Thanks to virtual assistants on your smartphone, you can make weather-appropriate wardrobe choices before even leaving your bed. So convenient, yet so annoying when that promised sunshine dissolves into a rain shower 10 minutes after leaving the house.
Weather forecasting data comes from different sources. NASA’s Earth-sensing mission AQUA and other satellite-based sensors provide water-related data on ocean currents, sea ice and water vapor. Unmanned reconnaissance vehicles collect weather system information at altitude.
Scientific American describes how phased-array and dual-polarization beam radar scan through slices of atmosphere and work in tandem with Doppler for rain cloud statistics. There’s a lot of data to mine and specialized systems deliver weather intelligence to many different users.
Science Alert shows that it is still difficult to predict the weather with accuracy. This is partly due to the huge data sets generated, but National Geographic also notes that weather is more than just local or isolated; changes in one area spread around the globe to impact weather in other areas. These ripples, called “the butterfly effect” after the notion that a butterfly flapping its wings on one side of the world can create weather patterns in another, are the reason why it is so difficult to predict the weather in advance.
Curiosity describes how MIT meteorologist Edward Lorenz coined the phrase “the butterfly effect” when researching chaos theory and weather forecasting technology. Chaos theory is an interdisciplinary mathematical theory that states that within the apparent randomness of complex systems, there are underlying patterns and feedback loops. Lonez found that rounding up only one of 12 data variables in a computer simulation to three decimal places had a profound effect. With such a tiny change, he completely altered two months of weather modeling.
According to the Fractal Foundation, chaos theory plays a part in unpredictable weather. Since weather forecasting relies on analysis and pattern recognition on data gathered from radar, satellite and remote sensing, chaos makes a big impact.
Supercomputing is already helping to process data faster. The Weather Network explains that in 1989, processing power could perform 2.4 billion operations per second; in 2016, the U.S. National Weather Service modeling took place on computers capable of running 5.78 quadrillion operations per second.
Researchers are using machine learning to optimize processing power for unpredictability and large data sets. Weather prediction using machine learning helps refine algorithms for data analysis and pattern recognition for accurate, long-term forecasts. It has successfully been used to spot tropical cyclones, atmospheric rivers and other weather events, says Nature. Meanwhile, Interesting Engineering predicts that training computers with existing data sets will help meteorologists use weather forecasting technology more efficiently.
For example, Science Daily reports that Penn State researchers trained computers to automatically recognize the type of cloud formations seen in satellite images that predict severe weather events. Since this pattern identification happens faster than a human could observe, the software can nudge meteorologists to pay attention and issue warnings much earlier.
University of Maryland researchers used a reservoir computing approach for machine learning to improve accuracy. The reservoir computing technique essentially “learns” the dynamics of a chaotic system. The Weather Network describes how they trained their programs to learn the chaotic nature of weather systems. Through repeated analysis and training, the program was able to push the prediction accuracy further than previous attempts.
Soon we’ll be able to leave the umbrella at the front door with confidence.
Since its inception, The National Digest has been dedicated to providing authoritative and thought-provoking insights into trending topics and the latest happenings.