Chapter 7 Vector layers

Last updated: 2021-01-16 22:24:50


Our aims in this chapter are:

  • Become familiar with data structures for vector layers: points, lines and polygons
  • Examine spatial and non-spatial properties of vector layers
  • Create subsets of vector layers based on their attributes
  • Learn to transform a layer from one Coordinate Reference System (CRS) to another

We will use the following R packages:

  • sf
  • mapview
  • stars

7.1 Vector layers

7.1.1 What is a vector layer?

Vector layers are essentially sets of geometries associated with non-spatial attributes (Figure 7.1). The geometries are sequences of one or more point coordinates, possibly connected to form lines or polygons. The non-spatial attributes come in the form of a table.

Geometry (left) and non-spatial attributes (right) of vector layers (

Figure 7.1: Geometry (left) and non-spatial attributes (right) of vector layers (

7.1.2 Vector file formats

Commonly used vector layer file formats (Table 7.1) include binary formats (such as the Shapefile) and plain text formats (such as GeoJSON). Vector layers are also frequently kept in a spatial database, such as PostgreSQL/PostGIS.

Table 7.1: Common vector layer file formats
Type Format File extension
Binary ESRI Shapefile .shp, .shx, .dbf, .prj, …
GeoPackage (GPKG) .gpkg
Plain Text GeoJSON .json or .geojson
GPS Exchange Format (GPX) .gpx
Keyhole Markup Language (KML) .kml
Spatial Databases PostGIS / PostgreSQL

7.1.3 The sp package

The first R package to establish a uniform vector layer class system was sp, released in 2005. Together with rgdal (2003) and rgeos (2011), the sp package dominated the landscape of spatial analysis in R for many years.

The sp package defines 6 main classes for vector layers (Table 7.2):

  • One for each geometry type (points, lines, polygons)
  • One for geometry only and one for geometry with attributes
Table 7.2: Spatial data structures in package sp
Class Geometry type Attributes
SpatialPoints Points -
SpatialPointsDataFrame Points data.frame
SpatialLines Lines -
SpatialLinesDataFrame Lines data.frame
SpatialPolygons Polygons -
SpatialPolygonsDataFrame Polygons data.frame

We are not going to use sp in this book, but the newer sf package (Section 7.1.4). However, if you are going to work with spatial data in R it is very likely you will encounter sp in forums, books, or other packages, so it is important to be aware of it.

7.1.4 The sf package

The sf package (Figure 7.2), released in 2016, is a newer package for working with vector layers in R, which we are going to use in this book. In recent years, sf has become the standard package for working with vector data in R, practically replacing sp, rgdal, and rgeos.

Pebesma, 2018, The R Journal (

Figure 7.2: Pebesma, 2018, The R Journal (

One of the important innovations in sf is a complete implementation of the Simple Features standard. Since 2003, Simple Features been widely implemented in spatial databases (such as PostGIS), commercial GIS (e.g., ESRI ArcGIS) and forms the vector data basis for libraries such as GDAL. The Simple Features standard defines several types of geometries, of which seven are most commonly used in the world of GIS and spatial data analysis (Figure 7.6). When working with spatial databases, Simple Features are commonly specified as Well Known Text (WKT). A subset of simple features forms the GeoJSON standard.

The sf package depends on several external software components (installed along with the R package27), most importantly GDAL, GEOS and PROJ (Figure 7.3). These well-tested and popular open-source components are common to numerous open-source and commercial software for spatial analysis, such as QGIS and PostGIS.

`sf` package dependencies (

Figure 7.3: sf package dependencies (

Package sf defines a hierarchical class system with three classes (Table 7.3):

  • Class sfg—a single geometry
  • Class sfc—a geometry column, which is a set of sfg geometries + CRS information
  • Class sf—a layer, which is an sfc geometry column inside a data.frame with non-spatial attributes
Table 7.3: Spatial data structures in package sf
Class Hierarchy Information
sfg Geometry type, coordinates
sfc Geometry column set of sfg + CRS
sf Layer sfc + attributes

The sf class represents a vector layer by extending the data.frame class, supplementing it with a geometry column. This is similar to the way that spatial databases are structured. For example, the sample dataset shown in Figure 7.4 represents a polygonal layer with three features and six non-spatial attributes. The attributes refer to demographic and epidemiological attributes of US counties, such as the number of births in 1974 (BIR74), the number of sudden infant death cases in 1974 (SID74), and so on. The seventh column is the geometry column, containing the polygon geometries.

