An old joke claims that meteorology is the only profession where you can be wrong half of the time and still get paid.
It’s a true and sometimes frustrating reality that weather forecasts are not always 100 percent accurate. Sometimes this is due to a weather system that changed rapidly or unexpectedly. But incorrect predictions can also result from errors, inefficiencies, or lack of quality data in weather forecasting models.
Using computer models to simulate and predict the weather, known as Numerical Weather Prediction, is still not a perfect science, but recent advancements in computing technologies combined with the growing availability of weather-related data has served to dramatically improve the accuracy of forecasts. Today’s researchers are increasingly investing in powerful supercomputing systems in an effort to quickly uncover patterns in historical data, and merge those with current observations to predict what might happen in the future.
Meteorologists have realized that taking a data-centric approach to weather prediction can be more impactful. A staggering amount of data goes into every weather prediction, including historical observations as well as current factors like barometric pressure, temperature, wind speed, dew points, and more. Researchers are leveraging cutting-edge computing tools to rapidly analyze all of this data, and artificial intelligence (AI) has begun to become part of their process.
Deep learning, a branch of AI, is one technique that is showing promise in the field of weather prediction. Deep learning allows researchers to process, analyze and enact on extremely large data sets by leveraging a series of trained algorithms that can learn and make predictions based on past data. Deep learning techniques have already been proven successful in areas like image and speech recognition and natural language processing, but now they are making their way into the weather and climate field as well.
There are some challenges associated with using deep learning algorithms for weather prediction, and these techniques are still in a stage of early adoption for this field. Weather patterns are comprised of a complex number of data points, making weather prediction a highly data- and compute-intensive exercise. Also, deep learning algorithms are only as effective as the inputs they are trained on, making data labeling a crucial component of this technique.
However, deep learning is increasingly taking hold across a variety of industries, and is offering researchers new ways of predicting the future. For example, a paper featured in the latest edition of the book “Hybrid Artificial Intelligent Systems” found that an architecture based on deep learning demonstrated an improved ability to predict the accumulated daily precipitation for the next day. Using autoencoders to capture the non-linear relationships between attributes and multilayer perceptron for the prediction task, the architecture was able to forecast the daily accumulated rainfall in a specific meteorological station in Colombia, and outperform other approaches that were historically used for this task.
This is a very positive development for the weather and climate industry, because increasing the accuracy of weather forecasting has far-reaching effects on many aspects of human life and business. For example, agricultural companies can better determine whether a crop will produce, or a utility company can plan ahead for repairs in an area they know will be effected by a storm. With more accurate predictions to rely on, agencies responsible for disaster prevention can make better decisions and react more quickly to weather events to keep people and property safe.
Weather forecasting has been a persistent challenge since the dawn of civilization, but thanks to high performance computing tools, forecasts are getting better and more reliable than ever before. Using cutting-edge techniques like deep learning, researchers are uncovering new ways to streamline weather prediction, helping to drive better decision-making and improve safety across all aspects of human life.
For more insights into the ground-breaking computing technologies that are driving transformation in the weather and climate industry, please follow me on Twitter at @Bill_Mannel.