No, There Is Not An App For That
The portability and versatility of smartphones has undoubtedly revolutionized our society. We have access to so much information in real time within the palms of our hands. We also have access to a library of millions of apps, including useful tools, sources of entertainment, and silly fads. Some of these apps include weather forecast apps, and I believe too many people are relying too heavily on these apps, and it's getting them into trouble.
Before I go any further, I should clarify what exactly is meant by "weather app", because this is a very broad term that could refer to many different things. Specifically, I will be discussing weather forecasting apps, which are standard on almost every smartphone. I'm not talking about third party apps (such as the ones published by local news affiliates). In fact, many third party apps are very helpful, because they relay critical weather information to the end user that might otherwise be missed. This also does not refer to Wireless Emergency Alerts (WEAs). These are high priority and life-saving alerts that every phone should receive (if you've disabled all WEAs on your phone, you may come to regret that decision someday for that might be your only way of knowing a tornado, or other hazardous weather, is headed your way).
So, what is it about these generic weather apps that is so bad? For one thing, the forecasts they produce are derived from weather model output. This means that you're listening to a machine that cannot think like a sentient human can. A human forecaster can better account for uncertainty and look at the larger scale picture, such as the underlying physical processes that influence the temperature, the moisture content, the precipitation chances, the winds, and the like. The automated (machine-produced) forecast is directly rooted in statistics; i.e. it is not directly rooted in the physical processes. Sure, the numerical model outputs are based in physics, but those outputs should never be taken as definitive truth because those are just a machine's best guess of what the future will be based on the numbers that were plugged into an algorithm.
That said, these models are pretty capable of accurately predicting things like temperature and wind up to 6 hours in the future. Sometimes their performance remains decent deeper into the future. However, once you get past 48 hours into the future, even the most intricate models quickly lose their accuracy. So, anything that takes these model outputs at face value is not likely to be accurate either.
I remember an uncle of mine was planning to take a trip to South America and wanted to know what the weather was going to be like. This trip was planned for January, which is the southern hemisphere's summer. The forecast apps he was using were telling him radically different things with temperature forecasts ranging from 20 F to 60 F. Part of this was the fact he was looking for a forecast 14+ days in advance, and anyone who has ever looked at raw model output knows anything depicted more than 10 days can be absolutely ludicrous and nonsensical. So, I put my meteorology background to good use. I recognized this was the summertime, and weather patterns experience very little change during the summer. Usually, the weather you see in June will be what you see in July and in August (if you're in the Northern Hemisphere). All the high temperatures will be roughly the same and all of the low temperatures will be roughly the same. If only the automated weather apps knew this... Then they could have forecasted something reasonable. As it turns out, my own forecast was more helpful and also more accurate, because I am a human and humans can think. Machines cannot. This reminds me of Professor Falken's quote from the 1983 War Games film:
"General, you are listening to a machine. Do the world a favor and don't act like one."
Alright, so missing the temperature by 10 degrees doesn't seem like a huge deal. A smartphone user checks the app everyday, so they'll always get that more accurate 6 hour time window for their temperatures, right? Well, actually no.
Usually, the only items of interest are the high temperature, the low temperature, and the chance of precipitation. What is the high temperature? It is the maximum temperature recorded in a 24-hour timeframe. Normally, this occurs in the afternoon. Although, you can have days where the high temperature is recorded in the morning or even in the middle of the night. This is especially common during the fall, winter, and spring months. I remember one November, the app on my own device said the high temperature was going to be in the mid 70s and the low in the mid 40s. What my device didn't tell me was that high temperature would occur in the morning and then a massive cold front would sweep through by noon. And, as it turns out, a lot of people around my college campus went out in shorts that morning only to be surprised by temperatures in the lower 40s when the afternoon came around (plus a 30-40 mph northerly wind).
