Introduction

Urban flooding has affected most developed cities due to rapid urbanization and climate change. Urban flooding is the accumulation of flood water within a city or town when the inflow of stormwater exceeds the capacity of the drainage system. Intense rainfall, which is becoming more frequent due to climate change, adds to the risk of urban flooding. The excess rainwater can flow into the basement of the infrastructures and block the sewer lines, therefore causing sewer backups. Cities with older sewer systems and large population, for example NYC, are particularly prone to urban flooding and the problems arise from it such as sewer overflows in residential areas (Lane et al., 2013).

In this research project, we focused on demographic and geographic factors that may exacerbate the risk of sewer overflows. The demographic factors we included are population density and median income. Population density refers to the number of people per square mile of land area. Some studies have shown a correlation between urban flooding and population density (Framing the challenge of urban flooding in the United States, 2019). We predicted that higher population density would be related to more sewer backups. Median income is the median household income for each census tract. Urban flooding has been shown to impose a greater impact on low-income communities (Haque et al., 2020). Based on this study, we predicted that low-income neighborhoods would have more sewer backups, potentially due to aging and deteriorating infrastructure.  

The geographic factors we assessed are land use, imperviousness, and elevation. Urban growth occurred rapidly over the past century; this process increased imperviousness and changed land use. Impervious surfaces, such as buildings, roads, and parking lots, increase surface runoff that can be otherwise absorbed by the natural landscape. More open space is replaced by developed areas where people reside and work in high numbers. The increase in land-use change can increase the incidence of flood inundation (Dammalage and Jayasinghe, 2016). Elevation is defined as the height above sea level. Climate change has caused urban flooding to occur more frequently at low elevations. Because gravity drainage flow networks use gravity to drain the water away, the rate of the drainage depends on the elevation difference between receiving and outpouring sites. In areas with low elevation, elevation differences can slow or reverse the rate of drainage (Habel et al., 2020). Additionally, studies have shown climate change results in sea-level rise, which increased the number of “nuisance floods” for areas with low elevation (Strauss et al., 2016; Sweet and Marra, 2014). 

APPROACH

In order to test our predictions, we investigated the number and the location of sewer backups from 311 call data from NYC OpenData. NYC311, developed by the NYC Department of Environmental Protection, allows property owners to report a sewer backup in the building. After a call is made, NYC311 records the date and location of the call.  

In this project, we evaluated the relationship between 311 calls and various parameters, including population density, median income, imperviousness, elevation, flood zone, and land use. Population density and median income data are from the US Census (US Census Bureau, 2011). Imperviousness and land use data are from the National Land Cover Database (MRLC, 2016). Elevation data is from The USGS National Map (USGS), and flood zones are from the FEMA National Flood Hazard Layer from the Flood Map Service Center (FEMA). We compared the values of these attributes at the 311 call data locations with a random spatial distribution to identify the potential impact of different factors on the number of sewer backups. We calculated the minimum, maximum, mean, and standard deviations of each attribute at the 311 call locations and across the random spatial distribution. To determine whether there was a significant difference between the 311 call locations and random locations, we compared the two datasets using Student’s t-tests. Student’s t-tests are a statistical test used to determine whether a statistical difference exists between two groups. 

Sanborn Fire Insurance Maps from the website of the Library of Congress were reviewed in an attempt to identify the age of the pipes at different parts of the city. This source, however, does not provide us with a specific age of the sewer pipes. We also lack prior research studies on the correlation between age and the risk of sewer backups. 

FINDINGS

311 calls by borough

First, we grouped our 311 call location data points by their boroughs. From 2011 to 2020, after filtering the data to only 311 calls that were for standing water, we found a total of 1,401 311 calls. Of these calls, 411 were from Brooklyn, 397 were from Queens, 284 were from Manhattan, 162 were from the Bronx, and 149 were from Staten Island.

Staten Island has the highest number of calls per capita. Manhattan has the highest number of calls per square mile.

Figure 1. Map of 311 data points and randomized data points divided by borough.
Figure 2. 311 calls per capita for each borough.
Figure 3. 311 calls per square mile for each borough.

Relationship between 311 calls and map parameters

Population Density

Next we looked at each attribute that may be related to 311 calls. We first looked at the population density of the areas that the points were located in. We predicted that population density can have a significant impact on increasing the risk of sewer backups. For population density, we found significant differences between the two datasets for NYC overall and in all boroughs. 311 calls tend to locate in areas with higher average population density. . This suggests the likelihood of 311 calls occurring in a higher population density area was statistically significant.

