Analysis of Persistent Bias and Suggested Improvements in Forecasting Temperature Patterns over Canaan Valley in West Virginia with the National Blend of Models (NBM)

Weather prediction over complex terrain consistently poses one of the most significant challenges to forecasters and numerical models. One particularly unique location in this regard is Canaan Valley, West Virginia, with a floor elevation higher than any other large valley east of the Mississippi River. It is especially susceptible to cold air drainage, in which cooler, denser air sinks into the valley, forming a cold pool with temperatures much lower than the surrounding valley rim. The National Blend of Models (NBM), a critical foundation to operational forecasting, does not resolve these cold pool events and instead equates valley floor temperatures with those over surrounding higher elevations. This issue leads to a significant warm bias that exceeds 15ºC in the most extreme cases. A preliminary study connected this warm bias to non-linear quality control (NLQC) within the UnRestricted Mesoscale Analysis (URMA), which serves as ground truth for NBM bias correction. This research expands on this hypothesis by presenting four new case studies of Canaan Valley temperature patterns, taking place on 22 January, 4 February, 5 May, and 2 June 2023. Three out of four cases featured a pronounced cold pool at the valley floor, with warm biases observed in both NBM output and URMA analyses. It is shown that a low weight in NLQC was a likely cause of the warm bias, as there were no other recorded quality control issues, and lower values of the weighting function are directly associated with strong differences from the background field. To improve this discrepancy and subsequent NBM biases, algorithms within NLQC should allow for higher weights over complex terrain regions prone to strong temperature gradients.