Pemodelan Statistical Downscaling Berbasis Machine Learning Untuk Prediksi Curah Hujan Ekstrem Di Tiga Provinsi Sumatera Bagian Utara

Marpaung, Rafael A M (2026) Pemodelan Statistical Downscaling Berbasis Machine Learning Untuk Prediksi Curah Hujan Ekstrem Di Tiga Provinsi Sumatera Bagian Utara. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5003221133-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Perubahan iklim global telah memicu peningkatan frekuensi dan intensitas cuaca ekstrem, termasuk kejadian hujan lebat yang memicu banjir bandang pada November 2025 di wilayah Aceh, Sumatera Utara, dan Sumatera Barat. Untuk mengantisipasi dan memitigasi risiko bencana tersebut, diperlukan informasi curah hujan beresolusi tinggi di tingkat lokal. Penelitian ini bertujuan untuk memodelkan hubungan antara variabel atmosfer skala global dari reanalisis ERA5 dengan curah hujan harian observasi melalui pendekatan Statistical Downscaling. Variabel prediktor yang digunakan meliputi Total Precipitation (TP), Mean Sea Level Pressure (MSLP), serta Sea Surface Temperature (SST) yang berfokus pada dinamika suhu di perairan Jawa. Metode Principal Component Analysis (PCA) diaplikasikan untuk mereduksi dimensi data, di mana tiga komponen utama pertama terbukti mampu mempertahankan sekitar 87% hingga 90% variansi informasi prediktor. Selanjutnya, prediksi curah hujan dilakukan menggunakan algoritma Machine Learning, yakni Support Vector Regression (SVR) dan Random Forest (RF). Hasil evaluasi menunjukkan bahwa variabel TP memiliki korelasi positif tertinggi terhadap curah hujan aktual. Selain itu, model SVR yang dioptimasi dengan kernel terbukti lebih presisi dibandingkan RF dalam menangkap interaksi non-linier iklim dan anomali hujan ekstrem. Keunggulan ini ditandai dengan perolehan nilai Root Mean Square Error (RMSE) dan Mean Absolute Error (MAE) yang secara konsisten lebih rendah di ketiga wilayah penelitian. Model Statistical Downscaling yang optimal ini diharapkan dapat menjadi landasan preskriptif bagi sistem peringatan dini hidrometeorologi di Sumatera bagian Utara.
==================================================================================================================================
Global climate change has triggered an increase in the frequency and intensity of extreme weather, including heavy rainfall events that led to flash floods in November 2025 across Aceh, North Sumatra, and West Sumatra. To anticipate and mitigate such disaster risks, high-resolution local rainfall information is highly required. This study aims to model the relationship between large-scale atmospheric variables from ERA5 reanalysis and observed daily rainfall through a Statistical Downscaling approach. The predictor variables utilized include Total Precipitation (TP), Mean Sea Level Pressure (MSLP), and Sea Surface Temperature (SST) focusing on the thermal dynamics of the Java Sea. Principal Component Analysis (PCA) was applied for dimensionality reduction, wherein the first three principal components successfully retained approximately 87% to 90% of the predictors' variance. Subsequently, rainfall predictions were carried out using Machine Learning algorithms, specifically Support Vector Regression (SVR) and Random Forest (RF). Evaluation results indicate that the TP variable exhibits the highest positive correlation with actual rainfall. Furthermore, the kernel-optimized SVR model proved to be more precise than RF in capturing non-linear climate interactions and extreme rainfall anomalies. This superiority is indicated by consistently lower Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values across all three study regions. This optimal Statistical Downscaling model is expected to serve as a prescriptive foundation for hydrometeorological early warning systems in Northern Sumatra.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Curah Hujan, GCM, Random Forest, Statistical Downscaling, Support Vector Regression., ERA5, Extreme Rainfall, Random Forest, Statistical Downscaling, Support Vector Regression
Subjects:	H Social Sciences > HA Statistics > HA30.3 Time-series analysis H Social Sciences > HA Statistics > HA30.6 Spatial analysis H Social Sciences > HA Statistics > HA31.3 Regression. Correlation. Logistic regression analysis. Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.5 Principal components analysis. Factor analysis. Correspondence analysis (Statistics)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Rafael A. M. Marpaung
Date Deposited:	14 Aug 2026 07:36
Last Modified:	14 Aug 2026 07:36
URI:	http://repository.its.ac.id/id/eprint/144351

Actions (login required)

View Item