The issue of icing has been around for decades in aviation industry, and while notable improvements have been made in the study of the formation and process of icing, the prediction of icing events is a challenge that has yet to be completely overcome. Low level icing prediction, particularly in complex terrain, has been bumped to the back burner in an attempt to perfect the models created for in-flight icing. However, over the years there have been a number of different, non-model methods used to better refine the variables involved in low-level icing prediction.

One of those methods comes through statistical analysis and modeling, particularly through the use of the Classification and Regression Tree (CART) techniques. These techniques examine the statistical significance of each predictor within a data set to determine various decision rules. Those rules in which the overall misclassification error is the smallest are then used to construct a decision tree and can be used to create a forecast for icing events.

These CART techniques were used in an examination of icing events in the White Mountains of New Hampshire, specifically on the summit of Mount Washington. The Mount Washington Observatory (MWO), which sits on the summit and is manned year around by weather observers, is no stranger to icing occurrences. In fact, the summit sees icing events from October all the way until April, and occasionally even into May.

The research presented for this poster examined the icing events that occurred from October 2010 through April 2011 in an attempt to generate decision tree models with a high predictive accuracy in classifying icing events on the summit. Although it was an exploration into a somewhat unconventional method of classifying these events, the results showed potential and in the future may be a very useful tool in the prediction of icing events in complex terrain.