Cropping a raster from s3, masking using exactextract

crop.r

library(exactextractr)
library(raster)
library(stars)
library(sf)

# define quad bounds using a GeoJSON string
quad_bounds <- st_read("{\n  \"coordinates\": [\n    [\n      [\n        [\n          -98.97075, \n          46.7021\n        ], \n        [\n          -98.6657, \n          47.44555\n        ], \n        [\n          -97.48445, \n          47.27325\n        ], \n        [\n          -97.7895, \n          46.5298\n        ], \n        [\n          -98.97075, \n          46.7021\n        ]\n      ]\n    ]\n  ], \n  \"type\": \"MultiPolygon\"\n}\n")

# create a stars proxy for the s3 object
scene <- read_stars('/vsis3/bucket-name-here/dataGlobal/landsat/8/031/027/LC08_L1TP_031027_20161107_20170219_01_T1/LC08_L1TP_031027_20161107_20170219_01_T1_TCT.tif',
                    proxy=TRUE)

# crop the stars proxy to the bbbox of the quad (cropping to the quad itself takes forever)
quad <- st_as_stars(scene[st_bbox(quad_bounds)])

# convert to raster
quad.r <- as(quad, 'Raster')

# plot to check
plot(quad.r[[1]])
plot(st_geometry(quad_bounds), add=TRUE)

# compute coverage fractions (0-1)
cov_frac <- coverage_fraction(quad.r[[1]], quad_bounds)[[1]]

# generate a mask based on pct of cell that must be within quad
mask <- (cov_frac >= 0.5)