Physical Simulation of High-resolution Satellite Images for Fractal Cloud Models

Based on fractal models for the horizontal distribution of cloud density, Landsat-type (i.e., 30 m resolution) radiance fields were simulated within the Nonlocal Independent Pixel Approximation (NIPA), an improved version of the Independent Pixel Approximation (IPA) that uses only the local optical thickness, Scale-by-scale analyses of liquid water variability inside stratus clouds indicate scale-invariance over three decades, from =I0 m to =lO km. A simple two-parameter fractal cascade model reproduces the observed variability, thus capturing the rich turbulent structure in cloud density, hence optical thickness. [PA-based radiation fields of these models preserve scaling properties of fractal cloud models. at least for small moments; however Landsat cloud scenes show a characteristic scale (200-300 m) below which radiance fluctuations are much smaller. This is shown to be the effect of physical smoothing by horizontal photon transport. As a convolution of IPA field with gamma-type smoothing kernel, NIPA emulates this radiative smoothing and produces realistic Landsat-type images. Their statistical verisimilitude is checked with multifractal analyses. The simulations are graphically illustrated and compared with a real Landsat scene.