You are on page 1of 1

SST-295-03

Social Boundaries in the Physical World


Introduction to GIS
Spatio-Temporal Patterns at Grinnell College
Professor Carter Mark Root-Wiley
December 17, 2008
ABSTRACT
Until the mid-to-late-20th century, social analyses ignored the inherent qualities of time and space (Urry
RESULTS
In order to see how the acadmic buildings, social locations, and dormitories effect
CONCLUSIONS
As shown by the Results section, there are multiple traits
1985:20). The study of time-geography, however, points to the face that peoples’ movements are influenced by
students’ distribution on campus, there must be a baseline with which to compare the patterns that can be linked to spatio-temporal distributions of
the activities they do and the order in which they do them. In short, movement is not random and a distribution
found in the data. Figure 4 shows what the expected directional distribution and mean center movement on Grinnell College’s campus. Place of residence is
of activity is based on many factors of both time and space.
of distribution of campus life would be were students to spend their time distributed evenly especially correlative. However, other macro trends can be
Hence, it should be theoretically possible to observe “aggregate phenomena” over time and within space.
across all parts of campus with dormitories and academic buildings on them (see the Types of found and are likely attributable to the overarching reason
Using this lens of aggregate phenomena in the context of Grinnell College, one observes that student movement
Analysis section below for a description of these and other methods used for this study). This people come to Grinnell College: education.
patterns are affected by what dorm they live in, what classes they take (and by extension what their major is),
is compared to the actual directional distribution and mean center of all the data points on or As a final way of visualizing the data focusing on campus
what activities they participate in, and who they associate with. Additionally, one may expect to find exaggerated
very near the college campus. residence, Figure 10 shows the directional distributions and
patterns of movement at Grinnell compared to the “outside world” because everyone is primarily a student
One immediately observes a few important differences between the two ovals representing mean centers for all three campuses as well as the campus as a
(Huisman and Forer 1998).
distribution of theoretical and actual students on campus . First, the baseline case is more whole. In this figure, the direction of movement is even more
This project sought to find some initial patterns of movement through time and space on the campus of
compact and located directly above the Campus Center. However, the distribution based on apparent than previously observed in the density or point
Grinnell College. To quote Marx, this study searches for “the concrete concept [that] is a synthesis of many
the GPS data is more diffuse and is centered relatively closer to the Southern, academic side of visualizations of the data. The fairly narrow ovals for East and
determinations, thus representing the unity of diverse aspects” (qtd. in Sayer 1985: 65).
campus. As more students (and participants in the study) live in the Northern half of campus North campus greatly contrast to the nearly-circular

