About Prediction Accuracy

How accurate are predictions? Can Tracking and Recovery team members just drive en masse to the predicted touchdown and wait for the payloads to land? Definitively, the answer is NO!

Long Range Predictions

Using the READY forecast wind data center it is now possible to get predicted winds for as long as 16 days prior to a flight. I've found that the accuracy of these long range predictions is somewhat erratic.

Follow THIS LINK to a page where I saved 14 sequential days of predictions for a flight. You can scroll through them and see how the path of the prediction deviates from one day to the next.

The start of semi-accurate predictions usually occurs around 6 days prior to flight. And beginning 3.5 days prior to flight the accuracy really starts to improve.

The best way to reduce the reliability of a prediction is to have a major frontal passage scheduled to move through your flight area at the same time you plan to fly. At times the models will predict the front has already passed through your flight area. Other times the models will think it hasn't reached your area. The air flow on one side of the front is going to be substantially different from that on the other side, especially in the troposphere. So, if the predictions for your area include a nasty deep front being in the area on flight day, keep an eye on this factor. I usually warn folks that my predictions are more susceptible to error because of the front. This causes them to adopt a more skeptical confidence level about prediction accuracy and has them fanning out to cover a larger possible landing site.

The Evening Before Flight

As a rule, the predictions for the evening prior to flight are fairly accurate. The predicted vector from launch site to touchdown is usually within a several of degrees of the actual flight. The distance is usually fairly close too.

On flight day the difference in the prediction from the preceding evening is usually only a few miles. This isn't written in stone but is often the case.

Statistically Speaking

If you examine our records (Flight History) you will discover that predicted touchdown locations averaged 14.2 miles from the actual site of the landing. The standard deviation is 9.7 miles. That's a bit fairer as it gives far less weight to the anomalous run away balloons that overshot their predicted touchdowns by as much as 57.7 miles.

It is very reasonable to place a dot on the map where the prediction has the payload systems landing, and draw a circle 20 miles in diameter around this spot. It is highly probable that the payloads will arrive within the bounds of this circle.

See the page on Confidence levels too.