Forecasting beauty: Student, alums make model to predict vivid sunsets, sunrises

By Liam Jackson
December 15, 2015

UNIVERSITY PARK, Pa. — As a student photographer working for Penn State, Jacob DeFlitch saw firsthand how difficult it could be to find a good sunset or sunrise.

“I would regularly take photos of buildings and sites around campus like Beaver Stadium, Old Main and the Nittany Lion Shrine. I would always do my best to get out for sunset photos, but it was sometimes frustrating when it was too cloudy or too clear, and there was no sunset,” said DeFlitch, a native of North Huntingdon.

DeFlitch, who graduated in May 2015 with a bachelor of science in meteorology and now works as a meteorologist for AccuWeather, had the idea to create a model that would predict just how vivid a sunset would be — much like models used to predict precipitation or temperature. He partnered with two friends — Ben Reppert, who also graduated with a bachelor of science in meteorology in 2015 and is now a research assistant with the Department of Meteorology, and Stephen Hallett, a sophomore majoring in meteorology with a knack for computer programming — to make that model a reality.

After collaborating for months, the trio debuted their new product, SunsetWx, on Nov. 18. It’s a free online tool that will help photographers and anyone else looking for a nice-looking sunrise or sunset.

What makes a beautiful sunrise or sunset

Sunrises and sunsets are complicated phenomena, and the sky has to have the right balance of meteorological factors to result in bright purples, reds and pinks instead of a dull blue.

“You can have a lot of high clouds, but if you have low clouds, that just takes away from the quality of the sunset. It also matters how thick your high clouds are. If you have a continuous deck of clouds, no light is going to get through. But if you have thin clouds like cirrus wisps, that’s one thing that goes into making a good sunset,” said Hallett, who is from Delaware County, outside of Philadelphia.

To figure out which of the hundreds of possible meteorological data were relevant factors to predict sunsets, the trio tapped into their knowledge of meteorology gained through courses at Penn State.

“Jake and I took an upper-level course on atmospheric radiation during our junior year, and the concepts we learned there had a lot of applicability to this project,” said Reppert, who hails from Bethlehem. “We also took a synoptic meteorology course that looks at the big picture of forecasting, which we needed to do to predict sunsets. When you have a strong cold front clearing half the country, there’s a good chance it will result in a nice sunset for a lot of people.”

They identified more than 20 relevant factors in all, with moisture, pressure and cloud cover serving as the core of the algorithm on which SunsetWx runs.

Once they knew the important factors, they had to create a way to collect the data in real time and manipulate it into a user-friendly format. This is where Hallett’s experience with computer programming came into play. A few years earlier, Hallett began learning a computer language called Grid Analysis and Display System (GrADS), which provides a way to create visualizations, such as maps, of meteorological data.

“When I first saw an image created with GrADS, it created a visualization that I didn’t know how to create on my own, and it was very visually appealing. It let me quickly see what was going on in the atmosphere, so I thought, ‘Okay, let’s see if I can recreate that,’” he said.

Hallett emailed Ryan Maue, who works on weather models with the company WeatherBell, and Maue showed Hallett how to get data and convert it into a format that GrADS could read. After learning and practicing advanced GrADS techniques, Hallett developed an algorithm that would pull in data collected by the National Oceanic and Atmospheric Administration (NOAA).

Next, the team had to create a website that would let users easily view the sunrise and sunset prediction data in real time. To do this, DeFlitch, Reppert and Hallett contacted one of Hallett’s friends, Justin Lowery, a graphic and Web design student at the Art Institute of Charlotte, who was “heavily experienced with the Linux operating system,” said Hallett. Lowery helped manipulate data through the SunsetWx website servers so that the website could display real-time and location-specific information.

The SunsetWx website contains maps that show how vivid a sunset or sunrise should be across the continental U.S., Canada and Mexico. The system outputs color-based qualitative data known as “units of vividness” rather than numerical data — by looking for warm colors on the map, users can identify hotspots where the right meteorological conditions for a vivid sunset will likely occur.

“It’s our hope that this tool will help people get out and enjoy beauty of nature around us,” said DeFlitch. “There are many useful and important weather models focused on storms and other potentially unsettling weather phenomena. We wanted to create something that the public could use on an everyday basis and enjoy.”

Bright beginnings

The model has garnered much national media attention in its first few weeks of existence, including being featured on Slate and Good Morning America. But, the team noted, it is a work in progress that occasionally outputs incorrect results if it receives incorrect data. However, it is right the majority of the time, and the team continues to meet regularly to adjust its algorithms and make tweaks to improve their model and website.

As they continue refining their work, Reppert, Lowery, Hallett and DeFlitch have all experienced a sense of fulfillment as they applied their knowledge to make a useful tool for the general public.

“People can easily find forecasts for the next day’s weather, but something like SunsetWx hasn’t existed before. We’ve been able to combine the knowledge of how meteorology and the weather work to figure out what makes a good sunset. Being able to give that product to people is something we can be proud of,” said Reppert.

SunsetWx sunrise visualization model

Example of one of the sunrise prediction maps from the SunsetWx website. Warm colors correspond to the likeliness that people can see a vivid sunrise in a given area.

map visualization of the sunset vividness prediction

Example of a sunset prediction generated by the SunsetWx algorithm and website. Warm colors correspond with likeliness of vivid sunsets in a given area. 



(Media Contacts)

Last Updated January 19, 2016