Short-Range Structure of Clouds Studied by High Resolution Photography From the Surface

Abstract:

Clouds exhibit structures at a wide range of length scales. Passive radiometry from satellite shows structure on scales of tens to thousands of kilometers, but there is much structure at short spatial scales not resolved by satellite imagery. Here we use a commercial camera having high spatial resolution (~20 µrad) and high dynamic range (16 bits in each of three color channels) in narrow field-of-view (20 mrad, 110 mrad), zenith-looking mode from the surface, to examine clouds at the scale of centimeters to a few hundred meters, focusing on non-precipitating single-layer clouds during daytime. Up-looking photography of clouds from the surface affords the further advantage, relative to satellite imagery, of black background (space) with contributions to radiance only from blue sky (Rayleigh scattering), aerosols, and clouds, permitting reconstruction of observed radiance by radiation transfer modeling. Contrast between cloudy and cloud-free sky is enhanced in Red/(Red + Blue), RRB, image Figure 1, but no unique value of RRB distinguishes a pixel as cloud vs. cloud-free. Short-range variability is characterized by the autocorrelation length scale, which is not uncommonly as short as a few meters; longer range variability, such as cloud characteristic size, separation distance, and cloud spatial organization, is also characterized. Scene reconstruction yields the 2D distribution of cloud optical depth; spatial inhomogeneity is attributed mainly to horizontal variation in vertical motion of the air and resultant condensation or evaporation associated with upward or downward motion, respectively. Alternative approaches to calculation of the radiative influence of such clouds from the autocorrelation structure of the cloud field are examined.

Figure 1. RGB image of zenith sky at New York City, May 22, 2015, (field of view 21 mrad corresponding to 56 m at cloud altitude 2.6 km) showing broken single-layer cloud; corresponding RRB image; and autocorrelation of RRB image.