Prediction of Air Temperature on Runway 10 Juanda Airport Using Hybrid LSTM

Abstract

Abstract – Global climate change increases the frequency of extreme weather, impacting various sectors, including aviation. Accurate weather prediction becomes crucial to ensure the safety of human activities, including aviation operations. This study aims to predict air temperature variables on Runway 10 at Juanda Airport using a Long Short Term Memory (LSTM) architecture stacked with a Gated Recurrent Unit (GRU) architecture, named Hybrid LSTM. The data used in this study is air temperature (per minute) obtained from the Automatic Weather Observing System (AWOS) in (.csv) format. Testing was conducted for short-term predictions and comparisons were made between Hybrid and non-Hybrid models. The test results show that the LSTM-GRU architecture produced the lowest evaluation values for predicting temperature with an MSE of 0.0181, MAE of 0.0814, RMSE of 0.1345, and MAPE of 0.29% using a batch size of 64 and 20 epochs. This indicates that the developed model is suitable for predicting the next short-term period (5 minutes). For future research, model development is needed by adding features or adjusting hyperparameters to accurately predict the long term.
Keywords: Predict, Air Temperature, Automatic Weather Observing System (AWOS), LSTM, GRU