Child pages
  • Guenter Final Assignment


Introduction to Geographic Information Systems (UEP 0232-01)

Barbara Parmenter

Fall 2013


Finding ways in GIS to operationalize the idea of Neighborways in Somerville



My project focused on the idea of implementing Neighborways in Somerville. Neighborways are residential streets that are especially designed for low volumes and speeds for auto traffic. They intent to facilitate children playing on the streets, allowing them to bike/ walk more safely to school and enhance the interconnectivity of possible families’ destinations that could easily be reached by biking or walking. The idea is based on an init iative by UEP lecturer Mark Chase and two of his former students (

To be more specific, the goal of my work was to determine which areas in Somerville should have a priority in initiating this project. I came up with two different approaches in how to operationalize the idea of Neighborways in GIS, and give hereby a more specific proposal for prioritization . Each of these concentrated on a different dimension. The first focused on the walkability around schools, hereby taking into account among others crash data on pedestrian and bicycle accidents in these areas. The second analysis took more into consideration the feasible aspects of Neighborways in Somerville. For this I tested different criteria as one-way streets and sidewalks width. Therefore, the two analyses can be distinguished by looking at either socio-demographic necessity or the infrastructural potentiality.

As a result, I wanted to show that there are different ways in prioritizing areas when looking for the potential conversion of residential streets into Neighborways. Whether these two approaches can be integrated remains to be seen. It is also a question of scale and political will. Either one concentrates on certain areas where Neighborways are most needed ( area focus) or on the low-hanging fruits that could be converted most easily ( individual-streets focus).

In the following I will give an overview of my work. This will not only include the interim outcomes and problems that I ran into while doing my analysis, but also provide information on the steps that are necessary to make. Hereby, I want to facilitate the use of my approach and also make it easier to make alterations within the approach.  

H:\Final Project_Neighborways\Maps\Families_in_Somerville_Blocks.jpg H:\Final Project_Neighborways\Maps\Families_in_Somerville_Tracts.jpg The general picture          In order to get a first impression of Somerville and approach the idea of Neighborways I was interested in where children and families actually live. Therefore, I looked for socio-demographic data as children and family density. My goal here was to identify in which areas Neighborways would make most sense because of families and children being, ultimately, the potential users. I did this analysis by, both looking at the block (for higher accuracy) and the tract level (for higher visibility). Anyhow, I did not consider the tract level merely because of providing more visibility but also because of giving a more complete picture than the block level. For some blocks there was no data available on both family and children characteristics. This was partially probably due to green spaces or industrial use covering the whole area. However, I doubt this is true for all the areas.  The idea here was to give a higher priority to areas of higher children density and high percentage of families. I calculated the children density because the mere population quantities themselves do not consider the spatial dimension and, thus fail to give accurate information.  Instead, density is a quantity unit per area. Also, higher percentages of families (comparing it to the number of total households) was presumably a better indicator than merely looking at family quantities (again, a spatial relation would be missing). Making use of family density instead of density per se as a criterion is also mentioned in Giles-Corti et al.’s study on ‘School Sites and the Potential to Walk to School’ (Giles-Corti et al. 2011: 549).

As can be seen on the maps we get a first notion of where more families and children within Somerville live. Clearly, there are more and less children- and family-populated areas. With the exception of only a couple of blocks in the West of Somerville the most dense children block are located in the East. Here, we find quite a dense and coherent area of children density and high percentages of families. 

H:\Final Project_Neighborways\Maps\Children_Density_Blocks.jpg Having mapped as well family percentages as children density was useful in order to figure out that a high percentage of families does not necessarily imply a high density of children (as can be seen for example for the area of Ten Hills). However, quite often these two categories do correlate and show same patterns.

H:\Final Project_Neighborways\Maps\Children_Density_Tracts.jpg





Walkability in School Areas

  1. Mapping Schools in Somerville

H:\Final Project_Neighborways\Maps\Schools.jpg Based on the idea that Neighborways serve as connectors between neighborhoods and places of interest I identified schools as the most important place for children and families, since this is, supposedly, the place daily frequented by children. Thus, making routes to school safer should be a top priority when looking for potential Neighborways (c.f. City of Somerville).