INTRODUCTION
than the Southern, this likely means that students spend a great deal of time going to classes distribution of South Campus. It also appears that East
(Grinnell is a college after all), studying, and doing extracurricular activities – behaviors residents’ patterns of movement may be denser than North or
conceived of as “activity programs” in the time-geography literature (Huisman and Forer 1998) South residents’.
– on the Southern half of campus. Just like the previous Figures, Figure 10 has both strengths
Grinnell College’s campus (Figure 1) is notable both for its Because the GPS unit records locations at a relatively steady interval, there are more points and weaknesses for what it can show us about movement
Figure 3 Figure 10
loggias and its three distinct campuses on the Northwestern, where people spend large amounts of time. Hence, observed patterns are not purely spatial but through time and space. Clearly, there is no one effective way
Northeastern, and Southeastern sections of the campus. also temporal. In this case, the effects of time and space cannot easily be separated, but as to analyze the GPS data collected in this study (or any other for
Over the past decade, the campus has changed enormously predicted by time-geography, the reason people come to Grinnell – education – clearly affects that matter). A multifaceted approach is warranted for any
with the addition of East Campus, the new Athletic and their spatio-temporal distribution. study of this type of data.
Having established that activities and other characteristics of students affect geographic
Fitness Center, and the Joe Rosenfield Center. In addition to
new buildings, some buildings have changed their primary trends, I analyzed the data with a focus on campus residence. Figure 4 shows all of the on-campus data points taken during the study. In order LIMITATIONS
use. With the opening of the Joe Rosenfield Center, the to get a general idea of what trends I might expect to find, I color-coded points based on which Inherent in the technology and methods of this study are a few shortcomings that should be considered
Forum (formerly the center of student life on campus) was ones came from study participants on each campus. As seen in Figure 5, people from each alongside the conclusions.
converted to a computer lab, and Quad and Cowles changed campus, unsurprisingly, spend the most time on their own campus. However, while First, the GPS units used in the study have an accuracy of +/- 3m when outside and up to +/- 20m inside.
from dining halls for South and North Campus to student visualizing only points, it is hard to see what other patterns may exist. Additionally, one This can lead to stationary units “wandering” even when they are not moving. Additionally, units struggle to
housing and event space, respectively. cannot tell where particularly heavy areas of overlap are for the different campuses. pick up a signal inside, so buildings on campus (especially Noyce) are underrepresented in the GPS units’
All of these changes affected the patterns in which people In order to understand the density of each individual group of campus residents, a series results. Second, data was collected in mid-November when the temperatures were dropping quickly relative to
moved about campus. No longer did the “jocks” eat at of four kernel density maps were created (Figures 6-9). Figure 6 shows the density of all the previous weeks. A few respondents even mentioned that they would have been outside and visited more
Cowles and the “hippies” at Quad. The Forum “beach,” recorded data points (those shown in Figure 4). This shows hot spots that are frequented by places had it been warmer. Hence, results of this study must be taken in the context of late fall/early winter.
formerly home to bocce and drum circles now lies dormant many people on campus. Figures 7-9 show the individual density maps for East, North, and Third, while construction has been a constant presence on campus for all four years of the senior class’s
every spring. East Campus blurred the North Campus-South South Campus, respectively. Like in the map of points, each campus has a high concentration time at Grinnell, the path between East and South campus (crossing 8th Avenue) was obstructed during most of
Campus dichotomy. near their respective dormitory. the study. Therefore, many people may have been less likely to go between the two campuses, and the paths of
Many believed that the introduction of the Joe Rosenfield However, other patterns show up. For instance, it appears that South campus residents those who did are different than from what one would expect on a campus free of construction.
Figure 1
Center would alleviate some of the perceived North (Figure 9) frequent the Forum and Burling Library more than other campus’s residents. Fourth, like many studies, participants were likely to fall prey to a few biases. Although the response rate
Campus-South Campus divide, but the divide can only be Additionally, one can see which high density places on the all-campus map in Figure 6 are was high, it is still likely that those who chose to enter the study may represent a different demographic group
limited so much due to the inherent distances between caused by residents from a certain campus (such as the cluster near Burling) and which are on campus than those who did not. Also, while the researcher attempted to not tell the participants too much
campuses. Furthermore, as people tend to live near their caused by students from all campuses (ARH). about the study’s focus before the units were carried, the participants may have altered their paths during the
friends, many people find few reasons to go to campuses other than the one they live on. day because they knew their movements were being recorded. Specifically, the points of one path had to be
This research sought to find out where on-campus residents move about on campus and whether this is TYPES OF VISUAL ANALYSES altered to remove a message “written” with the GPS unit.
Points and Lines – Points and lines are the most basic visualization of GPS data used for this
affected by which campus they live on, their major, and other characteristics. I hypothesize, that South, North,
and East Campus residents all have different “boundaries” on campus within which they spend most of their Figure 4 study. Points and lines can both be visualized with the same color or with a variety of FUTURE DIRECTIONS
time and will infrequently move between campuses. And while there is no data from previous eras at Grinnell colors based on a certain data field such as a person’s campus of residence (see Figure This study has the potential to spawn a great deal of future research. First, the
College, one can learn where movement occurs and why it occurs, and begin to understand how changes to the 5) or their major. Lines (as routes) data has yet to receive the level of in-depth analysis it deserves. Much of the data collected in the survey has yet
campus may have affected the patterns of everyday student life. can also be measured by length, to be visualized and statistically analyzed. It seems likely that there may be other patterns of movement or
although in the case of this study, the correlative variables waiting to be discovered. Second, due to the intensity of data collection, the data set is

METHODS
GPS “artifacts” (see Caveats section) were too great to make this a meaningful value. relatively small. However, the methods of this study are easily replicable and could be reused in order to add to
See Figures 2, 4, and 5. this data set or create a comparable one for spring, summer, or fall. Third, as highlighted by Kwan and Ding
(2008), time-geography often lacks a narrative approach despite the seemingly inherent narrative in any
Kernel Density – The kernel density function is a complex way of determining the density of “activity program.” A more thorough ethnographic study of movement on this campus would doubtlessly add
For this study, a stratified sample was used to select 25 students points at any one place on a map. Kernel density creates a smooth surface that can be and built on to the conclusions of this study.
each from North and South and 24 from East campus. 74 students used to approximate density on any part of the map. When mapping density, the more

