New model and features
30.01.2026
After months of work, the new version of the Sunsethue model has launched into a testing phase. With a higher resolution and more metrics, the goal is to improve the accuracy of the forecasts and provide more detailed information. Give it a try by clicking on one of the sunrises or sunsets below.
An update to the ray-tracing algorithm
The new model includes a whole rewrite of the code to trace rays through the atmosphere to obtain the quality metric. It is now more accurate, more extensible and more performant. This added performance means that I was able to increase the resolution of the model, and add a whole bunch of new insightful metrics.
Horizon clouds
The horizon clouds metric tells you how many clouds there are expected to linger at the horizon during sunrise or sunset in the direction of sunrise or sunset. This will help you to determine if the sun will be visible right after sunrise or right before sunset. Beautiful sunrises and sunsets can occur with a covered horizon, but seeing the sun go down may be more mesmerizing for some.
Uncertainty and quality potential
A couple years back, Sunsethue had an uncertainty metric which tried to give more insight into how likely the forecast is to be right. A high uncertainty means that the forecast may change drastically. The quality potential metric will help you understand what the upper bound of the quality is. The potential of a sunrise or sunset is the quality of said sunrise or sunset in case there would be no clouds that block it.
Cloud-layer specific forecasts
For many photographers it is important to know which clouds will catch light during sunrise or sunset. This is already taken into account in the quality calculations, but it is now also split up per cloud layer. You will now be able to see how many low, middle and high clouds are forecasted in the direction of sunrise or sunset. Besides this, the quality metric is also split up into these layers, to better understand which clouds are likely to catch light.
Challenges and the future
One big challenge has always been the accuracy of the weather data that is used as input for the model. The model might be getting better and better, but if the input data is wrong, the prediction is likely to be too. In the future it might be good to use better (regional) weather models as input. That’s why the new code structure has the ability to add other models built in. Currently the new model will remain in a testing phase for some time. You can help evaluate the model by rating past sunrises or sunsets, or by leaving feedback through the feedback box on the sunrise or sunset page. I’m also always more than happy through a mail or a quick call! During the testing phase will be available for free for everyone during the testing phase. In the future, some of these features might be moved to a paid “expert” subscription, to help fund the increased costs of this new model. The quality and cloud cover metrics will always remain free, and the new model will eventually replace the current forecasts (on the website, in alerts, and through the API) after the evaluation phase.
What do you think about this?
Give your opinion, or voice your concerns below.
© 2026 - Sunsethue (v0.10.2-beta)
-
Made in Europe
