When you plan laundry, a beach trip, or a BBQ this week, the weather app is probably your first go-to. Yet, satisfaction with its accuracy often falls short. This leads us to ponder: why are weather apps so unreliable?

Even professionals like Rob Thompson from UK Reading University share this frustration. He recently experienced a night of dryness that unexpectedly turned into morning rain, illustrating a common concern. Typically, our complaints center around unforeseen rain or snow.

Our expectations of weather apps and actual weather conditions significantly contribute to this issue. However, this isn’t the sole complication. The complexity of weather systems combined with the vast amount of data required for local forecasts makes accurate predictions extremely challenging.

Thompson acknowledges that some apps have struggled with accuracy in the UK lately. This is partly due to the unpredictable nature of summer rainfall, he explains. Convection rain happens when sunlight warms the ground, causing hot, moist air to rise, cool, condense, and form isolated showers. This differs greatly from the large-scale weather fronts influenced by pressure changes that dominate other seasons.

“Imagine boiling water. You can estimate how long it will take to boil, but predicting where the bubbles will form is impossible,” Thompson states.

A similar phenomenon occurs in North America and continental Europe. However, weather forecasting tends to be a localized endeavor, so let’s concentrate on the UK to better understand why pinpointing the exact timing of weather events is so difficult.

In general, forecasting for specific towns or villages can imply an unrealistic degree of precision.

“I’m in my mid-forties. In my career, there’s no way to predict shower clouds to the extent that rain hits my village of Sinfield while missing Woodley just three miles away,” says Thompson. Apps might claim to forecast two weeks ahead, but he finds that incredibly optimistic.

The two-week forecasting limit has long been established, and accuracy tends to diminish beyond that. Some researchers are using AI and physical models to extend predictions over a month, but managing vast global data while refining local forecasts remains a challenge for weather apps.

Though Thompson utilizes weather apps, he feels nostalgic for an era when TV forecasts provided context. Meteorologists had the time and tools to explain weather fronts, detailing the certainty of rain between specific times, along with the likelihood of showers within those windows. Such nuances are crucial. In contrast, a weather app may indicate a 50% chance of rain at 2 PM and 3 PM, losing subtleties that can lead to frustration even when the data is accurate.

If you inquire about the weather in Lewisham at 4 PM and are informed of heavy rain that doesn’t materialize, it may seem like an error. Yet, wider forecasts could highlight missed opportunities due to unpredictable fronts. These predictions come with margins of error, not outright failures.

One truth is clear: app developers are often reluctant to address these challenges, choosing instead to maintain the facade of absolute accuracy. Both Google and AccuWeather did not respond to New Scientist, while Apple declined to comment but requested an interview. The Met Office similarly chose not to engage but stated, “We are constantly exploring ways to enhance our app’s forecasts and provide more weather insights.”

The BBC also refrained from comment but noted that over 12 million users appreciate the Weather app’s straightforward interface, highlighting the extensive thought and user-testing that informs its design to balance complex information with user comprehension.

Striking this balance is challenging. Even when data is flawless, simplifying information leads to the inevitable loss of detail. Many weather conditions are condensed into a few symbols, each carrying subjective meaning. For instance, at what point do clouds replace the sun symbol with white or gray clouds?

“If you and I formulate an answer and then ask our mothers for their interpretations, we might not get the same response,” Thompson explains. This compromise opens the door for ambiguity and disappointment.

Other issues persist as well. Some predictors intentionally introduce a bias, making apps slightly pessimistic about rain probability. In his research, Thompson identified a “wet bias” across multiple apps. Users who experience shining sun often find that more frustrating than those caught in an unexpected shower. As a gardener, this often aggravates me.

Meteorologist Doug Parker from the University of Leeds emphasizes that many apps save on costs by leveraging freely available global forecast data rather than fine-tuning region-specific models.

For instance, some obtain data from the US government’s National Oceanic and Atmospheric Administration (NOAA). Raw global data may suffice for predicting large cyclones but falls short when considering localized rain forecasts, like at Hyde Park during lunch on a Monday.

Parker notes that when estimating the likelihood of flash floods in Africa—often a matter of life or death—some apps simply lack necessary data. He mentions several free forecast products with questionable reliability regarding Kenyan rainfall radar, stating, “It’s misleading since Kenya lacks comprehensive rainfall radar.” While satellite radars occasionally pass overhead, they don’t provide full data coverage. Without knowing the origin or reliability of these forecasts, users face significant uncertainty.

In contrast, the Met Office’s app utilizes refined models and rigorous post-processing to enhance UK weather predictions, drawing from the organization’s substantial human expertise. The app team crafts a distilled yet accurate representation of weather data through a thorough process.

“Presenting model data is a vast area of focus at the Met Office. They have a dedicated team for it,” Thompson remarks. “It’s essentially its discipline.”

Creating a weather forecast model involves integrating a huge volume of real sensor data and executing it on a supercomputer, a demanding task. Yet, this extensive work corresponds to realities we may not fully perceive. Current forecasts are better than ever and continue to improve. Our ability to predict weather today’s standards was unimaginable just decades ago.

Much of the frustration we experience with weather apps originates from misalignments in expectations regarding accuracy at a local level, oversimplified data representations, and the rising demands of a busy populace that often overlooks scientific nuances.

Parker notes that as meteorological capabilities have advanced over the decades, public expectations have swiftly adjusted, leading to an ever-increasing demand for accuracy. “Will people ever be satisfied?” he questions. “I doubt it.”