[ferret_users] Inaccurate area weights calculation

[Marco Steinacher <steinacher@xxxxxxxxxxxxxxxx>]

Dear Ferret Users,

I'm struggling with a problem related to the inaccurate calculation of
weights for area-weighted operations like @ave and @din in Ferret when
using a very coarse grid.

The grid has 36x36 equal-area cells and it is regular spaced in
longitude (10deg). In the NetCDF file the coordinates of the center
(lat_t) and of the edges (lat_u using the 'edges' attribute) is given
(see below).

When looking only at the grid cells at one longitude (eg. i=1) the
weight of each grid cell should be 1/36 = 0.028. However, the weights
calculated by Ferret differ largely. Obviously Ferret calculates the
weights as

  w1 = (lat_u[j+1]-lat_u[j])*dlon*cos(lat_t[j])

instead of

  w2 = ( sin(lat_u[j+1]) - sin(lat_u[j]) )*dlon

(Here, dlon is the spacing of the grid cells in longitude and, of
course, everything is converted to radians for the calculation). For j=1
we get the following (normalized) weights

  w1 = 0.038
  w2 = 0.028.

w2 is correct but the weight calculated by Ferret is more than 30% higher!

I'm wondering if anybody has experienced similar problems and if there
is any workaround to force Ferret to use the correct area-weights for
@ave etc. Any comments and suggestions on this issue are very welcome!

Many thanks,
Marco


PS: I know that I can use var[x=@sum,y=@sum]/var[x=@ngd,y=@ngd] to get
the arithmetic average which is equal to the weighted average in this
specific case. Nevertheless, I would prefer to be able to use @ave, @din
etc. as well.

--------------------------------------------------------

double lat_t(lat_t) ;
                lat_t:long_name = "Latitude (T grid)" ;
                lat_t:units = "degrees_north" ;
                lat_t:minimum = -76.4637972626189 ;
                lat_t:maximum = 76.4637972626188 ;
                lat_t:edges = "lat_u" ;

double lat_u(lat_u) ;
                lat_u:long_name = "Latitude (U grid)" ;
                lat_u:units = "degrees_north" ;
                lat_u:minimum = -90. ;
                lat_u:maximum = 89.9999991462264 ;

lat_t = -76.4637972626189, -66.4435356908988, -59.4415682140651,
        -53.6639424853861, -48.5903778907291, -43.9829631303587,
        -39.7090165530684, -35.6853347126521, -31.8554308240259,
        -28.1786427405299, -24.6243183521641, -21.1684488458325,
        -17.7915905730076, -14.4775121859299, -11.2122714176497,
         -7.98355614555541, -4.78019184719916, -1.59175417658591,
          1.5917541765859,   4.78019184719916,  7.98355614555541,
         11.2122714176497,  14.4775121859299,  17.7915905730076,
         21.1684488458324,  24.6243183521641,  28.1786427405299,
         31.8554308240259,  35.6853347126521,  39.7090165530684,
         43.9829631303587,  48.5903778907291,  53.6639424853861,
         59.4415682140651,  66.4435356908988,  76.4637972626188 ;

lat_u = -90,               -70.8118635462791, -62.7339555492672,
        -56.4426902380793, -51.0575587310186, -46.2382573073202,
        -41.8103148957786, -37.6698869643296, -33.7489885958886,
        -30,               -26.387799961243,  -22.8853804761586,
        -19.4712206344907, -16.1276202131608, -12.8395884069042,
         -9.5940682268606,  -6.37937020844281, -3.18473853672041,
         -3.18055468146e-15, 3.18473853672041,  6.3793702084428,
          9.59406822686046, 12.8395884069041,  16.1276202131608,
         19.4712206344907,  22.8853804761586,  26.387799961243,
         30,                33.7489885958886,  37.6698869643296,
         41.8103148957786,  46.2382573073202,  51.0575587310186,
         56.4426902380793,  62.7339555492672,  70.8118635462791,
         89.9999991462264 ;

-- 
***************************************
Marco Steinacher

Climate and Environmental Physics
Physics Institute, University of Bern
Sidlerstrasse 5, CH-3012 Bern

Phone ++41 (0)31 631 34 02
Fax   ++41 (0)31 631 87 42
steinacher@xxxxxxxxxxxxxxxx
http://www.climate.unibe.ch/
***************************************