As can be seen on the map we already see where most schools are located. This correlates quite well with the results above on children density and percentage of families. There is a certain nucleus of schools located in Eastern Somerville.

However, having mapped as well socio-demographic data as schools we do not have any more qualitative data on how walkable these areas are and whether there is a potential in converting streets into Neighborways. Thus, I made use of Giles-Corti et al.’s study on the potential to walk to school, hereby gaining information on the actual walkability in the surrounding areas of schools.

  1. Creating Pedsheds

For the first steps, I followed very much the network analysis Giles-Corti et al. did. I used information on street connectivity and traffic exposure in school neighborhoods creating in order to create so called Pedsheds. The sheds will be created by, firstly, calculating “a ratio of the pedestrian network area to the maximum possible area within a defined distance based on Euclidian […] distance.” (Giles-Corti et al. 2011: 546) Because Somerville is a small city I reduced the catchment area to only 400 meters or 0.25 miles (in contrast Giles-Corti et al. used a Pedshed of 2km), since this is also considered to be the walkable distance within a 5-minute walk. (c.f. Pedshed). Hereby, I also avoided too many overlappings that were created by using default breaks as 1km or more. I intended to do this, because originally I wanted to generate not only Pedsheds but also Cyclesheds. However, I gave up on this idea when becoming aware that I wanted to deal with smaller areas.

Instead of using the political boundary as a limitation of the data I created a 1mile-buffer around the area of Somerville in order to have comparable school catchment areas that would not arbitrarily end at the boundary of Somerville.

Giles-Corti et al. say the WSA/AA ratio has to be equal or above 0.6 in order to speak of a good catchment area. H:\Final Project_Neighborways\Maps\Pedsheds.jpg As we can see most school areas have a WSA/AA ratio of around 0.6 in Somerville. However, as we can also see some schools as the ‘East Somerville Community School’ or the ‘Albert F. Argenziano School’ show only ratios of 0.38 and 0.44, thus being way below the other ones. These differences should be more elaborated on throughout the following stages.

  1. Measuring Vehicular Traffic Exposure in School Areas

In the next step I used the category ‘street classes’ as a proxy in order to figure out the vehicular traffic exposure in the catchment areas (just as Giles-Cort et al. did). First, I tried to make use of the information on ‘Average Annual Daily Traffic’ that is conveyed in the ‘EOT_Roads’ layer. This would have been helpful, since the categories for differentiating between street classes is not very precise. Also, I believed that relying on categories such as local and non-local streets would not be very helpful, since most streets in Somerville are actually local ones. However, making use of the ADT data did not work as I hoped it would, because most residential streets were missing data (only given for major arterials in Somerville like Somerville Avenue, Broadway etc.).

By differentiating between local and non-local streets one can see that the higher the ratio the higher the exposure to vehicular traffic. There are clearly differences within our analysis. Some school areas show a much higher exposure to vehicular traffic (highest around 1.20) than others (lowest around 0.40). In conclusion, using a differentiation between local and non-local streets proofed unexpectedly successful. This may partly be due to massDOT’s good classification system that differentiates between different arterials, collectors and local streets. Thus, I was able to be sure about actually using only local streets that “provide access to abutting land with little or no emphasis on mobility.” (MassDOT). This is exactly the type of streets I was looking for in this analysis on Neighborways.

  1. Revision

As a next step Giles-Corti et al. created an index of the information mentioned above by making use of reverse coding (when necessary) and collapsing the scores into deciles. Put together one would have equally covered information on walkability and vehicular traffic exposure.

However, I did not deem the two dimensions used for the analysis as sufficient, since both do not cover any more qualitative information on the walking experience. A potential shortcoming is also mentioned by Giles-Cort et al.: “The school-specific walkability index included only two-subcomponents and may be strengthened by additional sub-components.” (Giles-Corti et al. 2011: 549) Also, one has to take into account that most streets in Somerville are actually ‘walkable’ in a certain sense (in providing sidewalks). This leads to the assumption that Somerville already has a, comparably, dense walking network. Nevertheless, this does not mean by necessity that the walking experience would be pleasant and safe, especially for children who are highly exposed to any sort of vehicular threat. For parents the walking environment and the safety of the children are supposedly the more decisive factors than merely having a dense walking network. Because of these reasons, I chose to include into my analysis two more criteria that covered a more qualitative dimension of walking instead of merely looking at the potential to walk.


