library(tidyverse)
library(RColorBrewer)
library(tidycensus)
library(ggrepel)
## Load TOPS data ----
## To load TOPS data for the whole state for crashes involving bikes and pedestrians):
## Step 1 - download csv from the TOPS Data Retrieval Tool with the query: SELECT * FROM DTCRPRD.SUMMARY_COMBINED C WHERE C.CRSHDATE BETWEEN TO_DATE('2023-JAN','YYYY-MM') AND LAST_DAY(TO_DATE('2023-DEC','YYYY-MM')) AND (C.BIKEFLAG = 'Y' OR C.PEDFLAG = 'Y') ORDER BY C.DOCTNMBR
## Step 2 - include RACE1 and RACE2 for download in preferences
## Step 3 - save the csv in the "data" directory as crash-data-download_2023.csv
TOPS_data <- as.list(NULL)
for (file in list.files(path = "data/TOPS/", pattern = "crash-data-download")) {
message(paste("importing data from file: ", file))
year <- substr(file, 21, 24)
csv_run <- read_csv(file = paste0("data/TOPS/",file), col_types = cols(.default = "c"))
csv_run["retreive_date"] <- file.info(file = paste0("data/TOPS/",file))$mtime
TOPS_data[[file]] <- csv_run
}
rm(csv_run, file, year)
TOPS_data <- bind_rows(TOPS_data)
## clean up data ----
TOPS_data <- TOPS_data %>%
mutate(date = mdy(CRSHDATE),
age1 = as.double(AGE1),
age2 = as.double(AGE2),
latitude = as.double(LATDECDG),
longitude = as.double(LONDECDG)) %>%
mutate(month = month(date, label = TRUE),
year = as.factor(year(date)))
retrieve_date <- max(TOPS_data %>% filter(year %in% max(year(TOPS_data$date), na.rm = TRUE)) %>% pull(retreive_date))
# Injury Severy Index and Color -----
injury_severity <- data.frame(InjSevName = c("No apparent injury", "Possible Injury", "Suspected Minor Injury","Suspected Serious Injury","Fatality"),
code = c("O", "C", "B", "A", "K"),
color = c("#fafa6e", "#edc346", "#d88d2d", "#bd5721", "#9b1c1c"))
TOPS_data <- left_join(TOPS_data, injury_severity %>% select(InjSevName, code), join_by(INJSVR1 == code)) %>%
mutate(InjSevName = factor(InjSevName, levels = injury_severity$InjSevName)) %>%
rename(InjSevName1 = InjSevName)
TOPS_data <- left_join(TOPS_data, injury_severity %>% select(InjSevName, code), join_by(INJSVR2 == code)) %>%
mutate(InjSevName = factor(InjSevName, levels = injury_severity$InjSevName)) %>%
rename(InjSevName2 = InjSevName)
# add bike or pedestrian roles ----
bike_roles <- c("BIKE", "O BIKE")
ped_roles <- c("PED", "O PED", "PED NO")
vuln_roles <- c(bike_roles, ped_roles)
TOPS_data <- TOPS_data %>% mutate(ped_inj = ifelse(ROLE1 %in% vuln_roles,
INJSVR1,
ifelse(ROLE2 %in% vuln_roles,
INJSVR2,
NA)))
# bike or ped
TOPS_data <- TOPS_data %>% mutate(vulnerable_role = ifelse(ROLE1 %in% bike_roles | ROLE2 %in% bike_roles,
"Bicyclist",
ifelse(ROLE1 %in% ped_roles | ROLE2 %in% ped_roles,
"Pedestrian",
NA)))
TOPS_data <- left_join(TOPS_data, injury_severity %>% select(InjSevName, code), join_by(ped_inj == code)) %>%
mutate(InjSevName = factor(InjSevName, levels = injury_severity$InjSevName)) %>%
rename(ped_inj_name = InjSevName)
# Race names
race <- data.frame(race_name = c("Asian", "Black", "Indian","Hispanic","White"),
code = c("A", "B", "I", "H", "W"))
TOPS_data <- left_join(TOPS_data, race %>% select(race_name, code), join_by(RACE1 == code)) %>% rename(race_name1 = race_name)
