Television reporters watch as Hurricane Michael makes landfall along the Florida Panhandle on Oct. 10, in Panama City Beach. (Jabin Botsford/The Washington Post)

Meteorologists are taught that an accurate weather forecast conjures a vision of what Mother Nature is most likely to deliver. Except that’s not always true. Sometimes forecasts have a different message.

In practice, there are two kinds of predictions that forecasters can make about the weather.

Most days, meteorologists and computers analyze the available data to produce forecasts that describe what they think the weather is most likely to be at some point in the future — “most likely” being the operative words.

When extreme or dangerous conditions are a possibility, however, a different paradigm is needed. Prudence requires that people be alerted to something closer to a worst-case scenario. If the pieces fall into place in a fairly low-odds but imaginable way, how bad could the weather get? People need to know this so they can be ready. Just in case.

By definition, preparation means taking action in advance. That could mean rearranging schedules, buying supplies or, in the extreme, evacuating neighborhoods. And, of course, at the time action is required, the exact weather effects three or four days into the future are unknowable.

In modern weather parlance, this type of forecast is called a “reasonable worst-case” prediction. It’s not the absolute worst imaginable, but it’s a low-odds scenario that is reasonably possible.

Tornado warnings are a kind of reasonable-worst-case prediction. We know that most people in the warning area are not going to get hit by a tornado, and the chance of any one house under the warning being affected is very low. But most people intrinsically understand the danger of a tornado. They want to be warned about a potential catastrophic event, even if the odds of their house being affected are small — within reason.

The difference between most likely and reasonable-worst-case forecasts come down to the threshold you set for being wrong. The forecast for tomorrow’s temperature, for example, is taken to be the middle of something like a bell curve of possibilities. Half the time it should be warmer and half the time cooler than the prediction.

A reasonable-worst-case forecast, on the other hand, is unlikely to be exceeded. Often, 10 percent is set as the chance of a more dangerous outcome.

The muddled messaging occurs when most-likely forecasts and worst-reasonable forecasts compete for the public’s attention. Hurricanes bring the biggest challenges. The public is bombarded with messages with multiple thresholds that convey different levels of risk.

Early Monday morning, Oct. 8, 2018, Hurricane Michael was organizing in the Caribbean. The National Hurricane Center issued a hurricane watch for the Florida Panhandle, including the Panama City area. The hurricane watch meant that hurricane conditions were possible in North Florida, and the dangerous weather could arrive in about two days.

Meanwhile, daily forecasts for Panama City for that following Wednesday unhelpfully predicted winds from the southeast at 35 to 40 mph, with gusts to around 55 mph in the afternoon. That was the day the hurricane was expected to hit.

In fact, both an extreme worst-reasonable-case forecast and a mundane most-likely forecast can be right at the same time — at a scientific and statistical level. But they can never both be right at a practical, understanding-the-risk level.

The tragedy here is that the meteorologists at the National Weather Service and at private weather offices well understood the extreme threat to the Panama City area. The National Hurricane Center diligently posted warnings of extreme winds and a storm surge up to 13 feet in parts of North Florida. These alerts were based on the Hurricane Center’s reasonable-worst-case analyses of what might happen.

Meanwhile, the computer systems that crank out specific wind forecasts every day on websites and weather apps continued their merry task of issuing most-likely-based public forecasts that masked the threat. This is a problem for the whole weather industry.

The only solution is to quit making most-likely forecasts, which intrinsically minimize the threat, when extreme weather is a reasonable possibility. Since preparations by the government, businesses and individuals should be based on the reasonable-worst-case projection, anything else confuses the issue.

We tend to blame the public for not paying close enough attention to all of the information available when a storm is threatening. But with the current system, paying too much attention can be hazardous to a clear understanding of the threat. People should not need to know what information they should ignore.

Bryan Norcross is a hurricane specialist at WPLG-TV, Miami, and the Weather Channel.

More from Bryan Norcross

If forecasters really want to reach their audience, they should consider the art of numbers

Facebook and Google have a responsibility during disasters

Hurricane forecasts are great — as long as the storms behave

It’s time to rethink our unintelligible weather terminology

The majesty and miracle of the 1969 moon landing ‘still boggles the mind’