During another storm, meteorologists rely on various forecasting models. As this time comes to an end, the experts take stock of what was done well and who didn’t leave them completely gobspacked.
Although Google Depmind’s Weather Lab started releasing forecasts in June, it was the best model for predicting Hurricane Track and Certainty this season, according to weather analysis and a senior researcher at the University of Miami. At that time, the American climate model – the global forecasting system – was performing very poorly.
The National Weather Service will release official data on each model’s performance in a few months, but these initial tests represent an opportunity to change storm forecasting. With the dramatic rise of AI-based models in plain view, it may be time to start ditching traditional, physics-based models.
“Going forward, it is safe to say that we will be more dependent on Google and other AI Weather models, which may improve in the coming years,” as it is better to improve the situation, “there is a situation of Houston meteorologist and journalist Eric Berger wrote ARS Technica.
The rise of AI prediction has begun
McNoldy’s analysis includes two charts: one that compares the accuracy of all 13 hurricanes named in the Atlantic basin this season, and one that compares the weather accuracy of all 13 hurricanes. Different colored lines represent different climate models, indicated by the legend on the right-hand line. Bottom line, the better that model is.
The GFS – referred to as avni in this example – is shown in orange all the way up the top of the charts. NOAA developed this model in the early 1980s, and the National Weather Service still uses an updated version as its primary forecasting system today.
“The gfs are terrible with their melissa weather, with an average track error of 5 days to over 500 miles [800 kilometers]they persisted with the outflow of the sea that they never finished,” according to Miami meteorologist and meteorologist and hurricane Sichael Lowry wrote in a recent post.
Unlike Google’s weather model, the GFS is based on traditional physics and advanced vendors. The difference between them is clearly prominent in these charts. The Google model is all the way down, showing superior performance to all other models tested – especially the GFS.
“The beauty of DeepMind and other similar data-driven models is that they are very fast when generating predictions compared to their traditional physics-based counterparts that require some of the most expensive and advanced algorithms in the world,” it is written. “Besides, these” intelligent “models with their own neural network structures have the ability to learn from their mistakes and correct them on-the-fly.”
Urgent need for better prognosis
Hurricane Melissa – which ravaged the Caribbean last week – is just one example of high sea surface temperatures. As climate change causes storms to become shorter and more damaging, it is important that forecasters have better tools to predict their paths and strength.
AI-based models can help predict global warming trends. DeepMind’s surprising haul certainly has their attention and could mark the start of a new era in storm forecasting.