Figure 4. Map of population density, 311 data points, and randomized data points.
Figure 5. Distribution of 311 data points and randomized data points based on population density.
Figure 6. Mean ± Standard deviation and p-value of 311 data points and randomized data points based on population density

Median Income

We then looked at the median income of the areas that the data points were located in. Median income may have an impact on the risk of sewer backups by affecting the quality of the infrastructures. For median income, we found significant differences between the two datasets for NYC overall and Brooklyn. Contrary to our prediction, 311 calls tended to locate in areas with higher median income. This suggests the likelihood of 311 calls occurring in higher median income was statistically significant.

Figure 7. Map of 311 data points and randomized data points based on median income.
Figure 8. Distribution of 311 data points and randomized data points based on median income.
Figure 9. Mean ± Standard deviation and p-value of 311 data points and randomized data points based on median Income

Imperviousness

We then can look at the imperviousness of the areas in which the data points are located. Imperviousness can increase the amount of surface runoff, thus increasing the risk of sewer backups. For imperviousness, we found significant differences between the two datasets for NYC overall and in Staten Island, Queens, and Manhattan.

311 data points in these three boroughs have a higher average imperviousness than the random spatial distribution. This suggests that the higher likelihood of 311 calls occurring in areas with greater imperviousness was statistically significant.

Figure 10. Map 311 data points and randomized data points based on imperviousness.
Figure 11. Distribution of 311 data points and randomized data points based on imperviousness.
Figure 12. Mean ± Standard deviation and p-value of 311 data points and randomized data points based on imperviousness.

Elevation

The elevation of the housing can affect its risk of sewer backups. For elevation, we found significant differences between the two datasets for Staten Island, Brooklyn, and Manhattan. In Staten Island, 311 calls tend to come from areas with lower elevation. Alternatively, in Brooklyn and Manhattan, 311 calls tend to come from areas with higher elevation.

Figure 13. Map of 311 data points and randomized data points based on elevation.
Figure 14. Distribution of 311 data points and randomized data points based on elevation.
Figure 15. Mean ± Standard deviation and p-value of 311 data points and randomized data points based on elevation.

Flood Zone

We also looked at the flood zone of each data point to determine if the risk of flooding correlates with the risk of sewer backups. We found that 311 data points are more likely to occur in a high-risk flood zone for NYC overall and in Staten Island, Queens, Brooklyn, and Manhattan in particular. For the Bronx, the percentage of points located at high flood risk zones is lower for 311 data points. Overall, this data shows 311 calls have a higher likelihood of occurring in a high-risk flood zone.

Figure 16. Map of 311 data points and randomized data points in each flood zone.
Figure 17. The percentage of the 311 data points and randomized data points in each flood zone across NYC.
Figure 15. The percent distribution of 311 data points and randomized data points based on flood zones in each borough.

Land Use

Lastly, we analyzed the type of land use in which that the data points were located. We found that 311 data points have a lower likelihood of being located in developed-low intensity areas for NYC overall and in all boroughs. We also found that 311 data points show a higher likelihood of occurring at a developed, high-intensity area for NYC overall and in Brooklyn and Manhattan.

Figure 16. Map of 311 data points and randomized data points in each type of land use.
Figure 17. The percentage of the 311 data points and randomized data points in each type of land use across NYC.
Figure 15. The percent distribution of 311 data points and randomized data points based on flood zones in each borough.

Historical data

We also predicted that the age of the sewer system can affect the risk of sewer backups. We looked at the time period when each borough began to develop. We found that, compared to other boroughs, urbanization of Staten Island and Queens began rather late.

Figure 6. Sanborn Fire Insurances map from Manhattan 1894.
Figure 7. Sanborn Fire Insurance Map from Brooklyn 1887.
Figure 8. Sanborn Fire Insurance Map from Staten Island 1885.

NYC

The construction Sewer system first began in 1849 during a series of cholera outbreaks. Between 1850 and 1855, the city laid 70 miles of sewer. By the mid to late 19th century, the network expanded throughout the developed section of the city. Most of these original sewage pipes are still in use today.

Manhattan

In the 19th century, Manhattan was classified as a manufacturing and industrial city. As the Sanborn Fire Insurance Maps showed, Manhattan was largely populated since the 1800s. After the Civil War, a growing number of immigrants from Europe chose to settle in Manhattan. Between 1800 and 1910, density in urban Manhattan tripled from 200 to 600 people per hectare.

Brooklyn

As the Sanborn Fire Insurance Maps have shown, in the late 1800s residents of Brooklyn mostly concentrated along the eastern coast. By the end of the 19th century, as a result of European migration, the population of Brooklyn reached 1 million people. In 1883, the opening of the Brooklyn Bridge brought a new wave of people into Brooklyn.

Bronx

In the early 19th century, the population rose from 1,755 to 3,023 as a result of Irish migration. During the construction of the High Bridge over the Harlem River (1837-48), the New York, Harlem Railroad (1841), and the Erie Canal (1825), more people choose to live along the shoreline of the Bronx.

