Why We Abandoned Satellite Imagery for an On-Device Procedural Sky

During the early stages of the Shipaton build, the author dedicated time to creating a weather satellite pipeline. However, upon reviewing the generated images, it became evident that the pipeline should be removed from the app's core. The primary focus of Amana, an iOS app, is to display a high-quality sky image that resembles the actual sky outside a user's window.

The author initially considered using the Himawari satellite, a geostationary weather satellite that captures the entire Earth disk every ten minutes. The data is publicly available, and the author's initial pipeline pulled raw bands from NOAA's AWS Open Data mirror, composited a true-color image, cropped it, and cached the result. While the Earth appeared round and blue, it was not suitable for the app's purpose.

The key issue was that satellites like Himawari view the Earth from above, capturing cloud tops and the ground. In contrast, the app aims to display the sky as seen from the user's perspective, with blue skies resulting from the atmosphere scattering sunlight toward the ground. The distinction becomes evident when observing clear days from both an orbital and ground-level perspective.

The author realized that the technical success of the pipeline did not guarantee the visual appropriateness for the app's intended outcome. This led to rethinking the approach, ultimately deciding to render the sky on the user's device using a physically-based procedural sky. Two main methods were considered: the MDLSkyCubeTexture from Apple's Model I/O and Hillaire 2020's LUT-based atmosphere model, which powers Unreal's SkyAtmosphere.

Both options required providing the user's current time, location, and solar elevation to generate accurate blue-to-amber transitions.

However, the author also acknowledged two important traps in this approach. First, SceneKit, a popular sky-rendering tool, was soft-deprecated at WWDC25, making it an unsuitable choice for new development. Second, the atmosphere models are clear-sky models, meaning they produce beautiful, cloudless skies regardless of actual weather conditions.

To address this, the final implementation combines a procedural atmosphere with cloud cover data from WeatherKit, creating a composite sky that better represents the user's actual surroundings.

In the end, the satellite pipeline, while initially deemed unsuccessful, remains part of the app as a contrast view. It showcases the weather conditions beneath the user's local sky, offering a unique perspective. Both the procedural sky and the satellite imagery are driven by the same clock and weather data, ensuring consistency between the views.

The author credits the Japan Meteorological Agency for the satellite imagery used in the app, emphasizing the importance of checking licensing terms before incorporating external data into a commercial product.