An Evaluation of Mesoscale-Model-Based Model Output Statistics (MOS) during the 2002 Olympic and Paralympic Winter Games

Abstract

The skill of a mesoscale-model-based model output statistics (MOS) system that provided hourly forecasts for 18 sites over northern Utah during the 2002 Winter Olympic and Paralympic Games is evaluated. The MOS system was developed using three winters (November–April 1998/99, 1999/2000, and 2000/01) of forecasts by the fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) and observations from Olympic venues and transportation corridors. MOS temperature, relative humidity, wind speed, and wind direction forecasts were considerably more accurate than those produced by the 12- and 4-km MM5 grids. A primary contributor to MM5 temperature and relative humidity errors was a systematic overprediction of surface temperature (i.e., a warm/dry bias) during persistent and nocturnal cold-pool events when corresponding errors in MM5 dewpoint temperature forecasts were not observed. MOS largely corrected for this temperature bias. MOS wind speed forecasts outperformed the 12- and 4-km MM5 forecasts by the largest margin at locations with the lowest wind speed variability. Raw model and MOS performance exhibited minimal sensitivity to variations in model initial and lateral boundary conditions (derived from the forecasts of either the National Centers for Environmental Prediction's Eta Model or the Aviation run of the Global Spectral Model). MOS temperature, relative humidity, and wind speed forecasts were equal to or more skillful than human-generated forecasts produced by the Olympic Forecast Team. The results illustrate that statistical techniques continue to improve upon purely numerical predictions even at high resolution. This is particularly true in a region of complex terrain where detailed characteristics of local topography and microclimates remain unresolved. It is recommended that traditional MOS or other statistical techniques based on high-density surface observations available from the MesoWest cooperative networks be used to improve gridded forecast products created by the National Weather Service Interactive Forecast Preparation System (IFPS) and other applications.