Wishlist.txt

$Revision$, $Date$
$URL$

Here's a list of capabilities we'd like to see added to the I3RC community Monte
Carlo code. If you implement any of these for your own work we'd love to have 
a copy to distribute. 

*) Code to compute the Legendre coefficients for a phase function that's been
supplied as a set of angle-value pairs. This would go into
Code/scatteringPhaseFunctions.f95. We've tried the naive approach (integrating
the product of each Legendre polynomial and the phase function over mu) but this
produces very noisy sets of coefficients.

*) A range of surface BRDF objects that conform to the interface used 
in Code/surfaceProperties.f95. Likely candidates include Cox-Munk for oceans and 
RPV and/or Li-Sparse-Ross-Thick for vegetated surfaces. Note that we may need to 
expand the interface in surfaceProperties.f95 to provide the albedo at a given 
incident angle along with the BRDF. 

*) Polarized radiative transfer. This is ambitious but quite useful to some. It 
would require, roughly: 
-) a polarized extension of the phaseFunction and phaseFunctionTable objects in 
the scatteringPhaseFunctions module
-) changes to the domain object in the opticalProperties module to reflect the 
phase matrices
-) changes to monteCarloRadiativeTransfer: the integrator object needs to tabulate
phase matrices instead of phase functions; individual photons need be be represented
by the full Stokes vector instead of a weight; polarization needs to change 
at each scattering event; polarization needs to be included in the local estimation 
for radiances; polarized results need to be reported
-) the driver programs need to report polarized results

*) A proper build system. Can we streamline the specification of
compilers, compiler flags, libraries, etc. across platforms? Should we
be using autoconf or something similar? Are we propagating configuration
information as gracefully as possible into subdirectories? Should we be
building libraries and linking against those? If so, how do we make the
module files available?

*) A testing system. In my imagination this looks like a script that runs a series of 
test cases, and another that compares results and timings between two sets if test cases.

*) More graceful initialization of the domain when using multiple processors in 
monteCarloDriver. At present, all processors read the domain file, and it would 
be better if only a single processor read that file and then distributed copies of the 
domain to all other processors. This would require adding code to opticalProperties, but 
complicated by the fact that each component of the domain has a unique phaseFuctionTable, 
and passing these around requires adding code to scatteringPhaseFunctions too. Note that 
we'd want to go through mulitpleProcess_*mpi.f95. 

Robert Pincus, University of Colorado/NOAA Earth System Research Lab
Robert.Pincus@colorado.edu