Susan Swinson

Project: Surface Analysis of School Allocation

Introduction

To explore Surface Analysis, we experimented with data from Centre County, PA. As one of the fastest
growing counties in the state, school enrollment is burgeoning and a new high school needs to be
appropriately sited to serve the four school districts. To do this, we had to analyze existing raster
datasets to determine a potential location for the new school.

Allocation Based on Straight Line Distance

The first step is to run the Spatial Analyst tool for a straight line distance analysis. This tool used the
locations of the four existing Centre County high schools and allocated cells based on their least distance
from each high school. For example, in Figure 1 below, every cell that is colored bright lime green (Zone
2) is closer to the school in that eastern zone than to any of the other three schools. This is the raster
equivalent of a proximity polygon (OSullivan, 2014).
Figure 1:
Euclidean
Allocation of the
four high
schools in
Centre County,
PA. Each
colored zone
represents areas
of the county
that correspond
most closely
distance-wise to
the high school
within that
zone.

For
comparison, I
also ran a Straight Line Distance Analysis of the area (see Figure 2). The Straight Line tool calculates the
distance from every cell to each school, measured as the crow flies, and then assigns the shortest
distance to the raster. The Allocation Tool is clearly based on the straight line distance calculations.

1
 Figure 2:
Straight Line
Distance
Analysis of
the four high
schools in
Centre
County, PA.
The distances
to each high
school are
determined
in a straight
line or as
the crow
flies.

One example to note from comparing these two different methods is a student who may be living
between the two schools near the center of the county. In Figure 2, we see an overlap in the light gold
coloration, indicating that the student could be within the attendance zone of either school. Figure 1,
however, provides a clearer picture in this particular instance as to which school the student should
attend (why, the one with the best science program, of course!)

In reality, straight line distances are not always useful. When we drive to school, unless we have some
type of futuristic hovercar, chances are we are making multiple turns as we follow various roadways or
political or topographical features. A political boundary example may be a student living near a county
border. The student may technically be closer to a school in the next county but as this school is not
within their county of residence, this sort of distance allocation would be inaccurate. We need to
combine multiple analyses to determine a more ideal output that takes all these variables into account.
One option for this would be to create an individual analysis for each school and then add these four
analysis grids together with the raster calculator.

Distance Analysis Over Roads (Part 1)

2
To begin an effective analysis of determining a suitable school site, we must first create a raster of the
road network. As mentioned previously, distances are not generally traveled in a straight line but
depend quite a bit on roads and other factors. Therefore we clearly need to take the road network into
account for this analysis to have any meaning. In order to do that, we must convert each set of roads
(Major and Local) into separate raster layers (MajorRds and LocalRds) using the Feature to Raster
conversion tool. The next step was to reclassify these rasters with major roads being given a value of
1, local roads being given a value of 2, and all remaining non-road areas being given a value of
100. (Because, as much fun as it would be, neither bus drivers nor parents are going to want to go off-
road in order to get students to school, though student drivers themselves may be another matter!)

Then I combined these two rasters and opted to use the values for the Major roads where any
intersections occurred, as I assume that such intersections will have at least some of the attributes of a
major road in that particular location. I set the non-road areas to display as NoData so the original DEM
can be seen underneath in non-road areas (see Figure 3).

Figure 3: The new Roads raster layer with major roads shown in red and minor roads shown in yellow.

Distance Analysis Over Roads (Part 2)

3
Now that our roads are weighted in terms of ease of travel, with major roads being quicker to travel
on than minor roads, we can repeat the previous distance analysis but this time taking the weighted
roads into account. The result clearly differs from the Allocation created in Figure 1.

Figure 4: New Euclidean Allocation of the four high schools in Centre County, PA based on the weighted roads
layer. Each colored zone represents areas of the county that correspond most closely distance-wise to the high
school within that zone after accounting for travel time on the road network. The thick black lines represent the
official school district boundaries and the thin black lines represent civil boundaries.