I had to experiment a lot in order to find accurate data that would give useful information on the quality of walking. The EOT_Roads layer has a huge abundance of aspects that are included. Anyhow, only few of them actually give useful information on using the streets by any other mode of transport than the car. I was only able to overcome this centeredness of the car by making use of the information on ‘width of sidewalks’. Adding to that, I used data on accidents in order to cover the safety dimension of walking within the school areas.  


  1. Using Information on Sidewalks and Accidents in order to include a more Qualitative Aspect of Walking

I presumed that merely looking at the width of sidewalks would not give any accurate information, since some of the streets show either a width of ‘0’ or very large numbers. Nevertheless, both sort of information are wrong in the sense that they do neither necessarily display a very bad infrastructure for walking nor a very good one. This is because wide sidewalks are supposedly to be found on arterial roads, no sidewalks are to be found in more residential streets that probably do not have any separation at all and, thus often probably allow for all modes of transport.

Hence, I came up with the idea to create another index of relating the width of the sidewalks to the actual right-of-way. By doing this, I hoped to cover more sufficiently the qualitative dimension of a walking experience, since sidewalks are put into the context of the street. Anyhow, I also see a potential shortcoming of this approach, because, again, there is no attention paid to the quality of the sidewalks or breaks in-between streets (e.g. for the connectedness of sidewalks). In addition, this approach assumes that when there is no sidewalk on a street the ratio, naturally, is also ‘0’, hereby adding to a more negative result of an area. As was already mentioned before, this causality does not necessarily hold true.

As can be seen the results do not vary that much. The range of sidewalk proportion on streets in the school areas is between about 0.19 and 0.33. However, taking into consideration the fact that the roadway is always the biggest portion of a street, also these minor variances in numbers can make quite a difference.  


The other qualitative aspect I wanted to integrate into the analysis was intended to cover the safety issue of streets. For this I was fortunate enough to find crash data on the Hubway Homepage (c.f. Hubway Data Challenge). In a first step I mapped only the two dimensions accidents with pedestrians (involving a car) and accidents with bicyclists (involving a car).

H:\Final Project_Neighborways\Maps\Accidents.jpg Since I did not make any further use of the dimension bikability in Somerville, I decided not to include accidents with bicyclists in my analysis. In order to generate a certain comparability, again, I referenced the data on the school areas by dividing it by square miles. As can be seen in the table below, the numbers are, generally speaking, fortunately not too high. This is due to the small numbers in accidents. Only some school areas show a ratio of more than 0.1, most areas lie below this number.











  1. Creating an Index on Walkability

In a last section I needed to put together all the information that I had collected throughout the last steps. Thus, I converted the polygon layers into raster files in order to, ultimately, put the information together and create a walkability index. Special attention needed to be dedicated to the coding of the raster layers. When reclassifying the raster files I reverse coded the crash data and the vehicle traffic exposure in order to have all data classified the same way. Since I made use of four criteria I created only five classes instead of ten as Giles-Corti et al. did in their analysis. Thus, I would create later on an index that has the same range (max. value 20), but is composed of four criteria instead of two (thus, min. value 4). The four reclassified raster layers look like the following (colorful, yet still useless):

The Pedshed

Local/ Non-Local Ratio

Sidewalk Proportion

Crash Ratio

H:\Final Project_Neighborways\Maps\Final_Map.jpg In a last step I put all these information together by making use of the ‘Raster Calculator’ tool. Hereby, one raster layer is created that conveys all the information. I decided not to weigh any factor higher or lower than another, because they all seemed equally important to me. However, if wanted this would have been an option at this point.

As a result we get a map that illustrates the walkability for school areas in Somerville. This was classified and color ramped appropriately so that spatial patterns would become more visible. It can be seen that there is a predominant West/ East-pattern. Although more schools are located in the East of the city, the results show that the walkability in these areas is worse than in the Western parts of Somerville. When we take into consideration factors as children density and percentages of families we can say that there is clearly an equity issue in Somerville in the sense that students going to schools in East Somerville have a disadvantage in comparison to students in West Somerville. This might partially be explainable by the fact that West Somerville is more residential than the East is. However, further research would need to be done in order to make correlations with socio-economic factors.

