36-613: Data Visualization

class: center, middle, inverse, title-slide

.title[
# 36-613: Data Visualization
]
.subtitle[
## Visualizing Spatial Data and Maps
]
.author[
### Professor Ron Yurko
]
.date[
### 10/3/2022
]

---

# How should we think about spatial data?

- Typically location is measured with __latitude__ / __longitude__ (2D)

- __Latitude__: Measures North / South (the "y-axis")

- Range is `$(-90^{\circ}, 90^{\circ})$`
  
  - Measures degrees from the equator `$(0^{\circ})$`
  
  - `$(-90^{\circ}, 0^{\circ})$` = southern hemisphere 
  
  - `$(0^{\circ}, 90^{\circ})$` = northern hemisphere 
  
--

- __Longitude__: Measures East/West (the "x-axis")

- Range is `$(-180^{\circ}, 180^{\circ})$`
  
  - Measures degrees from the prime meridian `$(0^{\circ})$` in Greenwich, England
  
  - `$(-180^{\circ}, 0^{\circ})$` = eastern hemisphere
  
  - `$(0^{\circ}, 180^{\circ})$` = western hemisphere

---

# Latitude and Longitude

---

# Map Projections

- Earth is a 3D object, but maps are 2D objects

- __Map projections__: Transformation of the lat / long coordinates on a sphere (the earth) to a 2D plane
  
- There are many different projections - each will distort the map in different ways.

- The most common projections are:

- [Mercator](https://en.wikipedia.org/wiki/Mercator_projection)
  - [Robinson](https://en.wikipedia.org/wiki/Robinson_projection)
  - [Conic](http://www.geo.hunter.cuny.edu/~jochen/gtech201/lectures/lec6concepts/Map%20coordinate%20systems/Conic%20projections.htm#:~:text=Conic%20projections%20are%20created%20by,a%20developable%20map%20projection%20surface.)
  - [Cylindrical](https://en.wikipedia.org/wiki/Map_projection#Cylindrical)
  - [Planar](http://www.geo.hunter.cuny.edu/~jochen/gtech201/lectures/lec6concepts/Map%20coordinate%20systems/Planar%20projections.htm)
  - [Interrupted projections](https://en.wikipedia.org/wiki/Interruption_(map_projection))

---

# Mercator Projection (1500s)

---

# Mercator Projection (Tissot indicatrix)

---

# Robinson Projection (Standard from 1963-1998)

---

# Robinson Projection (Tissot indicatrix)

---

# Winkel Tripel Projection (proposed 1921, now the standard)

---

# Winkel Tripel Projection (Tissot indicatrix)

---

# And many more... (see [xkcd comic](https://xkcd.com/977/))

---

## Visualizing spatial data on maps using [`ggmap`](https://cran.r-project.org/web/packages/ggmap/readme/README.html)

.pull-left[

```r
library(ggmap)
# First, we'll draw a "box" around the US
# (in terms of latitude and longitude)
US <- c(left = -125, bottom = 10, 
        right = -67, top = 49)
map <- get_stamenmap(US, zoom = 5, 
                     maptype = "toner-lite")

# Visualize the basic map
*ggmap(map)
```

- Draw map based on lat / lon coordinates

- Put the box into `get_stamenmap()` to access [Stamen Maps](http://maps.stamen.com/#terrain/12/37.7706/-122.3782)

- Draw the map using `ggmap()` to serve as base

]

.pull-right[
<img src="figs/Lec10/unnamed-chunk-10-1.png" width="100%" />

]

---

# Three main types of spatial data

1. __Point Pattern Data__: lat-long coordinates where events have occurred

2. __Point-Referenced data__: Latitude-longitude (lat-long) coordinates as well as one or more variables specific to those coordinates.

3. __Areal Data__: Geographic regions with one or more variables associated with those regions.

- Each type is structured differently within a dataset

- Each type requires a different kind of graph(s)

---

# Point-Pattern data

- __Point Pattern Data__: lat-long coordinates where events have occurred

- __Point pattern data simply records the lat-long of events__; thus, there are only two columns

- Again, latitude and longitude are represented with dots, sometimes called a dot or bubble map.

- The goal is to understand how the __density__ of events varies across space

- The density of the dots can also be visualized (e.g., with contours)

- __Use methods we've discussed before for visualizing 2D joint distribution__

---

# Point-Referenced data

- __Point-Referenced data__: Latitude-longitude (lat-long) coordinates as well as one or more variables specific to those coordinates

- Point-referenced data will have the following form:

```r
airports %>% dplyr::select(lat, lon, altitude, n_depart, n_arrive, name) %>% slice(1:3)
```

```
## # A tibble: 3 × 6
##     lat   lon altitude n_depart n_arrive name                        
##   <dbl> <dbl>    <dbl>    <int>    <int> <chr>                       
## 1 -6.08  145.     5282        5        5 Goroka Airport              
## 2 -5.21  146.       20        8        8 Madang Airport              
## 3 -5.83  144.     5388       10       12 Mount Hagen Kagamuga Airport
```

