AGU 2004 poster

1
ED43B-0939
Introductory Tools for Radiative Transfer
Models D.R. Feldman 1, V. Natraj 2, K. Le 2, and
Y.L. Yung 2 1. Caltech,
Department of Environmental Science and
Engineering, 1200 E. California Blvd. MC 150-21
Pasadena, CA 91125 626-395-6447 2.
Caltech, Division of Geological and Planetary
Sciences
Abstract Satellite data are currently so
voluminous that, despite their unprecedented
quality and potential for scientific application,
only a small fraction is analyzed due to two
factors researchers' computational constraints
and a relatively small number of researchers
actively utilizing the data. Ultimately it is
hoped that the terabytes of unanalyzed data being
archived can receive scientific scrutiny but this
will require a popularization of the methods
associated with the analysis. Since a large
portion of the complexity of the analysis is
associated with the proper implementation of the
radiative transfer model, it is reasonable and
appropriate to make the model as accessible as
possible to general audiences. Unfortunately,
the algorithmic and conceptual details that are
necessary for state-of-the-art analysis also tend
to frustrate the accessibility for those new to
remote sensing. Several efforts have been made
to have web-based radiative transfer
calculations, and these are useful for limited
calculations, but analysis of more than a few
spectra requires the utilization of home- or
server-based computing resources. We present a
system that is designed to allow for easier
access to radiative transfer models with
implementation on a home computing platform in
the hopes that this system can be utilized in and
expanded upon in advanced high school and
introductory college settings. This
learning-by-doing process is aided through the
use of the Community Radiative Transfer (CRT)
wiki which may be able to facilitate greater
interest in the field of remote sensing.
5
6
EXAMPLES FROM MODTRAN 4
PLATFORM CHOICES

• The presence of a wiki to instruct on how to run
  radiative transfer models will not be very useful
  for general audiences unless these audiences can
  run the radiative transfer models on standard
  platforms. The majority of radiative transfer
  codes that were surveyed for this project
  (SOURCE) have been designed to be compiled with
  commercial Fortran compilers and run on a Linux
  platform. However, since XX of personal
  computers used by college students majoring in
  science and engineering are either running a
  Macintosh or Windows, it is reasonable to orient
  the CRT wiki so that it focuses on distributing
• Stand-alone executable models built for Mac
  and Windows
• Code that runs by clicking in lieu of the
  command-prompt
• Code that easily facilitates batch runs.
• Some RT codes have these features. However, some
  RT
• There are clear difficulties associated with this
  approach because compiling Fortran code is
  entirely a non-trivial task, especially with
  freely-distributed compilers

One radiative transfer code that has been
utilized extensively over several decades is the
MODerate Resolution TRANsmittance code, MODTRANTM
xx. Since MODTRANs initial release, over 120
publications have utilized this code due to its
speed and comprehensive set of options. Several
wrapper packages have been released in order to
manage the arcane format for the input file that
MODTRAN requires. For the CRT project, a
freely-available wrapper has been released with
examples similar to those associated with the
routines associated with the AER RT suite of
codes (see Section (4)).
1
3
MOTIVATION
RADIATIVE TRANSFER MODEL DETAILS

• The measured quantity from a remote observation
  is a function of the specifics of the medium
  through which the light is propagating, the
  orientation of the viewer (viewing geometry), and
  the instrumental response function. Due to the
  diverse nature of these observations and the vast
  amount of complexity associated with achieving
  accurate results, a generalized monolithic
  radiative transfer routine is difficult to
  achieve and impractical to use. Consequently,
  several different classes of RT codes have been
  written and are generally organized according to
  the spectral resolution and spectral region of
  interest though actual solution methods vary
  dramatically.
• Nevertheless, a rubric can be developed so that
  the input specification file and execution of the
  code standardized where it can be and flexible
  where it needs to be. For atmospheric remote
  sensing problems, the standardized inputs may
  take the form of
• A matrix specifying the profiles of
  temperature, pressure, and radiatively active
  gases.
• The observer locations and viewing angle.
• Gray or non-gray surface properties.
• The solar function.
• The instrumental response function.
• Range of wavelengths under consideration.
• For treatment of scattering or where any of the
  above inputs require more specification, an
  inhomogeneous data structure may be a useful tool
  for allowing variable argument input length.
• Techniques used to solve the RT equation include
  discrete ordinates (where the phase function is
  expanded in Legendre polynomials and the angular
  integral is replaced by a quadrature sum),
  doubling-adding (where reflection and
  transmission functions are computed using
  repeated reflections between layers) and
  successive orders of scattering.

