Wednesday, 31 January 2024: 2:45 PM
317 (The Baltimore Convention Center)
When an aircraft flies through clouds at subfreezing temperatures, there may be accumulation of ice onto its airframe, known as airframe icing. With reference to the forecast icing potential (FIP) algorithm developed by the Inflight Icing Product Development Team of the Federal Aviation Administration’s Aviation Weather Research Program [1], the Hong Kong Observatory (HKO) has developed similar algorithms based on pilot reports over the China region using (i) global (ECMWF) numerical weather prediction (NWP) model elements and (ii) HKO’s regional NWP, namely the HKO-WRF [2, 3], model elements.
In addition, HKO explored using gradient boosting machine learning algorithm (XGBoost) to forecast icing occurrence with ECMWF and HKO-WRF NWP model elements using (i) pilot reports over the China region, and (ii) pilot reports over the China region together with special air reports in the Asia and Pacific Region.
Noting that risk level might be more relevant to the impact assessment during the planning phase in the aviation operations, HKO has tried to combine icing potential with icing occurrence probability to create icing risk level. Statistical analysis was performed to identify which combination of icing potential and icing occurrence probability would generate risk level that optimally associates with observed icing severity.
This presentation will describe the development of global/regional icing potential, machine learning models on icing occurrence probability, and icing risk levels based on icing potential and occurrence probability.
Reference:
[1]: McDonough, Frank, et al. "The forecast icing potential algorithm." 42nd AIAA Aerospace Sciences Meeting and Exhibit. 2004.
[2]: Hon, K.-K., 2020: Tropical cyclone track prediction using a large-area WRF model at the Hong Kong Observatory. Tropical Cyclone Research and Review, 9 (1), 67–74.
[3]: https://www.hko.gov.hk/en/wservice/tsheet/nwp.htm
In addition, HKO explored using gradient boosting machine learning algorithm (XGBoost) to forecast icing occurrence with ECMWF and HKO-WRF NWP model elements using (i) pilot reports over the China region, and (ii) pilot reports over the China region together with special air reports in the Asia and Pacific Region.
Noting that risk level might be more relevant to the impact assessment during the planning phase in the aviation operations, HKO has tried to combine icing potential with icing occurrence probability to create icing risk level. Statistical analysis was performed to identify which combination of icing potential and icing occurrence probability would generate risk level that optimally associates with observed icing severity.
This presentation will describe the development of global/regional icing potential, machine learning models on icing occurrence probability, and icing risk levels based on icing potential and occurrence probability.
Reference:
[1]: McDonough, Frank, et al. "The forecast icing potential algorithm." 42nd AIAA Aerospace Sciences Meeting and Exhibit. 2004.
[2]: Hon, K.-K., 2020: Tropical cyclone track prediction using a large-area WRF model at the Hong Kong Observatory. Tropical Cyclone Research and Review, 9 (1), 67–74.
[3]: https://www.hko.gov.hk/en/wservice/tsheet/nwp.htm

