To understand the nature of different cloud formation processes on Jupiter, we require a good characterization of the wavelength-dependent characteristics of these clouds. While this is generally possible for Earth-based or Hubble data, it is difficult to do with JunoCam, which is currently providing the highest resolution images of Jupiter’s atmosphere. This high resolution allows us to observe small-scale convective clouds, which play a vital role in transporting energy from deep within the planet, and are therefore important to study.

In this work, we use a deep neural network (DNN) to bridge the gap between high-resolution imaging and missing spectral information. Our model is trained to translate JunoCam images into synthetic versions that approximate how the same clouds would appear in Hubble Space Telescope filters. This enables us to estimate physical properties of small clouds, such as their composition and vertical structure, that were previously difficult to assess. We present our results from this model to characterize these small scale convective clouds on Jupiter to understand their properties and interpret their formation mechanisms.