Introduction

This project investigates the relationship between socioeconomic factors, including median household income and minority population data, to measure accessibility and identify disparities within Atlanta’s public transit system, specifically MARTA rail stations. The analysis employs a simulated park-and-ride travel method to Midtown Atlanta (work hub) to highlight potential inequities in transit access among different communities.

Task description

There are a few main components in this assignment - home location, road networks, transit network, and destination. We will simulate a journey that starts from the starting point (e.g., home), drives to the nearest MARTA rail station, transfers to MARTA rail transit, and finally arrives at Midtown station. The following is a list of tasks and data we need for this analysis.

Step 1. Download Required data from GTFS. Convert it to sf format, extract MARTA rail stations, and clean the stop names to delete duplicate names. Also extract the destination station.

Step 2. Download Required data from Census. Convert Census polygons into centroids and create a subset.

Step 3. Download Required data from OSM. Convert it into an sfnetwork object and clean the network.

Step 4. Simulate a park-and-ride trip (home -> closest station -> Midtown station).

Step 5. Convert what we did in Step 4 into a function so that we can use it to repeat it in a loop.

Step 6. Run a loop to repeat the function from Step 5 to all other home location. Once finished, merge the simulation output back to Census data.

Step 7. Finally, examine whether there is any disparity in using transit to commute to midtown.

Before we start, libraries first..

library(tidyverse)
## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.3     ✔ readr     2.1.4
## ✔ forcats   1.0.0     ✔ stringr   1.5.0
## ✔ ggplot2   3.4.3     ✔ tibble    3.2.1
## ✔ lubridate 1.9.2     ✔ tidyr     1.3.0
## ✔ purrr     1.0.2     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(tmap)
## Breaking News: tmap 3.x is retiring. Please test v4, e.g. with
## remotes::install_github('r-tmap/tmap')
library(units)
## udunits database from /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/library/units/share/udunits/udunits2.xml
library(sf)
## Linking to GEOS 3.11.0, GDAL 3.5.3, PROJ 9.1.0; sf_use_s2() is TRUE
library(leaflet)
library(dbscan)
## Warning: package 'dbscan' was built under R version 4.3.3
## 
## Attaching package: 'dbscan'
## 
## The following object is masked from 'package:stats':
## 
##     as.dendrogram
library(sfnetworks)
library(tigris)
## To enable caching of data, set `options(tigris_use_cache = TRUE)`
## in your R script or .Rprofile.
library(tidygraph)
## 
## Attaching package: 'tidygraph'
## 
## The following object is masked from 'package:stats':
## 
##     filter
library(plotly)
## 
## Attaching package: 'plotly'
## 
## The following object is masked from 'package:ggplot2':
## 
##     last_plot
## 
## The following object is masked from 'package:stats':
## 
##     filter
## 
## The following object is masked from 'package:graphics':
## 
##     layout
library(osmdata)
## Data (c) OpenStreetMap contributors, ODbL 1.0. https://www.openstreetmap.org/copyright
library(here)
## here() starts at /Users/katherinelosada/MSUA
library(tidytransit)
library(tidycensus)
## Warning: package 'tidycensus' was built under R version 4.3.3
library(leafsync)

epsg <- 4326

Step 1. Download Required data from GTFS.

# TASK ////////////////////////////////////////////////////////////////////////
# Download MARTA (Metropolitan Atlanta Rapid Transit Authority) GTFS data using `read_gtfs()` function and assign it to `gtfs` object
gtfs <- read_gtfs("google_transit.zip")
# //TASK //////////////////////////////////////////////////////////////////////



# =========== NO MODIFICATION ZONE STARTS HERE ===============================
# Edit stop_name to append serial numbers (1, 2, etc.) to remove duplicate names
stop_dist <- stop_group_distances(gtfs$stops, by='stop_name') %>%
  filter(dist_max > 200)

gtfs$stops <- gtfs$stops %>% 
  group_by(stop_name) %>% 
  mutate(stop_name = case_when(stop_name %in% stop_dist$stop_name ~ paste0(stop_name, " (", seq(1,n()), ")"),
                               TRUE ~ stop_name))