This is certainly a more realistic allocation, but does not seem to account for all the inconsistencies
between the straight-line distance allocation and the actual school districts. Rather, it appears that
school district boundaries tend to fall along civil boundaries, resulting in some allocated districts that do
not align well with actual school districts.

Allocation Based on Road Distances

4
My next task was to estimate the school-age population within each school district. This is important
information when determining the best location for a new school, as it would not be very useful to site
the school in an area with lower numbers of students.

By creating a raster representation and using Zonal Statistics, I calculated the number of school-age
children per civil division in Centre County as shown in Figure 5. This map shows that the State College
Area School District is the most populous district.

Figure 5: The density of school-age children (ages 5 to 17) per census block in Centre County, PA. Classified by
Natural Breaks (Jenks) with darker shades representing a higher number of school-aged children.

A map of the student population by school district rather than census block confirms this assessment
(see Figure 6).

5
Figure 6: The density of school-age children (ages 5 to 17) per school district in Centre County, PA. Classified by
Natural Breaks (Jenks) with darker shades representing a higher number of school-aged children.

The Zonal Statistics By Table tool displays the population of each school district:

Its interesting to note that the State College Area School District has more than twice the number of
students than the next most-populous district, Bellefonte Area School District. I think it makes it fairly
obvious, then, the general vicinity in which a new high school should be sited.

A map displaying student population according the Euclidean Allocated zones which were weighted
according to the road network serves to reinforce that assumption (see Figure 7).

6
Figure 7: Population of school-age children (ages 5 to 17) per school district in Centre County, PA, with darker areas
representing higher numbers of school-aged children.

And so...Where to Put a New School and What Should Be Its District?!

As we reasoned from our previous analysis, the State College Area School District already has more than
twice the number of students than other districts. Ideally, if we could divide the State College Area
District evenly without requiring any other districts to be redrawn, that would be the easiest and most
cost-effective method. But where shall we place our new school within the State College Area District?
Definitely near some fast-food joints who would appreciate all the lunchtime business, right? Well,
perhaps we should be a bit more specific so I performed a Kernel Density Estimation (KDE) to
determine precisely where the largest concentrations of school-aged children are located within the
county (see Figure 8).

7
Figure 8: Kernel Density Estimation of school-age children (ages 5 to 17) per school district and census block in
Centre County, PA, with darker areas representing higher densities of students.

The KDE shows that there a few clusters of students within the State College Area School District, but
the largest concentration seem to be located in the approximate center of the district (see Figure 9).

Figure 9: A zoomed-in view of the Kernel Density Estimation of school-age children (ages 5 to 17) in the State
College Area School District.

8
By drawing a new district along census block lines, we can site the new school to take a lot of pressure
off the existing school a sum of about 4,710 students, or approximately half of the districts current
9,578 (see Figure 10).

Figure 10: Statistics of the new school district created from select census blocks show that approximately 4,710
students aged 5 to 17 will be incorporated from the previous State College Area District. .

Figure 11: Areas highlighted in bright blue represent census blocks that would be appointed into the new school
district. A new school should be sited somewhere in the center of this new region and close to major roads, such as
the spot indicated by the red star.

Conclusion

In previous lessons, we explored how knowledge of underlying patterns and processes is critical to truly
understand the results of an analysis, as is a thorough understanding of various analysis methods such
as interpolation that may affect the outcome. In a similar vein, this lesson explored how a basic analysis,
such as our initial straight line distance, may tell us something about the data but may not effectively
present the entire picture. By delving deeper into an analysis and carefully examining any geographic
factors such as topography, road networks, specific populations, and so forth, we are very likely to end
up with a more refined and far more accurate analysis. It is key, however, to know where we are going
(with the data) before we can figure out how to get there.
9
References

O'Sullivan, David. (2014). GEOG 586: Geographic Information Analysis, Lesson 6: Surface Analysis. The
Pennsylvania State University World Campus. Accessed November 2013 at https://www.e-
education.psu.edu/geog586/l7.html

O'Sullivan, D., & Unwin, D. J. (2010). Geographic Information Analysis. (2nd ed.). Hoboken, New
Jersey: John Wiley & Sons, Inc.

10