The variable from a netcdf file comes out flipped

I have downloaded a nc file from

f=open.ncdf("file.nc")
[1] "file Lfile.nc has  2 dimensions:"
[1] "Longitude   Size: 1440"
[1] "Latitude   Size: 720"
[1] "------------------------"
[1] "file filr.nc has   8 variables:"
[1] "short ts[Latitude,Longitude]  Longname:Skin Temperature (2mm) Missval:NA"

I then wanted to work with the variable soil_moisture_c

A = get.var.ncdf(nc=f,varid="soil_moisture_c",verbose=TRUE)

I then plot A with image(A). I got the map shown below,I even transposed it image(t(a)) but that was changed to other direction,and not how it should be. Anyway,in order to know what is wrong,I used the netcdf viewer Panoply and the map was correctly plotted as you can see below.

Just as a complement to @mdsumner far better solution, here is a way to do that using library ncdf only.

library(ncdf)
f <- open.ncdf("LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040101.nc")
A <- get.var.ncdf(nf,"soil_moisture_c")

All you need is to find your dimensions in order to have a coherent x and y-axis. If you look at your netCDF objects dimensions, here what you see:

str(f$dim)
List of 2
 $ Longitude:List of 8
  ..$ name         : chr "Longitude"
  ..$ len          : int 1440
  ..$ unlim        : logi FALSE
  ..$ id           : int 1
  ..$ dimvarid     : num 2
  ..$ units        : chr "degrees_east"
  ..$ vals         : num [1:1440(1d)] -180 -180 -179 -179 -179 ...
  ..$ create_dimvar: logi TRUE
  ..- attr(*, "class")= chr "dim.ncdf"
 $ Latitude :List of 8
  ..$ name         : chr "Latitude"
  ..$ len          : int 720
  ..$ unlim        : logi FALSE
  ..$ id           : int 2
  ..$ dimvarid     : num 1
  ..$ units        : chr "degrees_north"
  ..$ vals         : num [1:720(1d)] 89.9 89.6 89.4 89.1 88.9 ...
  ..$ create_dimvar: logi TRUE
  ..- attr(*, "class")= chr "dim.ncdf"

Hence your dimensions are:

 f$dim$Longitude$vals -> Longitude
 f$dim$Latitude$vals -> Latitude

Now your Latitude goes from 90 to -90 instead of the opposite, which image prefers, so:

 Latitude <- rev(Latitude)
 A <- A[nrow(A):1,]

Finally, as you noticed, the x and y of your object A are flipped so you need to transpose it, and your NA values are represented for some reasons by value -32767:

A[A==-32767] <- NA
A <- t(A)

And finally the plot:

image(Longitude, Latitude, A)
library(maptools)
data(wrld_simpl)
plot(wrld_simpl, add = TRUE)

Edit: To do that on your 31 files, let's call your vector of file names ncfiles and yourpath the directory where you stored them (for simplicity i'm going to assume your variable is always called soil_moisture_c and your NAs are always -32767):

ncfiles
 [1] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040101.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040102.nc"
 [3] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040103.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040104.nc"
 [5] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040105.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040106.nc"
 [7] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040107.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040108.nc"
 [9] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040109.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040110.nc"
[11] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040111.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040112.nc"
[13] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040113.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040114.nc"
[15] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040115.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040116.nc"
[17] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040117.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040118.nc"
[19] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040119.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040120.nc"
[21] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040121.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040122.nc"
[23] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040123.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040124.nc"
[25] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040125.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040126.nc"
[27] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040127.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040128.nc"
[29] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040129.nc" "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040130.nc"
[31] "LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T173800Z_20040131.nc"

yourpath
 [1] "C:\\Users"

library(ncdf)
library(maptools)
data(wrld_simpl)
for(i in 1:length(ncfiles)){
    f <- open.ncdf(paste(yourpath,ncfiles[i], sep="\\"))
    A <- get.var.ncdf(f,"soil_moisture_c")
    f$dim$Longitude$vals -> Longitude
    f$dim$Latitude$vals -> Latitude
    Latitude <- rev(Latitude)
    A <- A[nrow(A):1,]
    A[A==-32767] <- NA
    A <- t(A)
    close.ncdf(f) # this is the important part
    png(paste(gsub("\\.nc","",ncfiles[i]), "\\.png", sep="")) # or any other device such as pdf, jpg...
    image(Longitude, Latitude, A)
    plot(wrld_simpl, add = TRUE)
    dev.off()
    }

The reason is that the NetCDF interface you are using is very low-level, and all you have done is read out the variable without any of its dimension information. The orientation of the grid is really arbitrary, and the coordinate information needs to be understood in a particular context.

library(raster) ## requires ncdf package for this file  
d <- raster("LPRM-AMSR_E_L3_D_SOILM3_V002-20120520T185959Z_20040114.nc", varname = "soil_moisture_c")

(I used a different file to yours, but it should work the same).

It turns out that even raster does not get this right without work, but it does make it easy to rectify:

d <-  flip(t(d), direction = "x")

That transposed the data and flipped around "x" (longitude), keeping the georeferencing from the original context.

Plot it up with a map from maptools to check:

plot(d)

library(maptools)
data(wrld_simpl)
plot(wrld_simpl, add = TRUE)

There are many other ways to achieve this by reading the dimension information from the file, but this is at least a shortcut to do most of the hard work for you.

You can also simply invert the latitudes from the command line using CDO:

cdo invertlat file.nc file_inverted.nc

来源：https://stackoverflow.com/questions/13700254/the-variable-from-a-netcdf-file-comes-out-flipped

标签

netcdf