Trees -- Time cost of different tree modelling approaches


Jack is right on, as is Greg with his comment about the roughness. In an annual simulation context, I think we're talking about grid calcs and not pretty pictures (rtrace, not rpict; a bunch of points, not a view) -- the latter being too expensive. So with some experimentation you could probably create some simplified geometry or even a pattern to approximate the effect of the trees with a lower overhead (in terms of modeling time as well as computation time).

To answer your question about a beginner's list, this radiance-general list -- in addition to the list archives -- is "open class", serving beginners and experts for over a decade. You will love it here.

- Rob G.


On Dec 6, 2011, at 7:32 PM, Jack de Valpine wrote:

Hi Tim,

Short answer is that I have not calculated the run time differences for the scenarios you suggest (at least not in recent memory). However I have done lots of visualizations with LOTS of trees (using instanced trees). If you have lots of trees you probably need to use instances to help manage scene memory. If on the other hand you are only trying to account for a few trees then perhaps you can get away with a relatively "flat" octree formation (eg using straight geometry not instancing frozen octrees).

Possibly pulling together this thread and your questions on trans materials, leaves on trees and annual simulations, I would offer a few thoughts though. I would suggest that having a trans material applied on your leaves could increase your simulation time pretty significantly depending on your run parameters, just think of all those trans calculations that are occurring as a ray is traced from one trans to another to another.... I would also suggest that for an annual (hourly) simulation this level of simulation detail is probably not that critical...

Others I am sure will offer their (greater) wisdom.


-Jack de Valpine

On 12/6/2011 6:25 PM, Tim Perry wrote:

Has anyone ever quantified the run-time differences between the following?
1) Creating a tree similar the pine tree Greg provides with the Radiance distribution and insert the octree into the scene multiple times.
2) Creating a tree similar to the Pine tree Greg provides, but insert the actual .rad definition into the scene multiple times before running oconv.
3) Creating a tree outline in 2-d and rotating it 90 degrees about the z-axis to get some tree-like effects.

Has anyone ever quantified how large the effect would be for an annual simulation to use the first or second option as compared to the third option?

Thanks for your time,

From: Greg Ward<>
To: Tim Perry<>; Radiance general discussion<>
Sent: Monday, December 5, 2011 5:48 PM
Subject: Re: [Radiance-general] Trees

Hi Tim,

You can do it with trans, and it will work with the 3-phase method. Here are some useful posts:

[Radiance-general] trans mat

[Radiance-general] trans dashboard widget

Hope this is enough to get you started!


From: Tim Perry<>

Date: December 5, 2011 5:34:27 PM PST
I would like to model trees outside buildings in my Radiance runs. I would like to model them transmitting about 20% of light and make the canopy 40% reflective.
    * Can I do this by using a translucent material? * If so, can you point me to a tutorial?
    * Will this be compatible with the 3-phase method of daylighting analysis? (I.e., daylight coefficient method with the 3-phase method and a BSDF for the window).

Thanks for any help,

Radiance-general mailing list

Radiance-general mailing list