Introduction
New York declared state of emergency on March 7, 2020 in response to the spread of SARS-CoV-2. Social distancing and lockdown measures soon followed this order to curtail the spread of the virus. Around March 20, a state-wide stay-at-home order was declared, and all non- essential businesses were ordered to close, including restaurants. As the measures stretched out through 2020, media and pest management authorities suggested increased sightings of rats, including aggressive behaviors during daylight hours, and in close proximity to people, as well as an increase of lethal competition between rats (e.g. muricide) and cannibalism in New York City (NYC) (as cited in Parsons et al., 2021).
A 2014 study estimated that there were around 2 million rats (±150 000) in New York City (Auerbach). These rats are commensal rats, which include Norway rats, roof rats, and house mice, and they live off humans and animals. They cost billions of dollars to society due to contaminated foods, the spread of disease-causing pathogens, infrastructure damage, and fires from electrical short-circuits caused by gnawing every year (Parsons et al., 2021; Bedoya-Pérez et al., 2021). Because urban rats depend on human as resources, their population is bound to be affected by the changes in human behavior following natural catastrophes. Thus, it is important to monitor and study changes in urban rodent’s population and behavior in order to effectively manage these pests.
This project aims to visualize rodent activity changes before and during the COVID-19 pandemic lockdown since March 2020, especially concerning restaurants closing and people staying home. It also looks at the amount of rodent sighting-related calls compare to other types of 311 complaints, as well as call volume distribution through different time periods of the day to further discuss what are some of the possible causes of this increase in rodent activity change.
Design Decisions
This was a continuation of one of my previous projects. The previous project merely used R Studio to create heat maps representing the rodent report, housing, and restaurant hotspots on the New York City map, which made it hard to see and understand. It was also in the form of a report, which was long and hard to read. Now knowing how to use geospatial data and software like QGIS, I decided to produce a series of infographics accompanied by short paragraphs, so that people can see the changes with more clarity. With more visuals than words, it will also benefit those who are less fluent in technical terms and attract audiences.
With these goals in mind, I decided to use simple and easily understandable color pallets, combine with shapes and lines to guide people’s eyes to navigate the main map. For the second part of the map, 2 other distinct colors were chosen to distinguish between its goals, and slightly longer paragraphs were written to explain the findings, especially when it comes to data analysis.
Methodology
Data Collection
After deciding to use the 311 Service Request dataset available on NYC OpenData website, the lockdown period was selected and the duration of the lockdown was calculated. The lockdown went from Mar 7, 2020 to Jun 23, 2021, which was 474 days. Then the same time period was subtracted from the day before the lockdown to determine the first date before the lockdown period, which was Nov 18, 2018.
Rodent-related 311 calls are classified into five categories: signs of rodents, conditions attracting rodents, rat sighting, mouse sighting, and rodent bites. For the purpose of this project, the complaint reasons were limited to only rat and mouse sightings so that the focus is narrowed.
The two datasets were then uploaded to OpenRefine so that all the irrelevant columns and null values can be removed. Because NYC Open Data display dates and time in the column, the column was also separated into two.
I then grouped all rodent sighting-related calls to 3 different time periods of the day: 12 am to 8 am, 9 am to 4 pm, and 5 pm to 11 pm. This allowed me to generate a line graph showing the trend of when people are reporting rodent sightings during the data, as well as running a Chi-squared test to see if there is a correlation between the rodent sighting-related complaints about these time periods before and during the lockdown.
In order to see the larger picture, the total amount of 311 calls collected during the 2 time periods was also collected. After finding out the top 9 types of requests for the entire dataset, the numbers for each of these categories were also collected before and during the lockdown. This allowed me to put the rodent sighting calls into context, producing a bar chart showing other changes besides this one category.
The housing and restaurant datasets were also found on NYC Open Data. The housing one was pretty complete, but the restaurant one also went through cleaning in R Studio because it was the inspection data. Since restaurant inspections occur every year and each restaurants can have multiple violations, repetitive restaurant names were being tripped so the remaining is a complete dataset of all the restaurants and their locations.
Mapping
I decided to scratch the heat map idea from the beginning because I know it’s not detailed enough to show people where and how are the changes occurring. The census tract on the other hand is small enough to display hot spots, but also big enough with boundaries to show zones.