ACKNOWLEDGEMENTS
were sent three emails asking them to sign up to carry a Global intense the hue, the denser the data points are. Kernel density has the ability to
Positioning System (GPS) unit for one full day. Of the 74 students, 48 illuminate patterns not easily seen when visualized as points. See Figures 6-9.
responded (a response rate of 65%) and 34 students agreed to carry
the units (an acceptance rate of 46% or 69% of those who responded). Mean Center – The mean center finds the central point of a distribution of points, lines or
Of the participants, all but one provided usable GPS data (that person polygons. See Figures 3 and 10. I am deeply appreciative of Professor Eric Carter for teaching me most of what I know about mapping and
never turned their unit on). spatial analysis and advising me on my data collection and analysis. I am also indebted to the Grinnell College
Both the response and acceptance rates were significantly higher Directional Distribution – The directional distribution creates an oval that contains points Anthropology Department for funding this study and providing me with the GPS units that made this all
than expected due to the day-long commitment of picking up, within one standard deviation of the mean center. Additionally, it shows the angle at possible.
carrying, and dropping off the GPS unit. These rates alone suggest which a distribution lies on a map. While kernel density is effective at showing spots of

REFERENCES
that the results may have at least some representativeness of the high intensity, directional distribution is good at comparing how tightly clustered a
campus. distribution is (the larger the oval, the less clustered the data) and the general direction Figure 5
Each participant was asked to come to the Grill (a Figure 2 of movement. See Figures 3 and 10.
centrally-located space on campus) the day before they were to carry
the GPS unit. At the Grill, participants signed a consent form and Huisman, Otto, and Pip Forer. 1998. “Computational agents and urban life spaces: a preliminary realisation of
were shown how to turn the GPS unit on. Upon waking up the next the time-geography of student lifestyles”. In 3rd International Conference on GeoComputation. University of
morning, participants turned the unit on and carried it with them in their bag or pocket until they returned the Bristol, United Kingdom http://www.geocomputation.org/1998/68/gc_68a.htm.
units to the Grill that evening. The GPS unit was set to its default tracking manner which attempts to balance
point-taking instead of plotting a point based on a certain distance traveled or amount of time passed. Kwan, Mei-Po, and Guoxiang Ding. 2008. “Geo-Narrative: Extending Geographic Information Systems for
Participants’ path data was then downloaded and mapped (Figure 2) and the GPS unit was distributed to Narrative Analysis in Qualitative and Mixed-Method Research.” The Professional Geographer. 60:443-465.
another person that evening. Data was collected on Monday, November 11 through Thursday, November 20,
Root-Wiley, Mark. 2008. "Grinnell College Student GPS Paths and Survey," unreleased data set.
and Monday, November 24. Hence, the results of this study are applicable for only weekdays and cold times.
Once each participant turned in the unit, they were asked to fill out an 18-question survey about where and
Sayer, Andrew. 1985. “The Difference that Space Makes.” Pp. 49-66 in Social Relations, Space, and Time. New
why they go places on campus. These answers were then joined to their GPS data and used to visualize the
York: St. Martin's Press.
dataset in an attempt to find correlations between geographic presence in a certain place and a person’s traits and
habits. Finally, maps made in ArcMap were used to visualize and analyze the data. The various methods of Urry, Gregory. 1985. “Social Relations, Space, and Time.” Pp. 20-48 in Social Relations, Space, and Time. New
visualization are outlined in the “Results” section. York: St. Martin's Press.
Each participant was assigned a unique ID number for purposes of anonymizing the data. Therefore, this
data can be shared with little risk to any of the participants. Furthermore, I attempted to only use mapping Yu, H. 2007. “Visualizing and Analyzing Activities in an Integrated Space-time Environment: Temporal GIS
Figure 6 Figure 7 Figure 8 Figure 9
techniques that could not be used to identify individual participants in the study. Design and Implementation.” In Transportation Research Board 86th Annual Meeting, Washington.

You might also like