Probabilistic Nowcasting of Low-Visibility Procedure States at Vienna International Airport During Cold Season

Abstract

Abstract-Airport operations are sensitive to visibility conditions. Low-visibility events may lead to capacity reduction, delays and economic losses. Different levels of low-visibility procedures (lvp) are enacted to ensure aviation safety. A nowcast of the probabilities for each of the lvp categories helps decision makers to optimally schedule their operations. An ordered logistic regression (OLR) model is used to forecast these probabilities directly. It is applied to cold season forecasts at Vienna International Airport for lead times of 30-min out to 2 h. Model inputs are standard meteorological measurements. The skill of the forecasts is accessed by the ranked probability score. OLR outperforms persistence, which is a strong contender at the shortest lead times. The ranked probability score of the OLR is even better than the one of nowcasts from human forecasters. The OLR-based nowcasting system is computationally fast and can be updated instantaneously when new data become available.

Forecasting Low-Visibility Procedure States with Tree-Based Statistical Methods

“…In the large setup lvp, cei, and rvr information is available also 10-min lags to forecast initialization and averaged for runway and airport, respectively. The variables dir N , dir SE , rr, and sza used in the standard setup are not included in the large setup (for more details to the standard predictor setup, see Kneringer et al 2018) conditional inference framework of Hothorn et al (2006b). These trees separate the variable selection and splitting procedure into two steps and do not suffer from a systematic tendency towards split variables with many possible splits or many missing values (Hothorn et al 2006b).…”

“…The first investigation of lvp, which is the relevant variable for airport operations, was conducted in a companion paper by Kneringer et al (2018). They used ordered logistic regression (OLR) models to produce lvp nowcasts at VIE for the cold season (September-March) with lead times up to 2 h and 30-min resolution.…”

“…They used ordered logistic regression (OLR) models to produce lvp nowcasts at VIE for the cold season (September-March) with lead times up to 2 h and 30-min resolution. To provide decision makers with more detailed information, we increase the temporal resolution of the predictions to 10 min and extend the predictor variable setup of Kneringer et al (2018) with temporally lagged and spatially averaged visibility information. During the cold season, the peak of lvp is in the early morning hours, coinciding with the airport's rush hours (Kneringer et al 2018).…”

“…To provide decision makers with more detailed information, we increase the temporal resolution of the predictions to 10 min and extend the predictor variable setup of Kneringer et al (2018) with temporally lagged and spatially averaged visibility information. During the cold season, the peak of lvp is in the early morning hours, coinciding with the airport's rush hours (Kneringer et al 2018). Since low visibility has most impact during these times, we generate lvp predictions in this paper only from September to March during 6 and 9 UTC.…”

“…Tree-based statistical methods are used as a flexible nonparametric alternative to the parametric OLR for generating the forecasts. The predictions of these models are compared amongst each other and to the forecasts of the OLR models of Kneringer et al (2018). Furthermore, the impact of the predictors on the forecasts is analyzed to provide information of the most important inputs for statistically based lvp nowcasts.…”