Prediksi Suhu Harian Model INA-NWP Menggunakan Metode Random Forest dan LSTM di Wilayah Surabaya

Kusuma, Ryanaldi Robby (2026) Prediksi Suhu Harian Model INA-NWP Menggunakan Metode Random Forest dan LSTM di Wilayah Surabaya. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221182-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Prakiraan suhu udara yang akurat sangat penting, namun model Indonesia Numerical Weather Prediction (INA-NWP) buatan BMKG sering menunjukkan bias sistematis underestimate terhadap data observasi permukaan di Surabaya. Penelitian ini bertujuan mengembangkan model koreksi bias Model Output Statistics (MOS) berbasis machine learning menggunakan metode Random Forest (RF) dan Long Short-Term Memory (LSTM) untuk parameter suhu maksimum dan minimum. Data yang digunakan meliputi prakiraan INA-NWP dan observasi suhu harian dari stasiun Maritim Perak II, AWS Sambikerep, dan Juanda untuk periode Maret 2021 hingga Maret 2022. Metode RF memodelkan hubungan non-linear secara independen, sedangkan LSTM mengekstraksi ketergantungan temporal dalam deret waktu. Kinerja model dievaluasi menggunakan Root Mean Square Error (RMSE) dan Improvement Percentage. Hasil eksplorasi mengonfirmasi bias underestimate pada INA-NWP, dengan rekor error tertinggi mencapai RMSE 6,2292. Setelah koreksi, kedua algoritma berhasil mereduksi bias secara signifikan. MOS-RF terbukti menjadi model terbaik untuk area terbuka Stasiun Juanda dengan capaian perbaikan tertinggi hingga 83,06%. Sementara itu, MOS-LSTM tampil lebih superior di kawasan pesisir Perak II dan area urban Sambikerep dengan tingkat perbaikan yang konsisten di kisaran 70% hingga 77%. Kesimpulannya, MOS-RF sangat tangguh meredam fluktuasi cuaca bebas di area terbuka, sedangkan MOS-LSTM lebih presisi menangkap pola memori dari laju pemanasan dan pendinginan bertahap di area pesisir serta urban.
=======================================================================================================================================
Accurate air temperature forecasting is crucial, yet the Indonesia Numerical Weather Prediction (INA-NWP) model developed by BMKG frequently exhibits a systematic underestimate bias against surface observation data in Surabaya. This study aims to develop a machine learning-based Model Output Statistics (MOS) bias correction model using Random Forest (RF) and Long Short-Term Memory (LSTM) methods for maximum and minimum temperature parameters. The data used include INA-NWP forecasts and daily temperature observations from the Perak II Maritime, AWS Sambikerep, and Juanda stations for the period of March 2021 to March 2022. The RF method models non-linear relationships independently, whereas LSTM extracts temporal dependencies within the time series. Model performance was evaluated using Root Mean Square Error (RMSE) and Improvement Percentage. Exploratory results confirmed the underestimate bias in INA-NWP, with the highest error record reaching an RMSE of 6.2292. After correction, both algorithms successfully reduced the bias significantly. MOS-RF proved to be the best model for the open area of Juanda Station, achieving the highest improvement rate of up to 83.06%. Meanwhile, MOS-LSTM was superior in the coastal area of Perak II and the urban area of Sambikerep, with consistent improvement rates ranging from 70% to 77%. In conclusion, MOS-RF is highly robust in dampening free weather fluctuations in open areas, while MOS-LSTM is more precise in capturing memory patterns of gradual heating and cooling rates in coastal and urban areas.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	INA-NWP, Koreksi Bias, Long Short-Term Memory, Model Output Statistics, Random Forest, Bias Correction, INA-NWP, Long Short-Term Memory, Model Output Statistics, Random Forest
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QC Physics > QC866.5 Climatology--Forecasting.
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Ryanaldi Robby Kusuma
Date Deposited:	30 Jul 2026 04:00
Last Modified:	30 Jul 2026 04:00
URI:	http://repository.its.ac.id/id/eprint/139813

Actions (login required)

View Item