How to create a beautiful map with R and geom_sf [part 2]

Last week in Mapping Texas Ports with R [part 1], we created a simple map of Texas ports with R, ggplot2, and geom_sf.

That map was really just a “rough draft.” It’s not terrible, but it didn’t look great either.

This week, we’re going to take that map and polish it up a little bit.

Let’s get started.

Run preliminary code

First, you’ll need to run some preliminary code.

This code is very similar to the code in part 1, with a few minor modifications (e.g., I made some modifications to the port names, etc).

#================
# import packages
#================
library(tidyverse)
library(sf)
library(ggspatial)
library(rnaturalearth)
library(tidygeocoder)
library(maps)
library(ggrepel)


#=============
# GET MAP DATA
#=============
world_map_data <- ne_countries(scale = "medium", returnclass = "sf")
state_map_data <- map('state', fill = TRUE, plot = FALSE) %>% st_as_sf()

class(world_map_data)
class(state_map_data)



#------------------
# CREATE SIMPLE MAP
#------------------
state_map_data %>% 
  filter(ID == 'texas') %>% 
  ggplot() +
    geom_sf()


#--------------------------
# DRAFT: Map of Texas Coast
#--------------------------
ggplot() +
  geom_sf(data = world_map_data) +
  geom_sf(data = state_map_data) +
  coord_sf(xlim = c(-100, -91), ylim = c(25,33))
  


#=====================
# CREATE LIST OF PORTS
#=====================

portlist = c('Port Brownsville, Texas'
            ,'Port Isabel, Texas'
            ,'Port Mansfield, Texas'
            ,'Port Corpus Christi, Texas'
            ,'Port Lavaca, Texas'
            ,'Port Freeport, Texas'
            ,'Texas City, Texas'
            ,'Port Galveston, Texas'
            ,'Port Houston, Texas'
            ,'Port Sabine Pass, Texas'
            ,'Port Arthur, Texas'
            ,'Port Beaumont, Texas'
            ,'Port of Orange, Texas'
            )


#geo_osm('Port of Texas City, Texas')

#--------------
# CREATE TIBBLE
#--------------
port_data = tibble(location = portlist)


#--------------------
# CREATE 'BRIEF' NAME
#--------------------
port_data %>% 
  mutate(location_brief = str_replace(location, ', Texas', '')) ->
  port_data



#---------------------------------
# CREATE EMPTY LAT, LONG VARIABLES
#---------------------------------
port_data %>% 
  mutate(lat = NA
         ,long = NA
         ) ->
  port_data
  

#inspect
head(port_data)


#------------------
# GEOCODE LOCATIONS
#------------------
for(i in 1:nrow(port_data)){
  coordinates = geo_osm(port_data$location[i])
  port_data$long[i] = coordinates$long
  port_data$lat[i] = coordinates$lat
}


#inspect
head(port_data)

You’ll need to run that code, because it has some of the building blocks that we need going forward.

After you run it, you can create our rough draft from part 1:

#--------------------------
# DRAFT: Map of Texas Coast
#--------------------------
ggplot() +
  geom_sf(data = world_map_data) +
  geom_sf(data = state_map_data) +
  geom_point(data = port_data, aes(x = long, y = lat), color = 'red') +
  coord_sf(xlim = c(-100, -92), ylim = c(25,33))

OUT:

Again … this is really rough around the edges, so to speak.

In the next step, we’ll make it look good.

Polishing up the Texas map

We’ll improve this in steps.

We’re going:

• to create a theme to modify the fonts and colors
• create an updated, themed plot
• add the state labels
• add the port names
• adjust the port name positions

Let’s go …

Create theme

Here, we’re going to create a “theme” that will format the plot elements of our chart.

Specifically, it will do things like:

• change the font for the text
• change the background color
• change the gridline color
• change the font size for the title, subtitle, and other text

To do this, we’re going to use the ggplot theme function, and change specific plot elements.

#-------------
# CREATE THEME
#-------------

mytheme <- theme(text = element_text(family = 'Avenir')
                 ,panel.grid.major = element_line(color = '#cccccc' 
                                                  ,linetype = 'dashed'
                                                  ,size = .3
                                                  )
                 ,panel.background = element_rect(fill = 'aliceblue')
                 ,plot.title = element_text(size = 32)
                 ,plot.subtitle = element_text(size = 14)
                 ,axis.title = element_blank()
                 ,axis.text = element_text(size = 10)
                 )

Notice that we're changing the color of panel.background to 'aliceblue'. That will make the color of the ocean on the map a light shade of blue.

Also note that we're saving this theme syntax as mytheme. That's one great thing about ggplot2 ... you can save your theme code with a name, and then re-use it for multiple plots.

Create 'themed' map of Texas ports with ggplot and geom_sf

Next, we'll apply our theme and create a themed map (i.e., a map that has updated colors, etc).

