# Algorithm development - Task 3

## Raman scattering - WP3100

The possibility to include Raman scattering has been added to the uvspec model. The traditional radiative transfer equation solvers within uvspec solves the monochromatic radiation transport. Raman scattering redistributes part of the scattered radiation to other wavelengths. The solution of the radiative transport equation must be changed accordingly to allow for the wavelength redistribution of photons. This has been achieved by treating Raman scattering as a perturbation. An iterative solution to the radiative transfer equation for the wavelength dependence has been adopted. This solution method includes the addition of an extra source term which requires a radiative transfer equation solver that can handle general source terms. A solver described by Kylling,1992, has been adopted to the task. This solver has been used for various transport problems, see for example Stamnes, 1991b. Examples of Raman scattering calculations are provided in the verification report libradtran_verification_report.pdf.

*WP status: closed*

## Polarization in MYSTIC - WP3200

#### Extend MYSTIC - WP3210

Polarization has been integrated into the MYSTIC Monte Carlo solver. The new algorithm combines methods described in various papers and books (Marchuk et al., 1980; Collins et al., 1972; Mayer, 1999, 2000; Marshak and Davis, 2005; Cahalan et al., 2005). It is a flexible code that can be run in forward and backward mode, in 3D, 1D plane-parallel and in spherical model geometry, and with arbitrary model atmospheres that may include molecules, aerosols and clouds. The algorithm includes the so called local estimate method for efficiency. Some results are presented here and in pm3_polarization.pdf and pm4_polarization.pdf.

Basic documentation is included in the libRadtran user guide, more detailed documentation will be included later. Validation is in progress. A model comparison with the polradtran solver shows a very good agreement for Rayleigh scattering and aerosol scattering for aerosols with moderate forward scattering peaks. For aerosols with enhanced forward scattering MYSTIC is superior to polradtran since polradtran is not able to handle such scattering phase functions.

*WP status: closed*

#### Include tools to calculate single scattering properties of aspherical particles - WP3220

To perform polarized radiative transfer simulations, single scattering properties of cloud particles are required. In scalar calculations the scattering process is described by the phase functions, the extinction cross sections and the absorption cross sections. For polarized calculations the single scattering properties are much more complex: The 4×4 phase matrices and the 4 component absorption and extinction vectors are required (for each possible particle size, shape and orientation).

The well-tested Mie-code by Wiscombe, 1980a,
which is already incuded in the
*libRadtran* package (Mie-tool), can be used to generate the data for spherical
particles, i.e. water droplets or some aerosol types.
*libRadtran* does not include a tool to calculate scattering properties
of aspherical particles. The first step in this work package is to
test freely available codes for computing scattering properties of
aspherical particles which can be randomly oriented or horizontally
aligned (which is often observed in cirrus clouds). Based on these
tests the best code will be implemented as a tool in the *libRadtran*
/ESAS-Light toolbox.
Parameterizations of scattering properties are not available for
polarized radiative transfer, so it is absolutely necessary to have
such a tool.

So far, two methods have been tested. The first is DDSCAT, a tool based on the discrete dipol approximation. It is appropriate for the thermal wavelength region but it does not converge for large size parameters (clouds in the UV/Vis range). The second is the T-matrix method. The tools be Mishchenko have been tested. The T-matrix method is applicable for rotationally symmetric particles like cylinders and spheroids. The method also does not converge for larger size parameters. In order to simulate ice particle scattering a geometrical optics tool is required, such a tool has not yet been tested.

A comprehensive collection of tools to compute optical properties of particles is given at http://www.t-matrix.de/. Links to download the tools and their documentation are provided. At the moment it is probably more appropriate to let libRadtran users choose their own tool from this webpage, rather than providing one with the libRadtran package, since we can not judge which of the tools (more than 100!) is the best.

