Modeling dust-induced cirrus cloud formation using CALIOP data and machine learning Algorithms

Cirrus clouds play a key role in the Earth's radiative balance, and their microphysical properties are influenced by aerosol particles, particularly mineral dust. This study models the relationship between dust presence and cirrus cloud formation over Central Asia using CALIOP lidar data and MERRA-۲ reanalysis data from ۲۰۰۶ to ۲۰۲۱. Using the LightGBM machine learning algorithm, we predicted the Cirrus_index (target variable) based on predictor variables including the Dust_index, atmospheric temperature, relative humidity, and pressure profiles. To focus on significant cloud-aerosol interactions, only data where both indices were greater than or equal to ۱ were used. The model predicted cirrus cloud formation patterns in the presence of dust particles with considerable accuracy, achieving a coefficient of determination (R۲) of ۰.۷۴, a Root Mean Square Error (RMSE) of ۴.۴۸, and a Mean Absolute Error (MAE) of ۲.۷۷. These findings indicate that machine learning methods are efficient tools for analyzing complex cloud-aerosol interactions and can contribute to improving climate modeling by explaining a significant portion of cirrus cloud variability when sufficient dust is present.

کلیدواژه ها:

Cirrus cloud ، dust ، machine learning ، CALIOP ، LightGBM