Pengembangan Model Estimasi Curah Hujan Berbasis XGBoost Menggunakan Data Himawari-8/9 dan Variabel Atmosfer di Jawa Timur

Wisnawa, Gede Gangga (2026) Pengembangan Model Estimasi Curah Hujan Berbasis XGBoost Menggunakan Data Himawari-8/9 dan Variabel Atmosfer di Jawa Timur. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6016241001-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (32MB) | Request a copy

Abstract

Provinsi Jawa Timur adalah salah satu wilayah yang rentan terhadap bencana hidrometeorolog di Indoensia. Hampir 50% dari seluruh kejadian bencana di Jawa Timur adalah banjir dengan potensi kerugian material yang signifikan. Meskipun Numerical Weather Prediction (NWP) lazim digunakan, tantangan biaya operasional dan kebutuhan perangkat keras berkinerja tinggi membuka peluang bagi pendekatan machine learning. Penelitian ini bertujuan mengembangkan model estimasi curah hujan berbasis algoritma Extreme Gradient Boosting (XGBoost) dengan mengintegrasikan data Brightness Temperature (BT) dan Brightness Temperature Difference (BTD) dari satelit Himawari-8/9, indeks iklim global (SSTA Nino 3.4 dan DMI), indeks monsun, serta data elevasi. Model dikembangkan menggunakan arsitektur multi-stage (klasifikasi biner, multi-kelasifikasi, dan regresi) dan dioptimalkan melalui Bayesian Optimization. Hasil penelitian menunjukkan bahwa optimasi hyperparameter meningkatkan performa model secara signifikan, menghasilkan nilai Root Mean Square Error (RMSE) sebesar 6,68 mm/3jam, Mean Absolute Error (MAE) 3,08 mm/3jam, dan koefisien korelasi (R) sebesar 0,48. Dalam aspek klasifikasi, model mencapai Critical Success Index (CSI) sebesar 0,49 pada tahap multi-kelasifikasi, yang secara signifikan mengungguli produk global seperti GSMaP, IMERG, dan ERA5. Analisis topografi menunjukkan degradasi performa seiring meningkatnya ketinggian wilayah akibat mekanisme hujan orografik hangat yang sulit dideteksi sensor inframerah. Secara keseluruhan, model XGBoost yang dikembangkan terbukti lebih akurat dan adaptif terhadap karakteristik lokal di Jawa Timur, menjadikannya alat yang efisien untuk mendukung manajemen risiko hidrometeorologi.
====================================================================================================================================
East Java Province is one of the regions in Indonesia most vulnerable to hydrometeorological disasters. Nearly 50% of all disaster events in East Java are floods, which carry the potential for significant material losses. While Numerical Weather Prediction (NWP) is commonly utilized, challenges regarding operational costs and high-performance hardware requirements provide opportunities for machine learning approaches. This research aims to develop a rainfall estimation model based on the Extreme Gradient Boosting (XGBoost) algorithm by integrating Brightness Temperature (BT) and Brightness Temperature Difference (BTD) data from Himawari-8/9 satellites, global climate indices (SSTA Nino 3.4 and DMI), monsoon indices, and elevation data. The model was developed using a multi-stage architecture (binary classification, multi-class classification, and regression) and optimized via Bayesian Optimization. The results indicate that hyperparameter optimization significantly enhanced model performance, yielding a Root Mean Square Error (RMSE) of 6.68 mm/3h, a Mean Absolute Error (MAE) of 3.08 mm/3h, and a correlation coefficient (R) of 0.48. In terms of classification, the model achieved a Critical Success Index (CSI) of 0.49 in the multi-class stage, significantly outperforming global products such as GSMaP, IMERG, and ERA5. Topographical analysis revealed performance degradation as elevation increased, attributed to warm orographic rainfall mechanisms that are challenging for infrared sensors to detect. Overall, the developed XGBoost model proved to be more accurate and adaptive to local characteristics in East Java, serving as an efficient tool to support hydrometeorological risk management.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	XGBoost, Himawari-8/9, Estimasi Curah Hujan, Jawa Timur, XGBoost, Himawari-8/9, Rainfall Estimation, East Java.
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QC Physics > QC866.5 Climatology--Forecasting. Q Science > QC Physics > QC925 Rain and rainfall
Divisions:	Faculty of Civil Engineering and Planning > Geomatics Engineering > 29101-(S2) Master Thesis
Depositing User:	Gede Gangga Wisnawa
Date Deposited:	27 Jan 2026 02:25
Last Modified:	27 Jan 2026 02:25
URI:	http://repository.its.ac.id/id/eprint/130576

Actions (login required)

View Item