Structure of an `sf` object (

Figure 7.4: Structure of an sf object (

Figure 7.5 shows what the layer in Figure 7.4 would look like when mapped. We can see the outline of the three polygons, as well as the values of all six non-spatial attributes (in separate panels).

Visualization of the `sf` object shown in Figure \@ref(fig:nc-geometry-column)

Figure 7.5: Visualization of the sf object shown in Figure 7.4

7.2 Vector layers from scratch

7.2.1 Overview

As mentioned above (Table 7.3), the sf package defines a hierarchical system of data structures, composed of three classes, from simple to complex: sfg, sfc and sf. In this section, we are going to create an object of each of those thress classes, to learn more about them.

7.2.2 Geometry (sfg)

Objects of class sfg, i.e., a single geometry, can be created using the appropriate function for each geometry type:

  • st_point
  • st_multipoint
  • st_linestring
  • st_multilinestring
  • st_polygon
  • st_multipolygon
  • st_geometrycollection

from coordinates passed as:

  • numeric vectors—POINT
  • matrix objects—MULTIPOINT or LINESTRING
  • list objects—All other geometries

The seven most commonly used Simple Feature geometry types are displayed in Figure 7.6.

Simple feature geometry (`sfg`) types in package `sf`

Figure 7.6: Simple feature geometry (sfg) types in package sf

Of those seven types, the GEOMETRYCOLLECTION is more rarely used and more difficult to work with. For example, the Shapefile format does not support geometries of type GEOMETRYCOLLECTION. You may wonder why does it even exist. One of the reasons is that some spatial operations may produce a mixture of geometry types. For example, the intersection (Section between two polygons may be composed of points, lines, and polygons (Figure 7.7).

Intersection between two polygons may yield a `GEOMETRYCOLLECTION`

Figure 7.7: Intersection between two polygons may yield a GEOMETRYCOLLECTION

Let’s create some sfg geometries to see the principles in action. For example, we can create a point geometry object named pnt1, representing a POINT geometry, using the st_point function as follows:

Printing an sfg object in the console gives its WKT representation:

Note that the class definition of an sfg (geometry) object:

is actually composed of three parts:

  • "XY"—The dimensions type (one of: "XY", "XYZ", "XYM" or "XYZM"). In this book, as in most cases of spatial analysis in general, we will be working only with two-dimensional "XY" geometries.
  • "POINT"—The geometry type (one of the geometry types: "POINT", "MULTIPOINT", etc.)
  • "sfg"—The general class (sfg = Simple Feature Geometry)

Here is another example of creating an sfg object. This time, we are creating a POLYGON geometry named a, using function st_polygon. (Don’t worry if the expression is unclear: we learn about using list in Chapter 11).

Again, printing the object shows its WKT representation:

while class reports the dimensionality, geometry type, and general class, in that order:

The polygon is displayed in Figure 7.8:

An `sfg` object of type `POLYGON`

Figure 7.8: An sfg object of type POLYGON

Let’s create another POLYGON, named b:

The second polygon is shown in Figure 7.9.

Another `sfg` object of type `POLYGON`

Figure 7.9: Another sfg object of type POLYGON

The c function, when given sfg geometries, combines those geometries into one. For example, combining two POLYGON geometries results in a single MULTIPOLYGON geometry:

What type of geometry do you think c(a, b, pnt1) is?

Keep in mind that c always returns a single geometry, composed of all the shapes in its input. This is different from collecting the geometries into a geometry column, where the geometries are kept separate, which is done using function st_sfc as shown below (Section 7.2.3).

The multipolygon we created is shown in Figure 7.10:

An `sfg` object of type `MULTIPOLYGON`

Figure 7.10: An sfg object of type MULTIPOLYGON

A new geometry can be calculated applying various functions on an existing one(s). For example, the following example calculates the intersection of a and b, which is a new geometry hereby named i. We are going to learn about st_intersection, and other geometry-generating functions, in Chapter 8.

