A reference quality model for surface emissivity and backscattering from the microwave to the infrared

Agenda

18 October

1. Welcome (Steve English and Catherine Prigent)

2. The PARMIO model

2.1. Overview of the status of the PARMIO model (Emmanuel Dinnat)

The status of the PARMIO model has been presented by Emmanuel Dinnat, with an overview of the model features and modular structure. The available options for the different model modules were listed. Improvements to the model performed during the existence of the ISSI group include an update of dielectric properties in the infrared (Stuard Newman), and a frequency tuned foam emissivity (Maggie Anguelova). The code and repository status were presented.

2.2 Comparison of PARMIO with passive microwave observations and modification resulting from these comparisons (Lise Kilic)

Lise Kilic then presented the adaptation of the PARMIO model, for the development of the fast passive microwave emissivity code. Systematic comparisons have been conducted between PARMIO simulations and satellite observations (SMAP, AMSR2, GMI, and ATMS). Differences were noted mostly over cold ocean and under high wind speeds. Parameters are selected to limit differences with satellite observations. An adjustment is proposed to the foam coverage and emissivity.

2.3 Validation of PARMIO with WindSat Brightness Temperatures (Thomas Meissner)

Validation of PARMIO with WindSat Tb is suggested by Thomas Meissner, in the framework of the preparation of the Copernicus Imaging Microwave Radiometer (CIMR). In addition to have an absolute calibration, Windsat is a very good proxy to CIMR, covering frequencies from 6 to 37 GHz, and full polarimetry from 10 to 37 GHz. With its local ascending node at 18:00, collocation with SMAP observations at 1.4 GHz is facilitated.  Other US instruments suitable for CIMR preparation in the future were also presented (WSF-MWI and COWVR).

2.4 PARMIO comparisons with other modeling activities (Fuzhong Weng)

Fuzhong Weng presented modeling activities and comparison with PARMIO in China.  Within the Advanced Radiative Transfer Modeling System (AMRS), an ocean two scale roughness model has been developed that provides the Polarized Bidirectional Reflectivity Distribution Function (pBRDF), suitable for vector radiative transfer model. Surface Stokes emissivity vector, monostatic and bistatic normalized radar cross section and bistatic can be calculated from the BRDFs. Surface emissivity and backscattering results are compared with PARMIO, with reasonable agreements, except for some phase opposition with wind direction. The code will be made available to the community soon.

2.5 Results related to active microwaves with PARMIO (Lise Kilic)

To initiate the discussion on the PARMIO application to active microwaves, Lise Kilic showed some preliminary comparisons between backscattering GMFs and PARMIO simulations, at L, C, and Ku bands. Different wave spectra were tested. Significant discrepancies are observed between the simulations and the GMFs, and no consistent explanation is provided so far.

2.6 Scatterometer developments (Ad Stoffelen)

Ad Stoffelen presented recent improvements in wind retrievals with scatterometers, insisting on the accuracy of the GMF functions and on the stability of the ASCAT instruments. Compared to the high quality estimation of the wind vectors with scatterometers, ECMWF winds appear to be biased, and a correction scheme is being developed.

2.7 Results related to the IR with PARMIO (Stu Newman)

Extension of PARMIO to the infrared has been explored by Stuart Newman. An infrared dielectric model has been implemented and tested in PARMIO.  The two-scale model simulations converge to the geometric optics limit at high frequency, as expected.  Preliminary results with the foam model seem to indicate higher values than literature results. Comparisons with the current RTTOV IR emissivity model (IREMIS) show close agreement with PARMIO when neglecting the multiple correction terms in IREMIS.

3. The SURFEM-OCEAN model

3.1 Development of the SURFEM-OCEAN fast model based on PARMIO (Lise Kilic and Carlos Jimenez)

The SURface Fast Emissivity Model for ocean (SURFEM-Ocean) based on PARMIO is presented by Lise Kilic and Carlos Jimenez, along with its evaluation. SURFEM-Ocean is based on artificial Neural Network (NN) to reproduce much faster the results obtained with PARMIO. It extends the FASTEM frequency coverage from 0.5 to 700 GHz. It includes the jacobians, the tangent linear model, and its adjoint for an efficient use in NWP applications.

3.2 Initial evaluation of the fast model (Lise Kilic)

Simulations with SURFEM-Ocean have been compared to multiple satellite data, with satisfactory results.

3.3 Update on the RTTOV SURFEM-OCEAN implementation (Emma Turner and James Hocking)

SURFEM-Ocean has been incorporated in RTTOV by Emma Turner and James Hocking, and Emma Turner presented the initial tests, as compared to FASTEM. It will be distributed in the next version of RTTOV (13.2) within the next weeks.

19 October

4. What is next?

4.1 How to maintain PARMIO? (Emmanuel Dinnat + all)

The PARMIO code is already available to the team through a repository (GitHub). Other repository possibilities have been suggested (at JCSDA, at the UK Met Office), but the current solution is judged satisfactory. The code is written in fortran. It does not require large ancillary files (only a few small ascii files and the source code). Before being made publicly available, the code licensing is to be checked, and possibly modified (action Emmanuel Dinnat). The documentation has already been improved over the last months, and will be updated whenever necessary.