Identifying easily convertible, potential residential streets in Somerville

As another approach towards converting residential streets in Somerville into Neighborways I followed an easier way by merely looking at the existing infrastructure. I simply asked where the existing infrastructure is most unpleasant and simultaneously most feasible to be converted into Neighborways. This approach was kind of different to the one illustrated before, because I had to find ways to operationalize the idea of convertibility without relying on a given nucleus as schools. I tested different criteria in order to reduce the data step-by-step.

H:\Final Project_Neighborways\Maps\Local_Non_Local_Streets.jpg It was obvious how to start reducing the data. Since I only wanted to deal with residential streets, only those ones with low traffic volumes, I made again use of the category local streets. As was already mentioned before, by definition only streets classified as ‘local’ ones have a higher prioritization of access than mobility. 

H:\Final Project_Neighborways\Maps\Street Operation.jpg Still quite a lot of streets remained. Therefore, I made use of the category ‘Street Operation’ in order to select out all the streets that have two-way traffic. I deemed this a promising step, because by this I could highly reduce the data quite accurately. In my perception it seemed more feasible to convert one-way streets into Neighborways because the vehicular traffic levels on residential one-way streets undoubtedly has to be smaller than on two-way streets.

As can be seen on the map, this step proofed successful. By selecting out all two-way streets I got rid of half of the streets. Since there were still a lot of streets available I thought that Neighborways definitely have a good potential in Somerville, since most streets are actually supposed to be neighborhood-friendly and made for slow-traffic.

However, going further from here proofed to be difficult, because I had very little information on what Neighborways additionally require and how to convert these by making use of GIS. Thus, I tested information as curbs or speed limits because I supposed these would further reduce my data accurately. My thought was: the more curbs, the more difficult to level the street (which would be an ideal condition for Neighborways, since there ought to be no clear separation between different modes of transport). This intention proofed unsuccessful, because the data on curbs did not seem to have a high quality. The classification was not very accurate in the sense that curbs existed when there were not any curbs to be expected.  The same held true for speed limits. Although I found information on some streets, most of them were classified to not have any speed limits. I could have extrapolated the data, but the effort seemed not very promising because I supposed most of the streets I was dealing with had a speed limit anyways.

Eventually, I came across the idea to make use of the right-of-way which is the total width of street (including sidewalks and shoulders). I noticed that at a certain point the sidewalks were becoming smaller and smaller whereas the roadway always kept a minimum width. This fact held true for a width around 40 feet, because most streets have a minimum width of 40 feet.

Starting from that point, the sidewalks became less wide than 5 feet which definitely does not provide a nice walking environment anymore. This holds especially true when the sidewalks are crammed with parking cars, electric poles or garbage cans, as can be seen on the exemplary photos below. Here, the sidewalks work more as an alibi than serving actual pedestrian purposes. Supposedly, sidewalks are not used at all on these streets because they are too narrow and it would be inconvenient to use these (because of obstacles). Hence, I would call these sidewalks unused parts of the street, because they do not serve the purpose they are made for. If not they tend to be used by residents (as is the case for garbage cans) or car drivers (expanded parking zones). 

In order to make this more palpable I took some screenshots of examples for streets with either a right-of-way around 40 feet (1) or streets with a right-of-way below that (2). One can clearly see the differences in the streetscape with the first row of images allowing for a more or less pleasant walking environment and the second making walking a nuisance.

Partridge Avenue, Magoun Square

Everett Avenue, East Somerville

Hawthorne Street, Davis Square

Springhill Terrace, Spring Hill

Autumn Street, East Somerville

Franklin Avenue, East Somerville


In the following I will provide different maps, showing what happens if I select certain widths below the minimum of 40 feet. This will illustrate how much the data gets reduced by following this approach.