# Create a transfer table
gtfs <- gtfsrouter::gtfs_transfer_table(gtfs, 
                                        d_limit = 200, 
                                        min_transfer_time = 120)
## Registered S3 method overwritten by 'gtfsrouter':
##   method       from  
##   summary.gtfs gtfsio
# NOTE: Converting to sf format uses stop_lat and stop_lon columns contained in gtfs$stops.
#       In the conversion process, stop_lat and stop_lon are converted into a geometry column, and
#       the output sf object do not have the lat lon column anymore.
#       But many other functions in tidytransit look for stop_lat and stop_lon.
#       So I re-create them using mutate().
gtfs <- gtfs %>% gtfs_as_sf(crs = epsg)

gtfs$stops <- gtfs$stops %>% 
  ungroup() %>% 
  mutate(stop_lat = st_coordinates(.)[,2],
         stop_lon = st_coordinates(.)[,1]) 

# Get stop_id for rails and buses
rail_stops <- gtfs$routes %>% 
  filter(route_type %in% c(1)) %>% 
  inner_join(gtfs$trips, by = "route_id") %>% 
  inner_join(gtfs$stop_times, by = "trip_id") %>% 
  inner_join(gtfs$stops, by = "stop_id") %>% 
  group_by(stop_id) %>% 
  slice(1) %>% 
  pull(stop_id)

# Extract MARTA rail stations
station <- gtfs$stops %>% filter(stop_id %in% rail_stops)

# Extract Midtown Station
midtown <- gtfs$stops %>% filter(stop_id == "134")

# Create a bounding box to which we limit our analysis
bbox <- st_bbox(c(xmin = -84.45241, ymin = 33.72109, xmax = -84.35009, ymax = 33.80101), 
                 crs = st_crs(4326)) %>% 
  st_as_sfc()

# =========== NO MODIFY ZONE ENDS HERE ========================================

Step 2. Download Required data from Census

# TASK ////////////////////////////////////////////////////////////////////////
# Specify Census API key whichever you prefer using census_api_key() function
tidycensus::census_api_key(Sys.getenv("Census_API"))
## To install your API key for use in future sessions, run this function with `install = TRUE`.
# //TASK //////////////////////////////////////////////////////////////////////



# TASK ////////////////////////////////////////////////////////////////////////
# Using get_acs() function, download Census Tract level data for 2022 for Fulton, DeKalb, and Clayton in GA.
# and assign it to `census` object.
# Make sure you set geometry = TRUE.

# Required data from the Census ACS:
#  1) Median Household Income (name the column `hhinc`)
#  2) Minority Population (%) (name the column `pct_minority`)
# Note: You may need to download two or more Census ACS variables to calculate minority population (%). "Minority" here can refer to either racial minorities or racial+ethnic minorities -- it's your choice.

census <- get_acs(geography="tract",
                  state="GA",
                  county=c("Fulton", "DeKalb", "Clayton"),
                  variables = c(hhinc = 'B19013_001', totpop = 'B03002_001', black = 'B03002_004', asian = 'B03002_006', hispanic = 'B03002_012'),
                  year = 2022,
                  survey = "acs5", 
                  geometry = TRUE, 
                  output = "wide",
                  progress = FALSE,
                  ) %>%
  mutate(hhinc = hhincE, pct_minority = (blackE+asianE+hispanicE)/totpopE)
## Getting data from the 2018-2022 5-year ACS
## Downloading feature geometry from the Census website.  To cache shapefiles for use in future sessions, set `options(tigris_use_cache = TRUE)`.
# //TASK //////////////////////////////////////////////////////////////////////



# =========== NO MODIFICATION ZONE STARTS HERE ===============================
census <- census %>% 
  st_transform(crs = 4326) %>% 
  separate(col = NAME, into = c("tract", "county", "state"), sep = ", ")
## Warning: Expected 3 pieces. Missing pieces filled with `NA` in 600 rows [1, 2, 3, 4, 5,
## 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, ...].
# Convert it to POINT at polygon centroids and extract those that fall into bbox
# and assign it into `home` object
home <- census %>% st_centroid() %>% .[bbox,]
## Warning: st_centroid assumes attributes are constant over geometries
# =========== NO MODIFY ZONE ENDS HERE ========================================

Step 3. Download Required data from OSM.

