dt_aggregate: Spatial aggregation of data.table.raster

In ldemaz/dtraster: Raster operations with data.table

Description

Spatial aggregation of data.table.raster

Usage

dt_aggregate(xdt, nr, nc, fact, funexpr, dxy, outname, dropid = TRUE)

Arguments

nr	Number of rows in data.table.raster
nc	Number of columns in data.table.raster
fact	aggregation factor
funexpr	Aggregation function for data.table to evaluate (see details)
dxy	Resolution of input data.table.raster (square grids only)
outname	Character vector for value to be aggregated
dropid	Keep or remove cell number ID in output data.table.raster?
dt	Input data.table, with x, y coordinates

Details

This function was created to allow more complicated aggregations, such as weighted mean aggregation. Functions need to be passed in as an expression, complete with the names of the variables to evaluate. E.g. for a data.table.raster with variables x and w, then pass in funexpr = parse(text = "weighted.mean(x, w)"). Aggregation will always be done by variables ac and ar, which indicate the aggregation groupings. Don't name one of your variables x, because as.data.table.raster creates a x coordinate, and the function will fail if it detects a duplicate column name.

Value

An aggregated data.table.raster

Examples

r1 <- raster(xmn = 10, xmx = 20, ymn = 10, ymx = 20)
res(r1) <- 0.25
set.seed(1)
r1[] <- sample(1:10, ncell(r1), replace = TRUE)
r2 <- raster(r1)
set.seed(3)
r2[] <- runif(ncell(r2), min = 0, max = 1)
s <- stack(r1, r2)
names(s) <- c("v", "w")  # selecting name v, rather than x.
rdt <- as.data.table.raster(s, xy = TRUE)
p4s <- CRS(projection(r1))
plot(dt_to_raster(rdt, p4s))
vw <- dt_aggregate(x = rdt, nr = nrow(r1), nc = ncol(r1), fact = 5,
                   funexpr = parse(text = "weighted.mean(v, w)"),
                   dxy = xres(r1), outname = "muw")
v <- dt_aggregate(rdt, nr = nrow(r1), nc = ncol(r1), fact = 5,
                  funexpr = parse(text = "mean(v)"),
                  dxy = xres(r1), outname = "mu")
plot(dt_to_raster(vw, p4s))
plot(dt_to_raster(v, p4s))
plot(aggregate(r1, fact = 5))
plot(aggregate(r1, fact = 5) - dt_to_raster(v, p4s)) # r inferno
plot(round(aggregate(r1, fact = 5) - dt_to_raster(v, p4s), 10)) # same

# speed test
r3 <- raster(xmn = 10, xmx = 20, ymn = 10, ymx = 20)
p4s <- CRS(projection(r3))
res(r3) <- 0.001
set.seed(1)
r3[] <- sample(1:10, ncell(r3), replace = TRUE)
names(r3) <- "v"  # selecting name v, rather than x.
rdt <- as.data.table(r3, xy = TRUE)  # time penalty to convert to data.table
system.time(v <- dt_aggregate(rdt, nr = nrow(r3), nc = ncol(r3), fact = 10,
                              funexpr = parse(text = "mean(v)"),
                              dxy = xres(r3), outname = "mu"))
# user  system elapsed
# 11.328  16.006  35.346
system.time(r4 <- aggregate(r3, fact = 10))
# user  system elapsed
# 2.240   1.183   4.159 # raster aggregate still much faster
dtr <- dt_to_raster(v, p4s)
plot(dtr)
plot(r4)
plot(round(dtr - r4), 10)

ldemaz/dtraster documentation built on May 20, 2019, 6:44 p.m.