My University of Utah Atmospheric Sciences colleague Dr. Timothy Garrett and his team are embarking on a citizen science project to classify snowflakes and train computer algorithms to improve snowfall prediction. Tim provides a summary of this effort, including how you can participate, below.
Snowflake ID
A British expression for a phony excuse is “the wrong kind of snow”, a phrase originally coined by British Railways to justify winter train delays. For residents of Utah, such dry humor might be lost — snowflake type affects our enjoyment of wintertime sports. The key distinction that is usually made is between fluffy “aggregates”, or assemblages of pristine six-sided snow crystals, and denser “graupel”, small hail-like pellets that grow when a single snowflake collides with millions of tiny cloud droplets as it falls, and these freeze on its surface. Studies have shown that determining which type of snow forms in clouds is crucial for accurate weather and climate prediction. Because graupel tends to fall faster than aggregates, getting the difference right affects forecasts of where geographically snow will land first as well as the speed of the water cycle.
To help add insights into this difficult problem, Dr. Tim Garrett and his team at the University of Utah have launched a project that recruits citizen scientists to help build a snowflake classification algorithm.The Multi Angle Snowflake Camera or MASC is a new meteorological instrument that was developed at the University of Utah to automatically photograph snowflakes in freefall from three different angles.
 |
| Examples of snowflake images captured by the Multi-Angle Snowflake Camera (MASC) |
Because the MASC has subsequently been sold to research teams around the world, hundreds of millions of snowflake images have been collected in locations as far flung as Greenland, the Antarctic, the Alps, South Korea and right here in Utah at Alta Ski Area. The issue now is one of so many snowflakes to classify, so little time. To help automate the classification process, about 5000 snowflake images have been uploaded to Zooniverse.org, a citizen science web portal owned and operated by the Citizen Science Alliance. The Zooniverse Snowflake ID webpage first educates users in snowflake identification and then allows users to classify as many snowflakes as they feel up to. The classification process is designed so that anyone aged 8 and up can participate, learning more about snowflakes in the process.
The goal of the project is to “teach” an artificial intelligence computer algorithm to be able to automatically classify any snowflake it is presented. Scientists in the US and internationally can then use this sophisticated tool to accurately classify the millions of snowflakes images that they have been collecting, supporting their studies into the weather conditions that affect snow development. Some day, with refined weather and climate models, it may be possible to better predict “the right kind of snow”
You can check out the project and learn how to contribute at https://www.zooniverse.org/projects/fitch09/snowflake-id.
No comments:
Post a Comment