Fast Raster Summary and Manipulation

Description

Fast alternatives to several relatively slow 'raster' package functions. For large rasters, the functions run from 5 to approximately 100 times faster than the 'raster' package functions they replace. The 'fasterize' package, on which one function in this package depends, includes an implementation of the scan line algorithm attributed to Wylie et al. (1967) <doi:10.1145/1465611.1465619>.

Details

The DESCRIPTION file:

Index of help topics:

cat_to_val              Convert a Categorical Raster to a Value Raster
crosstabDT              Speedy Raster Cross-tabulation
fasterizeDT             Polygon Rasterization Using Numeric, Factor, or
                        Character Fields
freqDT                  Speedy Raster Value Frequency Tabulation
rasterDT-package        Fast Raster Summary and Manipulation
subsDT                  Speedy Raster Value Substitution
zonalDT                 Speedy Zonal Statistics

Fast alternatives to several relatively slow raster package functions. For large rasters, the functions run from 5 to approximately 100 times faster than the raster package functions they replace.

Author(s)

Joshua O'Brien

Maintainer: Joshua O'Brien <joshmobrien@gmail.com>

Convert a Categorical Raster to a Value Raster

Description

Use a categorical raster's RAT to convert it to a continuous raster

Usage

cat_to_val(r, which = 2)

Arguments

Value

A continuous raster with each category level in r replaced by its corresponding value.

Author(s)

Joshua O'Brien

Examples

r_cat <- raster(matrix(c(2, 2, 2, 1), ncol = 2))
levels(r_cat) <- data.frame(ID = c(1, 2),
                            VAL1 = c(0.1, 200),
                            VAL2 = c(33, 44))

## Second column of RAT is used by default
r_con1 <- cat_to_val(r_cat)
as.matrix(r_con1)

## Use 'which=' argument for conversion to another RAT column
r_con2 <- cat_to_val(r_cat, which = "VAL2")
as.matrix(r_con2)

Speedy Raster Cross-tabulation

Description

A fast data.table-based alternative to raster::crosstab().

Usage

crosstabDT(x, y, digits = 0, long = FALSE, useNA = FALSE)

Arguments

Value

Either a table or a data.table recording the frequency of each combination of raster values.

Author(s)

Joshua O'Brien

Examples

r <- raster(nc = 5, nr = 5)
r[] <- runif(ncell(r)) * 2
s <- setValues(r, runif(ncell(r)) * 3)
crosstabDT(r, s)

rs <- r/s
r[1:5] <- NA
s[20:25] <- NA
x <- stack(r, s, rs)
crosstabDT(x, useNA = TRUE, long = TRUE)

Polygon Rasterization Using Numeric, Factor, or Character Fields

Description

A front end for fasterize::fasterize(), fixing several of its infelicities.

Usage

fasterizeDT(
  x,
  raster,
  field = NULL,
  fun = "last",
  background = NA_real_,
  by = NULL
)

Arguments

Details

Unlike other functions in this package, fasterizeDT() does not use data.table to speed up its computations. Instead, it is a wrapper for fasterize::fasterize(), intended to address several of that function's limitations.

Most importantly, fasterizeDT() takes care to properly handle rasterization operations in which either the template RasterLayer or the selected polygon feature field is a factor. Specifically, it always returns a raster whose type (numeric or factor) and levels (if a factor) match that of the spatial polygon attribute indicated by its field argument. Second, when field specifies an attribute of class "character", fasterizeDT() automatically converts it to a factor and returns a factor raster. In this, it is unlike both fasterize::fasterize() and raster::rasterize(). Finally, unlike fasterize::fasterize(), fasterizeDT() accepts as inputs either sf::sf() objects or sp::SpatialPolygonsDataFrame objects.

Value

A raster of the same size, extent, resolution and projection as the supplied raster template. Unlike fasterize::fasterize(), fasterizeDT returns a raster of the same type as the data in the column of x selected by the field argument.

Author(s)

Joshua O'Brien

Examples

## Load example polygons and prepare a template raster
if (require(raster)) {
SPDF <- shapefile(system.file("external/lux.shp", package = "raster"))
llratio <- 1/cos(pi * mean(coordinates(SPDF)[, 2])/180)
rr <- raster(extent(SPDF),
             resolution = c(llratio * 0.01, 0.01),
             crs = proj4string(SPDF))

## An integer-valued field produces a numeric raster
rInt <- fasterizeDT(SPDF, rr, field = "ID_2")
plot(rInt, col = colorRampPalette(blues9)(12))

## A character-valued field returns a factor raster
rFac <- fasterizeDT(SPDF, rr, field = "NAME_2")
if (require(rasterVis)) {
    levelplot(rFac)
}
}

Speedy Raster Value Frequency Tabulation

Description

A fast data.table-based alternative to raster::freq().

Usage

freqDT(x, ...)

## S4 method for signature 'RasterLayer'
freqDT(x, digits = 0, value = NULL, useNA = c("ifany", "no", "always"), ...)

## S4 method for signature 'RasterStackBrick'
freqDT(
  x,
  digits = 0,
  value = NULL,
  useNA = c("ifany", "no", "always"),
  merge = FALSE,
  ...
)

Arguments

Author(s)

Joshua O'Brien

Examples

r <- raster(nrow = 18, ncol = 36)
r[] <- runif(ncell(r))
r[1:5] <- NA
r <- r * r * r * 5
freqDT(r)

freqDT(r, value = 2)

s <- stack(r, r*2, r*3)
freqDT(s, merge = TRUE)

Speedy Raster Value Substitution

Description

A fast data.table-based alternative to raster::subs().

Usage

subsDT(x, dict, by = 1, which = 2, subsWithNA = TRUE, filename = "", ...)

Arguments

Value

A RasterLayer object.

Author(s)

Joshua O'Brien

Examples

r <- raster(ncol = 10, nrow = 10)
r[] <- round(runif(ncell(r)) * 10)
df <- data.frame(id = 2:8, v = c(10, 10, 11, 11, 12:14))
x <- subsDT(r, df)
x2 <- subsDT(r, df, subsWithNA = FALSE)

df$v2 <- df$v * 10
x3 <- subsDT(r, df, which = 2:3)

s <- stack(r, r*3)
names(s) <- c("first", "second")
x4 <- subsDT(s, df)
x5 <- subsDT(s, df, which = 2:3)

Speedy Zonal Statistics

Description

A fast data.table-based alternative to raster::zonal().

Usage

zonalDT(x, z, fun = sum, na.rm = TRUE)

Arguments

Value

A data.table with a summary value for each zone.

Author(s)

Joshua O'Brien

Examples

r <- raster(ncols = 10, nrows = 10)
r[] <- runif(ncell(r)) * 1:ncell(r)
z <- r
z[] <- rep(1:5, each = 20)
## for big files, use a character value rather than a function
zonalDT(r, z, "sum")

## for smaller files you can also provide a function
zonalDT(r, z, mean)
zonalDT(r, z, min)

## multiple layers
zonalDT(stack(r, r*10), z, "sum")