What can we Learn From High Resolution Digital Photography of Clouds?

Wednesday, 17 December 2014: 4:45 PM
Stephen E Schwartz, Brookhaven National Lab, Upton, NY, United States, Daniela Viviana Vladutescu, New York City College of Technology of the City University of New York, Electrical and Telecommunications Engineering Technology Department, New York, NY, United States, Antonio Aguirre, New York City College of Technology of the City University of New York, Applied Mathematics Department, New York, NY, United States and Clement Li, City College of the City University of New York, New York, NY, United States
Commercially available digital cameras provide an unprecedented opportunity for detailed study of cloud structure. Key attributes of such cameras include large number of pixels, (e.g., 3456 x 4608) yielding rich detail of spatial structure, high spatial resolution (e.g., 30 µrad, corresponding to 30 mm for a cloud at 1 km height), and high dynamic range (16 bit in each of three color channels). These attributes permit detailed examination of spatial structure and temporal variability of the influence of clouds on the radiance field. Photography of clouds from the surface looking upwards affords the further advantage, relative to satellite imagery looking downward, that the background is black (space) with contributions to path radiance only from blue sky (Rayleigh scattering), aerosols, and clouds, without complication of surface-leaving radiance. Here we present preliminary results from measurements at Long Island, NY, in summer 2014. The camera was pointed vertically, typically with field of view 22 x 29 mrad (cf. solar diameter 9.3 mrad), corresponding to 22 x 29 m at 1 km. Even at this scale there is no uniquely determined cloud fraction. Cloud fraction defined as the fraction of pixels that encompass at least some cloud (DiGirolamo and Davies, JGR, 1997) is found to be highly dependent (several tens of percent) on threshold and on resolution, which can readily be artificially degraded by pixelating the image). Likewise, in contrast with findings of Sachs, Lovejoy, and Schertzer (Fractals, 2002) no unique fractal dimension appears to be associated with clouds, the retrieved value being dependent on approach and averaging method, Figure 1.

Figure 1. Upper left, color image of zenith sky (2048 x 2048 pixels; 1 pixel = 6.3 µrad) at Upton, Long Island, NY (40.87˚ N; 72.89˚ W; 33 m MSL), July 3, 2014, 0803, local standard time). Lower left, two different cloud thresholdings, corresponding to cloud fraction 0.48 and 0.68. Right, squared magnitude of Fourier transform (FFT) of gray-scale image of cloud at left versus inverse radial distance; colors indicate density of points; green line and slope denote fit to points for which log(inverse radial distance) ≤ 0.1.