In annakat/spectrolab: Class and Methods for Spectral Data

library("spectrolab")
knitr::opts_chunk$set(echo = TRUE)

The problem

Instruments may measure different spectral ranges using different sensors, resulting in abrupt "jumps" in the reflectance or radiance data. In such cases, the regions between sensors need to be matched, i.e. spliced together. Unmatched spectra collected with a 3-sensor instrument, such as the SVC HR1024, may look like this:

# Path to raw (unmatched) spectra
path_raw = system.file("extdata/svc_raw_and_overlap_matched_serbin/SVC_Files/",
                       package = "spectrolab")

# Read spectra as reflectance and radiance
reflect_raw  = read_spectra(path = path_raw, type = "target_reflectance")

radiance_raw = read_spectra(path = path_raw, type = "target_radiance")

# Sensor overlaps marked with vertical dashed lines
lwd = 0.5
cex = 0.7
par(mfrow = c(2, 1))

plot(reflect_raw, main = "Reflectance",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)
abline(v = c(990, 1900), col = "red", lty = 2, lwd = lwd)

plot(radiance_raw, main = "Radiance",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)
abline(v = c(990, 1900), col = "red", lty = 2, lwd = lwd)

The solution

Use the function match_sensors to splice the sensor overlap regions. You must pass the boundary between sensors using the splice_at argument. It is critical that you get those bands right (or very close) and every instrument (even from the same vendor) is different.

You can use plot_interactive zoom into a particular spectral region and decide what the splice_at values should be. You can also use the function guess_splice_at to estimates these bands for you but keep in mind that these are guesses.

# Spectrolab's guess of what the splice bands are.
# However, you should also visually inspect the spectra to determine what the
# boundaries between sensors are.
splice_bands_guess = guess_splice_at(reflect_raw)
splice_bands_guess

# Finally, if you know what those sensor bounds should be (say, they're given by
# the manufacturer), just use those numbers instead of spectrolab's guess.
splice_bands = c(990, 1900)

# Match the reflectance and radiance data
reflect_matched = match_sensors(x = reflect_raw, splice_at = splice_bands,
                                interpolate_wvl = c(5, 1))

radiance_matched = match_sensors(x = radiance_raw, splice_at = splice_bands,
                                 interpolate_wvl = c(5, 1))

lwd = 0.5
cex = 0.7
par(mfrow = c(2, 1))

plot(reflect_raw, main = "Reflectance",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(reflect_matched, col = "red", add = TRUE,
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(radiance_raw, main = "Radiance",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(radiance_matched, col = "red", add = TRUE,
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

And we can check the results from spectrolab's match_sensors against SVC's proprietary matching algorithm.

path_moc = system.file("extdata/svc_raw_and_overlap_matched_serbin/SVC_Files_moc/",
                       package = "spectrolab")

reflect_moc  = read_spectra(path = path_moc, type = "target_reflectance")

radiance_moc = read_spectra(path = path_moc, type = "target_radiance")

lwd = 0.5
cex = 0.7
par(mfrow = c(2, 1))

plot(reflect_moc, main = "Reflectance", col = "black",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(reflect_matched, col = "red", add = TRUE,
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(radiance_moc, main = "Radiance", col = "black",
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)

plot(radiance_matched, col = "red", add = TRUE,
     lwd = lwd, cex.main = cex, cex.lab = cex, cex.axis = cex)