Wind-dependent Variables: Predict Wind Speeds of Tropical Storms Hosted By Radiant Earth Foundation

1 week left
$13,000

Predict Wind Speeds of Tropical Storms

Overview


Tropical cyclones, which include tropical depression, tropical storms, hurricanes, and major hurricanes, are one of the costliest natural disasters globally. Hurricanes can cause upwards of 1,000 deaths and $50 billion in damages in a single event, and have been responsible for well over 160,000 deaths globally in recent history.

An accurate diagnostic model for tropical cyclones is critical for disaster preparedness and response. According to the National Oceanic and Atmospheric Administration (NOAA), storm damage models approximate risk using an exponential or power of wind speed. It is therefore essential that storm forecasters be able to objectively and consistently estimate the maximum sustained surface wind speed, or intensity, of tropical cyclones.

Today, forecasters primarily rely on adaptations of a satellite image-based classification method, known as the Dvorak technique, to predict wind speed. These techniques involve visual inspection of images and are limited by human subjectivity in assessing complex cloud features. There is a vital need to develop automated, objective, and accurate tropical cyclone intensity estimation tools from readily available satellite image data.

Recently, the NASA IMPACT team developed an experimental framework to investigate the applicability of deep learning-based models for estimating tropical cyclone intensity in near-real time. While the current model shows that Convolutional Neural Networks (CNNs) can capture key patterns in the satellite imagery of storms to estimate wind speed, we seek to improve the accuracy for operational applications.

Task


The goal of this challenge is to estimate the wind speeds of storms at different points in time using satellite images captured throughout a storm’s life cycle and the temporal memory of the storm. Radiant Earth Foundation has worked with the NASA IMPACT team to assemble a data set of tropical storm imagery, which includes single-band satellite images at a long-wave infrared frequency and corresponding wind speed annotations. Improving initial wind speed estimates from satellite imagery could mean significant improvements in short-term storm intensity forecasting, risk approximation models, and disaster readiness and response.

If the winning solution of this competition performs better than the existing model running on Hurricane Intensity Estimator, the model will be replaced with credit given to the winner.


Competition End Date:

Feb. 2, 2021, 11:59 p.m. UTC

Place Prize Amount
1st $6000 and $8000 Azure credit
2nd $4500 and $6000 Azure credit
3rd $2500 and $4000 Azure credit

This challenge is convened by our friends at Radiant Earth Foundation.

With generous support from:

Convening Sponsor

Gold Sponsors

Silver Sponsors

Azure Credit Sponsor

Technical Supporter


Banner image courtesy of the Centers for Disease Control and Prevention.