Undefined local variable with regrid2.crossSection.get_latitude_wts_bnds

[zshaheen]

@chengzhuzhang Uncovered this error.

A quick glance at the get_latitude_wts_bnds() function shows that wts is only declared in the if statements, thus an error on the line return (wts, bnds).
Maybe we need to initialize wts to None?

Your thoughts @dnadeau4

Traceback (most recent call last):
  File "acme_diags_driver.py", line 79, in <module>
    run_diag(parameter)
  File "/Users/zhang40/anaconda2/envs/acme_diags_env/lib/python2.7/site-packages/acme_diags-0.1-py2.7.egg/acme_diags/driver/zonal_mean_2d_driver.py", line 199, in run_diag
    mv1_reg = mv1_p.crossSectionRegrid(lev_out, lat_out)
  File "/Users/zhang40/anaconda2/envs/acme_diags_env/lib/python2.7/site-packages/cdms2/avariable.py", line 1217, in crossSectionRegrid
    fromlat, newLatitude, fromlevel, newLevel)
  File "/Users/zhang40/anaconda2/envs/acme_diags_env/lib/python2.7/site-packages/regrid2/crossSection.py", line 116, in __init__
    lat_wts_bndsIn = get_latitude_wts_bnds(latIn)
  File "/Users/zhang40/anaconda2/envs/acme_diags_env/lib/python2.7/site-packages/regrid2/crossSection.py", line 678, in get_latitude_wts_bnds
    return (wts, bnds)
UnboundLocalError: local variable 'wts' referenced before assignment

[doutriaux1]

@zshaheen can you provide the in and out files please? (link to file is enough)

[zshaheen]

This is the model file on aims4 that breaks it:

/export_2/zhang40/tests/acme_output_test/20161118.beta0.F1850COSP.ne30_ne30.edison_ANN_climo.nc

Here's one that we've been using forever that works:

/space/golaz1/ACME_simulations/20160520.A_WCYCL1850.ne30_oEC.edison.alpha6_01/pp/clim_rgr/0070-0099/20160520.A_WCYCL1850.ne30_oEC.edison.alpha6_01_ANN_climo.nc

Let me know if you want me to try and rerun any changes to cdms in acme diags.

Thanks guys :)

[doutriaux1]

thhanks @zshaheen do you know what the target grid is?

[zshaheen]

We're regridding the model to the level and latitude of the obs. So we're doing:

model.crossSectionRegrid(obs.getLevel(), obs.getLatitude())

The observation data is located here on aims4:

/space1/test_data/obs_data_20140804/

[zshaheen]

@doutriaux1
We also see that for the working model data (20160520.A_WCYCL1850.ne30_oEC.edison.alpha6_01_ANN_climo.nc) we have the dimensions for T to be 1:

        float T(time, lev, lat, lon) ;
                T:mdims = 1 ;
                T:units = "K" ;
                T:long_name = "Temperature" ;
                T:cell_methods = "time: mean" ;

While for the non-working model data (20161118.beta0.F1850COSP.ne30_ne30.edison_ANN_climo.nc), we have the dimensions for T to be 9:

        float T(time, lev, lat, lon) ;
                T:mdims = 9 ;
                T:units = "K" ;
                T:long_name = "Temperature" ;
                T:cell_methods = "time: mean" ;
                T:cell_measures = "area: area" ;

---

This is the ncdump of the obs we're using (/space1/test_data/obs_data_20140804/MERRA_ANN_climo.nc):

	float T(time, lev, lat, lon) ;
		T:info = "function clmMonLLLT: contributed.ncl" ;
		T:time_op_ncl = "Climatology: 20 years: 1990-2009" ;
		T:missing_value = 1.e+20f ;
		T:valid_range = 250.4688f, 251.875f ;
		T:missing_value_original = -32767s ;
		T:_FillValue_original = -32767s ;
		T:units = "K" ;
		T:long_name = "Air Temperature" ;
		T:_FillValue = 1.e+20f ;

[golaz]

What is mdims? Is CDMS actually reading mdims? mdims is definitely not the dimensions of the variable 'T'; they are given by the dimensions time, lev, lat, lon.

[doutriaux1]

@golaz I agree, I wonder what mdims is. @dnadeau4 is that something you use internally?

[doutriaux1]

I'm getting closer, any gaussian grid with an odd number of latitudes seems to fail. seems like _regrid.gridattr(nlat, grid_type) expects even number of latitudes

[doutriaux1]

ok both odd gaussian axes and generic latitude axes are now fixed at: #154
As soon as @zshaheen and jill confirmed it fixes things for them, i will merge.

[zshaheen]

@doutriaux1 It's all good to merge into master.
Thanks again for the help!