A method for retrieving cloud optical depth ( τ c ) using a UCSD developed ground-

based Sky Imager (USI) is presented. The Radiance Red-Blue Ratio (RRBR) method is

motivated from the analysis of simulated images of various τ c produced by a Radiative

Transfer Model (RTM). From these images the basic parameters affecting the radiance

and RBR of a pixel are identified as the solar zenith angle (SZA), τ c , solar pixel an-

gle/scattering angle (SPA), and pixel zenith angle/view angle (PZA). The effects of these

parameters are described and the functions for radiance, I λ (τ c ,SZA,SPA,PZA) , and the

red-blue ratio, RBR(τ c ,SZA,SPA,PZA) , are retrieved from the RTM results. RBR, which

is commonly used for cloud detection in sky images, provides non-unique solutions for τ c , where RBR increases with τ c up to about τ c = 1 (depending on other parameters) and

then decreases. Therefore, the RRBR algorithm uses the measured I meas

λ

(SPA,PZA) , in

addition to RBR meas (SPA,PZA) to obtain a unique solution for τ c . The RRBR method is

applied to images of liquid water clouds taken by a USI at the Oklahoma Atmospheric

Radiation Measurement program (ARM) site over the course of 220 days and compared

against measurements from a microwave radiometer (MWR) and output from the Min

[ MH96a ] method for overcast skies. τ c values ranged from 0-80 with values over 80

being capped and registered as 80. A τ c RMSE of 2.5 between the Min method [ MH96b ]

and the USI are observed. The MWR and USI have an RMSE of 2.2 which is well within

the uncertainty of the MWR. The procedure developed here provides a foundation to test

and develop other cloud detection algorithms.

Using the RRBR τ c estimate as an input we then explore the potential of using

tomographic techniques for 3-D cloud reconstruction. The Algebraic Reconstruction

Technique (ART) is applied to optical depth maps from sky images to reconstruct 3-D

cloud extinction coefficients. Reconstruction accuracy is explored for different products,

including surface irradiance, extinction coefficients and Liquid Water Path, as a function

of the number of available sky imagers (SIs) and setup distance. Increasing the number

of cameras improves the accuracy of the 3-D reconstruction: For surface irradiance,

the error decreases significantly up to four imagers at which point the improvements

become marginal while k error continues to decrease with more cameras. The ideal

distance between imagers was also explored: For a cloud height of 1 km, increasing

distance up to 3 km (the domain length) improved the 3-D reconstruction for surface

irradiance, while k error continued to decrease with increasing decrease. An iterative

reconstruction technique was also used to improve the results of the ART by minimizing

the error between input images and reconstructed simulations. For the best case of a nine

imager deployment, the ART and iterative method resulted in 53.4% and 33.6% mean average error (MAE) for the extinction coefficients, respectively.

The tomographic methods were then tested on real world test cases in the Uni-

versity of California San Diego’s (UCSD) solar testbed. Five UCSD sky imagers (USI)

were installed across the testbed based on the best performing distances in simulations.

Topographic obstruction is explored as a source of error by analyzing the increased error

with obstruction in the field of view of the horizon. As more of the horizon is obstructed

the error increases. If at least a field of view of 70 ◦ is available for the camera the accuracy

is within 2% of the full field of view. Errors caused by stray light are also explored by

removing the circumsolar region from images and comparing the cloud reconstruction to

a full image. Removing less than 30% of the circumsolar region image and GHI errors

were within 0.2% of the full image while errors in k increased 1%. Removing more than

30 ◦ around the sun resulted in inaccurate cloud reconstruction. Using four of the five

USI a 3D cloud is reconstructed and compared to the fifth camera. The image of the fifth

camera (excluded from the reconstruction) was then simulated and found to have a 22.9%

error compared to the ground truth.