Science derived from remote sensing measurements
aboard satellite platforms is currently
data-rich. Each satellite instrument produces
several megabytes per second of calibrated
measurement data that have the potential to yield
useful scientific results, especially in light of
the novelty of many of these instruments.
However, the ratio of data to researchers is
quite large, and with the advent of new
instruments, this ratio only stands to grow
larger. For NASA, much effort has been devoted
to the dissemination of the data through online
data gateways (SOURCE), but the technical nature
of the data being disseminated presents a barrier
to greater community participation. In
particular, understanding the nature of the
measurements being made, and thus having a grasp
of radiative transfer, may prevent those who are
new to or curious about the field of remote
sensing from being able to conduct science on
their own personal computers. At its heart,
radiative transfer is a deceptively simple
endeavor which seeks to solve for radiance as a
result of its propagation through a medium as
seen in the Fundamental Equation of Radiative
Transfer where I is the radiance, J is the
emission/scattering term, k is the absorption
coefficient, and s is the pathlength. The
multi-dimensional nature of and the computational
requirement for solutions in all but the most
elementary instances are challenges for the
classroom setting. At the collegiate level, it
is hoped that courses that cover radiative
transfer concepts rigorously will also be able to
educate students with the actual tools needed to
perform remote sensing science.
Figure 4 Similar to Figure (2) but using the
MODTRAN 4 code.
7
EXAMPLES FROM THE LIDORT SUITE
One of the most popular RT codes is DISORT, which
is a general and flexible package that could be
used in a wide variety of atmospheric
applications. LIDORT incorporates the good
features of DISORT in addition, it is designed
to give more accurate results for a much larger
range of solar and viewing geometries by taking
into account the curvature of the atmosphere.
Further, weighting functions are generated
simultaneously with the radiances by using
analytic differentiation of the RT equation.
LIDORT has been used for ozone profile retrievals
from OMI on EOS-AURA and GOME-2 on the METOP
series, among others.


8
VISUALIZATION
The visual representation of the results of RT
codes is crucial even for the most elementary
analysis. Several visualization packages exist
and are commonly-utilized in the remote sensing
community including Matlab (and its open-source
counterpart Octave) and IDL . These support
advanced plotting capabilities with
straightforward syntax.
2
4
WIKI DESCRIPTION
EXAMPLES FROM AER CODES

• The aim of the Community Radiative Transfer (CRT)
  project is to facilitate the appropriate
  execution of radiative transfer models so that
  novel science can be performed using the vast
  amounts of unexplored remote sensing data. One
  easy tool that can be used to achieve such a task
  is a wiki. A wiki is a web-based set of pages
  that can be edited by most if not all users with
  the intent to introduce participants to intricate
  subject matter and resolve technical and
  conceptual problems therein. Many wikis exist
  for anything from general encyclopedias to
  technical how-to guides. The CRT wiki is
  technical and meant for an audience focused on
  understanding radiative transfer and how it
  relates to remote sensing data analysis.
• The wiki is a realization of the DokuWiki package
  (SOURCE) and is organized into 4 main sections
  (referred to internally as namespaces)
• Administrative
• Conceptual
• Forum
• Technical
• Many concepts are shared with Wikipedia, though
  access restriction is employed so that while
  pages are viewable by all, only registered users
  can edit pages. The registration process is very
  simple and can be done in 1 minute. The wiki is
  located at
• http//www.gps.caltech.edu/drf/wiki

Several different radiative transfer codes have
been developed by AER, Inc. for modelling
infrared spectra, and broad- and narrow-band
fluxes, and heating rates (SOURCE). The CRT wiki
introduces some of these codes focusing on the
line-by-line code LBLRTM and the broadband flux
and heating-rate code RRTM. LBLRTM has been the
subject of at least 20 publications while RRTM
has been the subject of at least 21 publications.
However, the inputs to the codes remain somewhat
arcane, especially for the former code. As part
of this project, a series of wrapper routines is
introduced that adhere to the input details
discussed in Section (3).
Figure 5 Derivation of the AMFs for sample GOME
viewing scenes over western Pennsylvania and the
central North Pacific. The panels show scattering
weights for the clear-sky and cloudy fractions of
the scenes as determined with the LIDORT
radiative transfer model. Adapted from Fig. 2,
Martin et al. (2002).
9
Conclusions

• There is much remote sensing data that is left
  unscrutinized and a community-oriented,
  accessible approach to radiative transfer may be
  able to address the data deluge.
• The Community Radiative Transfer (CRT) wiki has
  been developed to facilitate knowledge regarding
  the implementation of RT codes. The ultimate
  goal for the project is to port RT codes to the
  most commonly-used operating systems for personal
  computers.
• Preliminary efforts have been made regarding
  commonly-used RT codes.
• While it is desirable for RT codes to output text
  files of their results, visualization is a
  central part of analysis using radiative
  transfer. Several plotting packages exist and

Figure 3 Broadband heating rates calculated
from the longwave RRTM code for a clear-sky
tropical model atmosphere. The color of the line
denotes the heating rate of a particular spectral
band as indicated in the legend on the right.
Figure 2 Comparison between measured and
modeled spectra for the CAMEX mission (SOURCE)
using the LBLRTM code. The top panel shows an
overlay of the two spectra while the bottom panel
details the difference.
10
References

• PLACEHOLDER

Figure 1 Screenshot of the CRT wiki index page.
Page layout and searching is intended for ease
of use.
Special thanks would like to be extended for the
technical support from Wing-Ning Yung, the RT
team at AER, Inc, and Lex Berk of Spectral
Sciences, Inc.
Research supported by the NASA Earth Systems
Science Fellowship, (06-ESSF_06R-87)