Comprehensive approach to the high-level cloud research: polarization lidar, meteorological data, satellite spectroradiometer, and machine learning

Cloud cover has a crucial impact on Earth's radiation budget by regulating the inflow of solar energy to its surface. High-level clouds (HLCs) cover up to 50% of the Earth's surface [1] and play a major role in weather and climate formation [2]. They reflect solar radiation, cooling the atmosphere, and enhance the greenhouse effect by trapping infrared radiation [3]. Numerical models often neglect the microstructure of HLCs, reducing the accuracy of weather and climate forecasts. These clouds consist mainly of ice crystals whose size, shape, and orientation depend on meteorological conditions. HLC particles are often horizontally oriented. Their flat facets reflect visible radiation specularly, as crystal sizes exceed the light wavelength. This causes anomalous scattering, including backscattering.

Studying the HLC microstructure is challenging: contact instruments disturb crystal orientation, while satellites may miss small, dynamic structures. Polarization laser sensing (PLS) is an effective remote method to determine the size, shape, and orientation of HLC particles. It is based on analyzing the backscattering phase matrix, which carries microstructural information [5], and is implemented in the high-altitude matrix polarization lidar (HAMPL) [6] at Tomsk State University. From 2009 to 2024, specular HLCs were recorded in 22% of HAMPL sessions. They occurred more often from May to September at 5–11 km altitude.

Interpreting PLS data requires information on atmospheric state: temperature, humidity, and wind. Russia’s aerological station data are limited: radiosondes are launched twice daily, and interstation distances often exceed 100 km. Additional data are provided by atmospheric reanalyses. ERA5 (ECMWF) offers vertical meteorological profiles since 1940 with 1-hour and 0.25×0.25° resolution [6]. MERRA-2 (NASA) covers data since 1980 and includes extra parameters like cloud ice mass fraction and cloud cover score, with 1–3 h time steps and 0.5×0.625° resolution [7]. Another key HLC data source is the MODIS satellite spectroradiometer, which provides global cloud monitoring and covers the entire Earth every 1–2 days.

Analysis of HLC parameters was conducted in preparation for developing a machine learning-based software tool. Input data will include atmospheric meteorological parameters to predict the geometric and optical properties of HLCs, as well as assess their dynamics and likelihood of occurrence. Accounting for HLC spatial heterogeneity, including ice particle orientation, in atmospheric models will improve representation of radiative processes and enhance the accuracy of weather and climate forecasts.

The work was performed with the financial support of the Russian Science Foundation, Grant No. 24-72-10127.

1. Ali S., et al. // Atmos. Chem. Phys. 2022. V. 22, № 12. P. 8321–8342.

2. Heymsfield A.J. // Meteor. Monogr. 2017. V. 58. P. 2.1–2.26.

3. Tarasova T.A. // Radiative properties of cirrus. Moscow, Nauka, 1989. P. 169–176.

4. Kaul B.V. Doctoral Dissertation in Mathematics and Physics. Tomsk, IAO SB RAS, 2004.

5. Bryukhanov I.D., et al. // Opt. Atm. Okeana. 2024. V. 37, No. 2. P. 272–279.

6. Copernicus Climate Data Store. URL: https://cds.climate.copernicus.eu.

7. Tao J., et al. // The Cryosphere Discuss. 2019. V. 13. № 8. P. 2087–2110.