Queens

During the 1800s, the area remained mostly agricultural. In the late 19th century, the construction of the subway lines and opening of The Queensboro Bridge allowed the population of Queens doubling in a decade, from less than 500,000 in 1920 to more than one million in 1930.

Staten Island

In the late 1800s, the Sanborn Fire Insurance Map only documented the coastal areas of Staten Island. Before the 2000s, Staten Island was less densely populated compared to other boroughs. It wasn’t after the construction of the Verrazano-Narrows Bridge, that the population doubled from 221,991 in 1960 to 443,728 in 2000.

Rainfall

Precipitation can contribute to sewer backup by overwhelming the capacity of the sewer line. For precipitation, we found that the months with the most rainfall somewhat overlap with the months with the most 311 calls. June and August experienced the highest percentage of 311 calls. 22.58% in June and 22.51% in August (Figure 10). On average NYC experiences the highest average monthly precipitation in May and July. This indicates that 311 calls tended to occur in times with higher monthly rainfall.

Figure 9. The percentage of 311 data points for each month.
Figure 10. Average Monthly Rainfall

Discussion

Population Density

The student’s t-test results support our hypothesis that the number of sewer backups are significantly related to the population density. One of the explanations for this is that the more people using the sewer lines, the more likely sewer backups to occur. To further understand this correlation, we would need to evaluate similar datasets captured from other cities across the U.S. As urbanization continues and the population continues to increase, sewer backups are likely to happen more often.

Median Income

Our result does not consistently support our hypothesis that lower median income correlates with more sewer backups. After looking at previous studies, which employed statistical analysis and questionnaire surveys to collect information on the demographics of the surveyed population (Satterthwaite et al. 2007), future research might include the income information of the households that made the 311 calls, instead of looking at the median income of the community.

The level of income inequality might contribute to the reason why only in some boroughs the number of sewer backup are affected by median income. Queens, Bronx, and Staten Island have lower levels of income inequality, thus income may be less likely to affect the number of sewer backups (Conway 2017).

Having more precise information on the household that made the income might lead to different conclusions. Studies have also shown that climate change are likely to increase the risk of flooding for low-income households, because affordable housing was concentrated in floodplains (Cusick 2020).

Imperviousness & Land Use

For imperviousness, our results support the hypothesis that higher imperviousness increases the risk of sewer backups. In case of heavy rain, imperviousness increases surface runoff causing sewer backups. For land use, our result does not correspond with our prediction that sewer backups are more likely to occur in developed, high-intensity areas. To better understand the relationship between land use and sewer backups, we could compare the 311 data from different periods of time, the time before the land-use change took place and the time after land-use took place (Dammalage and Jayasinghe 2016). In addition to affecting the risk of sewer backups, the increase in imperviousness can also trigger urban flooding when there’s heavy rain (Frazer 2005). Ways to mitigate risk from increased imperviousness include increasing green space and changing the building materials of buildings (Frazer 2005).

Elevation & Flood Zone

For elevation, our results support the prediction that sewer backups are more likely to occur in areas with low elevation. Our results do not support the prediction that sewer backups are more likely to occur at a high-risk flood zone. The two datasets have similar distributions across different flood zones for all of NYC. For some boroughs, elevation does not correlate with 311 calls. This might be due to the fact adaptation was taken at areas of low elevation and high-risk flood zone. For example, housing located at low elevation areas may use more pervious material and have a better-quality sewer system (EPA). Through our background research, we also found that climate change will likely lead to an increase in sea level, and low elevation areas along coasts are likely to experience a higher risk of flooding (Vitousek et al. 2017).

Conclusion

In the first stage of the report, we reviewed research on urban flooding and sewer backups, and we identified the factors that might affect the likelihood of sewer backups occurring in NYC. The factors that we focused on included: population density, median income, imperviousness, elevation, flood zone, and land use. 

We then evaluated how these attributes are related to locations of sewer backups. The dataset that we used is the 311 call dataset, which tells the location of the sewer backups. We compared this to a random spatial distribution across NYC. We categorized our data by boroughs and by NYC overall. First, we identified the minimum;  Q1 – quartile 1, the median of the lower half of the data set; Q2 – quartile 2, the median of the entire data set; Q3 – quartile 3, the median of the upper half of the data set; and the maximum. We then created box plots to visualize our data and identify the mean and standard deviation of each dataset for each factor. With this information, we used student’s t-tests to determine the significance between the datasets for each factor.  

According to our analyses, the factors that contribute to the number of sewer backups vary between boroughs:

For NYC overall, sewer backups correlate with population density, median income and imperviousness. 

Population density contributed most to the number of 311 calls about sewer backups. We found significant differences between the two datasets in NYC and for each of the boroughs. 

