WRF output grid is smaller than input definition

I am running some test forecasts using WRF 3.6.1. and am having some problems with my grids. Specifically, I am specifying a parent and nest domain each with 25 points for longitude and latitude (so that there is a point right at the middle of the inner domain and the parent grid points are coincident with inner grid points with a 3:1 nesting ratio).

I would expect an output of 25 by 25 points for each domain. however, I am only getting 24 by 24 points as output. In fact if I look at the geom_em or met_em netcdf files generated by WPS I already get only grids with 24 by 24 points.

What's going on?

Here is the geogrid section of my namelist.wps file:

&geogrid
parent_id         =   1,   1,
parent_grid_ratio =   1,   3,
i_parent_start    =   1,  9,
j_parent_start    =   1,  9,
s_we              =   1,  1,
e_we              =  25, 25,
s_sn              =   1,  1,
e_sn              =  25,  25,
geog_data_res     = '10m','2m',
dx = 18000,
dy = 18000,
map_proj = 'lambert',
ref_lat   =  53.526,
ref_lon   =  -113.529,
truelat1  =  53.526,
truelat2  =  53.526,
stand_lon =  -113.529,
geog_data_path = '/media/data/WRF3.6.1/data/geog/'
/

• Are all of the grids 24x24 or just the w and scalar grids? Is the u grid 25x24 and the v grid 24x25? I'm not familiar with WRFs grids but I have a hunch... – casey May 6 '15 at 20:18
• you are right. The XLONG, XLAT, T2, W variables are 24x24. The U and V are (as you have mentioned) 24x25 and 25x24 respectively. So it has something to do with grid staggering? – piyushnz May 6 '15 at 20:46

1 Answer

Your model grid is an Arakawa C grid:

Image from Skamarock et al. A description of the Advanced Research WRF Version 3

It appears the geogrid grid generation uses the 25x25 as the furthest extents of the grid (and this is backed up by the grid maximum needing to be a multiple of the grid ratio +1). As a result, only the u and v variables will see 25 grid points and each only in one direction, east/west and north/south respectively. All other variables are located horizontally on the $\theta$, or scalar points in the middle of the grid boxes and there will only be 24 of these points in your setup leaving most of your grids having a horizontal dimensions of 24x24.

• Exactly Right! If his start value had been 0 (instead of 1) then the end value would be 24. It is an inclusive extent range. – farrenthorpe May 6 '15 at 21:25
• yes, I believe you are right. Reading the documentation again, I see it says that the e_we and e_sn values do in fact refer to an unstaggered grid. The terminology for the centered coordinates is the mass grid. I think I am doing the right thing with my nested domains, but I might have to rethink how to set up my grids to align with a validation point (originally I was going to centre the grid on such a point, but the mass grid then is not aligned...) – piyushnz May 6 '15 at 21:39
• @farrenthorpe Thanks. I had to think for a minute as the model I use (CM1) does it the other way around where you specify nx and ny for the scalar/mass points and it uses nx+1, ny+1 for the u,v grids. – casey May 6 '15 at 23:05
• I can confirm this answer is correct. – milancurcic May 6 '15 at 23:56