quad_interpolate developer tests large interp residual

[hkershaw-brown]

Not sure of the test here, I ran the quad_interpoalte tests out of the box and plotted the results.

https://github.com/NCAR/DART/tree/quad_dev_test_play
also has gaussian pivot unit test.

[hkershaw-brown]

Get a bunch (15953) of 'large interp residual'

large interp residual            3 -0.78632070799036002      -0.78632070560813749       -2.3822225303504752E-009
 large interp residual            4   2.5516798628345159        2.5516798638937366       -1.0592207111415064E-009
 large interp residual            1  -888888.00000000268       -888888.00000000000       -2.6775524020195007E-009
 large interp residual            2  -4.5552459289018798E-002  -4.5552455967070679E-002  -3.3219481193613731E-009
 large interp residual            3 -0.78632070799036002      -0.78632070560813749       -2.3822225303504752E-009
 large interp residual            4   2.5516798628345159        2.5516798638937366       -1.0592207111415064E-009

 --------------------------------------
 Finished ... at YYYY MM DD HH MM SS = 
                 2026  4 15 12 49  8
 Program test_quad_irreg_interp v11.22.0
 --------------------------------------

./test_quad_irreg_interp | grep 'large interp residual' | wc -l
15953

15953

If I am counting correctly:
sampling grid size
integer, parameter :: nrx = 210
integer, parameter :: nry = 150

210×150 calls to quad_lon_lat_evaluate, which is 4 (one at each quad corner) x210x150 interp residual checks (=126000)

15953/126000 = 12%

DART/models/utilities/quad_utils_mod.f90

Lines 2271 to 2272 in acf0a14

	if(abs(interp_val - p(i)) > 1e-9) &
	write(*, *) 'large interp residual ', i, interp_val, p(i), interp_val - p(i)

Maybe 1e-9 is not large?

DART/models/utilities/quad_utils_mod.f90

Line 2266 in acf0a14

! When interpolating on dipole x3 never exceeded 1e-9 error in this test

maybe this comment is not true?

Maybe the massive negative value in the out of the box test means you'll always get large interp residuals with the quad bilinear interp.

Tangentially related CROCODILE-CESM/DART_interface#22 #994 and the note in the quad_utils issue #860 (comment) these messages about large interp residuals are not written to the dart log.

[hkershaw-brown]

idw
./test_quad_irreg_interp | grep 'large interp residual' | wc -l
0

Different shape o.

[hkershaw-brown]

Added qceff label because of out-of-bounds