Visualize urban growth

The General Directorate for the Cadastre of Spain has spatial information of the all buildings except for the Basque Country and Navarra. This data set is part of the implementation of INSPIRE, the Space Information Infrastructure in Europe. More information can be found here. We will use the links (urls) in ATOM format, which is an RSS type for web feeds, allowing us to obtain the download link for each municipality.

This blog post is a reduced version of the case study that you can find in our recent publication - Introduction to GIS with R - published by Dominic Royé and Roberto Serrano-Notivoli (in Spanish).


Package Description
tidyverse Collection of packages (visualization, manipulation): ggplot2, dplyr, purrr, etc.
sf Simple Feature: import, export and manipulate vector data
fs Provides a cross-platform, uniform interface to file system operations
lubridate Easy manipulation of dates and times
feedeR Import feeds RSS or ATOM
tmap Easy creation of thematic maps
classInt Create univariate class intervals
sysfonts Loading system fonts and Google Fonts
showtext Using fonts more easily in R graphs
# install the packages if necessary
if(!require("tidyverse")) install.packages("tidyverse")
if(!require("feedeR")) install.packages("feedeR")
if(!require("fs")) install.packages("fs")
if(!require("lubridate")) install.packages("lubridate")
if(!require("fs")) install.packages("fs")
if(!require("tmap")) install.packages("tmap")
if(!require("classInt")) install.packages("classInt")
if(!require("showtext")) install.packages("showtext")
if(!require("sysfonts")) install.packages("sysfonts")
if(!require("rvest")) install.packages("rvest")

# load packages

Data download

The download is done with the download.file() function that only has two main arguments, the download link and the path with the file name. In this case, we use the tempfile() function, which is useful for creating temporary files, that is, files that only exist in the memory for a certain time. The file we download has the extension *.zip, so we must unzip it with another function (unzip()), which requires the name of the file and the name of the folder, where we want to unzip it. Finally, the URLencode() function encodes an URL address that contains special characters.

# create a temporary file
temp <- tempfile()

# download the data
download.file(URLencode(val_link), temp)

# unzip to a folder called buildings
unzip(temp, exdir = "buildings_valencia") # change the name according to the city

Import the data

To import the data we use the dir_ls() function of the fs package, which can obtain the files and folders of a specific path while filtering through a text pattern (regexp : regular expression). We apply the st_read() function of the sf package to the Geography Markup Language (GML) file.

# get the path with the file
file_val <- dir_ls("buildings_valencia", regexp = "building.gml") # change the folder if needed

# import the data
buildings_val <- st_read(file_val)
## Reading layer `Building' from data source 
##   `E:\GitHub\blog_update_2021\content\en\post\2019-11-01-visualize-urban-growth\buildings_valencia\A.ES.SDGC.BU.46900.building.gml' 
##   using driver `GML'
## Simple feature collection with 36284 features and 24 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 720570.9 ymin: 4351286 xmax: 734981.9 ymax: 4382906
## Projected CRS: ETRS89 / UTM zone 30N

Data preparation

We only have to convert the column of the construction year (beginning) into a Date class. The date column contains some dates in --01-01 format, which does not correspond to any recognizable date. Therefore, we replace the first - with 0000.

buildings_val <- mutate(buildings_val, 
               beginning = str_replace(beginning, "^-", "0000") %>% 
                            ymd_hms() %>% as_date()
## Warning: 4 failed to parse.

Distribution chart

Before creating the maps of the construction years, which will reflect urban growth, we will make a graph of distribution of the beginning variable. We can clearly identify periods of urban expansion. We will use the ggplot2 package with the geometry of geom_density() for this purpose. The font_add_google() function of the sysfonts package allows us to download and include font families from Google.

#font download
sysfonts::font_add_google("Montserrat", "Montserrat")

#use showtext for fonts
# limit the period after 1750
filter(buildings_val, beginning >= "1750-01-01") %>%
 ggplot(aes(beginning)) + 
    geom_density(fill = "#2166ac", alpha = 0.7) +
  scale_x_date(date_breaks = "20 year", 
               date_labels = "%Y") +
  theme_minimal(base_family = "Montserrat") +
  labs(y = "",x = "", title = "Evolution of urban development")

Buffer of 2,5 km for Valencia

To visualize better the distribution of urban growth, we limit the map to a radius of 2.5 km from the city center. Therefore, we use the geocode_OSM() function of the tmaptools package to obtain the coordinates of Valencia in class sf. Then we project the points to the system we use for the buildings (EPSG: 25830). The st_crs() function returns the coordinate system of a spatial object sf. Finally, we create with the function st_buffer() a buffer with 2500 m and the intersection with our building data. It is also possible to create a buffer in the form of a rectangle indicating the style with the argument endCapStyle =" SQUARE ".

