Characterizing New York City's One-Way Streets

One-way signs are a common occurrence in many North American streetscapes. In New York City, they are particularly common — and as much part of the city’s image as its steel subway cars. The beginnings of a trend to designate streets as one-way can be found  as early as in the 1910s  (Gray, 1997). Since then, their number has been constantly growing.

Nevertheless, one-way streets are a contentious topic:  Advocates  often cite safety, capacity and convenience as major advantages of one-way streets (Stemley, 1998).  Opponents , on the other hand, argue that safety on one-way streets is negatively impacted by their tendency to encourage speeding and reckless driving (Badger, 2015).  

Contentious or not, though, one-way streets are omnipresent. We wanted to understand their presence in New York City better, particularly their spatial distribution, and the types of land uses that most commonly surround them. On this basis, we formulated two research questions:

(1) Are there community districts in NYC that have greater proportions of one-way streets?

(2) Do one-way streets tend to be particularly common for specific types of land uses?

Before we dive into our own analysis, we want to highlight a research project not too dissimilar to our own in terms of the types of data it is using, and the methods it is applying:

In their study,  Oliver et al.  (2007) compare the useability of different types of buffering methods for examining the influence of land use on walking patterns. To us, even though it has a different thematic focus, this study is nevertheless particularly interesting as it examines different ways of measuring land use around streets. 

According to the authors, the three most commonly used ways of defining spatial units to measure aspects of the build environment, such as land use, are:

(1) Using a predefined unit, e.g. census tracts, planning neighborhoods, ZIP code areas, et cetera.

(2) Working with circular buffers, e.g. around individuals’ addresses, or other points of interest.

(3) Working with polygon-based road network buffers, which create buffers from a starting point along the road network for a predefined distance, resulting in a single non-circular buffer around that point, based on the surrounding road network. 

A less commonly used method is buffering streets line segments themselves. In terms of land use, though, the authors argue that this method has the potential for many benefits — chief among them better accuracy. When analyzing land use related to streets — as is the case in the authors’ study, looking at walking patterns, and in our study, looking at relationships between land use and the distribution of one-way streets — this method can provide much more accurate and comparable results, the authors argue.

The authors go on to investigate in a case study how both polygon-based road network buffers and street line buffers can successfully interact to measure the effect of different types of land use on walking patterns.

Particularly worth highlighting in the context of our own analysis are the authors’ buffer size decisions in this case study: 

• “A 50 metre buffer was chosen to ensure that parcels along the selected roads would be included but that most parcels located further from the road (e.g. behind those adjacent to the road) would not be selected.”

• “A 100 metre buffer was too large as it often included such inaccessible parcels and could bias the weight of parcels whose depth away from the road is disproportionate with their "shop front" profile experienced by walkers along the road.” 

• “A 25 metre buffer was too narrow as it would sometimes miss properties set slightly back from the road along wide roadways or wide right-of-ways.”

Overall, this study provides a good basis for understanding the potentials of street line buffering when analyzing land use around those streets. We strongly agree with the study’s emphasis on the upsides of using this method in this particular context, which is why we are using a similar approach.

We downloaded the "LION Single Line Street Basemap" GDB from NYC Panning. It depicts a lot of information, including features that are not of interest to this project. The Geodatabase comes with a number of .lyr files, which essentially consist of only specific features from the overall LION file. We were interested in Street Centerlines, particularly such that two-way streets are not depicted by two separate lines but rather by a single line, as this makes the analysis later on much easier. The “Lion Streets - Generic.lyr” file provides this type of filtering.

We added “Lion Streets - Generic.lyr” to the map and performed a number of further cleaning steps using the “Select by Attribute” tool. Those steps included filtering out ferry routes, alleyways and paths (as our focus is only on actual streets), as well as filtering out the small number of street segments that did not have a traffic direction attribute assigned to them. The final layer was saved as “streets”.

In this new “streets” layer, we summarized traffic direction into two simple categories, “OW” for one-way and “TW” for two-way, and based the layer’s symbology on these two categories. Resulting is a city-wide overview of the street network, distinguishing one-way and two-way streets through color:

After gaining an initial overview of New York City’s one-way street network, we went on to quantify this distribution on a neighborhood level. With a Community Districts (CD) layer as a spatial unit representing the city’s neighborhoods, we used the “Summarize Within” function to calculate both the sum of the length of all one-way street segments in a given CD, and the sum of the length of all street segments in that CD. By dividing the former by the latter, and multiplying the result by a hundred, we receive a value representing the proportion one-way streets make up of a CD’s total street network. On this value, we visualized the following map:

RQ 1 Findings: Overall, a definite spatial pattern is visible in the distribution of one-way streets throughout New York City: Areas on the city’s outskirts have much less one-way streets than more central areas.

We downloaded land-use data at the lot-level as the file “MapPLUTO - Shoreline Clipped" shapefile from NYC Planning. Then, we changed symbology to Unique Values, using the LandUse field, which are integers from 1 to 11 representing different land-use categories. Next, we added the written categories as labels to the symbology.

Next, we created a 75ft buffer around all streets, calling this layer. Then, using the "Tabulate Intersection" tool, we calculated the distribution of land use categories per segment. The tool returns a table with multiple rows per street segment, with each row representing a land use category present within that segment, and the percentage this category takes up. The category with the highest percentage is the predominant category within that street segment.

The “Summary Statistics” tool helps us find the predominant land use in each segment. It looks at each unique SegmentID and finds the maximum-percentage-land-use for each. Resulting is a table we call with one row per street segment and a field showing the highest land use percentage for that segment. However, it does not include the corresponding land use itself. Consequently, another step is necessary: using the “Join Field” function, we add the Land Use category to the table of land use percentages, and called it streets_landuse_predominant.

