Fig. 5
From: Limitation of super-resolution machine learning approach to precipitation downscaling
MLPerfect and MLImperfect model output precipitation (at 12.5 km) when provided zero coarse precipitation input (at 100 km) on all grid points except a small precipitation perturbation of 0.5 mm/h at the three selected points (one in Papua New Gunia (PNG), second in South East Australia (SEA), and the third in Southern Ocean (SO) as shown in (a)), each point perturbation at a time and the orography input is unchanged. MLPerfect and MLImperfect model output precipitation when perturbed at the PNG point with orography unchanged are shown in (b) and (e), respectively. The MLPerfect and MLImperfect model outputs when perturbed at SEA and SO points are shown in (c), (f) and (d), (g), respectively. Maps are drawn using the Python Cartopy package (v0.24.1).
