This set of experiments is the same as for RAMI phase 1 and 2 and requires to simulate the radiative transfer regime in the red and near infra-red spectral bands for heterogeneous environments.
These scenes - which are structurally identical to those used for the discrete floating spheres test cases in the purist corner - are composed of a large number of non overlapping spherical objects representing the individual plant crowns, located over and only partially covering a horizontal plane standing for the underlying soil surface.
These objects, in turn, contained randomly distributed finite size disc-shaped scatterers characterized by the specified radiative properties (reflectance, transmittance), and the orientation of the normals to the scatterers followed exclusively a uniform distribution function. The radiative properties of the underlying soil were also specified (in this case a simple Lambertian scattering law). The particular values selected for these input variables represented classical plant canopy conditions. To address the needs of different RT models, we are providing both a statistical scene description, as well as, a file with the exact coordinates of every leaf in one of the spherical volumes. You may or may not make use of this information depending on the needs of your particular model.
The architecture of this scene (HET01) is precisely characterized as follows:
where the Leaf Area Index (LAI) is calculated as follows:
LAI = (# of leaves × one-sided area of single leaf) ⁄ (π × square of the radius of sphere)
An ASCII file with the radius (R), centre coordinates (Xc,Yc,Zc), and direction cosines (Dx,Dy,Dz) of every single leaf in a spherical volumes centered at 0,0,0 can be found here. This file (is ~ 2.2 Mbytes and) contains 49999 lines of format R Xc Yc Zc Dx Dy Dz and may serve as input to your scene creation process (provided that you are able to create multiple instances of its content, each one of which is then translated to the actual locations of the sphere centers in the scene. The coordinates (X, Y, Z) of the various sphere centers are as follows:
The underlying soil is occupying the entire lower surface of the scene, i.e., the area enclosed between the following coordinates: (−50.0, −50.0, 0.0), (−50.0, +50.0, 0.0), (+50.0, −50.0, 0.0), and (+50.0, +50.0, 0.0), where all values are in meters.
The tables below provide the details required to execute each of the experiments in this category. Every table is preceeded by the corresponding experiment identifier tag that is needed in the naming of the various measurement results files (see file naming and formatting conventions).
The absence of any absolute "truth" renders the evaluation of RT models more tricky but, nevertheless, the model discernability issue can be addressed through the usage of "credible solutions". As a matter of fact, it can reasonably be admitted that the latter correspond to the actual values that could be measured from an instrument with its intrinsic limited accuracy. We attempted to examine the issue of model discernability by establishing, for all scenarios concerned, what could be considered as a "credible solution" by estimating the arithmetic mean of every BRF value calculated within a subset of the participating three-dimensional Monte Carlo model results.
Overall six different 3D MC models were allowed to contribute to the establishing of the ``credible'' solution. These are: DART, drat, FLIGHT, Rayspread, raytran, and sprint3. The following rules were applied when choosing the exact name and number of 3D MC models that then contributed to the establishing of the ``credible'' solution (note that both the name and number of 3D MC models may change between exeriments and also between individual measurement types, as outline in the table here):
Once a suitable set of ``credible solutions'' are available the model discernability can then be analyzed by comparing the BRF values generated from individual models with those of the credible solution using a normalised Chi-square metric:
where the uncertainty estimator in the denominator (sigma) is simply expressed as a fraction of the credible BRF (at any given view angle).
The figure below displays the Chi-square metrics obtained in the red (x-axis) against those for the near-infrared (Y axis) using a value of 0.03 for the uncertainty f (equivalent to sensors having a 3% calibration accuracy):