In this study, we investigate the role of cloud radiative effects and latent heating in shaping the variability of the subtropical jet and polar front jet. To do this, we employ 3D (latitude, longitude, height) cloud radiative heating rate and latent heating rate data from the CloudSat 2B-FLXHR-lidar satellite product and the Global Precipitation Mission (GPM) Gridded Convective Stratiform (3GCSH) satellite product, along with the MERRA-2 reanalysis dataset. These datasets include vertically resolved daily air temperature tendencies resulting from longwave and shortwave radiation under clear-sky and all-sky conditions, as well as air temperature tendency due to moist processes.
We first validate the MERRA-2 products against the CloudSat/GPM products to ensure their accuracy in reproducing CRE and latent heating patterns compared to satellite measurements. Subsequently, we explore building machine learning models to predict the position and strength of the subtropical jet and polar front jet in both hemispheres. For this purpose, we utilize profiles of CRE and latent heating, as well as common indices of large-scale climate modes (such as MJO and ENSO), and memory from the jet streams with different time lags as input features to the models. Finally, we utilize explainable artificial intelligence methods and Granger Causality tests to interpret the model results and identify physically meaningful relationships between CRE and latent heating and the jet streams.

