Areal Data: Rasters

HES 505 Fall 2024: Session 8

Carolyn Koehn

Announcements

  • R server fix: run install.packages('sf') in the Console

  • Fall Graduate Proposal Showcase is tomorrow, 3-5pm, in the Student Union Building, Simplot AC Ballroom!

  • Any questions that came up during the homework?

Revisiting the Raster Data Model

  • Vector data describe the “exact” locations of features on a landscape (including a Cartesian landscape)

  • Raster data represent spatially continuous phenomena (NA is possible)

  • Depict the alignment of data on a regular lattice (often a square)

    • Operations mimic those for matrix objects in R

  • Geometry is implicit; the spatial extent and number of rows and columns define the cell size

Rasters with terra

  • syntax is different for terra compared to sf

  • Representation in Environment is also different

  • Can break pipes, Be Explicit

Rasters by Construction

Rasters by Construction

mtx <- matrix(1:16, nrow=4)
mtx
     [,1] [,2] [,3] [,4]
[1,]    1    5    9   13
[2,]    2    6   10   14
[3,]    3    7   11   15
[4,]    4    8   12   16
rstr <- terra::rast(mtx)
rstr
class       : SpatRaster 
dimensions  : 4, 4, 1  (nrow, ncol, nlyr)
resolution  : 1, 1  (x, y)
extent      : 0, 4, 0, 4  (xmin, xmax, ymin, ymax)
coord. ref. :  
source(s)   : memory
name        : lyr.1 
min value   :     1 
max value   :    16

Note: you must have terra loaded for plot() to work on Rast* objects; otherwise you get Error in as.double(y) : cannot coerce type 'S4' to vector of type 'double'

Rasters by Construction: Origin

  • Origin defines the location of the intersection of the x and y axes
  • Ideally, the origin is the cell “corner” closest to c(0, 0)
r <- rast(xmin=-4, xmax = 9.5, ncols=10)
r[] <- runif(ncell(r))
origin(r)
[1] 0.05 0.00
r2 <- r
origin(r2) <- c(2,2)

Rasters by Construction: Resolution

  • Geometry is implicit; the spatial extent and number of rows and columns define the cell size
  • Resolution (res) defines the length and width of an individual pixel
r <- rast(xmin=-4, xmax = 9.5, 
          ncols=10)
res(r)
[1] 1.35 1.00
r2 <- rast(xmin=-4, xmax = 5, 
           ncols=10)
res(r2)
[1] 0.9 1.0
r <- rast(xmin=-4, xmax = 9.5, 
          res=c(0.5,0.5))
ncol(r)
[1] 27
r2 <- rast(xmin=-4, xmax = 9.5, 
           res=c(5,5))
ncol(r2)
[1] 3

Predicates and measures in terra

Extending predicates

  • Predicates: evaluate a logical statement asserting that a property is TRUE

  • terra does not follow the same hierarchy as sf so a little trickier

Unary predicates in terra

  • Can tell us qualities of a raster dataset

  • Many similar operations for SpatVector class (note use of .)

predicate asks…
is.lonlat Does the object have a longitude/latitude CRS?
inMemory is the object stored in memory?
is.factor Are there categorical layers?
hasValues Do the cells have values?

Unary predicates in terra

  • global: tests if the raster covers all longitudes (from -180 to 180 degrees) such that the extreme columns are in fact adjacent
r <- rast()
is.lonlat(r)
[1] TRUE
is.lonlat(r, global=TRUE)
[1] TRUE
  • perhaps: If TRUE and the crs is unknown, the method returns TRUE if the coordinates are plausible for longitude/latitude
crs(r) <- ""
is.lonlat(r)
[1] NA
is.lonlat(r, perhaps=TRUE, warn=FALSE)
[1] TRUE
crs(r) <- "+proj=lcc +lat_1=48 +lat_2=33 +lon_0=-100 +ellps=WGS84"
is.lonlat(r)
[1] FALSE

Binary predicates in terra

  • Take exactly 2 inputs, return 1 matrix of cell locs where value is TRUE

  • adjacent: identifies cells adajcent to a set of raster cells

Unary measures in terra

  • Slightly more flexible than sf

  • One result for each layer in a stack

measure returns
cellSize area of individual cells
expanse summed area of all cells
values returns all cell values
ncol number of columns
nrow number of rows
ncell number of cells
res resolution
ext minimum and maximum of x and y coords
origin the orgin of a SpatRaster
crs the coordinate reference system
cats categories of a categorical raster

Binary measures in terra

  • Returns a matrix or SpatRaster describing the measure
measure returns
distance shortest distance to non-NA or vector object
gridDistance shortest distance through adjacent grid cells
costDist Shortest distance considering cell-varying friction
direction azimuth to cells that are not NA

Practice

  1. Create an empty raster (a raster with no cell values) with 5 rows, 5 columns, and a resolution of 2 units with a total of 25 cells. You may need to use the rast helpfile and check the argument descriptions.
  2. Fill the raster’s cells with random values (hint: remember how we reassigned the origin? Use another unary measure). Use a unary predicate to test if the raster has values. Plot your raster.
  3. Print the default origin of your raster.
  4. Reassign the values of cells adjacent to cell 12 to NA and plot. Try different cases if time allows.
  5. With some cells set to NA (from step 4), create and plot a raster showing the distance to the nearest non-NA cell.

Extra Practice

  1. Run the examples for costDist and gridDistance to see how those functions can be used.
  2. Load the wildfire_hazard_agg.tif data from the assignment03 folder. Use the data as the input for the distance function and plot the result. How might this be useful in your research?