how to validate a daylight simulation conducted in Radiance?

asked 2019-08-15 13:45:00 -0600

Wolfgang
41 ●1

updated 2019-08-18 11:52:07 -0600

Hey guys, I know Radiance V.5 has been tested by so many different institutions and theses as well. Even we had a question earlier that "was it tested against CIE 171:2006" and we had two dissertations that responded to the question. My question is how can I validate against field measurements while I'm not having and not enough funding to measure it. Is there any model database that I can use to validate it again? Is this type of validation acceptable for scientific journals?

I would rather prefer to have an experientially measured model with its mean daylight factor. I know it's the best metric to use but I think it's the simplest one since it is supposed to be carried under an overcast sky. Other methods especially the new indicator are dependant on the weather data so it would be harder to validate. I'm using Radiance in Rhino. Any help is highly appreciated. We are also open to collaboration for our scientific paper.

Thank you!

edit retag flag offensive close merge delete

add a comment

4

answered 2019-08-18 08:31:22 -0600

Andyrew

790 ●1 ●10 http://andy.mcneilorama.com/

I'll start by saying you should be realistic about how close a daylight simulation model and reality can be. Funded validations, where care is taken to measure surface reflectances, and monitor sky conditions, routinely have bias error of up to 15% and RMS error of 30-40%. If you're doing an ad hoc validation, point in time error of 50% or more should be expected.

https://dspace.lboro.ac.uk/dspace-jsp... https://www.sciencedirect.com/science...

You should know that there are two categories of error. First, model error, which is error related to reflectance and transmittance characteristics and geometry. More effort can help drastically in these cases.

The second source of error is in the representation of the sky. The sky in your simulation will never match reality. Perez skies generated from direct and diffuse measurements are always wrong. Even HDR captured luminance maps will have error of 10% (http://www.ibpsa.org/proceedings/BS20...).

If I had to do a quick and dirty validation, I'd use a camera with fisheye lens just outside the window ala Mehlika Inanici (https://faculty.washington.edu/inanic...) to capture sky and ground contribution to the window. Then I'd use illuminance sensors and maybe a second camera in the space to compare against simulated models.

Good luck!

edit flag offensive delete link

Comments

Thank you very much Andyrew, just one more question, is there any study that has the mean daylight factor measured experientially? or is there such a thing in CIE 171:2006?

I saw in the first references that interior illuminance and exterior data exist; but how to calculate the mean daylight factor under the overcast sky for it?

Wolfgang ( 2019-08-18 12:46:19 -0600 )edit

None that I know. Sorry.

Andyrew ( 2019-08-27 14:14:12 -0600 )edit

add a comment

1

answered 2019-08-21 11:53:23 -0600

Lars
11 ●1

updated 2019-08-21 13:52:03 -0600

Aaron Boranian

14250 ●86 ●24 http://bigladdersoftware.com/

Please be careful about the idea of using the daylight factor, whenever you measure something in a real building. The daylight factor assumes (and is calculated as such by software) a theoretical "overcast sky", which you will probably never find in nature. So by definition, you compare the results of different sky distributions - and one should start worrying if that leads to agreement. The one exception is cases where the distribution has no impact. Say, you building consists of a glass bubble, where it does not make any difference how the luminance is distributed over the sky hemisphere as long as the integral is correct. But for such cases, there would not be much to validate.

Often, as long as you know direct and diffuse illuminance from measurements, sunny sky conditions give better agreement. Here, you know that most of the illuminance is attributed to the sun direction, and not knowing the exact distribution of the "blue sky" has little impact. But beware - this changes immediately if a bright white cloud comes into sight, which adds some illuminance from its direction, but gets attributed to the entire hemisphere in your diffuse illuminance measurement.