Here, we're using ggplot() in combination with the geom_sf function to create the basic map with the country and state shapes.

Notice also that we're applying mytheme to the plot.

We're also making some modifications to the point sizes and the color of the land on the map. We're actually using geom_point twice. One is a semi-transparent point that identifies a plot location. The second use of geom_point is creating a fully opaque border around those points.

These are somewhat subtle design choices. They aren't hard to do, but you need to know a few tricks to understand how to execute them. Moreover, you really need to learn enough about plot design to realize that it might be a good idea to plot the data like this.

#-------------------------------------
# CREATE BASE PLOT: Map of Texas Coast
#-------------------------------------
land_color <- c('antiquewhite1')

base_plot <- ggplot() +
  geom_sf(data = world_map_data, fill = land_color, size = .4) +
  geom_sf(data = state_map_data, fill = NA, size = .4) +
  geom_point(data = port_data, aes(x = long, y = lat), size = 4, color = 'red', alpha = .15) +
  geom_point(data = port_data, aes(x = long, y = lat), size  = 4, shape = 1,  color = 'red') +
  coord_sf(xlim = c(-100, -90), ylim = c(25,33)) +
  mytheme

Next, we can plot the chart, base_plot by using print():

#---------
# SHOW MAP
#---------
print(base_plot)

OUT:

This already looks a lot better.

Notice that we've changed the land color and the ocean color. We changed the land color with the fill= parameter of geom_sf. We changed the ocean color with the panel.background theme element. Most of the other modifications were also made with the theme changes.

Create labels for state name data

Next, we're going to modify our state-level data to make some labels that we can add to the plot.

There's a few things we need to do. We need to change the state names (the ID variable) to title case.

We need to calculate the center of the state (where we want to add those state name labels), and add those centroid X and Y coordinates to the dataset.

And we also need to add some "nudge" variables that will enable us to move the labels a little away from the centroid, as needed.

All of this is a little complicated. Not terribly, but a little.

Notice though that we're mostly just using dplyr functions like mutate() and then some functions from the sf package that help us calculate the centroids.

#----------------------
# CHANGE STATE NAME
# change to "title case"
#----------------------
state_map_data %>% 
  mutate(ID = str_to_title(ID)) ->
  state_map_data


names(state_map_data)


#--------------------
# ADD STATE CENTROIDS
#--------------------
state_map_data %>% 
  mutate(centroid = st_centroid(geom)) ->
  state_map_data



#------------------------
# ADD X AND Y COORDINATES
#------------------------
statename_coords <- state_map_data %>% 
  st_centroid() %>% 
  st_coordinates() %>%
  as_tibble()

state_map_data %>%  
  bind_cols(statename_coords) %>% 
  select(ID, X, Y, centroid, geom) ->
  state_map_data



#----------------------------
# ADD OFFSETS FOR STATE NAMES
#----------------------------
state_map_data %>% 
  mutate(x_nudge = case_when( ID == 'Texas' ~ 1.3
                              ,ID == 'Louisiana' ~ -.6
                              ,ID == 'Mississippi' ~ 1.5
                              ,TRUE ~ 0
                              )
         ,y_nudge = case_when( ID == 'Texas' ~ .5
                              ,ID == 'Louisiana' ~ 1
                              ,TRUE ~ 0
                              )
         ) -> 
  state_map_data

From here, we'll use geom_text() to create some labels that we can add to our plot, which we'll save as state_names.

state_names <- geom_text(data = state_map_data
                    ,aes(x = X, y = Y, label = ID)
                    ,color = "#333333"
                    ,size = 4
                    ,fontface = 'bold'
                    ,nudge_x = state_map_data$x_nudge
                    ,nudge_y = state_map_data$y_nudge
                    )

And now we can plot:

#----------
# ADD NAMES
#----------
base_plot + 
  state_names

OUT:

Better.

We're getting close.

Add port names

Now, we'll add the port names.

First, let's just do a simple trial of this.

Draft of map with port names

Here, we'll just do a dry run and try to add the port names with geom_text().

#---------------
# ADD PORT NAMES
#---------------
base_plot + 
  state_names +
  geom_text(data = port_data
            ,aes(x = long, y = lat, label = location_brief)
            ,family = 'Avenir')

OUT:

Ok, I'll be honest. This is a f*#^ing mess.

We need to "nudge" those port names to new locations.

Move port name labels

Here we're going to move the labels to new positions, slightly offset from the actual port location.

To do this, we'll ultimately use geom_text_repel(), which adds text labels, but also repels those labels away from one another, so they do not overlap.

To make this work we first need to create some offsets.

Create label offests

Here, we're going to create some offset variables called x_nudge and y_nudge. These will eventually tell geom_text_repel() to "nudge" the text label away from the actual label location by a small amount in the x and y direction.

