R terra/raster: uploading a netcdf file alters resolution

Question

I have netcdf files (GPM IMERG precipitation) which are supposed to be at0.1x0.1 resolution like in the red box at the bottom here:

Well the pixel size is not exactly 0.1 on the image, but at least its height and width are the same.

However when uploading the netcdf via terra:: or raster:: the resolution slightly changes:

class       : SpatRaster 
dimensions  : 185, 97, 1  (nrow, ncol, nlyr)
resolution  : 0.09999998, 0.1000001  (x, y)
extent      : 34.2, 43.9, 62.3, 80.80002  (xmin, xmax, ymin, ymax)
coord. ref. : lon/lat WGS 84 
source      : _(1).nc:precipitationCal 
varname     : precipitationCal (Daily accumulated precipitation (combined microwave-IR) estimate) 
name        : precipitationCal 
unit        :               mm 
time (days) : 2000-06-04 

Any solution on how to transform the rasters in R back to their original resolution would be welcomed. Btw, does this has to do with rounding coordinates of the extent in R packages?

Answer 1

Can you point to that file? From what I can see, either the extent or the resolution shown on the website is wrong. Based on these numbers, it should be:

print((43.9000038 - 34.20000855) / 97, digits=10)
#[1] 0.09999995103

Perhaps for the website the resolution was computed using the center of the cells, so that you get

print((43.95 - 34.15) / 97, digits=10)
#[1] 0.1010309582

It is hard to be sure without the file. But I would put my money on the website being wrong.

---

With your file I also see this

f <- "__1_.nc"
r <- rast(f)
res(r)
#[1] 0.09999998 0.10000007

But that is correct in the sense that that does reflect the x and y coordinates written to the file. The x and y coordinates for the cells that are provided are not strictly regular (but nearly so). For example, here are the first and last three y coordinates:

62.3500022888184 62.4500083923340 62.5499992370605 ...  
80.5500030517578 80.6500091552734 80.7500152587891

From the file alone I cannot say whether these small decimals are intentional or stem from sloppiness, I assume the latter and in that case it indeed seems reasonable to round the extent:

ext(r) <- round(ext(r), 4)
r
#class       : SpatRaster 
#dimensions  : 185, 97, 1  (nrow, ncol, nlyr)
#resolution  : 0.1, 0.1  (x, y)
#extent      : 34.2, 43.9, 62.3, 80.8  (xmin, xmax, ymin, ymax)
#coord. ref. : lon/lat WGS 84 
#source      : __1_.nc:precipitationCal 
#varname     : precipitationCal (Daily accumulated precipitation #(combined microwave-IR) estimate) 
#name        : precipitationCal 
#unit        :               mm 
#time (days) : 2000-06-04