Low simulated illuminance levels under direct sun

Hi,

I am trying to develop a Radiance model of an actual building. I have
collected hourly global horiz. illum. on roof and vertical illum. on
facade (using a LI-COR sensor) and have been making hourly comparisons
between measured and simulated data. The building is surrounded by
several buildings which block much of the direct sun reaching the
facade - I've included the obstructions in the computer model.

While the simulated global vertical illuminance levels on the facade
are within +/-10% of those measured on site under diffuse light
conditions (much of the light reaching the facade is diffuse due to
heavy exterior obstructions), I am getting a high error when for
simulated vs actual illuminance under direct sun. Under direct sun,
illuminance levels are underpredicted by 30% in the computer model.

The site is Berkeley, CA, and I used the Oakland weather file, so the
location should be very close in terms of latitude/longitude and I
double-checked the facade orientation (8 deg east of South). I
excluded all of the exterior obstructions and compared to illuminance
levels one might expect under direct sun for that orientation using
daylight availability tables for San Francisco and the levels once
again seem low.

Here is a bit more info on the model:
CIE clear sky
-ab 5 -ad 1024 -as 256 -ar 1064 -aa 0.1
Max. dimension of the site/model modeled is 4,200 ft

Does anyone have any idea why illuminance levels under direct sun
might be underpredicted?

Thank you,

Krystyna

M.S. in Architecture student
Department of Architecture
University of California - Berkeley