The Regime Dependence of Degree Day Forecast Technique, Skill, and Value

Abstract

An investigation into the manner in which forecasters adjust their reliance on particular pieces of forecast information as the large-scale flow pattern evolves into different regimes, and the relationship between those adjustments and forecast skill and value is presented. For the cold season months (December?February) of the period 1 January 1973 through 31 December 1992, a total of three regime types (identified through cluster analysis) comprising 63% of the days were identified. A framework for investigating the weighting of pieces of forecast information, based upon multiple regression techniques, was applied to National Weather Service (NWS) degree day forecasts (constructed from the 12?24-h minimum and 24?36-h maximum temperature forecasts) for this period. It was determined that substantial changes in the usage of Model Output Statistics (MOS) by NWS forecasters have occurred with the advent of the improved numerical model guidance represented by the Limited Fine Mesh (LFM) MOS, and that these changes occurred in response to improvements in the longer-range forecasts (validating 24?36 h from the initial time). However, it was also shown that this increased weighting of MOS was situation dependent and that forecast skill and value were maintained under large-scale flow regimes in which MOS was less useful through significant adjustment of forecast technique. Overall, skills were found to be lowest for flows in which either the variability of the MOS weight was highest (reflecting uncertainty in its reliability) or in which limitations of that guidance were evident. These results are then related to earlier investigations concerning the relationship between forecast skill and experience.