# get the coordinates of Valencia
ciudad_point <- tmaptools::geocode_OSM("Valencia", 
                                      as.sf = TRUE)

#  project the points
ciudad_point <- st_transform(ciudad_point, st_crs(buildings_val))

# create the buffer
point_bf <- st_buffer(ciudad_point, 2500) # radius of 2500 m

# get the intersection between the buffer and the building
buildings_val25 <- st_intersection(buildings_val, point_bf)
## Warning: attribute variables are assumed to be spatially constant throughout all
## geometries

Prepare data for mapping

We categorize the year into 15 groups using quartiles. It is also possible to modify the number of classes or the applied method (eg jenks, fisher, etc), you can find more details in the help ?classIntervals.

# find 15 classes
br <- classIntervals(year(buildings_val25$beginning), 15, "quantile")
## Warning in classIntervals(year(buildings_val25$beginning), 15, "quantile"): var
## has missing values, omitted in finding classes
# create labels
lab <- names(print(br, under = "<", over = ">", cutlabels = FALSE))
## style: quantile
##      < 1890 1890 - 1912 1912 - 1925 1925 - 1930 1930 - 1940 1940 - 1950 
##         932        1350         947         594        1703        1054 
## 1950 - 1958 1958 - 1962 1962 - 1966 1966 - 1970 1970 - 1973 1973 - 1978 
##        1453        1029        1223        1158        1155        1190 
## 1978 - 1988 1988 - 1999      > 1999 
##        1149        1111        1244
# categorize the year
buildings_val25 <- mutate(buildings_val25, 
                          yr_cl = cut(year(beginning), 
                                       labels = lab, 
                                       include.lowest = TRUE))

Map of Valencia

For the mapping, we will use the tmap package. It is an interesting alternative to ggplot2. It is a package of functions specialized in creating thematic maps. The philosophy of the package follows the same as in ggplot2, creating multiple layers with different functions, which always start with tm_ *and combine with +. Building a map with tmap always starts with tm_shape(), where the data, we want to draw, is defined. Then we add the corresponding geometry to the data type (tm_polygon(), tm_border(), tm_dots() or even tm_raster()). The tm_layout() function help us to configure the map style.

When we need more colors than the maximum allowed by RColorBrewer, we can pass the colors to the colorRampPalette() function. This function interpolates a set of given colors.

# colours
col_spec <- RColorBrewer::brewer.pal(11, "Spectral")

# colour ramp function
col_spec_fun <- colorRampPalette(col_spec)

# create the final map
tm_shape(buildings_val25) +
              border.col = "transparent",
              palette = col_spec_fun(15), # adapt to the number of classes
              textNA = "Without data",
              title = "") +
 tm_layout(bg.color = "black",
  = "black",
           legend.outside = TRUE,
           legend.text.color = "white",
           legend.text.fontfamily = "Montserrat", 
            panel.label.fontfamily = "Montserrat",
            panel.label.color = "white",
   = "black",
            panel.label.size = 5,
            panel.label.fontface = "bold")

We can export our map using the function tmap_save("name.png", dpi = 300). I recommend using the dpi = 300 argument for a good image quality.

An alternative way to the tmap package is ggplot2.

# create the final map
ggplot(buildings_val25) +
     geom_sf(aes(fill = yr_cl), colour = "transparent") +
  scale_fill_manual(values = col_spec_fun(15)) + # adapt to the number of classes
    labs(title = "VALÈNCIA", fill = "") +
  guides(fill = guide_legend(keywidth = .7, keyheight = 2.7)) +
theme_void(base_family = "Montserrat") +
theme(panel.background = element_rect(fill = "black"),
      plot.background = element_rect(fill = "black"),
      legend.justification = .5,
      legend.text = element_text(colour = "white", size = 12),
      plot.title = element_text(colour = "white", hjust = .5, size = 60,
      margin = margin(t = 30)),
      plot.caption = element_text(colour = "white",
      margin = margin(b = 20), hjust = .5, size = 16),
      plot.margin = margin(r = 40, l = 40))

To export the result of ggplot we can use the function ggsave("name.png").

Dynamic map with leaflet

A very interesting advantage is the tmap_leaflet() function of the tmap package to easily pass a map created in the same frame to leaflet.

# tmap object
m <-   tm_shape(buildings_val25) +
              border.col = "transparent",
              palette = col_spec_fun(15), # adapt to the number of classes
              textNA = "Without data",
              title = "")

# dynamic map

Buy Me A Coffee

Dr. Dominic Royé
Dr. Dominic Royé
Researcher and responsible for data science