Right-of-way: 40 feet

Right-of-way: 37 feet

Right-of-way: 30 feet

Clearly visible, there are not that many differences anymore between a right-of-way of 37 or 30 feet. Therefore, we can assume that the critical point is at 40 feet. Having provided some images above, my hypothesis is that the less wide the street ( only residential streets), the less pleasant the sidewalk. Thus, it would make a lot of sense to start with local, one-way streets that have a right-of-way below 40 feet. It seems to be especially these streets that provide the least pleasant walking environments and that force people to walk on the roadway. Thus, we find here both premises holding true: most unpleasant conditions for walking and most feasibility of converting the streets into Neighborways. The latter is extremely important. Because both parties, pedestrians and cars might be ‘suffering’ from the current conditions, converting these streets into Neighborways would only be a logical step.  Because both parties currently use the space of the other without being supposed to do so, it would only make sense to adjust the design of the street to this, and hereby create a more pleasant and safe environment for all. 


In a last step these results could be matched with suggestions from the official Neighborways homepage. By having paid special attention to the right-of-way width, and hereby differentiating between existing conditions within local streets in Somerville, different treatments are necessary. It seems to me that most of the suggestions have a certain width of the street in mind when advocating for bumpouts, residential parklets, chicanes or crossing islands (c.f. Somerville Neighborways). While all these treatments might be very useful for residential streets that have a right-of-way of approximately 40 feet I doubt their appropriateness for street widths below that number.



As a result, I came up with two different approaches showing where and how Neighborways could be prioritized. The goal was to give a guideline on how to find potential Neighborways in Somerville with each analysis focusing on a different dimension. The ‘walkability in school areas’-analysis looked at the socio-demographic necessity within Somerville, hereby taking into consideration vehicular traffic exposure, sidewalk proportions and crash data. By looking at this, we found that there is actually a discrepancy in where children live (or rather, most students go to school) and the walkability around these schools. This holds true, since most schools in East Somerville showed a worse walkability in the surrounding area than the ones located in the West, though obviously more children live in the East of Somerville. Therefore, one might focus especially on these school areas when thinking about ways of implementing Neighborways. Both, structural need and demand would supposedly be biggest here.


The ‘infrastructural potentiality’-analysis focused more on the existing streetscape, hereby looking at street function, street operation and width of the right-of-way. The question here was not so much an equity or demand one but where Neighborways would be most logical or feasible to implement. As a result, we could observe that by changing the right-of-way as a last criterion of reducing GIS data, different types of residential streets were observable that probably also need to be addressed differently (40 feet wide vs. below 40 feet wide). As a very interesting outcome we saw that especially those streets seem to be, both, the most problematic and the most potential ones that have the narrowest right-of-ways. Quite counterintuitively these streets seem to be, regarding the infrastructure, the most pedestrian-unfriendly ones, since in every case the sidewalks are the neglected parts of the street when dealing with a scarcity of space.  


This analysis does not claim any sort of absoluteness. The intention was to try out different approaches and ways in GIS how to prioritize Neighborways. Hence, I do believe there ought to be some shortcomings in my analysis. However, both tested approaches proofed to be successful and revealed interesting information on current problems and future potentialities. It would be interesting to link the two approaches more closely and find ways in how to merge them. Also, it would be worth it to pay more attention to the connectivity between different places (as between different schools) by already existing safe routes (as by community paths or bicycle trails). An example for this would be the ‘Somerville East/ West Neighborway’ (c.f. Somerville Neighborways).









Giles-Corti et al. (2011): School site and the potential to walk to school: The impact of street connectivity and traffic exposure in school neighborhoods. In: Health & Place (17). 545-550.


Non-Scientific Resources

City of Somerville. Accessed on Dec 16, 2013.

Hubway Data Challenge. Accessed on Dec 15, 2013.

MassDOT. Accessed on Dec 15, 2013.

Somerville Neighborways. Accessed on Dec 15, 2013.

Pedshed. Walkable Urban Design & Sustainable Placemaking. Accessed on Dec 15, 2013.




Useful Academic Resources


Holbrow, Gabriel. December, 2010 Walking the Network. A Novel Methodology for Measuring Walkability Using Distance to Destinations along a Network: Case Study of Washington, D.C.   []

The first source stems from a former student of Tufts University. Holbrow created a walkability-index by putting together distance-raster layers from certain points like grocery stores, schools, theaters, etc. He set a standard of what he deemed as walkable (400 or 800 meter), created a corresponding raster layer with the distances calculated and weighted the destinations differently in order to get as a result a walkability score for each raster cell of a certain area. Thus, he was able to compare the walkability score of a chosen area with demographic data in order to see which groups are more or less likely to live in walkable or unwalkable areas.