Take the before lockdown dataset for rodent complaints as an example of my mapping process. After importing the datasets to QGIS, I used the census tract as my base layer. I then have the software showed all the dots representing each report and ran the Count Points in Polygon analysis so that it can calculate within each area of the census tract, how many rodents were spotted. This calculation then allows me to color each census area in NYC based on how many reports were in that area using a graduated color scheme, with red being the highest and light yellow being the lowest. I then normalized the data by deciding the number of reports in each polygon by the area (per square kilometer). However, since there are extreme outliers, I decided to use a more popped up color to show them since I don’t what it to interfere with the classification of the rest of the values.
After performing the same steps on during lockdown dataset, I joined the normalized values of rodent reports for both datasets into one with the original census tract file. I then perform a calculation by subtracting the before values from the during values, in order to show the changes in numbers in each area. Using white for 0, grey for negative values (decrease in rodent reports), and red or positive values (increase in rodent reports), distinctive colors for outliers, as well as grey lines for areas with no reports, I was able to complete my map to show rodent activity changes for before and during the lockdown.
I then used the same methods for the housing the restaurant maps, except there is no calculations involved.
Visualizing the Bigger Picture
There were 3053395 311 requests filed for the 474 days before the lockdown and 3481760 for during, so there is definitely an increase in people calling 311. A CSV files were created to visualize 9 different categories plus rodent complaints during each time period in Tableau. This is not a particularly hard graph to do, but it’s serves an important role as it showed patterns in the increased categories. The categories were picked from the top to bottom in terms of the entire dataset, so that the most popular complaints can be represented.
Another graph was created for the number of reports being filed during different periods of the day. A line graph was used to specifically show the trend.
User Testing
User testing contributed to a huge part of the project, especially informing the way that I designed the infographic posters. It was done right after the first draft of the poster, which was when it looked like this:
There are a few ideas that were on the page, and the main goal for this page was to use the page to tell a story of what the changes were and what may be some contributing factors. I did a think-aloud type user testing with 2 people: none of them are familiar with the topic of my project and both of them have some sort of experience with GIS and mapping. I developed a few questions to pick their heads, and the result was very useful.
Question 1: After looking at my project, can you describe what’s it about?
Question 2: Do the shapes and line on the graph provide any guidance?
Questions goal: See if the layout make sense and if they understand the narrative.
Findings: One person pointed out that the main introduction should be on the top instead of the upper left side, and pointed out that I am missing a legend. The other person said that the color on the map made sense, but even though lines do direct where they look at, they are kind of everywhere. They suggested that list out all the major areas on the left side horizontally instead of putting them everywhere, and believed that it would bring more structure to the page.
Question 3: Are there too many words on this page?
Question goal: I was really worried that there were too much explanations for people to read so they can’t focus on the visuals as much.
Findings: Both of them said that the contents were good and important for them to understand the story here, but they could be shortened with a neater layout.
Question 4: What parts of the story am I missing? What other information you wish to see?
Question goal: I knew this was a rough draft, and I wanted to put more information on there, but I want to see if what they are thinking is what I am thinking.
Findings: They were really interested in my ideas of compare these major changes to the locations of the restaurants and housing areas. They believed that it would add a lot to the narrative. They are also interested in running some actual data analysis to see if these changes are correlated with each other. One person who is more familiar with data analysis and R Studio offered some methods, and we tried running some tests on site. However, because my dataset is too large, it was impossible to do so.
Besides this formal user testing, many other people offered ideas through other revenues. My GIS professor pointed out that I should show the null values (areas with no rodent-sighting reports). My data visualization professor suggested that I look at the totally amount of reports as well as when are the reports being filed during the day. Some of my other friends also looked at my poster every time I made changes to provide feedbacks. All of these inputs were valuable and made huge, positive changes to the final product.
Findings
Map
Three major findings showed that there is very unclear relationship between the number of rodents related reports and the number of housing/restaurants at an area. For upper Manhattan, rodent sighting reports are moving from midtown to uptown, especially around the Central Park area. With a higher housing concentration for uptown than restaurants, it shows that rodents are moving away from closed restaurants for alternative food sources. As for the concentrated spot that occurred in the Bedford, Clinton Hill, and Crown Heights area, even though it’s surrounded by declined reporting, it is hard to tell if it has to do with the number of housings or restaurants. For Staten Island, the sightings moved to the out rim of the island. However, there is one interesting spot that has an high increase in reporting number, but low housing units. There is a high concentration of restaurants, but considering most of them were closed, it’s hard to say what caused the increase.
