3-mapping

<h1 class="center middle">Visualizing Spatial Data</h1>

<h3 class="center middle">(static and interactive maps)</h3>

---

# R has great map-making functionality!

---

# This map was created in R

![:imgcenterscale 80%](images/section-3/bivariate_switzerland.png)

.footnote[
[Source: Timo Grossenbacher](https://timogrossenbacher.ch/2019/04/bivariate-maps-with-ggplot2-and-sf/)
]

---

# This map was created in R

![:imgcenterscale 80%](images/section-3/london_cycle_map.png)

---

# This map was created in R

![:imgcenterscale 75%](images/section-3/rayshader.gif)

---

# What packages should you use?

---

# There are dozens of mapping-related packages in R

* But only a few are all-purpose

---

# All-purpose static mapping

* The `plot()` function
* {tmap}
* {ggplot2} (we won't cover {ggplot2})

---

# All-purpose interactive mapping

* {tmap}
* {mapview}
* {leaflet} (we won't cover {leaflet})

---

exclude:true

# Color palettes

* [{viridis}](https://cran.r-project.org/web/packages/viridis/vignettes/intro-to-viridis.html)

* [{RColorBrewer}](https://cran.r-project.org/web/packages/RColorBrewer/index.html)

* [{swatches}](https://github.com/hrbrmstr/swatches)

* [{ggthemes}](https://cran.r-project.org/web/packages/ggthemes/index.html)

---

# A lot of great packages for niche mapping needs

* {rayshader}
* {geogrid}
* {globe}
* {linemap}
* {cartogram}
* {cartography}
* {mapedit}
* {rasterVis}

---

background-image: url("images/background-old.jpg")
background-size: cover
class: center, middle, separator-slide

# Static mapping

---

# Take home messages

* I use `plot()` for a quick look at data

* I use {tmap} for everything else

* {tmap} has an insane amount of customization allowed, we will only touch the surface

---

# Why not `ggplot2`?

* I love `ggplot2` and use it every day

* But I find making maps significantly easier in `tmap`
  * Can do almost anything in `tmap`
  * Easy to include just fill, just borders, etc
  * Interactive views
  * etc...

---

# But I will show a ggplot-map at the end of this section

---

# `plot()`

---

# Yes `plot()` is not super exciting

![:imgcenterscale 80%](images/section-3/bored-kid.jpg)

---

# But `plot()` is great for a quick look at your data

---

# `plot()` has methods for vector and raster data

The packages need to be loaded to plot vector and raster data

---

# Some setup: load packages

No need to type with me, you'll practice in the exercise

```r
library(sf)     # read vectors
library(raster) # read rasters
library(tmap)   # mapping
```

---

# Some setup: read in NYC data

```r
boroughs <- read_sf("boroughs.gpkg")
schools <- read_sf("schools.shp")
canopy <- raster("canopy.grd")
```

---

# With vectors the default is to plot the attributes

```r
plot(boroughs)
```

![:imgcenterscale 80%](images/section-3/plotboroughs.png)

---

# I don't really like this default

![:imgcenterscale 50%](images/section-3/hanks.png)

---

# You can plot a single attribute

```r
plot(boroughs['Shape_Area'], main = "Area",
     pal = rev(heat.colors(5)))
```
![:imgcenterscale 60%](images/section-3/plotoneattribute.png)

---

# For a quick look at vectors, I often just want the geometry

* You can extract just the geometry with `st_geometry()`

* Then call `plot()` on the output

---

# `st_geometry()` and then plot

```r
st_geometry(boroughs) %>% plot()
```

![:imgcenterscale 60%](images/section-3/plotstgeometry.png)

---

# Or wrap the two functions

```r
plot(st_geometry(boroughs))
```

![:imgcenterscale 60%](images/section-3/plotstgeometry.png)

---

# Try to memorize this concept/code

```r
st_geometry(boroughs) %>% plot()
```

```r
plot(st_geometry(boroughs))
```

---

# To combine layers with `plot()` use `add = TRUE`

```r
plot(st_geometry(boroughs))
plot(st_geometry(schools), add = TRUE,
     pch = 16, col = "red", cex = 0.5)
```

![:imgcenterscale 50%](images/section-3/plotwithpoints.png)

---

# Careful with `add = TRUE`, only works with `st_geometry()`

```r
# This won't work!
plot(boroughs)
plot(schools, add = TRUE)
```

---

# Easy to use `plot()` with rasters

```r
plot(canopy) # a single band raster
```

![:imgcenterscale 100%](images/section-3/plotcanopy.png)