# TASK ////////////////////////////////////////////////////////////////////////
# 1. Get OSM data using opq() function and bbox object defined in the previous code chunk.
# 2. Specify arguments for add_osm_feature() function using 
#    key = 'highway' and 
#    value = c("motorway", "trunk", "primary", "secondary", "tertiary", "residential", 
#              "motorway_link", "trunk_link", "primary_link", "secondary_link", 
#              "tertiary_link", "residential_link", "unclassified")
# 3. Convert the OSM data into an sf object using osmdata_sf() function
# 4. Convert osmdata polygons into lines using osm_poly2line() function

osm_road <- opq(bbox = bbox) %>%
  add_osm_feature(key = 'highway', 
                  value = c("motorway", "trunk", "primary", "secondary", "tertiary", "residential", 
                  "motorway_link", "trunk_link", "primary_link", "secondary_link", 
                  "tertiary_link", "residential_link", "unclassified")) %>%
  osmdata_sf() %>% 
  osm_poly2line()
  
# //TASK //////////////////////////////////////////////////////////////////////


# TASK ////////////////////////////////////////////////////////////////////////
# 1. Convert osm_road$osm_lines into sfnetwork using as_sfnetwork() function
# 2. Activate edges
# 3. Clean the network using edge_is_multiple(), edge_is_loop(), to_spatial_subdivision(), to_spatial_smooth()
# 4. Assign the cleaned network to an object named 'osm'

osm <- osm_road$osm_lines %>% 
  select(osm_id, highway) %>%
  as_sfnetwork(directed = FALSE) %>%
  activate("edges") %>%
  filter(!edge_is_multiple()) %>%
  filter(!edge_is_loop()) %>%
  convert(.,sfnetworks::to_spatial_subdivision) %>%
  convert(.,sfnetworks::to_spatial_smooth) 
## Warning: to_spatial_subdivision assumes attributes are constant over geometries
# //TASK //////////////////////////////////////////////////////////////////////



# TASK ////////////////////////////////////////////////////////////////////////
# Add a new column named 'length' to the edges part of the object `osm`.
osm <- osm %>% 
  mutate(length = edge_length())
# //TASK //////////////////////////////////////////////////////////////////////

Step 4. Simulate a park-and-ride trip (home -> closest station -> Midtown station).

# =========== NO MODIFICATION ZONE STARTS HERE ===============================
# Extract the first row from `home` object and store it `home_1`
home_1 <- home[1,]
# =========== NO MODIFY ZONE ENDS HERE ========================================


# TASK ////////////////////////////////////////////////////////////////////////
# Find the shortest path from `home_1` to all other stations
# using st_network_paths() function.
paths <- st_network_paths(osm, from = home_1, to = station, type = "shortest") 
# //TASK //////////////////////////////////////////////////////////////////////

  
# =========== NO MODIFICATION ZONE STARTS HERE ===============================
# Using the `paths` object, get network distances from `home_1` to all other stations.
dist_all <- map_dbl(1:nrow(paths), function(x){
  osm %>%
    activate("nodes") %>% 
    slice(paths$node_paths[[x]]) %>% 
    st_as_sf("edges") %>% 
    pull(length) %>% 
    sum()
}) %>% unlist() 