It was decided that a database for aspherical particles should be inluded into the libRadtran package rather than a tool. However, this database can not be generated withing the ESASLight study. In order to test whether MYSTIC is capable of handling aspherical particles an example file has been created based on Kuik et al, 1992, who provide phase matrices for aspherical aerosols. This file can be used as an example for users who have other phase matrix data.

*WP status: closed*

#### Generate Database for Single Scattering Properties - WP3230

The objective of this task is to generate a database of single scattering properties of cloud particles to be used for polarized calculations. The first version of the database to be generated within the project will include only scattering properties of spherical particles. This can be generated using the already available Mie-tool. The database is easily extendable for aspherical particles provided that a good tool to calculate scattering properties of aspherical particles is available.

The new database may be used as input to MYSTIC (polarized) and to the fast vector code polradtran by Evans, 1991.

The format of the optical properties data has been revised in order to be able to represent highly peaked phase functions. This is sometimes not possible in terms of Legendre polynomials, which was used by the old format. The data is stored in a convenient netcdf format. New databases (including polarization) have been generated for liquid clouds and OPAC aerosols, and (without polarization) for ice clouds.

*WP status: closed*

## Extension of surface properties - WP3300

The BRDF is needed for the correction of view and illumination angle effects of the surface and is therefore an essential input parameter for simulating radiances.The most important effects of the BRDF are the sunglint of water bodies and the hot spot of surfaces with vegetation.

In this work package, the BRDF options of *libRadtran* have been
expanded so that
the operational MODIS BRDF/albedos data sets for land surface can be applied.
An example is shown here.
It was agreed that for water surfaces the BRDF model by Cox and Munk is appropriate and that no other model needs to be included.

As additional task (not in the proposal) a BRDF for water surfaces that includes polarization should be added. Mishchenko offers a code on his web page, which calculates the BRDF and reflectance matrices for polarized radiative transfer for rough surfaces and arbitrary reflective indices. The code was tested, it is written in Fortran 77 and has been implemented in libRadtran. Hasekamp et al. (2007) have derived functions for the BRDF of land and water surfaces which probably could be included in libRadtran.

*WP status: closed*

## Representation of aerosols -- WP3400

*libRadtran* includes the various aerosol standard
models as defined by Shettle, 1989. The standard models may be
modified by scaling aerosol visibility or the aerosol optical thickness,
or by modifying the asymmetry parameter, the single scattering albedo,
or the Angström coefficient.

In reality the mixture of different aerosol types can be
highly variable. To do a simulation of an aerosol mixture different to
the Shettle models, the user can generate an input file containing the
optical properties of this mixture.
The objective of this work package was to extend *libRadtran* in such a
way, that a more flexible treatment of aerosols is possible. The user
is now to define an aerosol mixture in the *libRadtran* input file
and the is then be done internally by the *libRadtran* program.
A database of scattering properties of aerosol basic types has been
calculated using the Mie tool. As input for the Mie tool, the
refractive index of the aerosol and its size distributions are
required. This data has been from the OPAC database
Hess, 1998, which includes the following
basic types: water soluble aerosol, soot, sea salt (accumulated and
coarse modes), mineral (nucleated, accumulated and coarse modes),
transported mineral, sulfate droplets.
*libRadtran* now allows the user to define arbitrary mixtures composed of these
basic aerosol types. The newly generated aerosol database includes phase
matrices, so that the depolarization by aerosols can be calculated
using one of the polarized RT solvers (MYSTIC or polradtran).

Ten standard aerosol mixtures (corresponding to the ones suggested in Hess et al. 1998) have been generated. These include for instance “continental average”, “maritime clean”, etc.

Some results using OPAC aerosols in combinations with MYSTIC and polradtran can be found in the presentations pm2_aerosolopac.pdf and pm3_polarization.pdf.

*WP status: closed*

## Further extensions - WP3500

#### Development of graphical tools - WP3510

A fully operational graphical user interface (GUI) for uvspec was demonstrated during the MTR-meeting. Since then more options have been added to the GUI. Furthermore, the plotting capabilities of the GUI has been improved.

