Visibility Forecasting in Meteorological Aerodrome Report using Facebook Prophet

Meteorological Aerodrome Reports (METAR) are vital in aviation operations because they provide real-time weather conditions of airports. The paper focuses on applying machine learning and time series methodologies in order to forecast visibility in METAR data. The study entails gathering and preparing METAR data, as well as using the Facebook prophet model to forecast visibility for the next few hours. In this paper, we used the prophet model to perform univariate and multivariate forecasting. The ARIMA and ARIMAX models are also performed in order to compare the performance of the prophet model. According to the findings, the prophet model for multivariate forecasting produces more promising results with RMSE of 2.48 which is the least value when compared to the RMSE calculated for the other models. The study also emphasizes the significance of taking seasonality and trend components into account when forecasting METAR data.