Special Edition: Inside Skylark’s Atmospheric Model

Positioning failures do not happen at the median of your error distribution. They happen at the tail, in the moments when the atmosphere stops cooperating. At scale, the tail stops being a statistical edge case and becomes something every customer encounters during the product’s lifetime. That's why it's critical to evaluate how your corrections provider handles the 99th percentile conditions.

What Happens Between Reference Stations

Atmospheric disturbances can cause sharp disruptions in the GNSS signal that change on the scale of kilometers and minutes. Unless your network RTK provider has a station every kilometer (they don’t), they need a solution to estimate what is happening between stations.

Most RTK services rely on geometric interpolation, which assumes the atmosphere changes smoothly between any two stations. With a dense network and typical conditions, this approach works well enough. But during a traveling ionospheric disturbance or a geomagnetic storm, geometric interpolation will fail.

Corrections providers that rely on geometric interpolation have no way of improving their performance beyond adding more stations. This brute force approach has diminishing returns, and it forces density-first providers to cut corners on station quality, siting, and maintenance to keep scaling.

The Solution: Atmospheric Modeling
Swift’s Skylark^TM Precise Positioning Service leverages a sophisticated atmospheric model instead of geometric interpolation. The model represents the atmosphere as one continuous field across the coverage area, abstracting the underlying network of stations. From that, it can predict the atmospheric state accurately at any point between stations to absorb roughly 95% of storm-induced delays and track disturbances as they propagate across continents.

Read the full blog post for:

• An analysis of the limitations of geometric interpolation

• A look under the hood of Skylark’s atmospheric model

• Additional benefits of combining a dense, high quality network with atmospheric modeling

• What to look for in a corrections provider

For real-time updates, follow us on LinkedIn

View The GNSS Fix Archive