PARMIO is a research code with many options, and parameter selection is necessary. Default parameter sets and test cases have to be prepared, along with the corresponding PARMIO results, to facilitate the accessibility to new users. One parameter set will correspond to the selection made for the SURFEM-Ocean fast model (action Lise Kilic), for passive microwave applications. Another parameter set will correspond to infrared applications (action Stuart Newman).

Only identified direct contributors to the code will be allowed to make changes and create a new branch to the code. Emmanuel Dinnat will moderate and validate these changes before implementation in the core code.  At this stage, requests for user support are expected to be rather limited, and Emmanuel Dinnat will handle them. If the situation evolves, alternative solutions will be envisaged (for instance, help of other institutions such as JCSDA or NWP SAF).

4.2 Science evolution of PARMIO

Differences between PARMIO and other models in the infrared were attributed to the fact that multiple reflectionswere not taken into account in PARMIO. The possibility to have multiple reflections in PARMIO will be tested (action Emmanuel Dinnat).

Thomas Meissner suggested to further test the sensitivity of the microwave emissivity modeling to the wind direction, especially for high wind speeds. This can be done with Windsat that provides full polarimetry. Fuzhong Weng also mentions possible problem with PARMIO with the wind direction, with phase opposition with their model.

Calculating full BRDF is of high interest in vector radiative transfer model, especially under scattering atmospheres. Discussion with Alan Geer and Patrick Erickson is encouraged to advance in that direction, in coordination with the atmospheric radiative transfer model. Fuzhong Weng and his group developed this capability, but they mention that computation time is high. Their experience will certainly be very valuable on that subject.

Significant progress has been done by the team, for passive microwaves and infrared applications. However, simulations of the active microwave responses with PARMIO do not reproduce well the observations. More efforts have to be conducted on that subject, in close relationship with the active microwave community. Recent developments suggest that improvement is possible, through a better description of the surface waves (see for instance, Plant & Irisov, A joint active/passive physical model of sea surface microwave signatures. JGR Oceans, 122(4), 3219-3239, 2017). Having a unified forward model for both passive and active microwaves would enable a consistent assimilation of both observations.

Possible improvements in the foam emissivity modeling have been mentioned by Maggie Angelova. In the microwave, adding the scattering effect above 90 GHz is to be tested.  In the infrared, the foam emissivity has to be tuned: changing the void fraction might help getting closer to the lab measurements.

 

Following the SURFEM-Ocean initiative for the passive microwaves, there is an interest to develop a fast emissivity model in the infrared (Stuart Newmand and James Hockings). Once the PARMIO model includes multiple reflections, a fast model could be computed and implemented in RTTOV, to replace IREMIS. That would provide optimized consistency from microwave to infrared ocean emissivity for assimilation.

Note that an ISSI team has been organized for improvement of the radiative transfer in the visible and near-infrared for NWP application (led by Ben Johnson and Antonella Benedetti). PARMIO is seen as the candidate for the ocean surface radiative transfer model in the visible.

The team recognizes that the PARMIO model uncertainties have not been systematically evaluated. This is expected to be a difficult task that would involve uncertainty propagation trough the different components of the model. Another more pragmatic approach consists in comparing the model simulations with observations, and this has been partly done for passive microwaves, with satellite observations.

5. Discussion on the ISSI experience

The ISSI experience was unanimously judged very productive, with two physical meetings in Bern (at least for some of us), and four intermediate on-line meetings. The initial objective of the working group is met, with a community code ready to be available to the community. The ISSI team facilitated open and efficient discussions on the different aspects of the model, to reach a consensus on the best way to progress.

During this period of pandemic with restricted exchanges, this working group maintained a good level of motivation to make progress at a good pace.

Given the success of this experience, the possibly to set up a new working group on radiative transfer modeling of another surface type (snow / ice) was discussed

Key outcomes of this ISSI Team:

Models / Codes

• A community model PARMIO (Passive and Active Reference Microwave to Infrared Ocean), to be made publicly available soon.
• A fast version of this model for passive microwaves, SURFEM-Ocean, for distribution with RTTOV 13.2 in November.
• Plans for the developments of a fast IR version of PARMIO

Publications, conferences

• A meeting paper in BAMS (English et al., Reference-quality emission and backscatter modeling for the ocean, doi: 10.175/BAMS-D-20-0085.1, 2020).
• A paper to be submitted to JGR Ocean within the next weeks (Kilic et al., Development of the SURface Fast Emissivity Model for Ocean (SURFEM-Ocean) based on the PARMIO radiative transfer model).
• Plans for a second meeting paper in BAMS (Emmanuel Dinna as a lead)
• Plans for a general PARMIO paper, illustrating its applicability to passive microwaves, active microwaves, and infrared (Emmanuel Dinnat as a lead)
• Plans for presentations to conferences (ITSC 2023, IGARSS 2023).