---

# If your raster has multiple layers (like an image) ...

`plot()` maps each layer separately

```r
plot(manhattan)
```

![:imgcenterscale 70%](images/section-3/manhattanplot.png)

---

# If your raster has a red, green and blue layer (like an image)

You can plot them together...

```r
plotRGB(manhattan)
```

![:imgcenterscale 50%](images/section-3/manhattan.png)

---

# Combine rasters and vectors with `plot()`

```r
plot(canopy)
plot(st_geometry(schools), add = T,
     pch = 16, col = "red", cex = 0.5)
```

![:imgcenterscale 90%](images/section-3/plotcanopyschools.png)

---

# For more sophisticated and fun static maps I use {tmap}

---

# {tmap}

---

# So much control with {tmap}!

```r
tm_text(text, size = 1, col = NA, root = 3, clustering = FALSE,
  size.lim = NA, sizes.legend = NULL, sizes.legend.labels = NULL,
  sizes.legend.text = "Abc", n = 5, style = ifelse(is.null(breaks),
  "pretty", "fixed"), breaks = NULL, interval.closure = "left",
  palette = NULL, labels = NULL, labels.text = NA, midpoint = NULL,
  stretch.palette = TRUE, contrast = NA, colorNA = NA,
  textNA = "Missing", showNA = NA, colorNULL = NA, fontface = NA,
  fontfamily = NA, alpha = NA, case = NA, shadow = FALSE,
  bg.color = NA, bg.alpha = NA, size.lowerbound = 0.4,
  print.tiny = FALSE, scale = 1, auto.placement = FALSE,
  remove.overlap = FALSE, along.lines = FALSE,
  overwrite.lines = FALSE, just = "center", xmod = 0, ymod = 0,
  title.size = NA, title.col = NA, legend.size.show = TRUE,
  legend.col.show = TRUE, legend.format = list(),
  legend.size.is.portrait = FALSE, legend.col.is.portrait = TRUE,
  legend.size.reverse = FALSE, legend.col.reverse = FALSE,
  legend.hist = FALSE, legend.hist.title = NA, legend.size.z = NA,
  legend.col.z = NA, legend.hist.z = NA, group = NA,
  auto.palette.mapping = NULL, max.categories = NULL)
```

---

# {tmap} syntax is similar to {ggplot2}

```r
# data set up           layer               
tm_shape(boroughs) + tm_polygons()
```

![:imgcenterscale 60%](images/section-3/tmapborough.png)

---

# A short-cut `qtm()`

Instead of `plot()` I often use this

```r
qtm(boroughs)
```

![:imgcenterscale 60%](images/section-3/tmapborough.png)

---

# Add multiple layers based on one input

```r
tm_shape(boroughs) + 
  tm_polygons() +   
  tm_dots(size = 2) +
  tm_text("BoroName", col = "red", size = 1.5)
```

![:imgcenterscale 100%](images/section-3/tmapboroughcentroidtext.png)

---

# Or multiple different layers using multiple shapes

```r
tm_shape(boroughs) + tm_borders() +
  tm_shape(schools) + tm_dots(size = 0.25)
```

![:imgcenterscale 50%](images/section-3/boroughs_and_schools.png)

---

# You can also save parts of the map and reuse

```r
mymap <- tm_shape(boroughs) + tm_polygons()
```

```r
mymap + tm_dots(size = 2) +
  tm_text("BoroName", col = "red")
```

![:imgcenterscale 90%](images/section-3/tmapboroughcentroidtext.png)

---

# Choropleth (color-coded map) based on a variable. So easy!

```r
tm_shape(boroughs) + tm_polygons("Shape_Area")
```

![:imgcenterscale 60%](images/section-3/tmapchoropleth.png)

---

# Or show an attribute with `tm_symbols()`

```r
tm_shape(boroughs) + tm_borders() +
  tm_symbols("Shape_Area", scale = 2)
```

![:imgcenterscale 60%](images/section-3/tm_symbols.png)

---

# Plot multiple variables at once

```r
tm_shape(boroughs) + tm_polygons(c("Shape_Area", "BoroName"))
```

![:imgcenterscale 100%](images/section-3/Plot_multiple_variables.png)

---

# `tm_shape()` can accept vector or raster

![:imgcenterscale 100%](images/section-3/cheering_minions.gif)

---

# Single-band raster with `tm_raster()`

```r
tm_shape(canopy) + tm_raster()
```

![:imgcenterscale 60%](images/section-3/tmapcanopy.png)