The result happens to be a GEOMETRYCOLLECTION, as demonstrated in Figure 7.7:

Figure 7.11 displays the GEOMETRYCOLLECTION named i:

An `sfg` object of type `GEOMETRYCOLLECTION`

Figure 7.11: An sfg object of type GEOMETRYCOLLECTION

7.2.3 Geometry column (sfc)

Let’s create two more point geometries (Section 7.2.2) named pnt2 and pnt3, representing two more points:

Geometry objects (sfg) can be collected into a geometry column (sfc) object. This is done with function st_sfc.

In addition to the geometries, a geometry column object also contains a Coordinate Reference System (CRS) (more information on CRS is given below, in Section 7.9) definition, specified with the crs parameter of function st_sfc. Four types of CRS definitions are accepted:

  • An EPSG code (e.g., 4326)
  • A PROJ4 string (e.g., "+proj=longlat +datum=WGS84 +no_defs")
  • A WKT string
  • A crs object of another layer, as returned by st_crs

Let’s combine the three POINT geometries—pnt1, pnt2 and pnt3—into a geometry column (sfc) object named geom. We will specify that the coordinates are lon/lat (WGS84), using the simplest of the three methods—an EPSG code (4326). More information on types of CRS, as well as where we can find the EPSG code of a particular CRS, is given in Section 7.9.2.

Here is a summary of the resulting geometry column:

The printout demostrates that other than the geometries themselves, and the derived properties of type, dimensionality, and bounding box, the geometry column contains the additional piece of information on the CRS.

7.2.4 Layer (sf)

A geometry column (sfc) can be combined with non-spatial columns, also known as attributes, resulting in a layer (sf) object. In our case, the three points in the sfc geometry column geom (Section 7.2.3) represent the location of the three railway stations in Beer-Sheva and Dimona. Let’s create a data.frame with several non-spatial properties of the stations, which we already worked with in Section 4.1.2, using function data.frame (Section 4.1.2):

Note that the order of rows in the attribute table must match the order of the geometries!

Now, we can combine the attribute table dat (data.frame) and the geometry column geom (sfc). This is done using function st_sf, resulting in a layer (sf):

7.2.5 Interactive mapping with mapview

Function mapview—which we are familiar with from Section—is useful for inspecting vector layers too. For example:

7.3 Extracting layer components

In Section 7.2 we:

  • Started from raw coordinates
  • Convered them to geometry objects (sfg) using a function such as st_point, st_polygon, etc. (Section 7.2.2)
  • Combined the geometries to a geometry column (sfc) using function st_sfc (Section 7.2.3)
  • Added attributes to the geometry column to get a layer (sf) using function st_sf (Section 7.2.4)

which can be summarized as: coordinates → sfgsfcsf.

Sometimes we are interested in the opposite “direction”. In other words, we sometimes need to extract the simpler components (geometry, attributes, coordinates) from an existing layer:

  • sf → geometry column (sfc)
  • sf → attribute table (data.frame)
  • sf → coordinates (matrix)

The geometry column (sfc) component can be extracted from an sf layer object using function st_geometry:

The non-spatial columns of an sf layer, i.e., the attribute table, can be extracted from an sf object using function st_drop_geometry:

The latter is analogous to opening an attribute table of a vector layer in GIS software, such as ArcGIS (Figure 7.12).

Attribute table in ArcGIS

Figure 7.12: Attribute table in ArcGIS

The coordinates of sf, sfc or sfg objects can be obtained with function st_coordinates. The coordinates are returned as a matrix:

In the case of a two-dimensional POINT layer, which is the most common use case of st_coordinates, the returned matrix has two columns X and Y as shown above. (With other types of geometries, the matrix has additional columns containing the information on grouping of point coordinates into shapes.)

7.4 Creating point layer from table

A common way of creating a point layer is to transform a table which has X and Y coordinate columns. Function st_as_sf can transform a table (data.frame) into a point layer (sf). In st_as_sf we specify:

  • x—The data.frame to be converted
  • coords—Columns names with the coordinates (X, Y)
  • crs—The CRS (NA if left unspecified)

Let’s take the rainfall.csv table as an example. This table contains UTM 36N (EPSG: 32636) coordinates in the columns named x_utm and y_utm:

The table can be converted to an sf layer using st_as_sf as follows:


  • The order of coords column names corresponds to X-Y!
  • 32636 is the EPSG code of the UTM 36N projection (Table 7.4)

The analogous operation in ArcGIS is the Display XY Data menu (Figures 7.137.15).

