John Metcalfe is CityLab’s Bay Area bureau chief, based in Oakland. His coverage focuses on climate change and the science of cities.
Weather geeks have built a highly accurate, predictive guide to daylight's best finishes.
Some folks are content to enjoy whatever sunset the day happens to provide. Others want to hunt down the very best sunset the weather promises and bask in its glorious hues of crimson, fire, and shining bronze.
Steve Hallett is one of the latter Sol-worshipers. “We all love a nice sunset,” he says, “but personally for me it helps me reflect on the events of the day and serves as a divider between the busier and calmer times of the day.”
So a little while ago, Hallett teamed up with a couple buddies from Penn State’s meteorology department, Ben Reppert and Jacob DeFlitch, and helped build a map that predicts vivid sunsets. Hosted at WX Sunset (“WX” is short for “weather” in Morse code), the tool uses the regularly updated NAM model to forecast fantastic sunsets in warm colors like red and yellow. Here’s the model run from last evening:
The main things the model considers are moisture, air pressure, and cloud cover. WX Sunset lays out some of the technical details:
After many trial runs and verifications, we weighted moisture the most. Our model ingests data from the high resolution, 4 km NAM, allowing us to account for values at all levels of the troposphere. Because of this, we took into account all levels of the troposphere from the surface, to as high as 200 millibars, and weighted each level accordingly, with the upper-levels having the most weight. Closer to the surface we reduced the weight substantially for moisture as it restricts the sun from refracting copious amount of light. Pressure, as well as the change in pressure over time was the next highest weighted factor as it helps recognize areas where cloud cover may dominate as well as FROPA’s. Finally, we included general cloud cover, to better account for regions where it is overcast, and help the model display it that way.
In a case study of the model’s simulation for November 17, the creators boast it got things mostly right. Here’s the blooming sky-fest in western New York, for example:
And a simmering horizon-fire in North Carolina:
The heavens displayed a classic beauty in a north-south ribbon over Oklahoma:
I checked the model’s veracity two days later on the 19th, and photos posted on social media mostly bear it out. Here’s that evening’s run:
Gorgeous scenes from Virginia correlate with a narrow band of beauty the model shows in the southwest part of the state. From Lynchburg:
Darlington, South Carolina, got pummeled with an “epic” purple-haze sunset close to the same band:
However, that day’s run slightly whiffed on the incredible sunset that blasted Vail, Colorado, or the below one from Danville, Pennsylvania—although, to its credit, in both cases it did forecast “hot spots” nearby.
Reppert says the map can be influenced by bad info from the NAM. “There are some days when the model that we are pulling our weather data from (the 4 kilometer NAM) is having a ‘rough day’, which will then reflect in the sunset forecast in the form of a poor forecast,” he says. “The phrase ‘garbage in, garbage out’ comes to mind here, and for nights like that, there is nothing we can control.”
As they improve their algorithm’s accuracy with photos from the ground, the group hopes to develop a text or email service and app for finding killer sunsets—“something that the everyday user can use,” says Hallett, “as well as the serious photographer.”