---

# A multi-layer raster with `tm_rgb()`

```r
tm_shape(manhattan) + tm_rgb()
```

![:imgcenterscale 60%](images/section-3/manhattan.png)

---

# Vector and raster

```r
tm_shape(manhattan) + tm_rgb()+
  tm_shape(boroughs) + tm_borders(col = "white", lwd = 2)
```

![:imgcenterscale 60%](images/section-3/manhattanwithoutline.png)

---

# Include a basemap in your map

Use a function from the companion package, {tmaptools}

```r
osmtiles <- tmaptools::read_osm(boroughs,
  type="stamen-terrain")
```

---

# Include a basemap in your map

```r
tm_shape(osmtiles) + tm_raster() +
  tm_shape(boroughs) +
  tm_borders(lwd = 2, col = "yellow")
```

![:imgcenterscale 60%](images/section-3/osm_tiles.png)

---

# Map extent is driven by the "master"

* First shape is master by default

* in tm_shape() can use `is.master = TRUE`

---

# Here is the default (extent based on raster)

```r
tm_shape(canopy) +
  tm_raster(title = "(percent canopy)", alpha = 0.75) +
  tm_shape(boroughs) + tm_borders(lwd = 2, col = "blue")
```

![:imgcenterscale 50%](images/section-3/preismaster.png)

---

# Force the extent to be the polygon borders

```r
tm_shape(canopy) +
  tm_raster(title = "(percent canopy)", alpha = 0.75) +
  tm_shape(boroughs, is.master = TRUE) + tm_borders(lwd = 2, col = "blue")
```

![:imgcenterscale 50%](images/section-3/tm_is_master.png)

---

# Using color palettes in `tmap`

* [`viridis`](https://cran.r-project.org/web/packages/viridis/vignettes/intro-to-viridis.html)
* [`RColorBrewer`](http://colorbrewer2.org/#type=sequential&scheme=BuGn&n=3)

---

# `palette_explorer()` is great

```r
tmaptools::palette_explorer()
```

![:imgcenterscale 75%](images/section-3/palette_explorer.png)

---

# Use a palette with the `pal` argument

```r
tm_shape(neighborhoods) +
  tm_polygons("shape_area", n = 5, pal = "Greens")
```

![:imgcenterscale 60%](images/section-3/neighborhoodsgreen.png)
---

# Use `-` to reverse the palette

```r
tm_shape(neighborhoods) +
  tm_polygons("shape_area", n = 5, pal = "-Greens")
```

![:imgcenterscale 60%](images/section-3/neighborhoodsminusgreen.png)

---

# Alter the map layout with `tm_layout()`

```r
tm_shape(neighborhoods) +
  tm_polygons("shape_area", n = 5, pal = "Greens") +
  tm_layout(legend.show = FALSE, frame = FALSE)
```

![:imgcenterscale 60%](images/section-3/neighborhoodsgreen_nolegend.png)

---

# Save your maps with `tmap_save()`

```r
mymap <- tm_shape(boroughs) + tm_polygons()
tmap_save(mymap, "mymap.png")
```

---

# It's easy to include more than one map in an image with `tmap_arrange()`

```r
# Create three maps
m1 <- tm_shape(boroughs) + tm_polygons()
m2 <- tm_shape(neighborhoods) + tm_polygons()
m3 <- tm_shape(schools) + tm_dots(size = 0.25)
```

---

# Arrange them on one image

```r
tmap_arrange(m1, m2, m3, nrow = 1)
```

---

# So many great other functions to explore

* `tmap_save()`
* `tm_layout()`
* `tm_style()`
* `tm_facets()`
* `tm_animation()`
* `tm_scale_bar()`
* `tm_compass()`

---

# As promised earlier, three slides on mapping with {ggplot2}

---

# {ggplot2} has a special layer for {sf} objects

* `geom_sf()`

---

# With `geom_sf()` no need to specify x and y

* Unlike `geom_line()`, `geom_points()` etc

---

# To create a choropleth use `aes()` with the fill argument

```r
library(ggplot2)
ggplot() + 
  geom_sf(data = boroughs, aes(fill = BoroCode)) +
  geom_sf(data = schools, color = "purple")
```

---

# You can also make nice maps with {ggplot2}

![:imgcenterscale 80%](images/section-3/bivariate_switzerland.png)

.footnote[
[Source: Timo Grossenbacher](https://timogrossenbacher.ch/2019/04/bivariate-maps-with-ggplot2-and-sf/)
]

