Create polygons representing bounding boxes for subgroups using sf

Question

I'm attempting to use the sf package and a piped/tidyverse workflow to generate bounding boxes based on groups defined in another column. I think it should work like below, but it seems like st_bbox is not respecting groups.

I expect to receive three polygon records that represent bounding boxes around proints from a, b, and c, but instead I receive three polygon records that represent bounding boxes for all points.

library(dplyr)
library(sf)

a <- data.frame(group=rep('a',100), lon=rnorm(100,11,.2), lat=rnorm(100,53,.2))
b <- data.frame(group=rep('b',100), lon=rnorm(100,11.5,.2), lat=rnorm(100,53.5,.2))
c <- data.frame(group=rep('c',100), lon=rnorm(100,12,.2), lat=rnorm(100,54,.2))
dat <- rbind(a,b,c)

pts <- dat %>% st_as_sf(coords=c('lon','lat'),crs=4326) 

pts %>%
  group_by(group) %>%
  summarize(geometry = st_as_sfc(st_bbox(geometry)))

This returns:

Simple feature collection with 3 features and 1 field
geometry type:  POLYGON
dimension:      XY
bbox:           xmin: 10.34313 ymin: 52.43993 xmax: 12.54254 ymax: 54.54012
epsg (SRID):    4326
proj4string:    +proj=longlat +datum=WGS84 +no_defs
# A tibble: 3 x 2
  group                                                                                     geometry
  <fct>                                                                                <POLYGON [°]>
1 a     ((10.34313 52.43993, 12.54254 52.43993, 12.54254 54.54012, 10.34313 54.54012, 10.34313 52...
2 b     ((10.34313 52.43993, 12.54254 52.43993, 12.54254 54.54012, 10.34313 54.54012, 10.34313 52...
3 c     ((10.34313 52.43993, 12.54254 52.43993, 12.54254 54.54012, 10.34313 54.54012, 10.34313 52...

Answer 1

One option is to use a nested dataframe using tidyr::nest and then purrr::map. I also used a wrapper function to simplify the map call

library(tidyverse)
box_sf <- pts %>% 
  group_by(group) %>%
  nest()

bbox_wrap <- function(x) st_as_sfc(st_bbox(x))
box_sf <- box_sf %>% 
  mutate(bbox = map(data, bbox_wrap))

This will give you a list of bounding boxes as a column of a dataframe. If you want to convert back to an sf object you can do this:

box_sf %>% 
  mutate(geometry = st_as_sfc(do.call(rbind, bbox))) %>% 
  select(-data, -bbox) %>% 
  st_as_sf()

Seems a little roundabout, I'm hoping to see a solution using group_by as you originally intended