Thursday, September 13, 2018

Tracking hurricanes with artificial intelligence



From Mapbox blog by Eric Gundersen

NASA and Development Seed are tracking Hurricane Florence using machine learning techniques, producing results six times faster than current capabilities.
Their neural network-based approach calculates hurricane strength and wind speed by monitoring live imagery as it’s delivered from weather satellites.
This allows NASA to create estimates hourly, a significant speedup from the usual six-hour cycle.

The eye of hurricane Florence
image : ESA / Alexandre Gerst

The primary factor for estimating a hurricane’s destructive potential is wind speed.
By creating faster, more reliable estimates of storm wind speeds, authorities may be able to make better decisions about moving people out of harm’s way and moving resources where they’re needed.
These decisions can help save both life and property.
The issue is growing in urgency: the 2017 hurricane season was the most destructive on record, claiming thousands of lives and causing an estimated $280 billion in damage.

This is going to be a crazy end to the week!
Eastern Pacific: #Paul #Olivia 
Western pacific: Typhoon #Mangkhut, Tropical Storm #27W, Invest 91W
AI vs. humans

Estimates of cyclone intensity rely upon the Dvorak technique, which matches satellite imagery of a storm to known patterns.
Once matched, it’s possible to estimate wind speed.
AI experts at NASA’s Marshall Spaceflight Center and Development Seed trained neural networks using historical hurricane imagery and classifications, allowing this workflow to be fully automated.

 The view of the Atlantic on Sept. 12. Florence on the right, bearing close to the US coast, Tropical Storm Isaac near the Lesser Antilles, and Hurricane Helene off the coast of Africa.
image : NOAA

 Looking at hurricane Florence through wave height
Black represents waves of about 7 meter waves. 

This allows data to flow directly from the GOES-16 weather satellite, to the NASA Cumulus framework running on AWS, to seamlessly generated predictions.

Although it’s currently running at six times the frequency of traditional prediction mechanisms, the system is theoretically constrained only by the bandwidth of its satellite source.


Available now :
The Hurricane Intensity Estimator, built with Mapbox, is running alongside data collected from human estimation and aerial flights.
NASA plans to continue to improve the prediction models.

Hurricane Florence’s location via the Google Crisis Map hurricane tracker

1960 U.S. Weather Bureau Hurricane Tracking Chart
courtsey of Geographicus


Links :

No comments:

Post a Comment