Dark Sky did not explain themselves with the barometer data thing. Their actual algorithm did radar-based extrapolation, meaning they used statistics to 'move' radar features like rain into the future. This is why Dark Sky always had a problem with forecasting new rainfall and was better at forecasting existing rain moving into your location based on wind speed.
They claimed something about barometer data but I don't honestly believe that they used it to improve their forecasts. I worked on this problem at different companies for many years, and with academia, and found it to be quite difficult. Possible, doable, but very difficult to demonstrate actual improvement. Significant academic work was done after Dark Sky made their claim, and the papers published used ensemble kalman filters among other on-device data-cleaning techniques that I do not believe Dark Sky implemented. So I doubt their barometer claims.
They claimed something about barometer data but I don't honestly believe that they used it to improve their forecasts. I worked on this problem at different companies for many years, and with academia, and found it to be quite difficult. Possible, doable, but very difficult to demonstrate actual improvement. Significant academic work was done after Dark Sky made their claim, and the papers published used ensemble kalman filters among other on-device data-cleaning techniques that I do not believe Dark Sky implemented. So I doubt their barometer claims.