Automated identification of cloud types at the ARM Southern Great Plains site
Gustafson, William I. — Pacific Northwest National Laboratory
Riihimaki, Laura Dian — CIRES | NOAA ESRL GML
Area of research
Different types of clouds have different radiative forcing. This means researchers quantifying the role of clouds on the global energy budget, as well as regional or global water cycles, need accurate classification of cloud types for Earth system models. A team of researchers, including scientists at the U.S. Department of Energy, used 13 years of data to develop an automated algorithm that identifies seven different cloud types at the Atmospheric Radiation Measurement (ARM) user facility observatory in the U.S. Southern Great Plains (SGP). The algorithm identified daily and seasonal cycles for different cloud types. The researchers also developed a method to identify fair-weather shallow cumuli clouds, a type of low cloud that is currently challenging to simulate in climate models.
Researchers often want to study only certain meteorological conditions, such as days with deep convection. The new cloud type analysis, called cldtype, assists researchers by narrowing down their search for particular cloud types. The fair-weather shallow cumuli cloud identification program, called shallowcumulus, further sub-categorizes the low clouds in cldtype to identify fair-weather cumuli. This was motivated by the Large-eddy Simulation ARM Symbiotic Simulation and Observation (LASSO) activity, which focuses on days with these clouds at the ARM SGP site. Overall, categorizing the different cloudy periods makes it easier for researchers to take advantage of ARM’s large suite of instrumentation.
Researchers used cloud observations from 1997 to 2009 collected at the ARM SGP observatory to generate an automated algorithm that classifies clouds into seven types: low clouds, congestus, deep convection, altocumulus, altostratus, cirrostratus/anvil, and cirrus. The researchers based this classification on the physical qualities of cloud top, cloud base, and physical thickness of cloud layers measured with millimeter-wavelength cloud radar and micropulse lidar.
Additionally, the researchers developed another algorithm to identify fair-weather shallow cumulus events using cloud fraction information collected from 2000 to 2008 with a total sky imager and ceilometer. The events identified automatically agreed closely with fair-weather shallow cumulus events identified manually. The automated analysis only missed six cases out of 70 possible events during the spring to summer seasons (May–August).