我有一个很大的面数据集，我想从每个面中减去一个唯一的缓冲区宽度，然后将缓冲区与原始面进行比较。我正在使用HydroSHEDS的HydroLAKES数据库https://www.hydrosheds.org/products/hydrolakes（Lake polygons shapefile，820 mb），并将此文件裁剪到加拿大/阿拉斯加/美国大陆的范围。裁剪到我感兴趣的区域后，有~966个，000行/观察。区分缓冲区从原来的多边形需要很长的时间（〉1.5小时在我的电脑上，即使并行处理;我有8个核心和32 GB内存），因此我想知道最有效的方法是什么.也许我不是有效的并行处理.乐意使用data.table包，以实现更快的结果，如果必要的.

library(dplyr)
library(sf)
library(parallel)
library(doParallel)

# generate dataset
hydrolakes_na <- read_sf("HydroLAKES_polys_v10_shp/HydroLAKES_polys_v10_shp/HydroLAKES_polys_v10.shp")
hydrolakes_na <- hydrolakes_na %>% filter(Country %in% c('Canada', 'United States of America'))
hydrolakes_na <- st_transform(hydrolakes_na, 3161) # using st_crs(3161)
ext_of_interest <- st_bbox(c(xmin=-2370387.108629, xmax=3349430.12726586, ymin=9663917.12113056, ymax=16719362.4659996), crs=st_crs(3161)) # desired spatial extent 
hydrolakes_crop <- st_crop(hydrolakes_na, ext_of_interest)

# inward buffer width (in meters)
set.seed(111)
hydrolakes_crop = hydrolakes_crop %>% mutate(inbuff_width = sample(100:1000, nrow(hydrolakes_crop)))

# without parallel processing (note that '-' preceding .$inbuff, to create inward buffer) 
hydrolakes_diff = hydrolakes_crop %>% st_buffer(-.$inbuff_width)
lake_diff = st_difference(st_geometry(hydrolakes_crop), st_union(st_geometry(hydrolakes_diff)))

# with parallel processing
cl <- parallel::makeCluster(4)
doParallel::registerDoParallel(cl)

lake_diff = foreach(i=seq_len(nrow(hydrolakes_crop)), .packages=c("sf")) %dopar% {st_difference(st_geometry(hydrolakes_crop[i,]), st_union(st_geometry(hydrolakes_diff[i,])))
}