Model Prediksi Suhu Maksimum Dan Minimum Indonesia-Numerical Weather Prediction Menggunakan Support Vector Regression Dan Long Short-Term Memory

Ichwani, Haikal Mauladana (2025) Model Prediksi Suhu Maksimum Dan Minimum Indonesia-Numerical Weather Prediction Menggunakan Support Vector Regression Dan Long Short-Term Memory. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003211142-Undergraduate-Theses.pdf Download (2MB)

Abstract

Kebutuhan akan prakiraan cuaca yang akurat mendorong Badan Meteorologi, Klimatologi, dan Geofisika (BMKG) telah mengembangkan model berbasis Numerical Weather Prediction yang dinamakan Indonesia-Numerical Weather Prediction (INA-NWP). Namun, model INA NWP seringkali menunjukkan bias sistematis dalam memprediksi suhu permukaan maksimum dan minimum. Penelitian ini menerapkan teknik statistical post-processing berbasis Machine Learning (ML) untuk meminimalkan bias dengan dua metode yaitu Support Vector Regression (SVR) dan Long Short-Term Memory (LSTM). Kedua model disusun menggunakan data luaran suhu maksimum dan minimum dari INA-NWP serta data observasi di tiga stasiun meteorologi di wilayah Surabaya. Hasil penelitian menunjukkan bahwa kinerja SVR secara umum konsisten lebih baik daripada LSTM dalam memprediksi suhu maksimum dan minimum di seluruh stasiun pengamatan. Hal ini ditunjukkan dengan nilai RMSE model SVR lebih kecil dan Improval Percentage yang lebih besar daripada model LSTM di dua lokasi stasiun Perak I dan Juanda. Khusus untuk lokasi Sambikerep pada model suhu minimum LSTM memiliki kinerja lebih baik daripada SVR. Temuan ini menunjukkan potensi besar NWP-ML untuk mengoreksi bias sistematis dan meningkatkan ketepatan informasi cuaca di Indonesia.
====================================================================================================================================
The demand for accurate weather forecasts has driven the Meteorological, Climatological, and Geophysical Agency (BMKG) to develop a Numerical Weather Prediction-based model known as the Indonesia-Numerical Weather Prediction (INA-NWP). However, the INA-NWP model often exhibits systematic biases in predicting surface maximum and minimum temperatures. This study applies statistical post-processing techniques based on Machine Learning (ML) to reduce such biases using two methods:Support Vector Regression (SVR) and Long Short-Term Memory (LSTM). Both models are developed using output data of maximum and minimum temperatures from INA-NWP along with observational data from three meteorological stations in the Surabaya area. The results show that SVR generally performs better than LSTM in predicting both maximum and minimum temperatures across all observation stations. This is indicated by lower RMSE values and higher Improval Percentage scores for the SVR model at two locations Perak I and Juanda. However, at the Sambikerep station, the LSTM model performs better for minimum temperature prediction than SVR. These findings demonstrate the significant potential of NWP-ML approaches in correcting systematic biases and improving the accuracy of weather information in IndonesiaThe demand for accurate weather forecasts has driven the Meteorological, Climatological, and Geophysical Agency (BMKG) to develop a Numerical Weather Prediction-based model known as the Indonesia-Numerical Weather Prediction (INA-NWP). However, the INA-NWP model often exhibits systematic biases in predicting surface maximum and minimum temperatures. This study applies statistical post-processing techniques based on Machine Learning (ML) to reduce such biases using two methods: Support Vector Regression (SVR) and Long Short-Term Memory (LSTM). Both models are developed using output data of maximum and minimum temperatures from INA-NWP along with observational data from three meteorological stations in the Surabaya area. The results show that SVR generally performs better than LSTM in predicting both maximum and minimum temperatures across all observation stations. This is indicated by lower RMSE values and higher Improval Percentage scores for the SVR model at two locations Perak I and Juanda. However, at the Sambikerep station, the LSTM model performs better for minimum temperature prediction than SVR. These findings demonstrate the significant potential of NWP-ML approaches in correcting systematic
biases and improving the accuracy of weather information in Indonesia.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	INA-NWP, Long Short-Term Memory, Prakiraan Cuaca, Statistical post processing, Support Vector Regression.
Subjects:	Q Science Q Science > Q Science (General) Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Haikal Mauladana Ichwani
Date Deposited:	01 Aug 2025 09:00
Last Modified:	01 Aug 2025 09:00
URI:	http://repository.its.ac.id/id/eprint/126132

Actions (login required)

View Item