Giles-Corti, Billie et al. 2011. School site and the potential to walk to school: The impact of street connectivity and traffic exposure in school neighborhoods, Health & Place, 17(2): 545-550. []

Giles-Corti et al. looked at the two criteria, street connectivity and traffic exposure, in school neighborhoods in order to create a walkability index. After that, the index was compared with data on in which areas children tend to walk to school regularly or don’t. Intuitively and proven by the research project, more children walking were found in those areas that showed, both, high street connectivity and comparably lower traffic volumes. 

Zhu, Xuemei; Lee, Chanam. 2008. Walkability and Safety Around Elementary Schools: Economic and Ethnic Disparities. American Journal of Preventive Medicine, 34(4): 282-290. []

Also, this paper is mostly concerned with equity issues, studying low-income and poverty populations. What the study shows is that low-income and Hispanic children are more likely to live in unsafe areas (being exposed to poor street environments) but also have some favorable neighborhood-level conditions. However, for my study I could mostly make use of the methodology used in this study. In order to measure the neighborhood-level walkability the researchers included an estimate of potential walkers (this is based on the percentage of students living within half a mile from school), pedestrian facilities (sidewalks and traffic-signal density), population density, street connectivity (street and intersection density), and land-use mix. At the other end of the spectrum, the neighborhood-level safety was measured by crime rates and traffic dangers such as traffic volumes, percentages of high-speed streets, and crash rates. (See table below)

Ackerson, K. J. 2005. A GIS Approach to Evaluating Streetscape and Neighborhood Walkability. Unpublished Master Thesis.


This last example is a thesis written by a student at University of Oregon. Ackerson tried to emphasize what is often neglected by researchers, namely the perspective of the micro-level. Thus, he not only used data on street networks but also paid attention to the streetscape characteristics that are an integral part of pedestrians daily walking experience but often forgotten by researchers focusing only on the larger-scale. These “pedestrian-level characteristics” are of crucial importance for people making decisions on their form of mobility.

What I found most interesting in this approach was, again, the methodology. Ackerson included in his research things like the walking surface or streetscape elements as trees or views. Thus, he not only paid attention to dimensions as safety or destinations but, even more so, on walking experiences themselves. This aesthetic dimension can also be found in other approaches (see ERTT – Easy Ride Travel Time methodology) that try to pay equal attention to functionality and enjoyability of a ride.


Useful Data

Data Layer





This layer covers mostly everything I did in the section on feasibility.

MassDOT (April 2012)

Layer contains fields as class or road types, but also useful information on sidewalk and right-of-way width. Also, there are fields on one-way or two-way traffic and number of existing curbs.

Schools    SCHOOLS_PT

This layer covers most of what I did in the section on walkability.

MassGIS (Oct 2012)

Layer contains the field grades and type of school if a further specification of the information is wanted (in case I wanted to make differences between kindergartens and high schools).

Open Space Somerville_Parks_05

City of Somerville/ Tufts University (July 2005)

Layer contains information on size of park (if a further specification based on the scale of the park is wanted)



MassGIS (Sept 2004)

Layer contains useful information on the trail status (existing, planned, potential etc.). However, the layer has not been updated since 2004 and is, thus, probably outdated.

Census 2010 Blocks CENSUS2010BLOCKS_POLY

TIGER/ Line Shapefile (2010)

Layer contains blocks in Somerville, MA. (layer used from M drive!)

Census 2010 Tracts CENSUS2010TRACTS_POLY

TIGER/ Line Shapefile (2010)

Layer contains tracts in Somerville, MA. (layer used from M drive!)

School Enrollment S1401

US Census Bureau (2010-2012 American Community Survey)

Table contains useful information on school enrollment, thus, specifying the type of school people attend

Household Size   QT-P11

US Census Bureau (2010 Census)

Table contains information on family/ non-family households, average family sizes (also available on the M drive)

Population            P1

US Census Bureau (2010 Census)

Table contains information on the total population and, thus, allows for questions as population density. (also available on the M drive)