Improving Meteorological Development Laboratory (MDL) Station-based Model Output Statistics (MOS) for Wind Speed and Wind Gust through Daily Bias Correction, Objective Weighting, and Blending

The Meteorological Development Laboratory (MDL) has been generating automated station-based Model Output Statistics (MOS) guidance since the late 1960’s as a means to remove bias from Direct Model Output (DMO). MDL MOS guidance for a particular model and weather element is generated by evaluating a seasonally dependent regression equation derived from a training sample usually encompassing several seasons. Further post-processing is then performed to ensure element and intra-element consistency. That said, the current MOS structure is not engineered to correct for its more recent short-term biases. Baars and Mass [Baars, J. A., and C. F. Mass 2005: Performance of National Weather Service forecasts compared to operational, consensus, and weighted model output statistics. Wea. Forecasting, 20, 1034–1047] have demonstrated that a more skillful MOS product can be created by (1) incorporating a fairly straightforward bias correction algorithm, (2) objectively weighting each MOS input, and (3) blending all inputs using each MOS input's objective weight.

The National Blend of Models (NBM) has already integrated this bias correction algorithm into its process by routinely bias correcting all gridded model data using the UnRestricted Mesoscale Analysis (URMA) as ground truth. For most weather elements bias correcting to URMA works quite well. However, URMA wind speed and wind gust analyses, which are derived from a disproportionate number of mesonet sites when compared to the number of METARs, has introduced a very noticeable low bias for both wind speed and wind gust especially in the vicinity of METAR locations. These biases are then propagated into the NBM by virtue of the NBM using URMA as its analyses. While URMA developers are aware of this problem and are actively working on a solution, in the interim, we plan to blend all MDL MOS wind speed and wind gust guidance. Each MOS input will be routinely bias corrected and weighted to its respective METAR location using the algorithm proposed by Woodcock and Engle [Woodcock, F., and C. Engel, 2005: Operational consensus forecasts. Wea. Forecasting, 20, 101–111]. The bias corrected values will then be objectively weighted and blended and then placed at the grid point nearest its respective METAR site. In this manner we stay true to the objective weighting methodology used in the NBM and simultaneously leverage the METAR observations and MOS guidance to more accurately predict wind speeds and wind gusts in the vicinity of airports where wind guidance is crucial to aviation operations.