Displaying XY data from CSV in ArcGIS (Step 1)

Figure 7.13: Displaying XY data from CSV in ArcGIS (Step 1)

Displaying XY data from CSV in ArcGIS (Step 2)

Figure 7.14: Displaying XY data from CSV in ArcGIS (Step 2)

Displaying XY data from CSV in ArcGIS (Step 3)

Figure 7.15: Displaying XY data from CSV in ArcGIS (Step 3)

Here is the resulting sf layer:

An interactive map, showing the spatial locations of the rainfall stations, can be created using mapview (Section 7.2.5). Here, we are using the additional zcol parameter to choose which attribute will be used for the color scale:

7.5 sf layer properties

7.5.1 Dimensions

An sf layer is basically a special type of data.frame, where one of the columns is a geometry column. Therefore, many of the functions we learned when working with data.frame tables (Chapter 4) also work on sf layers.

For example, we can get the number of rows, or features, with nrow:

and the number of columns (including the geometry column) with ncol:

or both with dim:

What is the result of st_geometry(rainfall)? st_drop_geometry(rainfall)?

7.5.2 Spatial properties

The st_bbox function returns the bounding box coordinates, just like for stars objects (Section

The st_crs function returns the Coordinate Reference System (CRS), also the same way as for stars objects (Section

Question: what is the difference between the two plots in Figure 7.16, created using the following expressions?

Two plotsTwo plots

Figure 7.16: Two plots

7.6 Subsetting based on attributes

Subsetting of features in an sf vector layer is exactly the same as filtering rows in a data.frame (Section 4.1.5). Remember: an sf layer is a data.frame. For example, the following expressions subset the rainfall layer:

Which meteorological stations are being selected in each of these two expressions?

Figure 7.17 shows the resulting subsets:

Subsets of the `rainfall` layerSubsets of the `rainfall` layer

Figure 7.17: Subsets of the rainfall layer

Subsetting columns in sf is also similar to subsetting columns in a data.frame, with one exception: the geometry column. The geometry column “sticks” to the subset, by default, even if we do not explicitly select it, so that the resulting subset remains an sf object:

In case we do need to omit the geometry column and get a data.frame, we can apply st_drop_geometry (Section 7.3) on the subset:

7.7 Reading vector layers

In addition to creating from raw coordinates (Section 7.2) and transforming a data.frame to point layer (Section 7.4), we often create vector layers by reading from a file or from a spatial database (Section 7.1.2)28. Reading a vector layer from a file or a database is done using the st_read function.

For example, the following expression reads the Shapefile of US county boundaries, named USA_2_GADM_fips.shp, from the course materials. In case the Shapefile is located in the working directory, we need to specify the name of just the .shp file, even though the Shapefile contains several other files (Table 7.1):

Let’s also read a GeoJSON file with the location of three particular airports in New Mexico:

7.8 Basic plotting

When plotting an sf object with plot, we get multiple small maps—one map for each attribute. This can be useful to quickly examine the types of spatial variation in our data. For example (Figure 7.18):

Plot of `sf` object

Figure 7.18: Plot of sf object

Plotting a single attribute adds a legend (Figure 7.19). The key.width and key.pos let us control the amount of space the legend takes and its placement, respectively:

Plot of `sf` object, single attribute with legend

Figure 7.19: Plot of sf object, single attribute with legend

Plotting an sfc or an sfg object shows just the geometry (Figure 7.20):

Plot of `sfc` object

Figure 7.20: Plot of sfc object

We can use graphical parameters to control the appearance of plotted geometries, such as:

  • col—Fill color
  • border—Outline color
  • pch—Point shape
  • cex—Point size

For example, the following expression draws county borders in grey (Figure 7.21):

Basic plot of `sfc` object

Figure 7.21: Basic plot of sfc object

Additional vector layers can be drawn in an existing graphical window using add=TRUE, similarly to the concept of layers in GIS software. For example, the following expressions draw both county and airports geometries (Figure 7.22). Note how the second expression uses add=TRUE:

Using `add=TRUE` in `plot`

Figure 7.22: Using add=TRUE in plot

We can also use add=TRUE to combine sfg or sfc geometries with rasters in the same plot. For example, let’s plot the rainfall layer on top of the rainfall.tif raster, which is an interpolated rainfall surface we met in Chapter 1. First we will read the raster file:

and then plot both layers (Figure 7.23):

`sfc` layer on top of a raster

Figure 7.23: sfc layer on top of a raster

Note that we need to use the additional argument reset=FALSE whenever we are adding more layers to a stars raster plot.

