I have a large rasterstack (s) with the following details:
class : RasterStack
dimensions : 510, 1068, 544680, 19358 (nrow, ncol, ncell, nlayers)
resolution : 0.08333333, 0.08333333 (x, y)
extent : -141, -52, 41, 83.5 (xmin, xmax, ymin, ymax)
coord. ref. : +proj=longlat +datum=NAD83 +no_defs +ellps=GRS80 +towgs84=0,0,0
names : Jan.1961.1, Jan.1961.2, Jan.1961.3, Jan.1961.4, Jan.1961.5, Jan.1961.6, Jan.1961.7, Jan.1961.8, Jan.1961.9, Jan.1961.10, Jan.1961.11, Jan.1961.12, Jan.1961.13, Jan.1961.14, Jan.1961.15, ...
time : 1961-01-01 - 2013-12-31 (range)
Doing something like:
writeRaster( s,"PP", overwrite=TRUE, format="CDF", varname="P", varunit="mm",
longname="totals", xname="lon", yname="lat",zname="time",
zunit="numeric")
takes more than 2 weeks to complete on my computer. How can I run this in parallel (may be via foreach loop and %dopar% command) to get the same results with shorter processing time?
sample data
s=brick(nrows=510, ncols=1068, xmn=-180, xmx=180, ymn=-90, ymx=90, crs="+proj=longlat +datum=WGS84", nl=193581)
dates=seq(as.Date("1961-01-01"), as.Date("2013-12-31"), by="day")
s<- setZ(s,dates)
NB: my true data is a rasterstack not brick.
You can try this code, but I did not really tested it on a big dataset. And I did not tested the ncecat part... I'll update it later, but you can try in the meantime.
wd <- "~/Bureau/Tmp"
# stack with 16 layers
nl <- 16 # 19358
s <- brick(nrows = 510, ncols = 1068,
xmn = -180, xmx = 180, ymn = -90, ymx = 90,
crs = "+proj=longlat +datum=WGS84",
nl = nl)
dates <- seq(as.Date("1961-01-01"), as.Date("2013-12-31"), by = "day")
s <- setZ(s, dates)
require(foreach)
require(doParallel)
cl <- makeCluster(4)
registerDoParallel(cl)
tmp <- foreach(i = 1:nlayers(s)) %dopar%
{
r <- raster::raster(s, i)
raster::writeRaster(r,
filename = paste0(wd,
"/PP_", formatC(i, width = 6, flag = "0")),
overwrite=TRUE, format="CDF", varname="P", varunit="mm",
longname="totals", xname="lon", yname="lat",zname="time",
zunit="numeric")
rm(r)
}
stopCluster(cl)
ppfiles <- list.files(wd)[grep("PP_", list.files(wd))]
system(paste0("ncecat ppfiles output.nc")