Will AI help weather forecasters sleep better?

Forecasters fret about making the accurate, timely predictions. Huawei’s Pangu AI model can help.

Gavin Allen: AI prediction tools like Huawei Cloud's Pangu weather model have turned out to be more accurate, precise, and quicker than traditional models. What’s your reaction to this new development? 
Florian Pappenberger: It’s a sudden improvement in skill which we haven't seen for a very long time. It revolutionizes the business of weather forecasting itself. 

Gavin Allen: And if it improves weather forecasting, what's the knock-on impact of that? 
Florian Pappenberger: If you’re planting crops, you may decide to do it today because tomorrow there's too much rain. If you walk out in the traffic, you may find there are snowplows in the city because it was ready for snow, thanks to the more accurate the weather forecast. 

So weather forecasting impacts every single aspect of our lives. Better weather forecasts will automatically improve all types of activities. 

Gavin Allen: What about preparing for natural disasters, such as extreme weather events? 
Florian Pappenberger: Definitely. If you want to produce early warnings for everyone, you will need better forecasts everywhere. 

Machine learning has a real advantage here. It’s actually quite cheap to run compared to conventional weather forecasts, which are really expensive. You need supercomputers, you need loads of scientists. Nowadays, with machine learning models, you can run forecasts on a fairly simplified infrastructure, so you can produce this everywhere. 

So, you will have warnings everywhere. That means you can save lives or can reduce the economic impact of an event. 

Gavin Allen: I'm British, I'm obsessed with the weather. Can I look forward to a day where we get 100% accuracy of the weather thanks to AI?
Florian Pappenberger: You will never have 100% accuracy of the weather; that's simply impossible. Our atmosphere is chaotic, our Earth system is chaotic. 

But there are degrees of accuracy. I can tell you that tomorrow, the weather will be between 1 degree and 20 degrees, or between 9.5 and 10.1 degrees. It's still an uncertain forecast – but far more accurate, better than in the past. That allows people to make better decisions, to plan better. I find this exciting. 

Gavin Allen: Are there other areas where this could transform what forecasting is able to do? 
Florian Pappenberger: Our mission is to produce the best weather forecasts in the world. That's what we're living for. 

To do that, I need open science, open data, an exchange of ideas, and an entire community coming together and freely trying to advance that field. 

So what I want to have is a really open atmosphere, an open discussion forum where we can really advance together. That's where private-sector companies can make a difference. 

Gavin Allen: And finally, when it comes to forecasting, what keeps you awake at night? 
Florian Pappenberger: We produce a weather forecast every day, and it has to be reliable. It has to go out the door at the same time every day, at a high level of quality. 

So for me, as a forecast producer who's trying to support member states of national weather services, that is the most important thing. It has to be on time and high-quality. If there's a tropical cyclone, I would like to predict that cyclone at exactly the time that my member-states, the people who make decisions based on the data I give them, can make the best use of it. 

That's probably what keeps me awake the most.