[Wrf-users] Losing sea ice in hinterp

[Don Morton]

Howdy, I'm a computer scientist who pretends to be
a weather modeller.

I'm having problems getting sea-ice into some experimental
runs I'm doing for the Alaska region, initialized by the
NAM 216 grid (AWIPAK).  

An example of the problems can be seen by going to

http://pileus.arsc.alaska.edu/weather/

click on "North Alaska CWA WRF Simulations" then

click on "Surface Pressure and Temperature" button on right,

then drag mouse down across forecast areas on the left.

You should see some pretty funky gradients up in the north!

I've been told by others that it appears the model isn't seeing
any sea ice, and this sure seems logical to me, so the following
describes how I've gone about narrowing down the problem.

Step 1 - use ncdump to look through the wrfout files, looking at
the XICE field.  The data shows very minimal presence of any ice.
I think the largest value I see is 0.25, but most values, even at
the top, are 0.

Step 2 - use ncdump to look at the XICE fields in the
wrf_real_input_em files.  I see essentially the same thing I
see in the wrfout files - very little sea ice.

Step 3 - Starting on the other end, use gribdump and wgrib to view
the AWIPAK input files.  I find fields "ICEC" and "ice_conc" which
appear to be the same (Parameter 91), and the higher numbered rows
(more northerly rows) seem to have a lot of 1's, suggesting reasonable
sea ice.

Step 4 - Looking at my degribbed files using "plotfmt.exe", the
graphical
output I get looks to me like the sea ice coverage is reasonable.  

http://www.cs.umt.edu/~morton/SEAICE/SEAICE.jpg    and
http://www.cs.umt.edu/~morton/SEAICE/LANDSEA.jpg

are graphics that lead me to the conclusion that the sea ice
fields are at least making it to the degribbed intermediate files.

Step 5 - Using "siscan" to view the files produced by "hinterp" 
I see that the SEAICE field shows a max value of 0.45. At this 
point, I conclude that hinterp ain't playing nice with the sea 
ice, as I would certainly expect to see a max value of 1.  See

http://www.cs.umt.edu/~morton/SEAICE/hinterp.txt

for the output of "siscan"

Step 6 - Look at the hinterp logs.  I only have some vague guesses about
what I'm looking at, but it appears that the processing is at least
proceeding.  I notice that the interpolation used is "nearest neighbor"
which the ARW Tutorial suggests is "not recommended" but, I don't see
how that approach would "melt" all my sea ice :)  I've pasted a small
excerpt from the hinterp log.  Again, I don't really know what I'm
looking at, so maybe there's something significant in here?

----------------------------------------
Doing masked interpolation for SEAICE  
INTERPOLATE_MASKED: Using source landmask field.
Masked interpolation using nearest neighbor value...
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   142  120
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   261  136
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   281  143
INTERPOLATE_MASKED: Bogus value of      0.000 used at point    84  150
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   313  150
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   245  177
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   195  179
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   208  180
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   219  180
INTERPOLATE_MASKED: Bogus value of      0.000 used at point   178  182
INTERPOLATE_MASKED: Bogus value of      0.000 used at point    23  207
----------------------------------------
MASKED INTERPOLATION SUMMARY:
  TOTAL POINTS IN GRID:            71456
  POINTS NEEDING VALUES:           34771
  POINTS NOT REQUIRED:             36685
  POINTS NEEDING FIX:              25827
  POINTS FIXED WITH OUT GRID:      25808
  POINTS FIXED WITH SRC GRID:          8
  POINTS FIXED WITH DEF VAL:          11
----------------------------------------
-----------------------------------------------------------


If anybody has some ideas or directions they could
point me in, I'd be very appreciative!

Thanks,

Don

---
Don Morton
Department of Computer Science
The University of Montana
Missoula, Montana
http://www.cs.umt.edu/~morton/