Share this article

print logo


I'll take a grain of salt with that weather app, please

Don Paul

I touched on the topic of weather apps quite some time ago here, but my colleague Bryan Norcross inspired me to come back to the often-undue high expectations placed on weather apps in a Washington Post column he recently wrote. (Norcross has been a fixture in Miami TV weather as well as on the Weather Channel for decades. At my age, I know decades.)

Weather apps can be a welcome addition to the arsenal of tools meteorologists use to communicate with the public. Many break down a forecast into hourly periods, with widely varying amounts of detail and accuracy. At the same time, they are automated, running off computer models with little if any human input. If you’ve never seen an hourly forecast — and so as not to focus on any TV station weather app — take a look at this typical breakdown from the Buffalo National Weather Service Forecast Office.

That is among the very best examples of an hourly forecast graph. It is based on multiple high resolution models running on supercomputers, and it is overseen by meteorologists at the office who can alter what they may view as an error in the making, as time allows. On a day such as this coming Sunday, the forecast process is less complicated and more likely to verify, even on an hourly basis.

Now, let’s look at this hourly forecast graph from the Tulsa National Weather Service Forecast Office for this midweek period on a day where there is a low risk for severe thunderstorms.

Even though this graph is based on the best available model input, this period in Tulsa involves greater uncertainties because of the nature of small-scale convective cells. For example, thunderstorms produce outflow gusts on the ground, which you often feel just ahead of the storm. These outflow gusts can create a boundary, which then moves away from the storm, lifting the moist and unstable air ahead of the weakening original storm and often generating new storms.

Models may be able to indicate conditions in which outflow boundaries could organize, but they cannot offer definition, location, direction of movement, and intensities for such potential storms. By the very nature of scattered convection in a moist, unstable air mass, there are uncertainties that can be poorly handled by the best hourly graphs and weather apps, even from the best sources.

This kind of hourly weather information clearly may be useful to members of the public, event planners, utilities, umpires and coaches, so long as their limitations are kept in mind.

These hourly forecast graphs have their flaws, as do human-massaged text forecasts, but they are not identical to the sea of weather apps that are freely available from the private sector. Here is the opening salvo from Bryan Norcross: “The rabbit hole begins with the app. Open any weather app, and you find a stew of good science and wishful thinking. We would all like modern science to tell us exactly when it is going to rain this afternoon and what the weather is going to be a week from now. In the end, many of those forecasts are pseudoscientific.”

There are graphic “slots” in these apps that are automatically filled. Not all of such slots are created equal in terms of certainty. In the private sector, such as local TV, staff meteorologists and weathercasters often lack the time to constantly tweak these automated forecasts in the apps offered by their stations. At some stations, the weather staff tries to adjust these forecasts a couple of times during a shift, particularly when major weather events are approaching. Yet the problem of automation from out-of-town computers that churn out the graphics and numbers can easily get beyond the time management of the one or two meteorologists working to prepare their on-air forecasts, weathercast graphics and social media output.

The problem with automated forecast apps is, as Norcross points out, computers have no sense of shame in damaging their credibility. They will fill those slots come hail or high water (oy!). Those of us who make our own forecasts and don’t rely on the apps most certainly have a sense of shame and feel a responsibility to make the most accurate forecast we can, as well as express the uncertainties as necessary.

Even during our seven-day forecasts, there is a graphical presentation that, if you’re not listening, implies higher certainty than is scientifically realistic. There are even situations in which confidence and certainty is actually higher further out in the seven-day than it is in the near term.

There are also some other weather apps we meteorologists refer to as “crap apps” — pardon the vulgarity. Take my built-in smartphone app, please. It's worthless.

But that could be a topic for another column. Suffice it to say, use them with caution.

I can’t improve upon Norcross’ quote in the Washington Post: “On television, a good weathercaster can blunt the misleading simplicity of the final forecast graphic with personality and explanations. Most apps, however, have no such opportunity or compunction about cranking out definitive forecasts in highly uncertain meteorological situations.”

• • •

Speaking of forecasts, here's a look ahead to weather for the Game of Games on Sunday, not based on app: The forecast continues to look favorable.

A ridge of high pressure over eastern Canada will bring dry conditions with abundant sunshine, meaning a high UV index even though it’s late September. The day will get off to a seasonably cool start, with readings in the low 50s early in the day rising to the upper 60s in the afternoon. Unlike last Sunday, a fairly light east-northeast breeze will not be a major factor in the kicking and passing game. The trailing warm front west of the ridge will not bring its threat of scattered showers until Monday.

There are no comments - be the first to comment