- The goal is to understand how the variable(s) (e.g., `altitude`) vary across different spatial locations

- Typically, the latitude and longitude are represented with dots, and the variable(s) are represented with size and/or colors

---

## Adding points to the map as usual

.pull-left[

```r
ggmap(map) +
* geom_point(data = airports,
             aes(x = lon, y = lat),
             alpha = 0.25)
```

- Display locations of airports using `geom_point()` layer, need to specify `data` for layer

- Currently viewing __point-pattern__ data

]

.pull-right[
<img src="figs/Lec10/unnamed-chunk-12-1.png" width="100%" />

]

---

## Altering points on the map (in the usual way)

.pull-left[

```r
ggmap(map) +
  geom_point(data = airports, 
             aes(x = lon, y = lat, 
*                size = sqrt(n_depart),
*                color = sqrt(n_arrive)),
             alpha = .5) +
  scale_size_area(breaks = sqrt(c(1, 5, 10, 50, 100, 500)), 
                  labels = c(1, 5, 10, 50, 100, 500), 
                  name = "# departures") +
  scale_color_distiller(palette = "Spectral") +
  labs(color = "sqrt(# arrivals)") +
  theme(legend.position = "bottom")
```

- Displaying additional variables through `aes`

]

.pull-right[
<img src="figs/Lec10/unnamed-chunk-13-1.png" width="100%" />

]

---

## Thinking about areal data

- __Areal Data__: Geographic regions associated with one or more variables specific to those regions

- Areal data will have the following form (example US states data from 1970s):

```r
state_data %>% dplyr::slice(1:3)
```

```
## # A tibble: 3 × 9
##   Population Income Illiteracy `Life Exp` Murder `HS Grad` Frost   Area state  
##        <dbl>  <dbl>      <dbl>      <dbl>  <dbl>     <dbl> <dbl>  <dbl> <chr>  
## 1       3615   3624        2.1       69.0   15.1      41.3    20  50708 alabama
## 2        365   6315        1.5       69.3   11.3      66.7   152 566432 alaska 
## 3       2212   4530        1.8       70.6    7.8      58.1    15 113417 arizona
```

- Need to match the region with the actual geographic boundaries

- Many geographic boundaries/features are stored as "shapefiles"

- i.e., complicated polygons
  
- Can contain the lines, points, etc. to represent any geographic feature

- Shapefiles are readily available for countries, states, counties, etc.

---

## Access shapefiles using `map_data()`

.pull-left[

```r
library(maps)
state_borders <- map_data("state") 
head(state_borders)
```

```
##        long      lat group order  region subregion
## 1 -87.46201 30.38968     1     1 alabama      <NA>
## 2 -87.48493 30.37249     1     2 alabama      <NA>
## 3 -87.52503 30.37249     1     3 alabama      <NA>
## 4 -87.53076 30.33239     1     4 alabama      <NA>
## 5 -87.57087 30.32665     1     5 alabama      <NA>
## 6 -87.58806 30.32665     1     6 alabama      <NA>
```

- For example: `map_data("world")`, `map_data("state")`, `map_data("county")` (need to install [`maps` package](https://cran.r-project.org/web/packages/maps/maps.pdf))

- Contains lat/lon coordinates to draw geographic boundaries

]

.pull-right[

- Typical workflow for plotting areal data (e.g., using states):

- Get state-specific data
  
  - Get state boundaries
  
  - Merge state-specific data with state boundaries (using `left_join()`)

```r
state_plot_data <- state_borders %>%
* left_join(state_data,
*           by = c("region" = "state"))
```
  
  
  - Then plot the data

]

---

## Create a choropleth map with `geom_polygon()`

.pull-left[

```r
state_plot_data %>%
  ggplot() + 
* geom_polygon(aes(x = long, y = lat,
*                  group = group,
*                  fill = Illiteracy),
*              color = "black") +
  scale_fill_gradient2(
    low = "darkgreen",
    mid = "lightgrey", 
    high = "darkorchid4",
    midpoint = 0.95) +
  theme_void() +
* coord_map("polyconic") +
  labs(fill = "Illiteracy %") + 
  theme(legend.position = "bottom")
```

]
.pull-right[
<img src="figs/Lec10/unnamed-chunk-18-1.png" width="100%" />

]

---

## Uniform size with [`statebins`](https://github.com/hrbrmstr/statebins)

.pull-left[

```r
library(statebins)
state_data$new_state <- 
  str_to_title(state_data$state)
*statebins(state_data = state_data,
          state_col = "new_state",
          value_col = "Illiteracy") +
  theme_statebins()
```

- Make all states equal in size

- Keeps spatial arrangement but focus is on the variable of interest displayed by the color

]

.pull-right[
<img src="figs/Lec10/unnamed-chunk-19-1.png" width="100%" />

]

---
class: center, middle

# Next time: Visualizing text data

__HW5 due Wednesday! Graphics Critique / Replication #2 due Friday Oct 7th!__