It has been tested that the programming framework the GUI is developed under, allows the GUI to be run under UNIX/Linux, Windows and Mac environments.

*WP status: ongoing*

#### Include I/O-functions for netcdf data format - WP3520

Detailed input data for radiative transfer simulations often
demands the use of global data sets. In order to simplify data
handling it is useful to have the possibility
to read input data from compact binary file formats and also to write
simulation results directly in such a format.
In this work package the program interface has been expanded, so that
various input parameters can be read from and the output stream can be
directed into binary files of the widely-used netCDF format. As
several open source programs for visualisation of netCDF data such as
*ncview* and *ferret* are available, this improvement
gives the possibility to quickly get overview of simulation results.

*WP status: closed*

#### Refraction - WP3530

Atmospheric refraction is very important for limb sounding instruments where the horizontal path through the atmosphere becomes large. For this reason refraction is implemented in the MYSTIC model, which can be operated in fully spherical geometry. Using Snells' law of refraction at each grid box boundary in the Monte Carlo model is calculated. This allows to compute direct radiances and transmission (without scattering). In order to implement refraction in a scattering atmosphere, the local estimate method needs to be modified which is not trivial, because at each scattering point one needs to know the direction of the photons that leave the atmosphere in the direction of the sun. This problem is currently being solved using an iterative method.

An example showing the transmission through the atmosphere along a limb path is shown here.

*WP status: ongoing*

#### DISORT improvements - WP3540

In the course of ESAS-Light we have identified a problem with DISORT: it requires all Legendre polynomials of the phase function for the second order intensity correction. For ice clouds this can be more than 20000(!) resulting in too large memory requirements. The other 1D solvers can only use the first nstr (usually 16 or 32) polynomials, and MYSTIC can use the phase function directly. DISORT is currently being modified so that is also uses the phase function directly, not the Legendre polynomials, for the 2nd order intensity correction. This change is very important since it saves a lot of memory and it allows to use the same input files of optical properties for MYSTIC and for DISORT.

Originally, we proposed to fix an instability of DISORT that has been identified several years ago. Since the instability occurs only in rare cases we decided to concentrate on the change in the 2nd order intensity correction before fixing the instability.

*WP status: closed*

#### Cleaning/restructuring of code - WP3550

*libRadtran* has been rapidly growing over the years. One
important task will
be to systematically ``overhaul'' the code to make sure that
everything is consistent,
to eliminate quick and dirty hacks, to introduce meaningful variable names, to
apply reusable functions as far as possible, etc.

The following improvements have been implemented:

- consistent treatment of ice cloud optical properties
- put reusable functions in sub-routines
- documentation of several functions (still ongoing)
- various small changes

*WP status: ongoing*

#### Miscellaneous - WP 3560

As outcome of task 1 (consolidation of ESAS-Light demonstration version requirements), we have defined a few more work packages:

##### Line-by-line interface

A python interface has been developed, that allows to compute absorption coefficients (line-by line based on HITRAN) as input for *libRadtran* radiative transfer simulations. It has been validated for the wavelength region from 10–12 micrometers.
More testing and validation is required to make sure that ARTS works correctly in the visible and NIR wavelength regions.

Some details are presented in pm2_lbl.pdf.

As a second tool for line-by-line calculations an interface for the MIRART code developed by Franz Schreier from DLR (IMF) has been provided.

*WP status: closed*

##### libRadtran webpage

A new libradtran webpage has been launched (www.libradtran.org). The layout has changed and there are some additions:

- A user area (platform where libRadtran users can contribute interesting applications, problems …)
- Basic usage instructions

*WP status: closed*

## Intermediate version of libRadtran - Task 3

A summary of new features can be found here.

## Complete libRadtran version - Task 3b

During this task the libRadtran version including all new modules developed within the ESASLight study will be completed. This version will include also detailed documentation. The validation of the new algorithms will be complete before this version will be released.

The final acceptance of all software shall be done at ESTEC, for Linux, Mac OSX and Windows environments.