Determining an Optimal Decay Factor for Bias-Correcting MOS Temperature and Dewpoint Forecasts

Abstract

Abstract Model output statistics (MOS) forecast relationships for temperature and dewpoint developed with least squares regression and put into operation by the National Weather Service (NWS) are unbiased over the sample period of development. However, short-term biases within that period can exist, and application of the regression equations to new data may produce forecasts with short- or long-term biases. Because NWP models undergo changes over time, MOS forecasts can be biased because of these changes, and also possibly because of local environmental changes. These biases can be largely eliminated. In the decaying average method, a “decay factor” is used. This value affects not only the short- and long-term bias characteristics, but also other accuracy measures of the forecasts. This paper shows how different values of the decay factor affect MOS temperature and dewpoint forecasts, and the range of factors that would be appropriate for bias correcting those forecasts. Biases and other quality measures are shown for both cool and warm season samples before and after various values of the decay factor have been applied.