There are generally two different types of scenarios that come to mind when the topic is faulty weather predictions.
The first is perhaps the more mundane: Phone apps that tell you it won't rain, but then it does.
The other, however, is more problematic. It involves scenarios in which natural catastrophes occur in places where people haven't received proper warning, causing deaths or injuries that could have been prevented.
We live in an age featuring the most highly developed technology in history. So how do things still go wrong?
Window of opportunity
First of all, we need some context. For the few things they get wrong, meteorologists and modern weather technology are still getting a lot right — more than they ever have before.
"Over the past 20 years, forecasting has improved in accuracy, so a five-day forecast is as good as a three-day forecast 10 years ago," Richard Allan, a professor of climate science in the meteorology department at the University of Reading, England, told DW.
That's due to the development of highly specialized supercomputers.
Meteorologists use these computers to analyze billions of data points gathered from satellite imaging in space, the atmosphere, and on-the-ground weather systems to facilitate something called "ensemble modeling."
Ensemble modeling allows meteorologists to map different scenarios based on the current state of the atmosphere and calculate how it will evolve based on something called the numerical weather prediction model, which uses equations to process current weather conditions.
Computer simulations are run up to 50 times to parse out an accurate forecast, said Allan, adding that there are sometimes only one or two likely scenarios from those 50 attempts.
But that's only when the time period being forecast is in the near future, he said.
"Because the atmosphere is chaotic, accurate regional forecasts are only possible for up to about a week, depending on the type of weather patterns," Allan said.
And meteorologists can have issues predicting events like sudden thunderstorms because they can be very localized, said Nigel Arnell, who is also a professor at the University of Reading's department of meteorology. Weather models operate on a grid in which meteorologists make predictions about one section of atmosphere based on adjoining sections.
So when a sudden thunderstorm hits in one section of that grid, meteorologists might miss the local atmospheric or topographic conditions that trigger them, or might predict them in a slightly different place, Arnell said.
No single model is perfect
Even if meteorologists are able to make fairly accurate predictions about the weather at the end of the day they're still just models, Karsten Haustein, a climate scientist at the University of Leipzig, told DW.
"No single model is ever going to be perfect," he said. "Models can be more or less accurate though. Some are known to perform better at certain lead times than others. Some are known to have specific biases that expert forecasters take into account."
Haustein said problems can arise when results from ensemble forecasts, for example, are misconstrued by laypeople who don't understand the forecasts' biases.
"Let's take last week's 'forecast' for temperatures higher than 40 degrees Celsius [104 degrees Fahrenheit, ed.] as an example," he said. "There was a tiny number of ensemble members in one weather forecast model that predicted these hot temperatures. Most experts knew that the model in question had a bias towards excessively hot temperatures. Plus, none of the other models showed the same. So clearly, this extreme scenario shouldn't have become a news item. Yet it did. Why? In my view, it wasn't a problem of 'incorrect models', but a problem of incorrect reporting."
This kind of misreporting of ensemble forecasts may result in some of the problems that lead to our day-to-day understanding of how the weather's going to look.
But it doesn't fully explain why some major weather events have catastrophic consequences and others of the same scale don't.
Predicting extreme weather doesn't prevent catastrophe
A year ago, massive floods in the Ahr valley region killed nearly 200 people in Germany.
That's despite the fact that the German Weather Service (DWD) predicted them ahead of time. Shortly after the flood event, German prosecutors launched a probe against a district chief in the region for "negligent homicide" after failing to warn residents soon enough.
This is just one example of what can happen when people aren't warned about dangerous weather far enough in advance. Countless others can be observed across the world, from Hurricane Katrina in the southern US in 2005, to the 2015 heat wave in India.
"Forecasting and prediction is only a part of the warning process," Nigel Arnell told DW. "You need to have good mechanisms in place to disseminate those warnings both to public authorities' infrastructure or service operators, and the exposed public."
He added that the authorities also need to know what to do when they receive such a warning, and that the public needs to take it seriously. In countries like Germany, where large inclement weather events like excess flooding or extreme heat don't happen on a regular basis, that's not always a given.
