[Radiance-general] sky definition part 2

[Zack Rogers]

Hello,

Gensky seems to be a hot topic these days.  My questions have to do with 
the turbidity factor and the accuracy of the gensky in producing a CIE sky.

My quest started today when I noticed in the man pages that for gensky a 
low turbidity factor corresponds to less particles in the atmosphere (1 
being a completely dark sky) and a higher turbidity means more particles 
in the air.  This caused me to question the way we had calibrated some 
skies to better match some measurements taken here in Denver.  I was 
wanting to adjust the turbidity factor to account for the high elevation 
here (higher elevations having less particles in the air).  However, I 
mistakenly adjusted the turbidity up since that was increasing my global 
horiz. illuminance, the expected outcome since higher elevations do have 
a slightly higher global illuminance.

What I discovered (which I would have found documented in the archives, 
Feb, 1995) is that the turbidity only adjusts the sky component and does 
not effect the solar component at all.  So when I was raising turbidity, 
I was increasing atmospheric particles (my sky component) while keeping 
exactly the same direct solar component.  I got the overall result I was 
looking for but looking deeper at the sun and sky components separately, 
I realized my sun was producing 2000fc less than it should and my sky 
was producing 1500fc more!  Here's my comparison for June 21, 40 lat, 
105 long, 12:00PM;

The CIE sky function tells me I need approximately 1486 fc diffuse horiz 
and 8817fc direct horiz for a total of 10303fc

Gensky w/ T=1.5
Sky = 203fc ; Sun = 6836fc ; Global = 7039fc

T=2.5
Sky = 948fc ; Sun = 6836fc ; Global = 7784fc

T=3.5
Sky = 1694fc ; Sun = 6836fc ; Global = 8530fc

T=4.5
Sky = 2438fc ; Sun = 6836fc ; Global = 9274fc

T=5.5
Sky = 3184fc ; Sun = 6836fc ; Global = 10020fc

Obviously T=5.5 is wrong but it does give the closest global value.

So my question now, or problem, is that I would like to be able to 
create true CIE skies but am not sure the best way how.

I started by determining what Turbidity gives me the desired sky 
illuminance.  This came out to be 3.22 for summer, 3.41 for equinox, and 
3.55 for winter.  I am not sure why the variation between summer and 
winter, it seems perhaps because more particles get trapped by cold 
pockets during the winter??

So, using T=3.22 gives me the desired sky component during the summer 
solstice but I still need to specify the correct direct component.  I 
did this by specifying the radiance with the -r option.  I avoided 
having to mess efficacy conversions by just doing it iteratively until 
my global illuminance was what I wanted.  A solar radiance -r value of 
9,250,000 and T=3.22 gave me very good results for the summer sky - 
1486fc sky ; 8819fc sun ; and 10305fc total to be exact. BTW changing 
the -r value does cause a very small increase in the sky component (as 
it should, but not sure if it is accurate).

So, I went to see how these values, calibrated to give CIE summer sky, 
would work for an equinox and winter sky.  I already knew the turbidity 
should be a little different under these times but was hoping the same 
solar radiance value would work reasonably well.  Here's those results 
(all with -r 9,250,000) ;

Equinox, 12:00PM
CIE tells me I need 1328fc sky ; 6894fc sun ; 8223fc global
Radiance gives me 1223fc (T=3.22) or 1330fc (T=3.41) sky ; 7003fc sun ; 
8333fc global

Winter Solstice, 12:00PM
CIE tells me I need 1036fc sky ; 3421fc sun ; 4458fc global
Radiance gives me 919fc (T=3.22) or 1041fc (T=3.55) sky ; 4114fc sun ; 
5033fc global
This sun produces 600fc more than it should!

So, these comparisons tell me that for an accurate CIE sky, I need to 
use a Turbidity somewhere between 3.22 to 4, depending on elevation and 
pollution levels, and that I need to determine separate solar radiance 
values for each sky.

This is a pain.  Why does gensky not produce a more accurate CIE sky 
model?  I guess Greg acknowledged this back in his 1995 post
"

Let me start by saying that I have little confidence in the sky or solar
luminance calculations of gensky.  They are based on some simple 
rule-of-thumb
calculations and mean sky measurements taken years ago.  If you are serious
about your sky model (and it appears that you are), you should insert 
your own
values for sky and solar luminance via the -b and -r (or -B and -R) options
to gensky.  This will override the turbidity calculation for zenith 
luminance,
which as you noted, does not affect solar luminance as it ought.

"
So, my suggestion is to update gensky to accurately implement the CIE 
sky model, or is it the general consensus that gendaylit based on the 
perez sky model is the more accurate model?  The CIE sky model has 
always been resonably accurate for me when comparing with both luminance 
and illuminance measurements - I am always looking for the most average 
sky though and not a completely accurate sky for a given point in time. 

Maybe one the main issues is that the turbidity factor needs to adjust 
the solar component as well as the sky, and the solar component seems 
defaulted at too low of a value for a CIE sky.  Furthermore, maybe it 
would be a good idea just to hide the turbidity factor completely and 
just let the user adjust the elevation and perhaps give a pollution 
factor.  When I created a CIE sky model for TracePro (a highly accurate 
forward ray-tracing program),  I gave it essentially the same inputs as 
gensky (lat long month day time skytype) but instead of a turbidity 
factor, which as I recall is not part of the CIE sky model equations, I 
did some studies on elevation impacts and created elevation factors. 
 The elevation factors were based on weather data from stations at 
altitudes varying from sea level up to 10,000ft.  The results showed 
that the diffuse component decreases roughly 15% at 5,000fc and the 
direct component increases 20% at 5,000fc compared to sea level (If 
anyone is interested in this elevation factor study let me know and I 
can e-mail you the graphs).

Anyways, I'm very interested in the groups thoughts on these issues. 
 Please correct me if I have misunderstood any of these results.  BTW - 
the "accurate" CIE values I reported were calculated with 2 different 
programs - one that descretizes the sky into 1deg by 1deg blocks and 
then summates and the more accurate TracePro macro program that 
descretizes the sky into as many rays as you want to trace.  Both of 
these give very similar results.

Thanks in advance!
Zack

PS It was great meeting everybody out in Berkeley - what a wonderful 
group of techno geeks we are!
PPS Sorry for the long e-mail

-- 
Zack Rogers
Staff Engineer
Architectural Energy Corporation
2540 Frontier Avenue, Suite 201
Boulder, CO 80301 USA

tel (303)444-4149 ext.235
fax (303)444-4304
www.archenergy.com