On post-pandemic short-term forecasting of quarterly air passenger traffic at Polish airports

Анотація

The COVID-19 pandemic caused a structural break in air passenger traffic dynamics, limiting the reliability of forecasting models based on long pre-pandemic time series. This study addresses the problem of short-term forecasting of quarterly total air passenger traffic at major Polish airports under post-pandemic market conditions. The research focuses on Warsaw Chopin Airport, Kraków-Balice, Gdańsk Lech Wałęsa Airport, Katowice-Pyrzowice, Wrocław-Starachowice, Poznań-Ławica, Rzeszów-Jasionka and Szczecin-Goleniów. The purpose of the article is to assess whether forecasting models based only on recent post-COVID data can provide reliable short-term predictions of passenger traffic. The empirical analysis uses quarterly statistical data published by the Polish Civil Aviation Authority. For most airports, the modelling period covers 2021–2024, while forecasts are generated for the four quarters of 2025 and compared with actual observations. For Szczecin-Goleniów Airport, due to the specific traffic dynamics, an additional modelling approach based on 2022–2025 data is considered. Two time-series forecasting methods are applied: ARIMA and ETS models implemented in the R environment. Forecast accuracy is evaluated using standard error metrics, including ME, RMSE, MAE, MAPE and MASE, as well as the Ljung-Box test and the corrected Akaike information criterion. The results show that post-pandemic quarterly passenger traffic at Polish airports demonstrates a stable seasonal pattern, although its scale and recovery dynamics differ across airports. ARIMA models generally provide better fit and forecasting performance for most airports, while ETS modelling is more appropriate in selected cases. The findings confirm that excluding pre-pandemic observations may improve short-term forecasting accuracy when the market has entered a new structural phase after COVID-19 and subsequent geopolitical disruptions. The study contributes to the methodological discussion on air transport forecasting by demonstrating the practical value of using post-break data windows instead of mechanically extending historical series across structurally incomparable periods.

Класифікація

Ідентифікатори

Схожі роботи