问题
I have a NetCDF file with rotated coordinates. I need to convert it to normal lat/lon coordinates (-180 to 180 for lon and -90 to 90 for lat).
library(ncdf4)
nc_open('dat.nf')
For the dimensions, it shows:
[1] " 5 variables (excluding dimension variables):"
[1] " double time_bnds[bnds,time] "
[1] " double lon[rlon,rlat] "
[1] " long_name: longitude"
[1] " units: degrees_east"
[1] " double lat[rlon,rlat] "
[1] " long_name: latitude"
[1] " units: degrees_north"
[1] " char rotated_pole[] "
[1] " grid_mapping_name: rotated_latitude_longitude"
[1] " grid_north_pole_longitude: 83"
[1] " grid_north_pole_latitude: 42.5"
[1] " float tasmax[rlon,rlat,time] "
[1] " long_name: Daily Maximum Near-Surface Air Temperature"
[1] " standard_name: air_temperature"
[1] " units: K"
[1] " cell_methods: time:maximum within days time:mean over days"
[1] " coordinates: lon lat"
[1] " grid_mapping: rotated_pole"
[1] " _FillValue: 1.00000002004088e+20"
[1] " 4 dimensions:"
[1] " rlon Size:310"
[1] " long_name: longitude in rotated pole grid"
[1] " units: degrees"
[1] " axis: X"
[1] " standard_name: grid_longitude"
[1] " rlat Size:260"
[1] " long_name: latitude in rotated pole grid"
[1] " units: degrees"
[1] " axis: Y"
[1] " standard_name: grid_latitude"
[1] " bnds Size:2"
Could anyone show me how to convert the rotated coordinates back to normal lat/lon? Thanks.
回答1:
I would use cdo for this purpose https://code.zmaw.de/boards/2/topics/102
Another option is just create a mapping between rotated and geographic coordinates and use the original data without interpolation. I can find the equations if necessary.
回答2:
NCO's ncks can probably do this in two commands using MSA
ncks -O -H --msa -d Lon,0.,180. -d Lon,-180.,-1.0 in.nc out.nc
ncap2 -O -s 'where(Lon < 0) Lon=Lon+360' out.nc out.nc
回答3:
I went through the CDO link as suggested by @kakk11, but somehow that could not work for me. Afte much research, I found a way
First, convert the rotated grid to curvilinear grid
cdo setgridtype,curvilinear Sin.nc out.nc
Next transform to your desired grid e.g. for global 1X1 degree
cdo remapbil,global_1 out.nc out2.nc
or for a grid like below
gridtype = lonlat
xsize = 320 # replace by your value
ysize = 180 # replace by your value
xfirst = 1 # replace by your value
xinc = 0.0625 # replace by your value
yfirst = 43 # replace by your value
yinc = 0.0625 # replace by your value
save this info as target_grid.txt and then run
cdo remapbil,target_grid.txt out.nc out2.nc
In my case there was additional issue that my variables did not have the grid information. so CDO assumed it to be regular lat-long grid. So before all the above-mentioned steps, I had to add grid information attribute to all the variables (in my cases all the variables ended with _ave) using nco
ncatted -a coordinates,'_ave$',c,c,'lon lat' in.nc
ncatted -a grid_mapping,'_ave$',c,c,'rotated_pole' in.nc
Please note that your should have a variable called rotated_pole in your nc file with the lat long information of rotated pole.
回答4:
There is also the possibility to do that in R (as the User is referring to it in the question). Of course, NCO and CDO are more efficient (way faster). Please, look also at this answer.
library(ncdf4)
library(raster)
nsat<- stack (air_temperature.nc)
##check the extent
extent(nsat)
## this will be in the form 0-360 degrees
#change the coordinates
nsat1<-rotate(nsat)
#check result:
extent(nsat1)
##this should be in the format you are looking for: -180/180
Hope this helps.
[edited]
来源：https://stackoverflow.com/questions/31549880/how-to-convert-a-rotated-netcdf-back-to-a-normal-lat-lon-grid