Short range structure of clouds has been examined by high resolution photography from the surface. Zenith-looking digital images were taken using a commercially available camera with angular resolution ~20 μ rad, field of view (FOV) ~20 × 30 mrad, corresponding, for a cloud at 1 km, to 20 mm resolution and 20 × 30 m FOV, much higher resolution than other imaging approaches, notably satellite imagery and the total sky imager. Such high resolution imagery aids understanding the processes responsible for cloud structure and radiative effects. Work to date has focused on developing and evaluating methods for characterization of cloud properties by digital image processing techniques: determination of dependence of cloud fraction on threshold and resolution, determination of fractal dimension by multiple techniques, and determination of autocorrelation properties. Emphasis thus far has been directed to optically thin clouds, which are poorly characterized observationally, and poorly represented in models, yet exert strong radiative influences, especially as light scattering is linear in optical depth. Preliminary analysis shows clouds exhibit much short-range structure on scales down to 1 meter. Histograms of (Red/Red + Blue) a measure of cloud contribution to radiance, commonly exhibit no clear break that would be indicative of a unique threshold for presence of cloud, resulting in strong dependence of cloud fraction on threshold, Figure 1. In contrast autocorrelation is nearly independent of intensity variable, threshold, and resolution, provided resolution is finer than autocorrelation length. The spatial autocorrelation function, together with the probability density function of the cloud quantity of interest, provides a statistical representation of cloud fields that can be used for radiation computation (Huang and Liu, Environ. Res. Lett., 2014) and cloud microphysics parameterizations. It is hypothesized as well that short-range cloud structure is a consequence of short-range turbulence of the atmosphere and thus that determination of this structure will lead to a means of characterizing atmospheric turbulence by remote sensing.
Figure 1. High resolution photograph of cloud at Upton NY (40.82°N, 72.87°W), 2014-07-03, 09:03 local time; north is at top of image. Shown are image in natural color and grayscale images of red channel, to enhance contrast, and (Red/Red + Blue), RRB, to permit thresholding. Yellow areas denote region exceeding threshold for three values of RRB threshold shown on probability distribution function (PDF) of RRB and corresponding cloud fraction. Panels at right show autocorrelation in central region of cloud (bounded by black square on natural-color image) as function of lag in pixels in original image, for which 1 pixel = 6 μ rad, and in East-West and North-South directions for resolution degraded by factors of 4 and 16.
Acknowledgment. Supported by the U.S. Department of Energy's Atmospheric Science Program (Office of Science, OBER) under Contract No. DE-SC00112704.
This page was last updated 2015-10-08.
Return to Stephen E. Schwartz Publications Page