The topic of my project is building a bike rental system in San Francisco. I choose this project because the San Francisco city government has proposed to promote biking as an alternative to driving, which is an important strategy in the city’s climate change action plan. I think bike rental is a great way of promoting biking, and I want to integrate the bike rental system with the current bus system in San Francisco by putting the bike rental stations close to the bus stops so that people can switch between these two different transportation modes conveniently.
Here is the structure of my project:
- Find 5 best locations to build bike rental stations in San Francisco.
- Analyze my proposed stations. Specifically, I would create buffers to get information within the buffers, create service areas to see which neighborhoods are covered by these stations and use OD matrix to see how the stations are integrated with the bus system.
- Analyze how the bike rental system would expand over time.
My first layout shows the percentage of workers who use bike as the main means of going to work in San Francisco. Apparently, biking is not a popular way of going to work in San Francisco, at least now.
- Median household income: I want to locate the stations where household income is low;
- Population density: I want to locate the stations where population density is high;
- Young population percentage: I think young people are more likely to ride bikes so I want to locate the stations where there are more young people (I define young population as people from 14 to 34 years old);
- Distance to the nearest bus stop: Since I want to integrate the bike rental system with the bus system, I want to locate the stations as close to the bus stop as possible.
This layout shows these 4 criteria in San Francisco.
Each criterion is given the same weight. Using hotspot analysis, I get the priority map to help me locate the rental stations.
However, the priority map only illustrates the high-priority rasters, and I still need to choose the specific locations from these rasters manually. To do this, I first check the buildings and facilities in each bunch of raster. I try to locate the rental station in an open space in the center of the area. For instance, in this map, I locate the station in the circle in the center, which seems to be an open space.
Then I use Google Map to get the specific address for the location I just chose.
By repeating this process, I get 5 specific locations to place the stations in the map.
By geocoding their addresses I show these proposed stations in the following map.
In this layout I create 1000-feet buffers for each station and extract information from the buffers. The young population percentage is relatively high within the buffers, but the percentage of workers who use bikes to go to work varies from station to station.
Here is the model I used to create the buffers.
This layout shows the service areas of each station within 3, 5 and 10 minutes’ walk. I checked Wikipedia and found the “preferred walking speed”, or the speed humans tend to walk at without extreme circumstances, is 1.4m/s, so I used this speed to calculate the service areas.
This layout shows the OD matrix within 10 minutes’ walk from proposed bike rental stations to bus stops. From any of the 5 rental stations, people can walk within 10 minutes to many bus stops, which means the bike rental system is well integrated with the bus system.
The last thing is using the time-based analysis to see how the bike rental system would expand over time until 2020. I have already proposed 5 rental stations to be built in 2013. For 2014, I would choose 5 new stations, using the same 4 criteria as before as long as one more criterion: the distance to the existing rental stations. We do not want the rental stations to be too close to each other. Each criterion is still given the same weight. After using hotspot analysis to get 5 locations for 2014, I merge these new locations with the 2013 locations, add a field indicating year in the attribute table and use the merged file in the analysis of 2015. By repeating this process and adding 5 new stations every year (it takes some time!), I finally get the movie in the presentation (See the movie in my presentation… I forgot to copy the original movie from the lab computer so I currently have no access to the original movie).
The main findings of this project include where to locate the bike rental stations, what the service areas of these stations are, how the system is integrated with the bus system and how the system would expand until 2020.
My project has several limitations. First, there might be disagreements on the criteria and weights I choose in the hotspot analysis. For instance, because I give each criterion the same weight, the distance to the existing rental stations only counts 1/5 in the hotspot analysis, so the southwest area in San Francisco never gets a chance to get high scores (they are far from the existing stations, but might not be competitive in other criteria), and the system never expands there even after several years. Another limitation is that in the time-based analysis, I assume some factors like median income and young population percentage to be constant, because it is both difficult and beyond the scope of this project to make projections about how these factors would change in the future.
Last words about data sources: The census trace demographic data are from ACS 2006-2010 found in Social Explorer. The San Francisco bus stops data are from SFMTA.