The chance of precipitation is another thing these forecast apps don't handle well. What these apps look at is the Quantitative Precipitation Forecast (QPF) produced by models and then apply statistical smoothing techniques to infer a probability of precipitation. What happens if the weather models never initiate any precipitation but show a pattern favorable for precipitation? In which case, the weather app will say the chance of precipitation is 0%, but that may not be true. More often than not, the reverse occurs; the models over produce precipitation, meaning the forecast apps often report precipitation probabilities that are exaggerated.
In the grand scheme of things, complaints about incorrect temperatures, misinterpreted forecasts, and inconsistent precipitation predictions aren't that big of a deal. What about things that are more impactful like severe weather episodes?
Some of the apps I've personally surveyed are actually pretty terrible at giving you sufficient advance notice about potentially hazardous weather, unless it is already occurring or about to occur (i.e. if you've already been placed under a watch or warning). I remember visiting a family friend up in Minnesota a while back and I was monitoring the potential hazardous weather. By checking the Storm Prediction Center's website every day, I could tell this friend there was potential, but they insisted on using their apps, which didn't even mention the possibility for severe weather until the morning of the threat. And, what did the app say? "Some storms may be severe." If presented with such a statement, there are a number of questions that should be asked. How severe will the storms be? What is the tornado risk? The straight-line wind risk? The hail risk? The timing? Will these storms produce flooding downpours? Will there be multiple rounds of storms? Is there potential for a high-end event (e.g. a tornado outbreak or derecho)? How confident are the forecasters in severe weather occurring? etc.
I remember when my location was placed under a Moderate Risk (Level 4 of 5) for severe storms one October day (yes, in October). There was a significant tornado risk, along with a substantial risk for hail and straight-line winds. I decided to ask the voice assistant on my phone if there was potential for tornadoes today. Guess what the answer was? "No, there's no tornado in the forecast." I repeated the same question for hail and wind, and got the same answer. Guess what happened later that day? The city I lived in got ravaged with a major hailstorm and the university I worked at nearly got hit by a tornado (the storm produced a tornado on the other side of town). Are these really the apps we want to people to entrust their lives with?
The other unfortunate reality (as of this writing) is that people are drifting away from local news sources and relying more on national outlets like The Weather Channel and AccuWeather. The local outlets will provide you with better information than a national source, and they will also provide you with better forecasts for your area. These national outlets don't have the manpower required to issue detailed forecasts for every geographic point in the United States (at least, not to the same quality and detail as the National Weather Service). Also, the apps published by local outlets tend to be informative when it comes to hazardous weather. Local TV meteorologists will often discuss a potentially hazardous weather event 3+ days in advance if confidence in such an event is high enough, which gets the attention of people and tells them they should be monitoring the situation more carefully. This heightened awareness could be the difference between surviving a catastrophe and being caught completely off guard. Being caught unprepared is how many fatalities and injuries occur.
I once directed a high school friend of mine to her local National Weather Service and said their forecasts were way better than the forecasts on her phone. That's because the National Weather Service has large teams of forecasters operating in local offices that produce forecasts (and alerts) 24 hours a day, 7 days a week, 365 days a year (or 366 days a year). These forecasts are free, and they usually provide enough information for the average person. Of course, that then raises the question, why don't these forecast apps just reproduce what the National Weather Service says? The answer is simple: money. These app developers and national media outlets want something unique so they can claim their service is proprietary and justify charging you money. Some apps may draw from a forecast produced by a team of humans, but those that don't just use information in a way that it wasn't intended to be used for.
Forecast model output is a tool, and tools only work well when used properly. These automated tools will get things right on occasion, and probably will improve with time, but (barring the invention of sentient AI) they will still lack important characteristics found in a well-trained and seasoned meteorologist.
Anyway, long story short, the next time you look at an app on a device, make sure you know what you're really looking at. You're looking at what a mindless machine asserts what the weather will be, and you're trusting that app to determine if and when the weather will turn dangerous (or even deadly). If you use these apps, just remember, you're listening to a machine.