Halo Identification Algorithm for Sky Images Produced in TSI Series

The radiative effects of cirroform clouds are challenging and varied. The goal of this project is to extract information about cloud composition, as well as about the spatial and temporal distribution of these clouds based on ice halo observations. Cirrus optical scattering behavior is heavily influenced by the types of ice particles, which may be present in many forms, including crystalline hexagonal habits in form of plates, pencils and prisms, hollow columns, bullets and bullet rosettes, and also as amorphous ice pellets, fragments, rimed crystals and others. If a majority of pristine crystal habits is present, the optical scattering behavior of the cirrus cloud gives information about the cloud particle types in form of ice halos, most frequently appearing as a bright ring of 22° radius around sun or moon. An important question emerges: How can we use ice halos to improve our knowledge about the composition of cirroform clouds and the conditions in the upper troposphere? One of the first tasks is to establish frequency of halo appearances across seasons and years, as well as geographically. Sky images have been collected for decades at several research facilities. We are using series of images produced by Total Sky Imagers (TSI) at Atmospheric Radiation Measurement (ARM) Climate Research Facilities in order to assess the presence of ice halos on these sites. Three locations are considered in this study: (1) Southern Great Plains (SGP) Central Facility[1], (2) Eastern North Atlantic (ENA) Graciosa Island, Azores [2], and (3) the North Slope Alaska (NSA) Facility [3]. These images have been produced and collected every 30 seconds over many years. An automated way of recognizing the presence of a halo became necessary.

We present an image-processing algorithm to automatically identify ice halos in TSI images. The TSI data present as jpg images of different sizes (over the years) and different quality. The images are rotated and cropped, non-sky parts masked to form a standard orientation. The radial brightness curve of four four sky quadrants surrounding the sun are analyzed for all three color channels. The radial brightness decay, the presence of the bright band, and the general sky conditions inform a skytype score that classifies the image as clear, partially cloudy, or cloudy, and assigns a halo probability. Parameters have been calibrated using a 200-image set representative of all sky types and locations. We will present our results on the frequency and distribution of ice halos for the three locations above. The data collection is still ongoing at time of abstract submission.

Data were obtained from the Atmospheric Radiation Measurement (ARM) Program sponsored by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division.

Sources:

1. Atmospheric Radiation Measurement (ARM) Climate Research Facility. 2000, updated hourly. Total Sky Imager (TSISKYIMAGE). 2000-07-02 to 2018-04-19, Southern Great Plains (SGP) Central Facility, Lamont, OK (C1). Compiled by V. Morris. Atmospheric Radiation Measurement (ARM) Climate Research Facility Data Archive: Oak Ridge, Tennessee, USA. Data set accessed 2018-06-07 at http://dx.doi.org/10.5439/1025309. 2018.
2. Atmospheric Radiation Measurement (ARM) Climate Research Facility. 2013, updated hourly. Total Sky Imager (TSISKYIMAGE). 2013-10-01 to 2018-05-28, Eastern North Atlantic (ENA) Graciosa Island, Azores, Portugal (C1). Compiled by V. Morris. Atmospheric Radiation Measurement (ARM) Climate Research Facility Data Archive: Oak Ridge, Tennessee, USA. Data set accessed 2018-06-07 at http://dx.doi.org/10.5439/1025309. 2018.
3. Atmospheric Radiation Measurement (ARM) Climate Research Facility. 2006, updated hourly. Total Sky Imager (TSISKYIMAGE). 2006-04-25 to 2018-04-11, North Slope Alaska (NSA) Central Facility, Barrow AK (C1). Compiled by V. Morris. Atmospheric Radiation Measurement (ARM) Climate Research Facility Data Archive: Oak Ridge, Tennessee, USA. Data set accessed 2018-06-07 at http://dx.doi.org/10.5439/1025309. 2018.