运行 raster::stackApply() 并行运行

Running raster::stackApply() function in parallel

我正在尝试并行化 示例。

我有一堆栅格,我想按一年中的一周汇总这些栅格。这是系列中的样子:

# create a raster stack from list of GeoTiffs
tifs <- list.files(path = "./inputData/", pattern = "\.tif$", full.names = TRUE)
r <- stack(tifs)

# get the date from the names of the layers and extract the week
indices <- format(as.Date(names(r), format = "X%Y.%m.%d"), format = "%U")
indices <- as.numeric(indices)

# calculate weekly means
r_week <- stackApply(r, indices, function(x) mean(x, na.rm = TRUE))

这是我使用 snowpbapply 进行并行化的尝试。

# aggregate rasters in parallel
no_cores <- parallel::detectCores() - 1 

tryCatch({
  cl <- snow::makeCluster(no_cores, "SOCK")
  snow::clusterEvalQ(cl, {
    require(pacman)
    p_load(dplyr
           ,rts
           ,raster
           ,stringr
           ,pbapply
           ,parallel)
  })
  parallel::clusterExport(cl = cl, varlist = list("r", "indices"))
  r_week <-  pbapply::pbsapply(r, indices, stackApply(r, indices, function(x) mean(x, na.rm = TRUE)), simplify = TRUE, USE.NAMES = TRUE, cl = cl)
  snow::stopCluster(cl)
}, error=function(e){
  snow::stopCluster(cl)
  return(e)
}, finally = {
  try(snow::stopCluster(cl), silent = T)
})

stackApply() 方法不采用簇参数,因此我试图将其包装在 pbsapply() 中。这 returns 出现以下错误:

<simpleError in get(as.character(FUN), mode = "function", envir = envir): object 'indices' of mode 'function' was not found>

我想我找到了使用 raster::clusterR() 方法的解决方法。虽然它不提供进度条。很高兴看到是否有人知道如何使用 snowpbapply 执行此操作。

tryCatch({
  system.time({
  no_cores <- parallel::detectCores() - 1
  raster::beginCluster(no_cores)
  myFun <- function(x, ...) {
    mean(!is.na(x))
  }
  r_week <- raster::clusterR(r, stackApply, args=list(indices = indices, fun = myFun, na.rm = TRUE))
  raster::endCluster()})
}, error = function(e) {
  raster::endCluster()
  return(e)
}, finally = {
  try(raster::endCluster())
})

尝试将 progress='text' 添加到 stackApply 参数。它在非并行版本中运行良好。祝你好运!