Forecast Verification for Denver International Airport

Aviation managers are often faced with many forecasting options to help make decisions. In winter, these forecasts are used to decide which snow alert level to declare and when to declare it. Snow alerts have significant impacts for airport operations, and for the customers who rely on them.

A recent study, led by NOAA GSL scientists, investigated the skill of four meteorological models in predicting snowfall for decision making at Denver International Airport.

This study was novel in its use of operational evaluation, which means it verified forecasts in the specific context of airport management. Most verification processes are general, but this study emphasized the importance of mimicking the context in which the forecasting is actually taking place. Operations and maintenance at DIA were regarded as the primary stakeholders throughout the study. The research team conducted interviews and shadowed staff at DIA to understand how the forecast information is used.

The study compared the accuracy of forecasts from the four models with the Meteorological Aerodome Reports (METARs). The models included were:

• Short Range Ensemble Forecast (SREF)
  • SREF is run at the National Centers for Environmental Protection, consisting of 26 members
  • Spatial Resolution: 16-km horizontal resolution
  • Time increments: 4 hours
•  High Resolution Rapid Refresh Ensemble  (HRRRE)
  •  HRRRE is real-time, cloud-resolving, convection-allowing forecasting system. As of June 2021, its experimental phase was concluded.
  • Spatial Resolution: 3-km
  • Time Increments: 4 hours
•  National Blend of Models  (NBM)
  • NBM is a nationally consistent suite of calibrated forecast guidance based on a blend of both NWS and non-NWS numerical weather prediction model data and post-processed model guidance
  • Spatial Resolution: Point forecasting (single predicted value)
  • Time Increments: 1 hour
•  Probabilistic Snow Accumulation Forecast (PSA) 
  • PSA is a text product produced by the Boulder WFO specifically for public- and private-based airport operations, with a goal of assisting decisions regarding airport operations during snow events.
  • Spatial Resolution: Point forecasting
  • Time increments: N/A

The amount of snow and start and end times of snowfall were tested from November 2018 to April 2019 as the first season, and December 2019 the April 2020 as the second season.

Results found that NBM had the most accurate predictions of snowfall timing. NBM also predicted fewer events and rarely produced high probabilities, unlike the other systems. The missed events and false alarms with HRRR and SREF were slightly larger than with NBM and PSA. . All products over-forecasted snow to some extent.

The results of the study have significant implications for both airport personnel and the Boulder National Weather Service Forecast Office. Operationally relevant verification results on timing and severity of snowfall is key to minimizing negative impacts at the airport. Accurate forecasts ensure customer and employee safety, airspace efficiency, and reduced costs. The timing of snowfall has significant implications for scheduling staff and resources, as well is in preparing for snow removal from runways.

The study can be quite helpful for aviation decision makers as they decide which forecast models to use. Next steps in the study include repeating the study with higher intensity snowfall events. Studies that explore this verification are crucial in making infrastructure more resilient to a changing climate.