TOPS_data <- left_join(TOPS_data, race %>% select(race_name, code), join_by(RACE2 == code)) %>% rename(race_name2 = race_name)
## make mutate TOPS_data
TOPS_data <- TOPS_data %>%
mutate(Year = year,
PedestrianInjurySeverity = ped_inj_name,
CrashDate = CRSHDATE,
CrashTime = CRSHTIME,
County = CNTYNAME,
Street = ONSTR,
CrossStreet = ATSTR) %>%
mutate(PedestrianAge = ifelse(ROLE1 %in% vuln_roles, age1, age2))
# add population census data ----
census_api_key(key = substr(read_file(file = "api_keys/census_api_key"), 1, 40))
county_populations <- get_estimates(geography = "county", year = 2022, product = "population", state = "Wisconsin") %>%
filter(variable == "POPESTIMATE") %>%
mutate(County = str_to_upper(str_replace(NAME, " County, Wisconsin", "")))
## generate county charts ----
county_focus <- unique(TOPS_data %>%
group_by(CNTYNAME) %>%
summarise(TotalCrashes = n()) %>%
slice_max(TotalCrashes, n = 8) %>%
pull(CNTYNAME))
TOPS_data %>%
filter(ped_inj %in% c("A", "K")) %>%
group_by(CNTYNAME, Year) %>%
summarise(TotalCrashes = n()) %>%
mutate(County = CNTYNAME) %>%
left_join(county_populations, join_by("County")) %>%
mutate(CrashesPerPopulation = TotalCrashes/value*100000) %>%
filter(County %in% county_focus) %>%
ggplot() +
geom_line(aes(x = Year,
y = CrashesPerPopulation,
color = str_to_title(CNTYNAME),
group = CNTYNAME),
size = 1) +
geom_label_repel(data = TOPS_data %>%
filter(ped_inj %in% c("A", "K")) %>%
group_by(CNTYNAME, Year) %>%
summarise(TotalCrashes = n()) %>%
mutate(County = CNTYNAME) %>%
left_join(county_populations, join_by("County")) %>%
mutate(CrashesPerPopulation = TotalCrashes/value*100000) %>%
filter(County %in% county_focus,
Year == 2023),
aes(x = Year,
y = CrashesPerPopulation,
label = str_to_title(County),
fill = County),
size=3,
min.segment.length=0,
segment.size = 0.25,
nudge_x=0.5,
direction="y") +
scale_color_brewer(type = "qual", guide = NULL) +
scale_fill_brewer(type = "qual", guide = NULL) +
scale_x_discrete(expand = expansion(add = c(0.5,0.75))) +
labs(title = "Drivers crashing into pedestrians & bicyclists per 100,000 residents",
subtitle = "Fatalities and Severe Injuries | 2017-2023",
x = "Year",
y = "Total crashes per year per 100,000 residents",
color = "County",
caption = paste0("crash data from UW TOPS lab - retrieved ",
strftime(retrieve_date, format = "%m/%Y"),
" per direction of the WisDOT Bureau of Transportation Safety",
"\nbasemap from StadiaMaps and OpenStreetMap Contributers")) +
theme(plot.caption = element_text(color = "grey"))
ggsave(file = paste0("figures/crash_summaries/counties_year.pdf"),
height = 8.5,
width = 11,
units = "in")
TOPS_data %>%
filter(County %in% county_focus) %>%
group_by(County, vulnerable_role) %>%
summarise(count = n()) %>%
ggplot() +
geom_col(aes(x = County,
y = count,
fill = vulnerable_role))
TOPS_data %>%
filter(County %in% "DANE") %>%
group_by(County, vulnerable_role, year) %>%
summarise(count = n()) %>%
ggplot() +
geom_col(aes(x = year,
y = count,
fill = vulnerable_role),
position = position_dodge())