7.9 Coordinate Reference Systems (CRS)

7.9.1 What are CRS?

A Coordinate Reference System (CRS) defines how the coordinates in our geometries relate to the surface of the Earth. There are two main types of CRS:

  • Geographic—longitude and latitude, in degrees
  • Projected—implying flat surface, usually in units of true distance (e.g., meters)

For example, Figure 7.24 shows the same polygonal layer (U.S. counties) in two different projection. On the left, the county layer is in the WGS84 geographic projection. Indeed, we can see that the axes are given in degrees of longitude and latitude. For example, we can see how the nothern border of U.S. follows the 49° latitude line. On the right, the same layer is displayed in the US National Atlas projection, where units are arbitrary but reflect true distance (meters). For example, the distance between every two consecutive grid lines is 1,000,000 meters or 1,000 kilometers.

US counties in WGS84 and US Atlas projectionsUS counties in WGS84 and US Atlas projections

Figure 7.24: US counties in WGS84 and US Atlas projections

7.9.2 Vector layer reprojection

Reprojection is the transformation of geometry coordinates, from one CRS to another. It is an important part of spatial analysis workflow, since we often need to:

  • Transform several layers into the same projection, so that they can be displayed one on top of the other (e.g., Figure 7.22) or so that they can be subject to a spatial operator (e.g., Figure 7.7)
  • Switch between geographic and projected CRS

A vector layer can be reprojected with st_transform. The st_transform function has two important parameters:

  • x—The layer to be reprojected
  • crs—The target CRS

Why don’t we need to specify the origin CRS in st_transform?

As mentioned above, the CRS can be specified in one of four ways, using an EPSG code, a PROJ4 string, a WKT string, or a crs object (Section 7.2.3). Where can we find EPSG codes or WKT definitions of different projections?

  • We can use CRS databases on the internet, such as or, to look up CRS definitions for a given country, of CRS name, using the search box.
  • We can use the make_EPSG function from the rgdal package in R, which returns a data.frame of CRS definitions. The data.frame can be filtered to locate the CRS of interest. For example, the following expressions return the information on CRS where the description contains the word “Israel”.

In this book, we are going to encounter five different projections: WGS84, Sinusoidal, UTM 36N, ITM and US National Atlas (Table 7.4). We are going to specify them (e.g., in st_transform) using their EPSG codes, which is the easiest method.

Table 7.4: Projections used in this book
Name Type Area Units EPSG code
WGS84 Geographic World degrees 4326
Sinusoidal Projected World meters -
UTM 36N Projected Israel meters 32636
ITM Projected Israel meters 2039
US National Atlas Projected USA meters 2163

Now, let’s see how we can use the st_transform function to repriject layers, in practice. The county and airports are currently in WGS84 (EPSG:4326) (How can we check?). Suppose that we would like to reproject those layers to the the US National Atlas Equal Area projection (EPSG:2163). The following expressions implement the reprojection:

The modified layers are shown in Figure 7.25. We can clearly see that the layer orientation has changes (also see Figure 7.24).

The `county` and `airports` layers in the US National Atlas Projection (EPSG=`2163`)

Figure 7.25: The county and airports layers in the US National Atlas Projection (EPSG=2163)

Examining the layer coordinates also shows that, indeed, the coordinates have changed with reprojecton:

Create a subset of county with the counties of New-Mexico, Arizona and Texas only, and plot the result (Figure 7.26).

Subset of three states from the `county` layer

Figure 7.26: Subset of three states from the county layer

7.10 Writing vector layers

Writing an sf object to a file can be done with st_write. Before writing the rainfall layer to disk, let’s calculate a new column called annual with the annual rainfall (Section 4.5):

In the last expression, why did we use st_drop_geometry(rainfall[, m]) as the first argument in apply, instead of rainfall[, m]?

The sf object can be written to a Shapefile with st_write, as follows:

The format is automatically determined based on the .shp file extension. To overwrite an existing file, use delete_dsn=TRUE.

  1. in Windows.

  2. For a complete list of vector formats that can be read with st_read, run st_drivers(what="vector").