This step essentially used the calculated percentages as join fields to retrieve the Land Use values that have the highest percentage. Because the percentage values have six digits after the decimal, they essentially act like unique identifiers, making them an appropriate join field. Nevertheless, we checked for duplicates and deleted the few that were found.

We joined "streets_landuse_predominant" and "streets" into a feature class called "streets_landuse" and visualized streets by predominant land use column. Next, we grouped the “null” and unavailable land use categories into a single new category called “99 - Land Use Unavailable," and then assigned symbology labels describing the land use to the all other numerical categories.

Via Select by Attributes on "streets_landuse," we located streets where direction is equal to one-way. We exported the selection as a new shapefile called  "streets_landuse_OW." We underlaid "streets_landuse_OW" with "streets" and changed their symbology in order to visualize.

Using the “Generate Tessellation” tool (and the community district boundary layer for the extent), we created a Hexagon layer covering NYC. We then used the “Tabulate Intersection” tool to calculate the shares of One-Way Street land use categories per hexagon. The tool returns a table with multiple rows per hexagon, with each row representing a land use category present within that hexagon, and the length of the street segment of that category within that hexagon. The category with the highest length is the predominant category within that hexagon.

The “Summary Statistics” then let us find the predominant land use in each segment. This looks at each unique GRID_ID (case field) and finds the maximum length for each. Resulting is a table with one row per hexagon, and a field showing the highest land use length for that segment. However, it does not include the corresponding land use itself. Consequently, another step is necessary: using the “Join Field” function, we added the predominant Land Use category to the table. This step essentially used the calculated lengths as Join Fields, to retrieve the Land Use values that have those same street lengths from the initially calculated table.

Next, we joined the table for predominant land use in each hexagon with the hexagon layer. One thing to note, is that not all hexagons have values for landuse, since not every hexagon intersects with One-Way Streets. We updated the new hexagons land use layer using the field calculator, to rename the fields to “99 - No One-Way Streets," and also changed the symbology to Unique Values based on land use. We opted to go with the same colors as as we did in the map above.

To produce some summary statistics, we used the “streets_landuse” layer, which includes both one-way and two-way streets. For any meaningful values, we need to look at both together.

We exported the attribute table of “streets_landuse” and, in a Python script, aggregated the data to a table showing the percentage one-way street segments and the percentage two-way street segments make up of the total streets segments for a given land use.  That way, we can compare if there is any land use were one way streets are predominant. Lastly, we visualized the table using the datawrapper.de website

RQ 2 Findings: The one land use for which one-way streets are more common than two-way streets is Multi-Family Walk-Up Buildings.

(1) Buffer Width

Our street buffer width excluded some streets that were too wide, as they didn’t reach the land use polygons. This was the reason for "null" values. But note that making the buffer wider would, in other cases, have led to land uses from the next street over being incorrectly included. So it was a judgment call.

(2) Only Taking the Predominant Land Use

We only took the predominant land use per street. Even if streets were very diverse in terms of land use, we only took the one occurring most in terms of percentage. This doesn’t always reflect the true land-use-landscape of a street.

(3) Street Filtering

We could have filtered streets in a more detailed analysis. These are, however, somewhat uninteresting since they are specialized streets. The analysis would be more refined if we got rid of those types of streets. That would, however, have required a much more complex data cleaning process, which was simply out of the scope of our project. 

(4) Land Use Unavailable

The second highest number of street segments after One- and Two-Family Homes, were segments for which no land use was available, either because of the aforementioned buffer issue, or because there was simply no land use recorded.

Ideas for future projects:

(1) Find a way to incorporate all land uses on each street segment into the analysis, not just the predominant one.

(2) Look at other patterns in the context of one-way streets; land use is just one example.

(3) Compare NYC to other cities around the world in terms of their one-way street distribution and characteristics.

Transferable Lessons:

(1) Studies of one-way street distribution and characteristics that yield new insights are very much possible; GIS is the way to tackle them.

(2) Simplified analysis can still lead to powerful insights, as long as the questions being asked are well thought-out.

Badger, E. (2015, April 17). Why one-way streets are bad for everyone but speeding cars. The Washington Post.  https://www.washingtonpost.com/news/wonk/wp/2015/04/17/why-one-way-streets-really-are-the-worst/ . 

Gray, C. (1997, December 7). Streetscapes/Readers' Questions; 1-Way Streets, a 1902 Building, a Notable Brownstone. The New York Times.   https://www.nytimes.com/1997/12/07/realestate/streetscapes-readers-questions-1-way-streets-1902-building-notable-brownstone.html .

NYC OpenData. (2023). Community Districts [Shapefile]. Department of City Planning.  https://data.cityofnewyork.us/City-Government/Community-Districts/yfnk-k7r4 . 

NYC Planning. (2023). LION Single Line Street Base Map (Version 23D) [Geodatabase]. Department of City Planning.  https://www.nyc.gov/site/planning/data-maps/open-data/dwn-lion.page . 

NYC Planning. MapPLUTO - Shoreline Clipped (Version 23v3) [Shapefile]. Department of City Planning.  https://www.nyc.gov/site/planning/data-maps/open-data/dwn-pluto-mappluto.page#mappluto . 

Oliver, L.N., Schuurmann, N., & Hall, A.W. (2007). Comparing circular and network buffers to examine the influence of land use on walking for leisure and errands. International Journal of Health Geographics, 6.  https://doi.org/10.1186/1476-072X-6-41 . 

Stemley, J.J. (1998). One-Way Streets Provide Superior Safety and Convenience. ITE Journal.  https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=3185c96965b69d4289dcaa18826f88cf9c67097a .