Notes on steps taken to create Greenland_5km_v1.1_SacksRev_c110629.nc
from Greenland_5km_v1.1.nc

Bill Sacks (wsacks@gmail.com)
June 29, 2011

The new file is identical to the old one except in the landcover
variable: Modified landcover to have value 0 [no_data] in some places
where it used to have value 1 [ocean], because some grid cells were
incorrectly classified as ocean (e.g., over Ellesmere Island);
however, the method for doing this means that many true ocean points
are now classified as no_data.

Note that the scripts fill_greenland.ncl and interpolate_greenland.ncl
can be found here:
https://svn-user-sacks.cgd.ucar.edu/cism_misc/fixing_greenland_data/trunk

---

Creating greenland mask map:

ncks -d latitude,58.,85. -d longitude,-93.,11. ~/from_glacier/political_masks/gl_ad0_5min.nc gl_ad0_5min_subset.nc

In R: Created mask equal to 1 over greenland, 0 over other countries:

> map <- read.ncdf('gl_ad0_5min_subset.nc')
> map$mask <- map$adminX_5min
> w <- which(map$adminX_5min$data != 1)
> map$mask$data[w] <- 0
> length(w)
[1] 64123
> write.ncdf('greenland_mask.nc', map)

Then removed degenerate dimensions:

ncwa -O -a level greenland_mask.nc greenland_mask.nc
ncwa -O -a time greenland_mask.nc greenland_mask.nc


---

Then extrapolated to greenland_mask_extrap.nc with the following; this
gives interpolated / extrapolated values of the mask in the ocean:

ncl fill_greenland.ncl

and reversed latitude with:

ncpdq -O -a -latitude greenland_mask_extrap.nc greenland_mask_extrap.nc

---

Did the following pre-processing of the original Greenland data:

Removed degenerate time dimension and just output lat & lon variables:

ncwa -O -a time -v lat,lon Greenland_5km_v1.1.nc Greenland_5km_v1.1_latlon_notime.nc

----

Then did the regridding: Regridded mask to 5km resolution:

ncl interpolate_greenland.ncl
ncrename -d ncl0,y1 -d ncl1,x1 greenland_mask_5km.nc
ncks -A -v lat,y1,lon,x1 Greenland_5km_v1.1_latlon_notime.nc greenland_mask_5km.nc

Added degenerate time dimension:

ncecat -O greenland_mask_5km.nc greenland_mask_5km.nc
ncrename -d record,time greenland_mask_5km.nc

----

Then used the mask to modify the landcover value: for points
classified as ocean (landcover=1) that are relatively far from
Greenland (mask_greenland < 0.666667), reclassify them as nodata
(landcover=0):

cp Greenland_5km_v1.1.nc Greenland_5km_v1.2.nc
ncks -A -v mask greenland_mask_5km.nc Greenland_5km_v1.2.nc
ncrename -v mask,mask_greenland Greenland_5km_v1.2.nc

ncap2 -O -s 'where(mask_greenland<0.666667 && landcover==1) landcover=0;' Greenland_5km_v1.2.nc Greenland_5km_v1.2.nc

ncks -O -x -v mask_greenland Greenland_5km_v1.2.nc Greenland_5km_v1.2.nc

---

Then added a note to the file:

ncatted -h -a note,landcover,c,c,'Bill Sacks (wsacks@gmail.com) (6-29-11): Modified landcover to have value 0 [no_data] in some places where it used to have value 1 [ocean], because some grid cells were incorrectly classified as ocean (e.g., over Ellesmere Island); however, the method for doing this means that many true ocean points are now classified as no_data' Greenland_5km_v1.2.nc


-----

10-3-11: 

To avoid name conflicts (because the original data creators may be
producing their own v. 1.2 of the data), I am renaming this file from
Greenland_5km_v1.2.nc to Greenland_5km_v1.1_SacksRev_c110629.nc.