New AI-based weather models are showing promise in hurricane forecasting. These AI weather models work by learning patterns from historical weather data.
This data is typically collected from satellites, buoys, planes, and other sensors. The AI algorithms then use these patterns to make predictions about future weather conditions, but they still have some drawbacks.
AI models are able to produce multiple projections in minutes, which is much faster than traditional physics-based models. This could allow meteorologists to produce more accurate ensemble forecasts, which show a range of possible scenarios and how likely they are to occur.
AI weather models offer several benefits over traditional physics-based models. First, they are much faster. AI models can produce multiple projections in minutes, while traditional models can take hours. Second, AI models are able to learn complex patterns in the data that traditional models may miss. This could lead to more accurate forecasts, especially for tropical storms and other extreme weather events.
These AI models are not yet able to quantify the uncertainty in their own predictions, which is a major weakness. This is known as the “black box” problem, and it is common to many machine-learning systems.
Another challenge is that AI models are trained on historical data, which means they may not be able to predict extreme or unusual weather events. Despite these limitations, AI models could revolutionize weather forecasting in the future. If they can be improved to address their weaknesses, they could make accurate forecasts more widely available and help people to better prepare for severe weather events.
Researchers are working on a number of ways to improve AI weather models. One approach is to develop new algorithms that can better quantify the uncertainty in the predictions. Another approach is to train AI models on larger and more diverse datasets, which could help them to learn to predict extreme or unusual weather events.
The sources for this piece include an article in Wired.