In BigelowLab/rasf: rasf

knitr::opts_chunk$set(echo = TRUE)

rasf

Tools for working with raster, terra and sf packages.

We have prepared the functionality in this package to operate on rasterized data from either the raster package or it's successor the terra package. Rather than using S3 class dispatch, handling the various inputs is done by simply testing that the input(s) inherit form either BasicRaster (raste package) or SpatRaster (terra package).

Requirements

Installation

remotes::install_github("BigelowLab/rasf", upgrade = FALSE)

Rasters

Here we view rasters as multi-layer images. We observe the convention that cells in raster may addressed by row and column number as well as cell number. Cell number is akin to a single address into a 2-d array. To these concepts we add index number which extends cell numbers into the 3-d layer space. As shown below, in the first layer cell numbers and index numbers are the same. In the next layer, cell numbers remain the same as in the first layer, but index numbers advance by the total number of cells in a single layer.

The raster_dim() function provides a layout for the shape of a multi-layer raster.

library(rasf)
library(dplyr)
library(terra)
library(raster)
S <- volcano_stack()
(shape <- raster_dim(S))
#  ncol   nrow  ncell nlayer nindex 
#    87     61   5307      3  15921

Raster point extraction

rasf provides functionality to extract points from multi-layer rasters...

index <- sample(shape[['nindex']], 100)
pts <- xyCellLayerFromIndex(index, S) %>%
  dplyr::select(x,y,layer)

# # A tibble: 100 x 3
#          x       y layer
#      <dbl>   <dbl> <dbl>
#  1 6478765 2667705     2
#  2 6479135 2667965     1
#  3 6479315 2667705     2
#  4 6479005 2667915     2
#  5 6479085 2667505     1
#  6 6479085 2667585     1
#  7 6479105 2667815     1
#  8 6479435 2667835     1
#  9 6478895 2667715     3
# 10 6478955 2667675     3
# # … with 90 more rows

values <- extractPts(pts, S)
head(values)
# [1] 139.6309 111.2851 137.6362 150.6019 122.4136 153.7757

Raster random points

Random points can be selected from a multi-layer raster. The user can specify a search 2-d polygon to limit the cells sampled. Missing values can be avoided as can specified points can be explicitly avoided (useful when sampling background points near observations.)

zlim <- raster_range(S)
p <- volcano_polygon()
pts <- randomPts(S, polygon = p)
par(mfrow = c(1,3))
for (i in seq_len(3)){
 plot(S[[i]], main = paste("Layer", i), range = zlim, axes = FALSE)
 plot(p, add = TRUE)
 with(pts %>% dplyr::filter(layer == i), points(x, y))
}

Closest available cells

Often in near shore settings (like estuaries) we would like to know what the sea state is from gridded datasets that might not provide nearshore data. So we look for the closest available cell.

Tables

Sometimes it is desireable to transform a table (data.frame or tibble) with coordinates from one CRS to another.

x <- data.frame(lon = c(-72, -63), lat = c(39, 46))
y <- project_table(x, 
                   from_names = c("lon", "lat"),
                   from_crs = "+init=epsg:4326",
                   to_crs = "+init=epsg:32619",
                   to_names = c("false_easting", "false_northing"))
#   lon lat false_easting false_northing
# 1 -72  39      240199.8        4321059
# 2 -63  46      964564.1        5111577

Hexbin

Finally, sometimes it is convenient to group scattered data into hexagonal bins. The binning process can produce the mean, median or count.

xyz <- hexbin_points()
hb <- st_hexbin(xyz$x, xyz$y, xyz$z)
hb <- dplyr::rename(hb, count = value)
pts <- sf::st_as_sf(xyz,
                    coords = c("x", "y"),
                    crs = "+init=epsg:4326",
                    agr = "identity")
plot(hb, asp = 1, reset = FALSE, axes = TRUE)
plot(pts, col = 'green', add = TRUE)