Novel Approaches in Predicting Imminent Weather
By: Teja Balasubramanian
A new wave arises. Computer programming and artificial intelligence are at the vanguard of science and technology. Whether it is writing algorithms that shorten pages and pages of calculations, accelerating healthcare progress, or augmenting the boundaries of communications, artificial intelligence is exponentially revolutionizing human progress. One particularly fascinating subdivision in which one can see the application of computer science and artificial intelligence is in the works of research scientist Dr. Anthony Wimmers. Dr. Wimmers has been a researcher at the Space Science and Engineering Center/Cooperative Institute for Meteorological Satellite Studies since 2003. Two years ago, Dr. Wimmers began working on utilizing artificial intelligence to gain insights about imminent weather, especially hurricanes.
In describing his projects, Dr. Wimmers says that his focus lies in meteorology — “short term prediction” or “real-time estimation” of impending weather hazards. This apparatus of using artificial intelligence in order to monitor weather patterns is, unsurprisingly, quite new. The prior state of the art techniques involved algorithms that read off distinct properties of a hurricane such as temperature differences, size of structures, or asymmetry. They were predominantly more direct analytical techniques. The presiding method implemented a satellite to approximate the intensity of a hurricane. Furthermore, a set of rules and flowcharts known as the Dvorak techniques, initially developed in the 1960s and 70s, are widely used for these predictions. At the time, this approach was quite groundbreaking as it allowed for quantities, such as the intensity of a hurricane, to be estimated to a degree of precision.
“There’s a payoff from looking just outside your discipline to come up with a more creative approach to the same problems that are facing the people in your field.”
However, these techniques are finally waning as they do not capture the more objective but complex characteristics required to truly comprehend these natural disasters. On example of such an evaluation is analyzing certain structures from images of a hurricane. Therefore, Dr. Wimmers felt as if this area of study needed innovation. In his research, Dr. Wimmers uses a device known as a geostationary satellite. Its primary job is to provide real time information on the weather and to observe patterns that connect to interesting weather. The geostationary satellite, along with artificial intelligence, is relatively advanced and provides sophisticated alternatives to the dominant old-fashioned methods.
Artificial intelligence offers a way to integrate objective and subjective components. “Artificial intelligence finds holistic patterns that connect to a desired prediction, instead of taking a traditionally reductionist approach,” says Wimmers. “For example, it can detect the spiral band forming around a hurricane, which only gets part of the way toward analyzing the hurricane. Artificial intelligence quickly builds a model that tells it apart.” This line of research is incredibly useful in not only comprehending the conditions that lead to a hurricane, but consequently creating real-time estimations so necessary precautions can be taken.
“Wisconsin has a long legacy and a lot of breadth in the field of satellite meteorology,” says Wimmers. The Space Science and Engineering Center was founded over 50 years ago and researchers at the university have been engaged in using weather satellites from the very beginning. Dr. Wimmers’ interest in studying hurricanes did not begin until he started working on his PhD. “Hurricanes are the most expensive type of natural disaster caused by weather,” he says. Studying techniques and programs to detect them and understand them seemed to be a natural fit for him. Dr. Wimmers’ loves the “teamwork between the different research efforts here” and the “encouragement to explore and try new approaches.”
“Sometimes the best approaches are not the ones that are laid out in front of you,” says Wimmers. “For example, when I started teaching myself about machine learning, I had just read things in the popular press. I thought it had a lot of potential. It wasn’t something that everyone was pointing me toward. There’s a payoff from looking just outside your discipline to come up with a more creative approach to the same problems that are facing the people in your field.”
“The reason that artificial intelligence is back in vogue is that there is an innovation in a field called ‘deep learning’,” he states. Deep learning is a knowledge subset of artificial intelligence in which predictive models are developed automatically by relating complicated and abstract patterns within the data. This is exactly how Dr. Wimmers is currently studying the distinctive evaluations of hurricanes. In this day and age, the implementation of technology for gathering knowledge is ubiquitous. Applying artificial intelligence could mean original and compelling results. Dr. Wimmers has expressed that he would be interested in working with students to research applications of his work to deep neural learning in the future.
