Criteria 1 limit in BrightSun2020CSDs.m

[cwhanse]

The calculation of c1_lim appears different than described in the JRSE paper.

csd-library/models/BrightSun2020CSDs.m

Lines 568 to 574 in f2d81b6

	z = (20:0.01:90)';
	% produce a linearly spaced correction factor as proposed by Larraneta
	% 2017. Note that they used three fixed bins whereas we smooth this
	% through interpolation
	c1_lim = flip([linspace(BrightSunCriteria.ghi_c1_big_zenith,BrightSunCriteria.ghi_c1_mean_diff,find(z==zenith_turn_around))';...
	linspace(BrightSunCriteria.ghi_c1_mean_diff,BrightSunCriteria.ghi_c1_small_zenith,length(z)-find(z==zenith_turn_around))']);

This calculation produces c1_lim that has a break at zenith=80, rather than at zenith_turn_around = 30 as described in Table 3 of the paper.

>> plot(z, c1_lim)
>> xlabel('z')
>> ylabel('c1\_lim')

Should there be a flipud(z) in the function somewhere? Flipping z produces the right picture:

>> plot(flipud(z), c1_lim)
>> xlabel('z')
>> ylabel('c1\_lim')

[JamieMBright]

Thanks for pointing this out @cwhanse .
This is not my expected behaviour here, i definitely wanted the flipud version.
I cannot fully investigate this at the moment as I am not allowed back in the office with all data.
I definitely struggled fine tuning this criteria though, and this would explain why.
I won't make changes until I can reliably test (as it works for now), but this will need investigating.