geometry

[mankoff]

Given the GEOM and GEOM_8x10 code on #91 (http://dev.edgcm.columbia.edu/attachment/ticket/91/GEOM_8x10.f), it appears that the budget pages are using GEOM (7.826x10).

To replicate for the Modern_Specified_SST run (average over years 1963-67) supplied with EdGCM 2.5 Beta, the values for

GLOBAL ANN SRFAIRTMP:
Global = 13.34, NH = 14.10, SH=12.58

Yet if I take the GEOM_8x10 dxyp values and try to calculate the SRFAIRTMP Global I get 13.4225

If I use GEOM dxyp values (7.826x10) then I get 13.339 (~ 13.34)

---

An alternate explanation is that the budget pages are correctly calculated with 8x10, but the netCDF maps are somehow produced on a 7.826x10 grid, and therefore they need to use the GEOM code and not GEOM_8x10 in order to replicate the budget page values.

It feels more likely that the budget pages are using the wrong code.

[mshopsin]

A check SumAndPD says that it does call the 8x10 or 7.826x10 code correctly and that geometry code itself is ok. The problem might be a bug in the geometry code or a difference between real*4 and real*8 or something else. Jean probably needs to get involved since she understands the diagnostics better than I do.

[mankoff]

summary of latest findings: no difference between dxyp outputs of 8 or 7.826 (except for 2 most polar cells at either hemisphere).

Assumption is that 7.826 is correct.

Furthermore, 8x10 has special code that runs from 3 to JM-2. This appears anomalous because there is no reason for cell 2 and JM-1 to be treated differently than any of the other non-pole cells.

More later as I work on this...

[mankoff]

It appears this isn't just something with the budget pages, but the model in general, hence the change in the title.