Identify the pixels and coordinates that are at a (set of) buffer distance(s) of the objects passed into coords. This is similar to rgeos::gBuffer but much faster and without the geo referencing information. In other words, it can be used for similar problems, but where speed is important. This code is substantially adapted from PlotRegionHighlighter::createCircle.

cir(
landscape,
coords,
loci,
allowOverlap = TRUE,
allowDuplicates = FALSE,
includeBehavior = "includePixels",
returnDistances = FALSE,
angles = NA_real_,
returnAngles = FALSE,
returnIndices = TRUE,
closest = FALSE,
simplify = TRUE
)

## Arguments

landscape Raster on which the circles are built. Either a matrix with 2 (or 3) columns, x and y (and id), representing the coordinates (and an associated id, like cell index), or a SpatialPoints* object around which to make circles. Must be same coordinate system as the landscape argument. Default is missing, meaning it uses the default to loci Numeric. An alternative to coords. These are the indices on landscape to initiate this function. See coords. Default is one point in centre of landscape.. Numeric vector of length 1 or same length as coords Numeric vector of length 1 or same length as coords. Default is maxRadius, meaning return all cells that are touched by the narrow ring at that exact radius. If smaller than maxRadius, then this will create a buffer or donut or ring. Logical. Should duplicates across id be removed or kept. Default TRUE. Logical. Should duplicates within id be removed or kept. Default FALSE. This is useful if the actual x, y coordinates are desired, rather than the cell indices. This will increase the size of the returned object. Character string. Currently accepts only "includePixels", the default, and "excludePixels". See details. Logical. If TRUE, then a column will be added to the returned data.table that reports the distance from coords to every point that was in the circle/donut surrounding coords. Default FALSE, which is faster. Numeric. Optional vector of angles, in radians, to use. This will create "spokes" outward from coords. Default is NA, meaning, use internally derived angles that will "fill" the circle. Logical. If TRUE, then a column will be added to the returned data.table that reports the angle from coords to every point that was in the circle/donut surrounding coords. Default FALSE. Logical or numeric. If 1 or TRUE, will return a data.table with indices and values of successful spread events. If 2, it will simply return a vector of pixel indices of all cells that were touched. This will be the fastest option. If FALSE, then it will return a raster with values. See Details. Logical. When determining non-overlapping circles, should the function give preference to the closest loci or the first one (much faster). Default is FALSE, meaning the faster, though maybe not desired behaviour. logical. If TRUE, then all duplicate pixels are removed. This means that some x, y combinations will disappear.

## Value

A matrix with 4 columns, id, indices, x, y. The x and y indicate the exact coordinates of the indices (i.e., cell number) of the landscape associated with the ring or circle being identified by this function.

## Details

This function identifies all the pixels as defined by a donut with inner radius minRadius and outer radius of maxRadius. The includeBehavior defines whether the cells that intersect the radii but whose centres are not inside the donut are included includePixels or not excludePixels in the returned pixels identified. If this is excludePixels, and if a minRadius and maxRadius are equal, this will return no pixels.

rings which uses spread internally. cir tends to be faster when there are few starting points, rings tends to be faster when there are many starting points. cir scales with maxRadius ^ 2 and coords. Another difference between the two functions is that rings takes the centre of the pixel as the centre of a circle, whereas cir takes the exact coordinates. See example. For the specific case of creating distance surfaces from specific points, see distanceFromEachPoint, which is often faster. For the more general GIS buffering, see rgeos::gBuffer.

## Examples

library(data.table)
#>
#> Attaching package: ‘data.table’#> The following object is masked from ‘package:raster’:
#>
#>     shiftlibrary(sp)
library(raster)
library(quickPlot)

set.seed(1462)

if (require(RandomFields)) {
# circle centred
ras <- raster(extent(0, 15, 0, 15), res = 1, val = 0)
middleCircle <- cir(ras)
ras[middleCircle[, "indices"]] <- 1
circlePoints <- SpatialPoints(middleCircle[, c("x", "y")])
if (interactive()) {
clearPlot()
Plot(ras)
Plot(circlePoints, addTo = "ras")
}

# circles non centred
ras <- randomPolygons(ras, numTypes = 4)
n <- 2
agent <- SpatialPoints(coords = cbind(x = stats::runif(n, xmin(ras), xmax(ras)),
y = stats::runif(n, xmin(ras), xmax(ras))))

cirs <- cir(ras, agent, maxRadius = 15, simplify = TRUE) ## TODO: empty with some seeds! e.g. 1642
cirsSP <- SpatialPoints(coords = cirs[, c("x", "y")]) ## TODO: error with some seeds! e.g. 1642
cirsRas <- raster(ras)
cirsRas[] <- 0
cirsRas[cirs[, "indices"]] <- 1

if (interactive()) {
clearPlot()
Plot(ras)
Plot(cirsRas, addTo = "ras", cols = c("transparent", "#00000055"))
Plot(agent, addTo = "ras")
}

# Example comparing rings and cir
a <- raster(extent(0, 30, 0, 30), res = 1)
hab <- gaussMap(a, speedup = 1) # if raster is large (>1e6 pixels) use speedup > 1
n <- 2
coords <- SpatialPoints(coords = cbind(x = stats::runif(n, xmin(hab), xmax(hab)),
y = stats::runif(n, xmin(hab), xmax(hab))))

# cirs
cirs <- cir(hab, coords, maxRadius = rep(radius, length(coords)), simplify = TRUE)

# rings
loci <- cellFromXY(hab, coordinates(coords))
cirs2 <- rings(hab, loci, maxRadius = radius, minRadius = radius - 1, returnIndices = TRUE)

# Plot both
ras1 <- raster(hab)
ras1[] <- 0
ras1[cirs[, "indices"]] <- cirs[, "id"]

ras2 <- raster(hab)
ras2[] <- 0
ras2[cirs2$indices] <- cirs2$id
if (interactive()) {
clearPlot()
Plot(ras1, ras2)
}

a <- raster(extent(0, 100, 0, 100), res = 1)
hab <- gaussMap(a, speedup = 1)
cirs <- cir(hab, coords, maxRadius = 44, minRadius = 0)
ras1 <- raster(hab)
ras1[] <- 0
cirsOverlap <- data.table(cirs)[, list(sumIDs = sum(id)), by = indices]
ras1[cirsOverlap$indices] <- cirsOverlap$sumIDs
if (interactive()) {
clearPlot()
Plot(ras1)
}

# Provide a specific set of angles
ras <- raster(extent(0, 330, 0, 330), res = 1)
ras[] <- 0
n <- 2
coords <- cbind(x = stats::runif(n, xmin(ras), xmax(ras)),
y = stats::runif(n, xmin(ras), xmax(ras)))
circ <- cir(ras, coords, angles = seq(0, 2 * pi, length.out = 21),
maxRadius = 200, minRadius = 0, returnIndices = FALSE,
allowOverlap = TRUE, returnAngles = TRUE)
}
#>     RFoptions