---
title: "Bird collisions at the Salt Lake Airpot"
author: "Andrew Heiss"
date: "`r Sys.Date()`"
output: 
  html_document:
    toc: true
    toc_float: true
---

Let's see if we can find any patterns in the number of bird collisions at the Salt Lake City airport. The Federal Aviation Administration (FAA) keeps track of every bird-airplane collision in its [Wildlife Strike Database](https://wildlife.faa.gov/) and makes its data available publicly available for free. Utah offers a subset of this data for Salt Lake City specifically in the [Utah Open Data Catalog](https://opendata.utah.gov/Public-Safety/Birdstrikes-At-Salt-Lake-City-Intl-Airport-2000-20/ez64-g6kd/about).

We'll use some basic functions from a few **tidyverse** packages (specifically **dplyr**, **ggplot2**, and **lubridate**).


# Look at the raw data

First we load the necessary libraries (this makes it so we can use functions like `mutate()` and `ggplot()`).

```{r load-libraries}
library(tidyverse)
library(lubridate)
library(scales)
```

Next we load the data. I downloaded this as a CSV previously from the [Open Data Catalog](https://opendata.utah.gov/Public-Safety/Birdstrikes-At-Salt-Lake-City-Intl-Airport-2000-20/ez64-g6kd/about).

We'll then clean it up a little by adding some columns to indicate the year and month of the incident and renaming some of the longer variables.

```{r load-data}
# Load raw data
bird_strikes_raw <- read_csv("data/Birdstrikes_At_Salt_Lake_City_Intl_Airport_2000-2014.csv")

# Clean up the raw data a little
bird_strikes <- bird_strikes_raw %>% 
  mutate(Date = mdy_hms(Date)) %>% 
  mutate(Year = year(Date),
         Month = month(Date, label = TRUE, abbr = TRUE)) %>% 
  rename(Phase = `Phase of Flight`,
         Cloudy = `Sky Condition`,
         Time = `Time of Day`,
         Cost = `Cost of Repairs`)
```

What does this dataset look like?

```{r view-data}
bird_strikes
```

Alternativey, click on `bird_strikes` in the "Environmnent" panel on the right to open the dataset in a new tab. Or, run `View()`:

```{r view-data-tab, eval=FALSE}
View(bird_strikes)
```


# Analyze and visualize the data

## Bird strike details

Let's look at some interesting patterns in the data. Which bird species gets hit the most often? Species information is included in the `Species` variable.

Things to try out:

- What happens if you remove `sort = TRUE`?

```{r species-count}
bird_strikes %>% 
  count(Species, sort = TRUE)
```

Find the count of bird strikes in each phase of flight. The `Phase` variable contains this information and indicates if a plane is taking off, landing, or doing something else. Make sure the list is sorted.

```{r phase-count}
bird_strikes %>% 
  count()
```

Find the count of bird strikes under different sky conditions. The `Cloudy` variable contains this information and indicates how cloudy it was 

```{r cloudy-count}

```

Find the count of bird strikes during different times of the day. The `Time` varialbe contains this information and indicates if it was during the day, at night, or in between.

```{r time-count}

```

## Bird strikes over time

Let's see if there are any patterns over time. Has the number of bird strikes increased since 2000?

Things to try:

- Try including `sort = TRUE` in `count()`. Why would you want to sort this list? Why would you not want to sort this list?

```{r count-year}
strikes_per_year <- bird_strikes %>% 
  count(Year)

ggplot(data = strikes_per_year, 
       mapping = aes(x = Year, y = n)) +
  geom_col()
```

Are there any months that are more or less deadly for birds (use the `Month` variable)? Why do you think this pattern exists?

```{r count-month}

```


```{r summarize-time-month}
birds_time_month <- bird_strikes %>% 
  filter(Time %in% c("Day", "Night")) %>% 
  count(Time, Month)
birds_time_month
```

Let's plot this so we can see patterns easier! What patterns do you see here? 

Change `position = "stack"` to `position = "dodge"`. What's different? Which plot is easier to interpret? Wny?

```{r plot-time-month1}
ggplot(data = birds_time_month, 
       mapping = aes(x = Month, y = n, fill = Time)) + 
  geom_col(position = "stack") +
  labs(x = "Month", 
       y = "Bird strikes",
       title = "Bird strikes per month",
       subtitle = "Autumn nights are bad for birds",
       caption = "Source: FAA Wildlife Strike Database")
```

Instead of dodging or stacking the bars, here we'll put each time period in its own subplot, or facet. 

Is this more interpretable or less interpretable than the dodged or stacked versions from above? Change `ncol = 1` to `ncol = 2`. What's different? Which is a better plot?

```{r plot-time-month2}
ggplot(data = birds_time_month, 
       mapping = aes(x = Month, y = n, fill = Time)) + 
  geom_col() + 
  labs(x = "Month", 
       y = "Bird strikes",
       title = "Bird strikes per month",
       subtitle = "Autumn nights are bad for birds",
       caption = "Source: FAA Wildlife Strike Database") +
  facet_wrap(~ Time, ncol = 1)
```

## Costs of bird strikes

Let's now see how expensive all these bird strikes are. First, let's summarize the total cost and average cost

Things to try:

- What happens if you remove `na.rm = TRUE`? Why does that happen?

```{r summarize-damages}
strike_damages <- bird_strikes %>% 
  summarize(total_damages = sum(Cost, na.rm = TRUE),
            average_damages = mean(Cost, na.rm = TRUE))
```

We can also calculate these summary statistics by groups of other variables. For example, we can look at the average and total damages by month.

```{r summarize-damages-month}
strike_damages_month <- bird_strikes %>% 
  group_by(Month) %>% 
  summarize(total_damages = sum(Cost, na.rm = TRUE),
            average_damages = mean(Cost, na.rm = TRUE))

ggplot(data = strike_damages_month, 
       mapping = aes(x = Month, y = total_damages)) + 
  geom_col() + 
  scale_y_continuous(labels = dollar) +
  labs(x = "Month",
       y = "Total damages",
       title = "Really expensive collisions happen in the fall?",
       subtitle = "Don't fly in August or October?",
       source = "Source: FAA Wildlife Strike Database")
```

What can we learn from this plot? What's going on in August and October? View the raw data in a new tab and see if you can find the super expensive incidents driving up these total amounts.

Summarize the total damages by some other variable, like cloud cover (`Cloudy`), time of day (`Time`), phase of flight (`Phase`), or something else in the dataset. Are there times, weather patterns, or phases of the flight that cause more or less expensive damage?

```{r summarize-damages-something-else}
# strike_damages_XXXX <- bird_strikes %>% 
#   group_by(SOMETHING_HERE) %>% 
#   summarize(total_damages = sum(Cost, na.rm = TRUE))
# 
# ggplot(data = strike_damages_XXXX, 
#        mapping = aes(x = SOMETHING_HERE, y = total_damages)) + 
#   geom_col() +
#   scale_y_continuous(labels = dollar) +
#   labs(x = "Month",
#        y = "Total damages",
#        title = "Something interesting",
#        subtitle = "Something else interesting",
#        source = "Source: FAA Wildlife Strike Database")
```