Graphs
For the comparison of other types complaints before and during the lockdown, the biggest increases for complaint types comes from noise complaints, especially for residential and street/sidewalk area, and the decreases are from categories like blocked driveway and street condition. This reflected the lockdown. As more people were staying home stead of going out, working from home instead of going to work, issues that are more connected to living conditions picked up. As rodent complaints are one of them, it make sense that its numbers increased during the lockdown.
For the breakdown of rodent sighting-related calls based on the time of the day, the pattern of calling time changed drastically during the lockdown compared to before.
Before the lockdown, most rodent sighting-related calls come from 12 am to 8 am, when people are asleep, but it becomes the lowest during the lockdown. 9 am to 4 pm when most people are awake and active during the day became the new peak, compared to before the lockdown when it declined as most people are working. At the end of the day between 5 pm to 11 pm is still the lowest amount of the 3 time periods of the day, but the lockdown still caused it to increase in number.
A Chi-squared test was also performed to see if there is a correlation between the rodent sighting-related complaints for the 3 time periods before and during the lockdown. With the p-value < 2.2e-16, which is smaller than 0.05, this means that there is a strong correlation, but no conclusion on whether it’s positive or negative.
Recommendations
I enjoyed doing this project and would say that there are major improvements compared to the similar one I did last semester. My skills in using QGIS helped me a lot, as well as the knowledge of organizing and presenting data visualizations that I learned in this class. However, there are a few things that could be done in the future to make it more scientifically solid.
The first one is limiting the lockdown period. Right now we are looking at 2 datasets each containing 474 days of data. These are big numbers, and as people pointed out, don’t reflect the lockdown that we went through. The harshest rule of lockdown was imposed in NYC for about 2 months, and that was when human activity really dropped to almost zero. As Professor Chris Sula pointed out, it could be more meaningful to look at this time period, since there might be a more distinct pattern emerging from it.
The second one is to run a spacial analysis regarding the locations of housing, restaurants, and rodent sightings. Right now, all the correlations are done just based on numbers, but this project could benefit from spacial analysis, and this may reveal more information on how these are connected.
The third recommendation is given by my classmate Micah Musheno. She pointed out in class that the sighting could also be connected to more trash being produced by households during the pandemic and the fact that building management and the city didn’t take care of them effectively. Of course, rodents feast on trash, so if we were to look at the increase in the number of trash being produced, where they were being kept, and then rodent sighting, we might discover correlations as well.
References:
311. “311 Service Requests from 2010 to Present”. 2010-2022. NYC OpenData, https:// data.cityofnewyork.us/Social-Services/311-Service-Requests-from-2010-to-Present/ erm2-nwe9
Auerbach, Jonathan. “Does New York City Really Have As Many Rats As People?”. Significance, vol 11, no. 4, 2014, pp. 22-27. Wiley, doi:10.1111/ j.1740-9713.2014.00764.x.
Bedoya-Pérez, Miguel et al. “Hunger Pandemic: Urban Rodents’ Boom And Bust During COVID-19”. 2020. Research Square Platform LLC, doi:10.21203/rs.3.rs-111434/v1.
Department of Health & Mental Hygiene. “311 Service Requests from 2010 to Present”. 2014-2022. NYC OpenData, https://data.cityofnewyork.us/Health/DOHMH-New-York- City-Restaurant-Inspection-Results/43nn-pn8j
The Department of Housing Preservation and Development (HPD). “Housing New York Units by Building”. 2022. NYC OpenData, https://data.cityofnewyork.us/Housing- Development/Housing-New-York-Units-by-Building/hg8x-zxpr
Parsons, Michael H et al. “Rats And The COVID-19 Pandemic: Considering The Influence Of Social Distancing On A Global Commensal Pest”. Journal Of Urban Ecology, vol 7, no. 1, 2021. Oxford University Press (OUP), doi:10.1093/jue/juab027.
Parsons, Michael H. et al. “Rats And The COVID-19 Pandemic: Early Data On The Global Emergence Of Rats In Response To Social Distancing”. 2020. Cold Spring Harbor Laboratory, doi:10.1101/2020.07.05.20146779.
The Department of Housing Preservation and Development (HPD). “Housing New York Units by Building”. 2022. NYC OpenData, https://data.cityofnewyork.us/Housing- Development/Housing-New-York-Units-by-Building/hg8x-zxpr