Rainfall Prediction from Himawari-8 Data Using the Deep Learning Method

Abstract

Changes in rainfall affect human activities and can cause natural disasters, such as floods and landslides. This research focuses on changes in extreme rainfall that result in natural disasters. Indonesia, as a country with a tropical climate, certainly has its characteristics regarding rainfall patterns, such as air temperatures that tend to be high, sunlight that occurs throughout the year, and low air pressure. The characteristics of the tropical climate will form patterns and allow natural disasters to occur. Losses due to natural disasters can be minimized if there is thorough preparation in dealing with the possibility of natural disasters. Thorough preparation in facing natural disasters is based on knowledge of predictions of when and where the natural disaster will occur. Changes in rainfall can be predicted based on past rainfall data. These data produce patterns, such as real-time intensity. The data used in this research comes from the himawari-8 satellite using a cloud dataset. The next stage is pre-processing, which is the cleaning and adjustment of data. The deep learning algorithms used in this research are long short-term memory (LSTM) and recurrent neural network (RNN) to manage time series data. In previous research, it has been recommended that LSTM and RNN algorithms be used for rainfall prediction. The system created in this research is a rainfall prediction model using python programming language analysis and system output in the form of software with the accuracy of the LSTM model reaching 92.62% and the RNN model reaching 89.38%.