This study helped us better understand which factors were related to sewer backups. This is important because sewer backups can cause damage to buildings and infrastructures as well as contamination and spread of waterborne diseases.   

References

311 Service Requests from 2010 to Present | NYC Open Data. data.cityofnewyork.us/Social-Services/311-Service-Requests-from-2010-to-Present/erm2-nwe9. https://data.cityofnewyork.us/Social-Services/311-Service-Requests-from-2010-to-Present/erm2-nwe9

“Adaptation Actions for Water Utilities.” EPA, Environmental Protection Agency, 19 Nov. 2020, https://www.epa.gov/arc-x/adaptation-actions-water-utilities

Bourman, Robert P.: The nature of flooding 2010, 52-53. https://ro.uow.edu.au/scipapers/708

Conway, Author Nick. “Inequality in New York City Neighborhoods, 1990-2015.” Nick Conway, 8 Aug. 2017, nickconwayblog.wordpress.com/2017/08/08/inequality-in-new-york-city-neighborhoods-1990-2015/. https://nickconwayblog.wordpress.com/2017/08/08/inequality-in-new-york-city-neighborhoods-1990-2015/

“Climate for New York Central Park, New York.” Climate in New York Central Park, New York, http://www.rssweather.com/climate/New%20York/New%20York%20Central%20Park/

Dammalage, T. L., & Jayasinghe, N. T. (2019). Land-Use Change and Its Impact on Urban Flooding: A Case Study on Colombo District Flood on May 2016. Engineering, Technology & Applied Science Research, 9(2), 3887-3891. doi:10.48084/etasr.2578 https://etasr.com/index.php/ETASR/article/view/2578

Cusick, Daniel. “Flood Risks to Low-Income Homes to Triple by 2050.” Scientific American, Scientific American, 1 Dec. 2020, www.scientificamerican.com/article/flood-risks-to-low-income-homes-to-triple-by-2050/.

“Federal Emergency Management Agency.” USAGov, https://www.usa.gov/federal-agencies/federal-emergency-management-agency

Framing the challenge of urban flooding in the United States. (2019). Washington, DC: The National Academies Press. https://www.nap.
edu/read/25381/chapter/1

Habel, Shellie, et al. “Sea-Level Rise Induced Multi-Mechanism Flooding and Contribution to Urban Infrastructure Failure.” Scientific Reports, vol. 10, no. 1, 2020, doi:10.1038/s41598-020-60762-4. https://www.nature.com/articles/s41598-020-60762-4

Haque, A. N., Bithell, M., & Richards, K. S. (2020, April 17). Adaptation to flooding in low‐income urban settlements in the least developed countries: A systems approach. The Geographical Journal, 186(3), 314-326. doi:10.1111/geoj.12348 https://rgs-ibg.onlinelibrary.wiley.com/doi/pdf/10.1111/geoj.12348

“Home: Library of Congress.” The Library of Congress, loc.gov/. https://loc.gov/

Lane, Kathryn, et al. “Health Effects of Coastal Storms and Flooding in Urban Areas: A Review and Vulnerability Assessment.” Journal of Environmental and Public Health, Hindawi, 30 May 2013, www.hindawi.com/journals/jeph/2013/913064/.

Frazer, Lance. “Paving Paradise: The Peril of Impervious Surfaces.” Environmental Health Perspectives, vol. 113, no. 7, 2005, doi:10.1289/ehp.113-a456. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1257665/

“National Flood Hazard Layer.” National Flood Hazard Layer | FEMA.gov, https://www.fema.gov/flood-maps/national-flood-hazard-layer

“National Geospatial Program.” The National Map, https://www.usgs.gov/core-science-systems/national-geospatial-program/national-map

National Land Cover Database (NLCD) 2016 | Multi … https://www.mrlc.gov/national-land-cover-database-nlcd-2016

Strauss, B. H., Kopp, R. E., Sweet, W. V. and Bittermann, K. Unnatural Coastal Floods: Sea Level Rise and the Human Fingerprint on U.S. Floods Since 1950. Climate Central Research Report, pp. 1-16 https://sealevel.climatecentral.org/uploads/research/Unnatural-Coastal-Floods-2016.pdf

Sweet, W. & J. Marra. 2014. State of Nuisance Tidal Flooding. National Oceanic and Atmospheric Administration. https://www.ncdc.noaa.gov/monitoring-content/sotc/national/2015/aug/sweet-marra-nuisance-flooding-2015.pdfel

Population growth exacerbates flood risk https://populationmatters.org/sites/default/files/population_growth_exacerbates_flood_risk.pdf

US Census Bureau Public Information Office. Census Bureau Releases 2011 American Community Survey Estimates – American Community Survey (ACS) – Newsroom – U.S. Census Bureau, 19 May 2016, https://www.census.gov/newsroom/releases/archives/american_community_survey_acs/cb12-175.html