Fig. 2
From: 3D volumetric tomography of clouds using machine learning for climate analysis
Simulated results. (A) Visualizations of 3D volumetric fields by MIP at side-view: A labeled test cloud, its estimation error, and uncertainty (normalized entropy). They increase at the cloud core. (B) Sample inferred probability distributions normalized by the MAP value. (C) Inferred results at 2000 voxels, which were randomly sampled across the test set. A high inferred normalized entropy (uncertainty) implies a possible large absolute error. Large errors of \({\hat{\beta }}\) rarely occur when the inferred entropy is low. (D1) The global horizontal irradiance (GHI) on the ground, under a partly cloudy sky31. (D2) The relative error (Eq. 13) is caused by an erroneous assumption of horizontally uniform clouds. (E) Uncertainty in \({\hat{\varvec{\beta }}}\) propagates to estimation of the cloud droplet effective radius \(r^\text{e}\). A value \(r^\text{e}>14\mu \text{m}\) is a precipitation trigger32, yielding rain and dramatic shortening of cloud life.
