Weather reports are rarely as accurate as we would like them to be — but now IBM has a plan to add a little more fact to forecasts. The company is combining its hyper-local weather models with global models from The Weather Company to create a better forecasting system.
The result is Deep Thunder, a program that’s built upon twenty years of weather research at IBM combined with more than forty years of experience from The Weather Company. In October 2015, IBM announced plans to purchase the digital component of The Weather Company as part of its commitment to the Internet of Things, leaving broadcast network The Weather Channel to operate independently.
IBM is reportedly creating new algorithms that will help analyze the enormous amounts of data that can be collected when the local and global models are working in sync, according to a report from Engadget. Petabytes of historical data are being fed into Deep Thunder to help prepare the system for live analysis.
The endgame for this plan is for IBM to be able to deliver relevant weather information to businesses. These reports are a little different from those that follow your daily news broadcast, as they’re designed to be actionable forecasts that relate to consumer behavior in relation to conditions.
For example, utility companies will be able to use the information to track which areas are set to be hit worst by a storm, and to plan their repairs accordingly. The system will also be able to inform retailers of how different weather conditions might affect buying patterns.
This program demonstrates how deep learning techniques can be implemented beyond such things as last year’s man vs. machine Go showdown. There’s obviously a place for well-publicized events that demonstrate the capabilities of this tech — but day-to-day applications like those IBM has outlined for Deep Thunder prove that these techniques are more than just fodder for research projects.
