Hacking the Atmosphere: AI-Powered Weather Disruption Is Here

People worldwide complain about the accuracy of weather forecasting, and the object of their abuse is often a long-standing publicly funded monopoly.

In Australia, this is the Bureau of Meteorology — commonly known as the BOM — which was founded in 1908 after the passage of the Meteorology Act.

Despite Australians grumbling whenever the BOM gets it wrong — and not noticing when it gets it right — the organization has enjoyed a total monopoly on Australian weather forecasting for more than 100 years until now.

Forecasting is now facing disruption from a number of AI and machine learning-powered startups, which are offering what they claim are more accurate forecasts, often on an hour-by-hour basis, for areas as small as a football field.

The agricultural industry, in particular, is interested. The Australian Broadcasting Corporation interviewed a Queensland farmer who said he had lost several crops due to wet weather, crops he said he wouldn’t have planted if the BOM had given him accurate forecasts.

"It's not just the amount the crop has cost you to grow and get to that point, which is quite a bit; it's the loss of sale," the farmer said.

"It's hard to keep the customers happy if you aren't consistent with your supply. And if we get [inaccurate] forecasts, it can really affect our ability to do that."

So, along with precision agriculture, it seems that another developing area in the world of Agtech is localized forecasting.

‘Specific microclimates’

This Queensland farmer turned to startup company Jane’s Weather, which claims to “adapt the best weather guidance from around the globe to a specific microclimate, empowering users with the most accurate outlook at your actual site.”

The company aggregates the top-performing global weather models — including those from the occasionally maligned BOM — into a consensus forecast and uses AM and ML to customize the data for a customer’s specific location.

“I'm amazed at the rapid pace by which we're seeing AI transform weather forecasting.”

As well as pitching its services to farmers, Jane’s Weather is also targeting the construction industry, offering “tailored guidance” for work sites which detail the hours when rain is likely, when winds are too strong for roofing, scaffolding and cranes, and when heat policies need to be enforced.

“Plan your crew and your operations with confidence,” is the pitch.

Another market is the renewable energy industry, and the company offers forecasts and insights on solar radiation, wind and extreme temperatures — all factors in de-risking supply and demand on the network.

Satellite constellation

Jane’s Weather is only one of the new generation of AI-powered weather forecasters, and there are also players in the U.S.

One of these is Tomorrow.io, which turned on data feeds from two radar satellites early this year and is pitching its solutions to a wide range of industries, from aviation and transport to retail, mining, sports and events.

The company plans to launch a constellation of satellites later this year. The satellites will collect data for a machine-learning algorithm that claims to accurately forecast "areas the size of an airport runway."

For organizations, Tomorrow.io offers a "resilience platform" that transforms forecasts into "operational excellence, ensuring proactive adaptation and seamless continuity."

There is also a global weather API, which organizations can plug into their systems, and an increasing array of satellite data.

Digital twins also have a role in this new sector. U.S. company Spire Global uses a version of Nvidia's Earth-2 digital twin to create its custom AI model.

Among its customers are companies in the financial sector, which use intelligence to help them get ahead in potential market movements and understand developments across other industries.

“Having spent my career in the field of weather, I'm amazed at the rapid pace by which we're seeing AI transform weather forecasting, especially when paired with proprietary data that can only be collected from space," Michael Eilts, general manager of weather and climate at Spire, said in a statement.

Possibly the biggest of these private weather companies is Switzerland-based Meteomatics, which has a weather model with over 1,800 parameters.

Describing itself as the "global leader in weather intelligence," Meteomatics is also collaborating with NVIDIA to use AI to power its hyper-local weather forecasts.

Disrupters rely on legacy

However, there is good news for Australia's BOM and other similar legacy weather forecasters worldwide.

One is that their historical and current data is invaluable for the new generation of private AI weather companies.

The second is that while the new AI-based forecasting seems more accurate on a short-term and localized basis, the old physically based models are still likely to be more accurate over longer periods.

In Queensland, the farmer who switched to the Jane's Weather app says it has been more accurate in forecasting daily rainfall levels. Still, the app needs to use BOM data combined with information from the private weather station on the farm with external data.

So, the farmer might curse the BOM as he screams at the sky when the forecast is wrong, but the old institution is still an essential part of forecasting's transformation even as the traditional is being disrupted. 

Image credit: iStockphoto/Grindi