A team of researchers has combined meteorological data and machine learning to be able to predict lightning. This technology has the ability to prevent the disastrous damage and deaths lightning can cause.
Contrary to the popular saying, lightning does strike twice. And aside from being magical to witness, it's also deadly. Lightning kills more people than tornadoes and hurricanes do, and it can ignite fires that cause hundreds of millions of dollars worth of damage. At the moment, the only real way we can predict lightning is once the storm clouds come rolling in. But that's about to change.
A team of researchers in Switzerland paired machine learning with meteorological data and came up with a simple and cheap system that can predict lightning strikes to the nearest 10 to 30 minutes within an 18 mile radius.
The researchers use a method called hindcasting. As opposed to forecasting, hindcasting is a way to test mathematical models. People put known inputs of past events into a model to see how well that output matches known results. If the model's output is the same as the known output, it's correct. In this case, the researchers used data about previous lightning strikes to create an algorithm that could make predictions about new lightning strikes.
To do this, the team taught their machine learning algorithm to recognize weather conditions that have led to lightning in the past. The four variables they used were: air pressure at station level, air temperature, relative humidity and wind speed.
After the algorithm learnt this information, it's predictions about potential lightning strikes were correct 80% of the time. That's pretty amazing for a first try.
Since it's based on existing data that can be obtained from any weather station, this AI simple and cheap to make. If you're interested creating your own lightning-predicting AI, check out the team's article in the Nature Journal. But until you have accurate predictions about where and when lightning will strike, make sure you stay safe.