Fork me on GitHub

Project Undergrad Team #5: 2022 MBTA Ridership and Precipitation, COSI 116A F23

Cindy Chi, Christina Lin, Sydney Cohen, Vu Le

Project-long Course Project as part of COSI 116A: Information Visualization, taught by Prof. Dylan Cashman, Brandeis University.

Motivation

In the beginning, our group chose an MBTA ridership dataset and wanted to look at the relationship between weather events and certain MBTA ridership. We started off with the goal of looking at how the 2022 MBTA Transit data changes with respect to month, season, and/or precipitation. Because the seasonal weather changes can be so drastic in Boston, especially with the cold and icy weather in the winter, we predicted that there would be some correlation between the number of people riding the MBTA and the month, season, and precipitation data. Creating visualizations that compare ridership to weather data would allow viewers to better understand how ridership relates to months, seasons, and precipitation. Additionally, it is crucial to understand factors that impact ridership in evaluating the accessibility and usage of the T, and also when looking at what changes can be made in the future to best offset seasonal and weather impacts, if needed.

Visualization

Ridership and Precipitation by Month 2022:

MBTA Rail Ridership by Month:

Monthly Total Precipitation (inches):

Demo Video

Visualization Explanation:

visualization explanation

This visualization, which combines a bar graph for MBTA rapid transit ridership, a line plot for average daily precipitation (in inches) at the monthly time scale shaded in transparent light blue, a linked pie chart for ridership breakdown by different MBTA lines, and a small graphic representation of precipitation level, is a powerful and comprehensive way to convey information about relationship between precipitation and ridership in the MBTA rapid transit system. The bar graph effectively captures ridership data, allowing for easy comparison between different lines or time periods. We first created a table that breaks down total MBTA ridership and total precipitation in inches by month, specifically focused in Boston. We also present a bar graph for ridership with the line plot of precipitation, and then we allow users to click on individual bars to further visualize the ridership breakdown of each line in a pie chart. We also added an additional graphical representation of precipitation, a cylinder showing the total precipitation level for a month. By comparing it to an average length Boston rat, which is around 7 inches, viewers can understand the water level in a more creative way, which connects back to the MBTA and Boston theme as well. This visualization is more relatable and provides some physical context to the precipitation levels. This approach aligns with the theory that incorporating a personal connection into data visualizations makes them more meaningful and engaging for our audience, further allowing them to have a better understanding of the complex relationship between ridership and water levels in the public transportation system MBTA.

Data Analysis

We started by taking the MBTA MBTA Monthly Ridership by Mode and Boston Weather Summary of the Month dataset We then filtered both of these datasets, starting by checking for missing data, and then selecting the data for January to December of 2022. For the MBTA Rail Ridership dataset, we specifically filtered for the mode to be Rail, since the data set also includes other modes of transportation like bus, ferry, and commuter rail. Because the Boston weather data also started out in pdf format, we manually put the total monthly precipitation values into a sheet with the month names, and then converted this to a json file. We were then able to take the datasets and use the month names to correlate the two variables. This allowed us to take the numerical precipitation and ridership data and then put it into our visualizations, including inputting them into line and bar graphs, a pie chart, and a table. It is important to note that the precipitation data set is missing the data for December, which meant that the water level visualization to specifically check for None values when evaluating the dataset. Additionally, we had to adjust the y-axis on the ridership bar graph in order to make the data show up for the month of January.

Task Analysis

original task table

Our visualization was primarily developed for consumption, especially allowing users to discover the way that weather events and month correlate with changes in MBTA ridership. We planned on having some parts that are more focused on presenting information as well, which would allow users to view information about speed restrictions, but we chose to cut this out in the end. This would have been more useful when looking at MBTA Rapid Transit times, since speed restrictions can impact both speed and ridership as well. However, because we did not end up looking at MBTA travel times, we also eliminated any analysis of speed restrictions. Our original task table is shown above.

