Biases in the Mesoscale Prediction of Ceiling and Visibility in Alaska and Their Reduction Using Quantile Matching

Abstract

Short-range (2 h) predictions of ceiling and visibility obtained from version 4 of the Rapid Refresh (RAPv4) model are evaluated over Alaska using surface meteorological station data. These forecasts tended to overpredict the frequency of aviation-impacting ceilings in coastal areas by as much as 50%. In winter, this overforecasting bias extends into the interior of Alaska as well. Biases in visibility predictions were more complex. In winter, visibility hazards were predicted too often throughout the interior of Alaska (+5%) and not often enough in northern and western coastal areas (−20%). This wintertime underprediction of visibility restrictions in coastal areas has been linked to the fact that the visibility diagnostic does not include a treatment for the effect of blowing snow. This, in part, results in winter IFR visibilities being detected only 37% of the time. An efficient algorithm that uses quantile matching has been implemented to remove mean biases in 2-h predictions of ceiling and visibility. Performance of the algorithm is demonstrated using two 30-day periods (January and June 2019). The calibrated forecasts obtained for the two month-long periods are found to have significantly reduced biases and enhanced skill in capturing flight rules categories for both ceiling and visibility throughout much of Alaska. This technique can be easily extended to other forecast lead times or mesoscale models.