the map. None of the participants seemed to have problems identifying the shape that was most commonly sighted in that area.



After zooming in to get the Bakersfield map, one participant was surprised to see a blank map, since he was expecting to see all the points displayed by default. He also wanted to see more of a direct mapping of the shapes on the timeline to the ones found elsewhere, asking, “Why can’t these dots look like these (referring to the shapes next to the checkboxes)?” All of the participants seemed to be drawn to the “flares” initially, possibly because they were the brightest icons, and maybe also because of their frequency. One participant suggested that larger icons be displayed on the zoomed out map to represent higher concentrations of that particular shape in an area, saying, “…If I wanted the number, I would click on the city to get the zoom in view.” This participant also noted that several shapes seemed very similar and that it would be better if these particular icons used very different colors to make it easier to distinguish between them.

The participants had some difficulties navigating through the data set to find any spatial/temporal correlations. For instance, there was some confusion about the arrows on both sides of the timeline, since two of the participants were expecting to see more data after clicking on those arrows. Also, all of the participants attempted to highlight date ranges by dragging and clicking; one of the participants also attempted to do so by clicking from one point to another. This participant also noted that because there was no information to indicate how date ranges could be highlighted, it may be difficult to discover how to do this. He did not initially expect this interface to act like a desktop interface [was assuming a web UI], where you would click and drag to highlight areas, or hold down a shift or control key while selecting.

Two out of the three participants wanted a way to compare the summary details of more than one data point at a time, but were concerned that the screen might get too cluttered. One user commented that she wanted to have a way of “saving” some of the data that she had looked at so she wouldn’t have to keep track of all of this information: “…It seems like there’s nothing on this tool to keep track of all the points and information. I would get a piece of paper, draw a map and number the points. I would want a corresponding number on a map point and details window to help compare and keep and track of details for each point.”

All of the participants remarked that the fish-eye timeline (see Figure 2. below) was “cool” and that they liked playing around with it and liked the fact that less clicking was required in the interaction. However, only one participant clearly preferred it to the pop-up timeline, noting that the pop-up timeline required too many clicks, was confusing and displayed redundant information. One participant preferred the pop-up timeline because it seemed less cluttered, noting, “It seems more precise to me, whereas with the fish-eye, it’s easy to lose my perspective.” One participant preferred the fish-eye for selecting a particular date, but all participants seemed to prefer the pop-up timeline for selecting date ranges.



Because we used a paper prototype, we were not able to observe the participants actual computer and mouse interaction with the fish-eye and pop-up timelines. This would be important in determining which timeline is actually easier to use and discover. It may also be useful to explore this visualization using an actual data set to see if more conclusive observations could be made using UFOvis.

Materials and Tools:
These usability tests were carried out using low-fidelity paper prototypes consisting of: several sheets of paper with printouts and hand-drawn facsimiles of the shape icons, maps, and timelines. We used transparency sheets and markers to show changes within the interface in response to the participants’ interactions with the interface. Notebooks and laptops were used to record the observation and analysis notes.

top

Future Work

Further implementation of this design is certainly needed for this project. We are planning to finish implementing the UFOvis' map views using the Yahoo Maps API since it seems to best meet the needs of UFOvis' functionality and interactive features. This mapping implementation will utilize an XML file as the data source and client-side JavaScripting to access the data and the Yahoo Maps API. We would also like to include features from the design revision such as the timeline and shape icons. It would be useful to do more user testing with the Yahoo! implementation including the new timeline.
Current Yahoo! Maps Prototype (in progress)

It would also be interesting to explore relating the size of an icon to density of sightings at the zoomed out level. Certainly, more user testing would be necessary to determine if this makes it easier for individuals to find areas of interest and whether there is any confusion between the different zoom and detail levels.

top

Tools

The initial design prototype was created using Adobe Illustrator, Photoshop and Microsoft Powerpoint. The usability testing was implemented using low-fidelity paper prototypes created from paper printouts of the maps with hand-drawn icons, transparencies, and markers. Notebooks and laptop applications were used to record observations and analyses of the tests.

We are currently working on implementing the revised design using the Yahoo! maps Flash API (link: work-in-progress prototype). Although TimeMaps was initially considered for implementation of this project, we decided to use Yahoo! Maps instead because it offered more flexibility and better documentation. Furthermore, TimeMaps seems to be less interactive and more animation-based, which does not suit the data exploration needs of UFOvis as well. Yahoo Maps! seems to meet our needs best: an application that can easily display zoomed-in and zoomed-out views, automatically map the given locations, and allow for more customizable interfaces.

top

Appendix

topic proposal - adobe PDF (.pdf)
powerpoint presentation 11/14/05 - powerpoint (.ppt)
usability test protocol - word doc
usability test notes - participant 1     participant 2     participant 3
initial design mockup - powerpoint (.ppt)
revised design mockup - html
Yahoo! maps design (in progress) html

top