Here, we're adding these variables with the dplyr::mutate() function, in combination with case_when, which allows us to conditionally create different offsets for different ports.

#----------------------------------------------
# CREATE X AND Y 'NUDGE' OFFSETS FOR PORT NAMES
#----------------------------------------------
port_data %>% 
  mutate(x_nudge = case_when( location == 'Port Brownsville, Texas' ~ 1.3
                             ,location == 'Port Isabel, Texas' ~ 1.3
                             ,location == 'Port Mansfield, Texas' ~ 1.5
                             ,location == 'Port Corpus Christi, Texas' ~ 1.5
                             ,location == 'Port Lavaca, Texas' ~ -1
                             ,location == 'Port Freeport, Texas' ~ 1
                             #,location == 'Port of Texas City, Texas' ~ 0
                             ,location == 'Texas City, Texas' ~ -1
                             ,location == 'Port Galveston, Texas' ~ 1
                             ,location == 'Port Houston, Texas' ~ -1.5
                             ,location == 'Port Sabine Pass, Texas' ~ .5
                             ,location == 'Port Arthur, Texas' ~ 1
                             ,location == 'Port Beaumont, Texas' ~ -.6
                             ,location == 'Port of Orange, Texas' ~ 1.6
                             ,TRUE ~ 0)
         ,y_nudge = case_when( location == 'Port Brownsville, Texas' ~ -1
                             ,location == 'Port Isabel, Texas' ~ 0
                             ,location == 'Port Mansfield, Texas' ~ .2
                             ,location == 'Port Corpus Christi, Texas' ~ 0
                             ,location == 'Port Lavaca, Texas' ~ .5
                             ,location == 'Port Freeport, Texas' ~ -.5
                             ,location == 'Texas City, Texas' ~ 0
                             ,location == 'Port Galveston, Texas' ~ -.5
                             ,location == 'Port Houston, Texas' ~ .8
                             ,location == 'Port Sabine Pass, Texas' ~ -.5
                             ,location == 'Port Arthur, Texas' ~ .1
                             ,location == 'Port Beaumont, Texas' ~ .6
                             ,location == 'Port of Orange, Texas' ~ .5
                             ,TRUE ~ 0)
  ) ->
  port_data

Ok. Let's try to plot again.

Plot map, with port labels and offsets

So finally, we're going to put everything together.

We're going to use the base plot that we created earlier and saved with the name base_plot.

We'll add the state names with the state_names object we created earlier.

And we'll use geom_text_repel() to add the port names. Notice that we're using the parameters nudge_x and nudge_y to pass in the offsets that we just created in the previous section. Ultimately, geom_text_repel() will add the labels with those offsets, and then use an iterative process to "repel" the names away from each other until they don't overlap.

Notice that we're also using using the labs() function to add a title and subtitle.

Ok, let's do it.

#==================
# CREATE FINAL PLOT
#==================
base_plot + 
  state_names +
  geom_text_repel(data = port_data
                  ,aes(x = long
                       ,y = lat
                       ,label = location_brief
                       )
                  ,family = 'Avenir'
                  ,nudge_x = port_data$x_nudge
                  ,nudge_y = port_data$y_nudge
                  ,segment.color = "#333333"
                  ) +
  labs(title = '13 Texas Ports'
       ,subtitle = 'Texas has over a dozen excellent ports, many of which are under-utilized')

OUT:

Alright!

This looks really pretty good.

There is probably a few other things that we might want to do here, but I'm very satisfied with this.

Notice that all of the port names are offset away from the points and none of them overlap.

To be honest, this is partially due to geom_text_repel() working it's magic, but it's also from a lot of trial and error from me manually modifying the offsets. It was a little challenging to get "just right," and really required a lot of iteration.

Final notes

Much of the code here was based off an example of how to create maps with the sf pacakge over at rspatial.org.

Their example was part of the inspiration for this tutorial series. I used their code as a starting point, although I heavily modified it to match my data and my map, as well as to match my particular programming style (for example, I used case_when to add the offsets).

If you're interested in creating maps in R programmatically, you should check out r-spatial.org.

Supply chain analytics will probably become important

To bring this back to my original motivation in part 1, I should note that it might be good to learn about geospatial data visualization.

For a variety of reasons, I think we're likely to have a lot more spatial information going forward ... from devices and sensors that will increasingly be added to tech products.

Additionally, with all of the supply chain reorientation happening right now, I think there will be more demand for fine-grained supply chain analytics. This tutorial doesn't cover everything you'd need to know ... not by a longshot. But it's something to keep in mind, and you might want to skill up.

Sign up to increase your data skills

If you want to skill up and increase your data science skills, sign up for our email list.

Every week, we publish free data science tutorials.

When you sign up for our email list, you’ll get all of our tutorials delivered directly to your inbox.

... we'll help you learn data science so you can take advantage of all of the opportunities that are emerging in the data industry.