Creating a BRDF file and using it as a bottom boundary
It is possible to use PlanarRad to generate a BRDF file that can then be used as input as the bottom boundary condition for another run. One reason to do this is to implement a layered media. Currently PlanarRad only supports vertically homogenous media, but a two layer system (for example) can be modelled by generating BRDF for the lower layer and then using that as the bottom boundary for the upper layer.
The key points are this:
- Generating the BRDF file requires 10 runs, one for each incident theta position of 0, 10, 20, 30, 40, 50, 60, 70, 80 and 87.5 degrees.
- For layers where the refractive index does not change substantially (most likely for layers where the embedding media is the same but constituents change) the runs for generating the BRDF file should use the 'non-existent' interface option 'None'.
- The separate tool 'reflectool' must be run from the command line to generate the BRDF file from the ten runs.
- Then to model the upper layer is a standard run with normal interface settings and uses the the BRDF file for the lower boundary.
- A simple test procedure to verify everything works OK is to do a model with two layers that have same optical properties and compare that to a single run for the depth of the two layers added together, the light field in the top half of the media should be identical.
This page gives a worked example for a model of two layers and demonstrates the verification that the output is correct.
Characterising the BRDF
To make a BRDF file requires that we have data on the full hemisphere of light reflected in every exitant direction for every incident direction, from a surface. In PlanarRad the standard directional discretisation for a hemisphere contains 217 quads, so the BRDF for one wavelength is a table of 217 x 217 numbers. To populate this table we can run PlanarRad with all the incident light in just one quad, and that will give us the 217 values, one for each exitant direction, for that particular incident direction. So on the face of it 217 PlanarRad runs would be required to characterise the whole BRDF. However, in almost all cases we can assume the BRDF is rotationally invariant.
The meaning of rotational invariance is this: Imagine you are looking at a surface from an angle and the sun is off to the right somewhere. You and the sun stay in the same place but the surface is rotated in a horizontal plane - it still looks the same. In PlanarRad currently the only way non-rotationally invariant results can be obtained is when using windblown air-water interface. Any runs using a flat or non-existent interface are rotationally invariant. The interface option of 'none' is what is normally required for a BRDF of a lower layer in a media. This means only 10 runs are required to characterise the BRDF, one for each incident theta row (zenith position) in the hemisphere. We can set the incident azimuth to zero for all runs and that will be enough information to characterise the BRDF.
Example model
For the example here we will generate a single wavelength BRDF for a 5m depth, attenuation c = 0.3, absorption a = 0.1, Petzold phase function, and bottom boundary black (reflectance = 0). Then we will input that as the bottom boundary for another run of 5m depth with the same optical properties. The result should be the same as a 10m run with black bottom boundary. We will verify that it is and then modify the upper layer to get a layered media result.
Runs to characterise the BRDF
The basic procedure is the same as described on this page: BRDF of a material with isotropic phase function, but with a few important changes. Assuming you are able to successfully follow the procedure on that page, do this:
1) Set the interface properties to 'None'. For script running the parameter is 'iface_type = absent'.
