Machine Learning Solution for Prevalent Aviation Challenge

Photo illustration airport by PIXABAY

This post is also available in: עברית (Hebrew)

Weather conditions such as fog and low stratus clouds over airports are responsible for severe flight delays translating into heavy costs for airlines. Determining the size of a fog patch over a large area is challenging, especially for satellite observations.

The US National Weather Service is using a new fog detection software developed by scientists at the University of Wisconsin-Madison Cooperative Institute for Meteorological Satellite Studies and NOAA to assist with fog warnings. 

The GOES-R software (known as a satellite product) uses machine learning techniques with near real-time data from weather satellites like NOAA’s GOES-East and -West to monitor conditions 24/7 and issue potential fog warnings.

Earlier fog detection systems would use comparisons between satellite data in the visible and infrared frequencies but were largely limited to conditions where no other clouds were present above the fog and were limited to nighttime use.

“To improve the system, we incorporated other atmospheric data used in Numerical Weather Prediction algorithms,” according to CIMSS. “This helps identify the type of fog over an area by evaluating every pixel in an image and generating a probability of the presence of fog and its intensity.”

The GOES-R Fog Product was designed based on needs from the aviation industry which has different requirements depending on the aircraft. Small airplanes for example, rely on a different set of rules when it comes to weather because they lack the advanced instrumentation of the larger airliners. 

The software identifies dangerous conditions for both types of aircraft and can be used to issue warnings around the country, avoiding costly delays and making air travel safer, according to