The Challenges of Predicting Severe Weather and the Role of AI
Incidences of severe weather are on the rise, making the importance of accurate forecasting more evident than ever. How AI may be able to outperform traditional weather models and save lives.
An Increasing Prevalence of Severe Weather
A report by the American Meteorological Society has recently confirmed a concerning trend of worsening severe weather. Last year alone (2022) had 18 severe weather events that ended up doing over a billion dollars worth of damage in the United States, according to the NOAA National Centers for Environmental Information. These 18 events alone would sadly also claim the lives of 474 individuals.
With no end in sight, we need every possible defense we can. One unlikely source of help may be Artificial Intelligence. Many of us have been surprised by the abilities of AI over the past few years. Especially with the publicity of AI models such as ChatGPT. But how can AI help us save lives and possibly even lessen property damage?
Current Forecasting Models
Considering how challenging predicting future weather conditions can be, our current prediction models are actually quite impressive. Many of these models work by performing complex physics computations to predict the weather based on current conditions.
This method requires immense amounts of data and a tremendous amount of computational power to compute that data. The NOAA, for example, uses a supercomputer with 5.78 petaflops of computing power. That computer is responsible for computing data from
- Weather Balloons
- Radar Towers
- Surface Observations
- Satellite Data
The Advantage of AI
Thanks to a comprehensive history of past weather data, we have a perfect starting point for generating AI models capable of accurately predicting future weather events. For example, by feeding an AI model past weather conditions and then showing how they changed over time, we can help it recognize meteorological patterns that may be beyond the scope of current prediction methods.
Speed
AI weather models also have an advantage when it comes to speed. This is vital as storm systems can be very dynamic and change rapidly. This speed advantage could mean quicker alerts and warnings for people in the path of severe weather. This is especially important for weather events such as tornadoes which can move and change direction quickly.
Computational Power
AI is not usually known for needing low amounts of computing power, but it may actually be more efficient when it comes to weather forecasting. This comes down to the fact that traditional weather forecasting requires immense numbers of complex calculations to simulate how weather events will unfold. This is slightly different from an AI solution which will instead look for patterns in weather data that can be used to predict how a storm will develop.
Cost
While cost may not be the most critical factor, it does allow for more accessibility. This is a surprising benefit of the AI models. For example, a Google Cloud example project shows how to build a model that outperforms traditional models for as little as one dollar.
Beyond Thunderstorms
Growing up in the Midwest, I usually correlate severe weather with thunderstorms and tornadoes. However, this is far from the only type of severe weather impacting humans. AI can be used to predict many different types of severe weather.
- Droughts — Properly predicting droughts can help us better prepare and account for them. Droughts can also bring along an increased fire risk. AI has shown promise in lowering this risk, too, by being able to detect wildfires through satellite imagery.
- Winter Weather — Winter weather can cause people to be stranded in potentially life-threatening temperatures. It can also lead to dangerous travel conditions.
- Flooding — Accurately predicting rainfall can also be essential for predicting where and when floods will occur. This could give us more time to evacuate or prepare for high waters.
- High and Low Temperatures — Both high and low temperatures can be lethal under the right circumstances. Advanced warnings can ensure we can prepare for these weather events before they hit.
The Future of AI Forecasting
While current AI models look highly promising, they must be thoroughly tested and vetted before they can be trusted to alert us to severe weather threats. Falsely predicting lousy weather can be inconvenient, but falsely predicting good weather can be deadly.
If you enjoyed this article, please let me know by leaving a clap or a comment! If you’d like to see more content like this, please check out my profile or leave a follow. Thanks!
