| Description |
The High-Resolution Rapid Refresh - Alaska (HRRR-AK) modeling system has been developed to provide high spatial (3 km horizontal) and temporal (0-36 hourly forecasts) guidance for weather conditions over Alaska. This study evaluated the experimental version of the HRRR-AK system available during the 2017 winter (December 2016 - March 2017) prior to its expected operational deployment by the National Centers for Environmental Prediction in early 2018. Of highest interest was to assess the model's ability to forecast the evolution, intensity, and timing of winter weather systems on the basis of surface pressure observations assimilated as part of its production cycle (e.g., from National Weather Service, NWS, stations) and those not assimilated (e.g., from USArray Transportable Array, TA, stations). Altimeter setting observations from the 100+ sites available from each of the NWS and TA networks are used to evaluate 138 complete 0- to 36-hour forecasts of altimeter setting initialized at 00 or 12 Universal Time Coordinated (UTC). More detailed examination of the forecasts using additional data assets was performed for the period 12-15 February 2017 during which a mid-tropospheric cut-off low over the Gulf of Alaska contributed to two distinct periods of strong downslope winds (southerly wind gusts in excess of 25 m s-1) in the lee of the Alaska Range near Fort Greely in the Tanana Valley. Throughout the 2017 winter, systematic differences in altimeter setting between the HRRR-AK analyses and the NWS assimilated (TA unassimilated) observations were small (large). Upon removal of these initial biases from each of the subsequent 1- to 36-hour altimeter setting forecasts at the observation locations, the NWS and TA observations were found to be equally valuable for assessing forecast skill. When aggregated over the entire state, forecast errors were highest during the 12-15 February period within which the Gulf of Alaska cut-off low and Fort Greely downslope windstorms transpired. While the HRRR-AK analyses were quite useful for diagnosing the synoptic and mesoscale conditions during this period, the model forecasts tended to generate more intense surface pressure features rotating around the cut-off low and underestimated the abrupt increase in temperature during the first downslope windstorm and the intensity of the downslope winds during both events. |
