Creating Probabilistic Flight Category Forecasts in Alaska using non-linear Multiple Discriminate Analysis

Aviation activities in Alaska are a critical means for transportation and supply chain management due to the vast distances between communities and the lack of surface transportation alternatives.  In addition, Anchorage is a major air cargo hub between Asia and North America.  Current Numerical Weather Prediction (NWP) models do not explicitly forecast cloud ceilings and visibility.  These parameters are derived through various post-processing algorithms and have significant biases, especially in marine areas and in snowfall.  In several areas of Alaska, current NWP does not adequately resolve the terrain, and in particular the narrow gaps, valleys, and fjords which are frequently used as primary general aviation flight routes, and/or are the locations of communities.  Linear Regression techniques are commonly used to provide flight category guidance at airport locations.  The latest version of this guidance available to NWS Alaska has been shown to have limited skill on average, with persistence as the “best” forecast.  Using a modified Multiple Discriminate Analysis (MDA) technique probabilistic Flight Category guidance was developed for a number of airports across Alaska.  This technique lends itself to non-linear categorical forecasts and has shown skill in other complex forecast problems.  Results from the MDA technique will be presented and compared with NWP and standard Linear Regression output.