I'm trying to plot all values in a Raster Brick for specific points. This is to create spectral plots for remote sensing data for specific pixels.
I can do this in a variety of ways, but they are quite clunky and slow (see example below). This is slow primarily because converting large raster files to a matrix is memory intensive.
Is there a better way to do this with baseR or tidy verse & or a built-in way to do this in one of the Raster / remote sensing packages?
Here's a reproducible example:
library (raster)
library (rgdal)
library (sp)
library (tidyr)
library (ggplot2)
library (dplyr)
##############################
### dummy data
##############################
coord1 <- c(50, 47, 45)
coord2 <- c(50, 51, 49)
frame <- data.frame(coord1, coord2)
coords <- coordinates(frame)
x = list(1, 2, 3, 4, 5)
y = list(1, 2, 3, 4, 5)
for (i in 1:length(x)) { # create clusters based on coords
set.seed(i+100)
x[[i]] <- rnorm(5000, mean = coords[, 1], sd = 1)
y[[i]] <- rnorm(5000, mean = coords[, 2], sd = 1)
}
obs <- data.frame(x, y)
names(obs) <- c('x1', 'x2', 'x3', 'x4', 'x5', 'y1', 'y2', 'y3', 'y4', 'y5')
coordinates(obs) <- c('x1', 'y1') # this creates a spatial points data frame
# create blank raster of study area
ex <- extent(c(45, 50, 48, 52))
target <- raster(ex, resolution = 0.5)
# create raster brick
r_list <- list()
for (i in 1:ncol(obs)) {
r_list[[i]] <- rasterize(obs, target, fun = 'count')
}
obs_frequency <- brick(r_list)
And here's one possible, but slow, solution
############################
### Example Solution
############################
vals_all <- obs_frequency[, ] ### this is very slow ###
vals_all <- data.frame(vals_all)
### sample values
points <- sample_n(vals_all, size = 5)
points <- t(points)
points <- data.frame(points)
points_tidy <- gather(points)
points_tidy$xval <- rep(1:8, times = 5)
### plot
ggplot(points_tidy, aes(x = xval, y = value)) + geom_line(aes(colour = key)) + theme_classic()
I have found a better solution to this using the raster::extract function. This samples values directly & avoids turning the whole raster brick into a memory-busting matrix.
It was worth noting that here, using a Brick is MUCH faster than using a Raster stack.
############################
### Extract values and plot
############################
### extract values
points <- c(49, 50, 51) #arbitrary points
pointvals <- raster::extract(obs_frequency, points) ##### USE THE RASTER::EXTRACT FUNCTION
### manipulate data structure
pointvals <- data.frame(t(pointvals))
points_tidy <- gather(pointvals)
points_tidy$xval <- rep(1:8, times = 3)
### plot
ggplot(points_tidy, aes(x = xval, y = value)) + geom_line(aes(colour = key)) + theme_classic()