Cloud classification from whole sky images

  • Cumulus clouds in Portland

What climatically meaningful information about clouds can a computer find in a sky image?

Project Abstract

Clouds and aerosols continue to contribute the greatest uncertainty to future climate model projections. As the land surface and atmosphere warm, more water will evaporate, resulting in higher atmospheric water vapor content. Since water vapor is a greenhouse gas, more water vapor will lead to enhanced warming. However, atmospheric water may condense into clouds. Clouds can result in either cooling or warming at the surface, depending on their altitude and composition (liquid, ice, or mixed phase). Low altitude (shallow) Cumulus clouds are generally quite bright and reflective, and have a net cooling effect on the earth surface and atmosphere. They can be difficult to observe from land-based or satellite-based instruments due to their warm temperature and transient coverage. Good old fashioned land-based visible observations are one of the most reliable and long-lived forms of observation. Unfortunately, we replaced most of our meteorologists with automatic cloud sensors in the mid-1990’s, and that record has not been maintained. Our work seeks to fill that void.  Digital sky images are relatively cheap and simple to capture, and contain valuable information on the presence of shallow cumulus clouds. We are working to create an image processing algorithm that can identify cloud boundaries and cloud types in total-sky images captured at the Atmospheric Radiation Measurement (ARM) climate research site near Lamont, OK. This capability will be helpful for creation of large eddy-resolving cloud models, and possibly for long-term cloud climatology observations.

More about this project

Screen Shot 2015-09-09 at 3.42.52 PM Jessi_Kleiss_small

This is a unique website which will require a more modern browser to work! Please upgrade today!