Radiative Transfer Modelling and remote sensing data analysis

Radiative transfer modelling

The purpose of the activity is the implementation and development of algorithms for the atmospheric radiative transfer modelling. We deal both with 'full physics' models (that are used as benchmark solutions and for specific case studies) and with 'fast' models (approximate but computationally efficient solutions suitable for the study of large datasets and for numerical weather and climate simulations). The accounted spectral range spans from microwaves to visible wavelengths, which include the treatment of bi-directional surface reflection and Rayleigh scattering. The inclusion of scattering layers accounts for both aerosols (of various types and mixtures) and clouds (liquid, mixed-phase, and ice). Moreover, the group collaborates with the European Space Agency for the development of the FORUM End-to-End simulator for performance consolidation on mission level, quantification of trade-offs impact on the mission products, and preparation of the user community for the mission exploitation.

Characterization of cloud and aerosol radiative properties

The group manages a large dataset of updated single scattering single particle properties of clouds and aerosols over the infrared and solar spectrum. The dataset is used in studies aimed at characterizing the radiative properties of scattering layers over the full radiative spectrum. We analyse remote sensing data from ground-based, airborne, and satellite platforms, to test the modelled optical properties of particle size distributions of spheres and complex shape crystals. Particle size distribution’s optical properties are parameterized for their usage in fast radiative transfer models within numerical weather prediction and climate models. The group collaborates with the Italian Space Agency at this regard.

Schema adapted from Rodgers, 2000: "Inverse Methods For Atmospheric Sounding"

Inversion Methods

Inversion methods allow to retrieve vertical profiles of the main atmospheric variables, as well as clouds and aerosols properties, from passive sensors at high spectral resolution. Our interest mainly focuses on infrared measurements, from ground-based and/or satellite platforms, from which cloud microphysical (i.e. particle size distribution, effective dimension) and optical (i.e. optical depth) are retrieved. The group collaborated with the IAA and INO of the Florence CNR for this topic.

Machine learning techniques and cloud statistics

This topic is investigated to develop automatic and semi-automatic algorithms (eg. support vector machine, principal components analysis) to be applied for cloud identification and classification from high spectral resolution remote sensing observations, especially in the infrared spectral region. The algorithms are designed both for data analysis and for operational use, as support for satellite missions and field measurement campaigns. The main focus is on the identification of thin ice clouds from infrared passive measurements.

Innovative methodologies are applied to remote sensing data routinely collected in the Antarctic Plateau, with focus on the identification and characterization of mixed phase clouds. This research is supported by the MIUR (Ministry of the Italian University and Research) through several projects.  Advanced statistical techniques are applied to broad datasets and archives, both in-house and third-party (eg. ASI, ESA, NASA).

Instantaneous precipitation estimate from conical (left) and cross-track (right) scanners. Mugnai et al., 2013, Nat. Hazards Earth Syst. Sci., 13, 1959-1981

Multisensor algorithms for precipitation estimate

Machine learning techniques are applied to multi-frequency (from visible to microwave) satellite data to provide precipitation estimation at the ground, with applications to hydrology, agrometeorology and the prevention of hydro-meteorological hazards.

Scientific responsible

Tiziano Maestri

Associate Professor

DIFA Members

Tiziano Maestri

Associate Professor

Michele Martinazzo

PhD Student

Research fellow

Teaching tutor

Federico Porcù

Associate Professor