---

# Keep in mind, there are other packages worth exploring

---

# Quick example of geogrid

![:imgcenterscale 55%](images/section-3/geogrid.png)

---

# Hillshading with rayshader

![:imgcenterscale 75%](images/section-3/rayshader.gif)

---

# Cartographic representations with {cartography}

![:imgcenterscale 60%](images/section-3/cartography.png)

---
class: center, middle, doExercise

# open_exercise(3) and work on activity 1-9 then stop

---

background-image: url("images/background-old.jpg")
background-size: cover
class: center, middle, separator-slide

# Interactive maps

---

# There are two packages I use for interactive maps

* {mapview} (for a quick interactive look)
* {tmap} (for a more polished interactive map)

---

# Why not RStudio's {leaflet}

* Main reason is it requires layers to be unprojected (we will discuss)

* There are work-arounds like `leafletCRS()`

* But {tmap} and {mapview} do a great job

---

# Both {tmap} and {mapview} also use the Leaflet JavaScript API

![:imgcenterscale 100%](images/section-3/leaflet.png)

---

# Since we're already talking about {tmap} let's start with {tmap}

---

# Remember this static {tmap} from earlier?

![:imgcenterscale 50%](images/section-3/neighpointtxt.png)

---

# Use `tmap_mode()` to change from static to interactive

```r
tmap_mode("plot") # default
my_map
```

![:imgcenterscale 50%](images/section-3/neighpointtxt.png)

---

# So easy to make it interactive

```r
tmap_mode("view") # make interactive
my_map # recreate plot
```

![:imgcenterscale 100%](images/section-3/tmap_view_nyc.png)

---

# Running `tmap_mode()` with no argument will give the current mode

```r
tmap_mode()
# current tmap mode is "view"
```

---

# `tmap` can do side-by-side linked plots

```r
tmap_mode("view")
tm_shape(boroughs) +
  tm_polygons(c("pop2018", "pop_change"),
    palette = "Oranges")
```
![:imgcenterscale 100%](images/section-3/tmap_linked.png)
---

# Save your interactive tmap with `tmap_save()`

```r
mymap <- tm_shape(boroughs) + tm_polygons()
tmap_save(mymap, "mymap.html")
```

![:imgcenterscale 35%](images/section-3/tmapsave.png)

---

# {mapview}

---

# Great for a quick interactive look at data

```r
library(mapview)
mapview(boroughs)
```

![:imgcenterscale 100%](images/section-3/mapview_nyc.png)

---

# Like {tmap}, a lot of customization allowed

```r
mapview(x, map = NULL,
  maxpixels = mapviewGetOption("mapview.maxpixels"),
  col.regions = mapviewGetOption("raster.palette")(256), at = NULL,
  na.color = mapviewGetOption("na.color"), use.layer.names = FALSE,
  values = NULL, map.types = mapviewGetOption("basemaps"),
  alpha.regions = 0.8, legend = mapviewGetOption("legend"),
  legend.opacity = 1, trim = TRUE,
  verbose = mapviewGetOption("verbose"), layer.name = NULL,
  homebutton = TRUE, native.crs = FALSE, method = c("bilinear",
  "ngb"), label = TRUE, query.type = c("mousemove", "click"),
  query.digits, query.position = "topright", query.prefix = "Layer",
  viewer.suppress = FALSE, ...)
```

---

# And easier than...

```r
tmap_mode("view")
tm_shape(boroughs) + tm_polygons()
```

---

# Multiple layers in one map with `list()`

```r
library(mapview)
mapview(list(boroughs, schools))
```

![:imgcenterscale 100%](images/section-3/mapview_nyc_schools.png)

---

# Alternative syntax for multiple layers

```r
mapview(boroughs) + mapview(schools)
```

```r
mapview(boroughs) + schools
```

---

# Color-code based on an attribute use the `zcol` argument

```r
mapview(boroughs, zcol = "Shape_Area")
```

![:imgcenterscale 75%](images/section-3/mapview_zcol.png)

---

# Also allows rasters

```r
mapview(canopy, alpha.regions = 0.4)
```

![:imgcenterscale 100%](images/section-3/mapview_raster.png)

---

# Saving your mapview interactive map with `mapshot`

Output is very similar to {tmap} (html file and folder)

```r
mymap <- mapview(boroughs)
mapshot(mymap, "mymap.html")
```

---

# For both static and interactive maps

* You can include in R markdown

* You can include in shiny application

* If you save with `tmap_save()` or `mapshot()` you can upload the files directly to a server

---
class: center, middle, doExercise

# open_exercise(3) and finish