These models are millions of lines of computer codes and they've been tweaking these models for the past ten years or so. With each tweak they like to try to get a half percent or one percent reduction in error. So the ingestion of this new data has made a very significant improvement in reduction of error.

Initial research shows the airborne sensor makes a 10 to 20 percent improvement in forecast error in numerical models and that's just with temperature.