# Replace zeros with a large value.
if (any(dist_all == 0)){
  dist_all[dist_all == 0] <- max(dist_all)
}

# Find the closest station.
closest_index <- which.min(dist_all)
closest_station <- station[closest_index,]

# Find the distance to the closest station.
closest_dist <- min(dist_all)

# Calculate how long it takes to traverse `closest_dist` 
# assuming we drive at 30 miles/hour speed.
# Store the output in trvt_osm_m.
car_speed <- set_units(30, mile/h)
trvt_osm_m <- closest_dist/set_units(car_speed, m/min) %>%  # Distance divided by 30 mile/h
  as.vector(.)
# =========== NO MODIFY ZONE ENDS HERE ========================================


# TASK ////////////////////////////////////////////////////////////////////////
# 1. From `osm` object, activate nodes part and
# 2. use `closest_index` to extract the selected path
paths_closest <- osm %>%
  activate("nodes") %>% 
  slice(paths$node_paths[[closest_index]])
# //TASK //////////////////////////////////////////////////////////////////////


# TASK ////////////////////////////////////////////////////////////////////////
# Use filter_stop_times() function to create a subset of stop_times data table
# for date = 2024-11-14, minimum departure time of 7AM, maximum departure time of 10AM.
# Assign the output to `am_stop_time` object
am_stop_time <- filter_stop_times(gtfs_obj = gtfs, 
                                  extract_date = "2024-11-14",
                                  min_departure_time = 3600*7, 
                                  max_arrival_time = 3600*10)
# //TASK //////////////////////////////////////////////////////////////////////



# TASK ////////////////////////////////////////////////////////////////////////
# 1. Use travel_times() function to calculate travel times from the `closest_station` 
#    to all other stations during time specified in am_stop_time. Allow ONE transfer.
# 2. Filter the row for which the value of 'to_stop_name' column 
#    equals midtown$stop_name. Assign it into `trvt` object.
trvt <- travel_times(filtered_stop_times = am_stop_time,
                     stop_name = closest_station$stop_name,
                     time_range = 3600,
                     arrival = FALSE, 
                     max_transfers = 1,
                     return_coords = TRUE) %>%
    filter(to_stop_name == midtown$stop_name)
# //TASK //////////////////////////////////////////////////////////////////////



# =========== NO MODIFICATION ZONE STARTS HERE ===============================
# Divide the calculated travel time by 60 to convert the unit from seconds to minutes.
trvt_gtfs_m <- trvt$travel_time/60

# Add the travel time from home to the nearest station and
# the travel time from the nearest station to Midtown station
total_trvt <- trvt_osm_m + trvt_gtfs_m
# =========== NO MODIFY ZONE ENDS HERE ========================================

Step 5. Convert Step 4 into a function

# Function definition (do not modify other parts of the code in this code chunk except for those inside the TASK section)

get_trvt <- function(home, osm, station, midtown){
  
  # TASK ////////////////////////////////////////
  # If the code in Step 4 runs fine,
  # Replace where it says **YOUR CODE HERE..** below with 
  # the ENTIRETY of the code in the previous code chunk (i.e., Step 4)

  home_1 <- home[1,]
  paths <- st_network_paths(osm, from = home_1, to = station, type = "shortest") 
  
  dist_all <- map_dbl(1:nrow(paths), function(x){
    osm %>%
      activate("nodes") %>% 
      slice(paths$node_paths[[x]]) %>% 
      st_as_sf("edges") %>% 
      pull(length) %>% 
      sum()
  }) %>% unlist() 

  if (any(dist_all == 0)){
    dist_all[dist_all == 0] <- max(dist_all)
  }

  closest_index <- which.min(dist_all)
  closest_station <- station[closest_index,]

  closest_dist <- min(dist_all)
  
  car_speed <- set_units(30, mile/h)
  trvt_osm_m <- closest_dist/set_units(car_speed, m/min) %>%  # Distance divided by 30 mile/h
    as.vector(.)
  
