Penentuan Turbulensi Berdasarkan Bilangan Richardson dari Data Historis Radiosonde Stasiun Meteorologi Kualanamu Menggunakan Machine Learning Tahun 2015 - 2018

Abstract

Support Vector Machine

Turbulence can occur on a small scale in a short period and often detected when there is a change in the wind direction and wind speed against altitude. Forecasting turbulence in atmospheric layer boundary is highly depends on Richardson Gradient Number. This research use upper air observation data (Radiosonde) from Stasiun Meteorologi Kualanamu from year 2015 to 2018 where observations at 00.00 UTC are 1,440 data and 12.00 UTC are 1,433 data. Specifically, using Richardson Gradient Number and Machine Learning approach through the Support Vector Machine method to determine and analyze the symptoms of turbulence and make a forecasting turbulence method. After conducting the research, it can be concluded that the symptoms of turbulence that occur at flight altitude cruising from 1,000 feet to 49,000 feet at 00.00 UTC as many as 13,158 symptoms (39.73% of the total atmospheric conditions studied), while at 12.00 UTC as many as 13,778 symptoms ( 41.80% of the total atmospheric conditions studied). In general, forecasting results with the Support Vector Machine method shows that the value of the percentage of error (ɛ) is close to 0 (zero), even in certain layers the error value is equal to 0 (zero) so that it shows a high approximation accuracy value. The result shows that the Support Vector Machine method has high accuracy that it can be used as one of the methods for forecasting turbulence in the future.