Overall, we wanted to allow users to better understand how recent weather patterns align with MBTA usage. In the end, we decided to focus on a variation of our original task number 2, replacing temperature with precipitation. We decided to look at Boston’s total monthly precipitation, rather than temperature, but still wanted to answer the same question of what weather is associated with the highest MBTA ridership, and weather it differs by line. This task started with the analytic task of filtering the data for ridership by line, and also correlating the precipitation and ridership data from multiple datasets. The search task was lookup, since we had the given months that we wanted to get the total monthly precipitation and the ridership for individual MBTA rail lines. For the analyze tasks, we focused on discovery, allowing users to look at our various linked visualizations to look at the data for individual months, or for groups of months.

The intended audience that we hope to reach is the people who run and manage the MBTA Rapid Transit system, since it is important for them to understand how their staffing and engineer needs (run more trains, deal with trains breaking down, help manage high traffic times within the stations) must change based on the weather. We also hope that our visualization is useful for MBTA Rapid Transit riders, since they may better understand when the T will be more or less crowded depending on the weather. Thus, the primary consumers are the public and MBTA employees.

Design Process

We decided on the key variables to investigate from our datasets: ridership by month, ridership by line, and precipitation levels. We started with a couple of sketches, using visualization channels like bar charts, node graphs, tree maps, line graphs, and pie charts to visualize our weather MBTA route maps. After discussing the strengths and weaknesses of each, we were able to choose the most approachable and creative way to combine our individual sketches into one comprehensive sketch in which our charts and graphs would effectively interact with each other and vividly present the dataset. First, we cleaned and filtered our datasets and organized them into csv and json files. Then, we implemented our first visualizations: the table, bar chart, and the general code for visualization.js and linking based on our previous experiences. Next, we added the pie chart showing the monthly ridership breakdown by line, followed by a line chart of monthly total precipitation. The code for the piechart was helpful in our next implementation of the water level visualizer, as both needed to listen to a mouse click event. A New York rat was added next to it as a reference for precipitation. The final step was linking the 5 visualizations together to achieve the interactive functionality. Regarding linking, the table is our starting basis. A mouse action on a table row will cause the other charts to reflect more specific information from that row. We chose to make it this way because a table is simple, self-explanatory, and holds all the necessary information together in a clean easy-to-read way. We then added colors and design attributes to our css file and cleaned up our formatting and text.

sketch 1 sketch 2 sketch 3

Conclusion

After some data wrangling and visualization techniques, we were able to display a table showing the MBTA's 2022 ridership by month, a pie chart for a breakdown of the months' ridership by line, a water level visualizer, a line chart for precipitation by month, and a bar chart representation of ridership by month. Some technical issues we ran into was trying to overlay a line graph over the bar chart. Although it wasn’t necessary to display the line graph, as the purpose of the feature was also represented by the total precipitation visualization, having the two graphs overlaid would have made it easier to look at whether there is any correlation between total ridership and total monthly precipitation. After evaluating our various visualizations, we cannot state that there is a correlation between total monthly precipitation and ridership. However, our displays do allow users to better understand how MBTA ridership differs by season and month, and also how precipitation in Boston varies by month. If we had more time to hone in on our skills, we could successfully incorporate the line graph. We could also add in more details to our visualizations to improve its aesthetic. We could incorporate a weather-like or MBTA-like theme for our page, and we could find a way to make our various visualizations more interconnected. Going forward, if we were to improve our current project, we would take all of these considerations into mind. Data-wise, we could also look for other variables and related datasets to further investigate the relationship between weather and MBTA. For instance, taking temperature, rain, the timing of sunrise and sunset, snow, or other major weather events into account would allow us to further examine the relationship between weather and MBTA usage. Additionally, it is important to recognize that there are many other variables that affect ridership and could be considered, such as social events, like the end of a concert nearby, big city events, or lockdowns.

Acknowledgments