  paths_closest <- osm %>%
    activate("nodes") %>% 
    slice(paths$node_paths[[closest_index]])
  
  am_stop_time <- filter_stop_times(gtfs_obj = gtfs, 
                                    extract_date = "2024-11-14",
                                    min_departure_time = 3600*7, 
                                    max_arrival_time = 3600*10)
  
  trvt <- travel_times(filtered_stop_times = am_stop_time,
                       stop_name = closest_station$stop_name,
                       time_range = 3600,
                       arrival = FALSE, 
                       max_transfers = 1,
                       return_coords = TRUE) %>%
      filter(to_stop_name == midtown$stop_name)
  
  trvt_gtfs_m <- trvt$travel_time/60
  
  total_trvt <- trvt_osm_m + trvt_gtfs_m
  # //TASK //////////////////////////////////////

  # =========== NO MODIFICATION ZONE STARTS HERE ===============================
  if (length(total_trvt) == 0) {total_trvt = 0}

  return(total_trvt)
  # =========== NO MODIFY ZONE ENDS HERE ========================================
}

Step 6. Apply the function for the whole study area

# Prepare an empty vector
total_trvt <- vector("numeric", nrow(home))

# Apply the function for all Census Tracts
# Fill `total_trvt` object with the calculated time
for (i in 1:nrow(home)){
  total_trvt[i] <- get_trvt(home[i,], osm, station, midtown)
}

# Cbind the calculated travel time back to `home`
home_done <- home %>% 
  cbind(trvt = total_trvt)

Step 7. Create maps and plots

Run the code below to generate thematic maps and plots

Write a short description of what you observe from the maps and plots

# Map
tmap_mode('view')
## tmap mode set to interactive viewing
tm_shape(census[census$GEOID %in% home$GEOID,]) + 
  tm_polygons(col = "hhinc", palette = 'BuPu') + 
  tm_shape(home_done) + 
  tm_dots(col = "trvt", palette = 'Oranges', size = 0.1)
tm_shape(census[census$GEOID %in% home$GEOID,]) + 
  tm_polygons(col = "pct_minority", palette = 'BuPu') + 
  tm_shape(home_done) + 
  tm_dots(col = "trvt", palette = 'Oranges', size = 0.1)
# ggplot
inc <- ggplot(data = home_done,
              aes(x = hhinc, y = trvt)) +
  geom_point() +
  geom_smooth(method = "lm", se = FALSE) +
  labs(x = "Median Annual Household Income",
       y = "Park-and-ride Travel Time from Home to Midtown Station") +
  theme_bw()

minority <- ggplot(data = home_done,
                   aes(x = pct_minority, y = trvt)) +
  geom_point() +
  geom_smooth(method = "lm", se = FALSE) +
  labs(x = "Minority Population (%)",
       y = "Park-and-ride Travel Time from Home to Midtown Station") +
  theme_bw()


ggpubr::ggarrange(inc, minority)
## `geom_smooth()` using formula = 'y ~ x'
## Warning: Removed 6 rows containing non-finite values (`stat_smooth()`).
## Warning: Removed 6 rows containing missing values (`geom_point()`).
## `geom_smooth()` using formula = 'y ~ x'

Results

Plot 1 visualizes the relationship between median household income and distance to the nearest MARTA station across different Census tracts. I expected a pattern suggesting that lower-income tracts are often farther from MARTA stations, indicating a potential access disparity. However, the tracts where the trvt is low (0-5) are clustered in a specific area with tracts varying in median house hold income. I would say there is no strong correlation between the two variables. Although not correlated to socioeconomic factors, it may showcase a different disparity on accessibility to public transit in Atlanta as only a small percentage of the city is in close proximity to MARTA rail stations. This highlights an issue in Atlanta’s transit infrastructure that needs to be addressed.

Plot 2 more clearly showcases the disparity explained previously as the tracts with larger percent minority tend to have a greater trvt score. There are still outliers with tracts of small percent minorities also having large travel times to MARTA rail stations. Furthermore, the clustered tracts with low trvt scores (0-10) also vary in percent minority ranging from 0% to 80% so it is hard to come up with a definite conclusion.

Pot 3 depicts a positive correlation between Park-and-ride Travel Time from Home to Midtown Station and median household income, which I was unable to see through the map. It also shows that trvt had a weaker relationship with the minority population, and this also contradicts what I said earlier.