CODEWITHSUNDEEP
rrasterna
MaeAntoinette
MaeAntoinette

Reputation: 303

How to Replace Raster Values Less than 0 to NA in R code

I am working with Landsat imagery to calculate net radiation. There are very low negative values (ex: -0.000003) that are in my converted reflectance raster layers. I want to make sure that my reflectances are 0 - 1 in order to reduce error in my future calculations.

How can I replace raster values less than 0 to "NA" in R. Similar to a raster calc function. I'm unsure how to provide an example, but I'm certain one of you could help me, yes?

Here is my equation for the reflectivity derived from Bastiaanssen et al. (1998)

Assuming, the pb1 indicates the reflectance for band 1 of Landsat, pi = 3.14..., lb1 = radiance for band 1, ESUN = exoatmospheric value for band 1, dr = the relative earth sun distance for the day of year. 

#Calculate reflectivity per band. QC: Always 0 to 1
pb1 = (pi * lb1)/(ESUN1 * cos(solzen) * dr)

After this raster is created, all I want to do is set the pb1 values that are less than 0 to NA.

Help?

Upvotes: 20

Views: 53275

Answers (5)

Robert Hijmans
Robert Hijmans

Reputation: 47706

With the terra package, the memory-safe way to do this is to use classify clamp, or ifel

library(terra)
r <- rast(ncol=10, nrow=10)
values(r) <- rnorm(100)
x <- clamp(r, 0, values=FALSE)
y <- classify(r, cbind(-Inf, 0, NA), right=FALSE)
z <- ifel(r < 0, NA, r)

These methods are memory-safe and you can also provide a filename argument so that you do not need to call writeRaster afterwards.

You can use app but that is a bit clumsy and much less efficient

a <- app(r, function(x) { x[x<0] <- NA; x })

Approaches like this are nice for interactive exploration with small data sets, but they should generally not be used in "production" code.

r[r < 0] <- NA

Upvotes: 9

Robert Hijmans
Robert Hijmans

Reputation: 47706

With the raster package, the memory-safe way to do this is to use reclassify or clamp

library(raster)
r <- raster(ncol=10, nrow=10)
values(r) <- rnorm(100)
x <- clamp(r, useValues=FALSE)
y <- reclassify(r, cbind(-Inf, 0, NA), right=FALSE)

Note the right=FALSE to not set cells that are zero to NA.

These methods are memory-safe and you can also provide a filename argument so that you do not need to call writeRaster afterwards.

Upvotes: 37

Andy Barbour
Andy Barbour

Reputation: 8893

raster::clamp is an easy and flexible way to do this. One can set everything above and/or below threshold values to that threshold, or by setting useValues=FALSE then values above/below are set to NA. For example, lower values only:

r <- raster(ncol=3, nrow=3)
values(r) <- seq_len(9)
x <- clamp(r, lower=3, useValues=FALSE)
values(x)
# [1] NA NA  3  4  5  6  7  8  9

upper values only:

x <- clamp(r, upper=6, useValues=FALSE)
values(x)
# [1]  1  2  3  4  5  6 NA NA NA

and both upper and lower values:

x <- clamp(r, lower=3, upper=6, useValues=FALSE)
values(x)
# [1] NA NA  3  4  5  6 NA NA NA

Note the difference if useValues=TRUE (the default):

x <- clamp(r, lower=3, upper=6)
values(x)
# [1] 3 3 3 4 5 6 6 6 6

Using raster_2.8-19 in this example.

Upvotes: 8

Leonidas
Leonidas

Reputation: 121

Another option is

pb1 <- raster::calc(pb1, function(x){x[x<0]<-NA; return(x)})

Upvotes: 0

eipi10
eipi10

Reputation: 93891

library(raster)

values(pb1)[values(pb1) < 0] = NA

Or, as suggested by @jbaums:

pb1[pb1 < 0] <- NA

If you want to keep the original raster object, remember to assign the original raster to a new object name before running the code above.

Upvotes: 32

Related Questions