Retrieving Cloud Characteristics from Ground-Based Daytime Color All-Sky Images

Long, C. N., Pacific Northwest National Laboratory

Cloud Distributions/Characterizations

Cloud Properties

Long, C. N., J. M. Sabburg, J. Calbo, and D. Pages, (2006): Retrieving Cloud Characteristics from Ground-based Daytime Color All-sky Images, JTech, 23, No. 5, 633–652.

Long, C. N., J. M. Sabburg, J. Calbo, and D. Pages, (2006): Papers of Note: Retrieving Cloud Characteristics from Ground-based Daytime Color All-sky Images, BAMS, 87, No. 6, 743–744.


Figure 1. Sky image (left) from 1300 LST Sept 4, 2004, and corresponding cloud decision image (right) denoting originally retrieved clear sky (blue), thin cloud (gray), and opaque cloud (white). Black denotes masked portions not counted in sky cover retrievals. Note the interpretation of a "cloudy" area near and below the sun in the cloud decision image when a virtually clear sky is evident in the sky image.


Figure 2. Total sky cover retrieval for Sept 4, 2004. The gray line is the original retrieval, the thin black line is the retrieval including the "first guess" adjustment of the sun circle area, and the black line is the final corrected result including all adjustments and smoothing.

An overview of sky imaging, along with efforts to infer cloud macrophysical properties from the resultant images, is given.

Given the rising costs associated with maintaining human sky observations, automated sky imagers are seeing increasing popularity. An overview of techniques for inferring cloud properties such as fractional sky cover, cloud aspect ratio, cloud brokenness, solar obstruction, and cloud type shows that an upswell in using sky imagery is also growing. In addition, advancements are being made in making more accurate retrievals. For instance, a problem with many sky imager systems is error in clear/cloud determination for the areas near the sun in the image, and along the horizon centered on the solar azimuth angle at low solar elevations. These areas of the image are naturally whiter than other parts of the cloudless sky in the image due to forward scattering. With no a priori knowledge of aerosol or haze loading that can be used in some way to predict an increased brightness, these pixels are often interpreted as “cloudy” in the sky imager retrievals when a human observer would label them as “cloudless.” Figure 1 is an extreme example of this type of sky cover retrieval error, where most of this day at the Pacific Northwest National Laboratory in Richland, Washington, exhibited clear but hazy skies, with two episodes of cloud incursion. We present a statistical methodology to correct for these errors given highly sampled (1-minute or better) sky cover retrievals. The original erroneous sky cover retrievals (upper gray line) and the more accurate results of this correction for retrieval error (lower black line) are shown in Figure 2 for the example day in Figure 1.

With advances in technology and algorithm development, sky imagers are increasingly used in recent years. This trend will likely continue, adding a powerful tool to surface-based cloud measurement efforts.