Faidah, Defi Yusti (2024) Pengembangan Metode Kalibrasi Peramalan Ensembel untuk Kejadian Ekstrem. Doctoral thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 06211760010002-Dissertation.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
06211760010002-Dissertation.pdf - Accepted Version Restricted to Repository staff only until 1 April 2026. Download (4MB) | Request a copy |
Abstract
Upaya pengembangan metode peramalan curah hujan masih terus dilakukan untuk memperoleh hasil yang lebih akurat. Terdapat beberapa metode untuk meramalkan curah hujan, mulai model tunggal, model hybrid hingga kombinasi beberapa model peramalan. Gabungan hasil ramalan dari beberapa model menjadi suatu nilai merupakan bentuk dari ramalan ensembel. Pada umumnya, hasil ramalan ensembel bersifat underdispersive atau overdispersive sehingga diperlukan kalibrasi. Bayesian Model Averaging (BMA) dengan distribusi Normal seringkali digunakan untuk mengkalibrasi prediksi ensembel. Akan tetapi karakteristik kejadian ekstrem tidak dapat didekati dengan distribusi Normal. Salah satu distribusi yang dapat menangkap observasi curah hujan ekstrem adalah distribusi Gamma dan Generalized Extreme Value (GEV) tipe II yaitu distribusi Frechet. Penelitian ini bertujuan mengembangkan metode kalibrasi yang melibatkan distribusi ekstrem yaitu Modified BMA-Gamma dan BMA-Frechet. Kajian teori dilakukan untuk menaksir parameter Modified BMA-Gamma dan BMA-Frechet. Implementasi Modified BMA-Gamma dan BMA-Frechet digunakan untuk mengkalibrasi ramalan ensembel curah hujan ekstrem pada masing-masing pola curah hujan di Indonesia. Hasil penelitian menunjukkan bahwa performansi modified BMA-Gamma dan BMA-Frechet dapat mengkalibrasi data curah hujan ekstrem dengan baik, ditunjukkan dari nilai bias yang kecil. Hal ini berarti hasil kalibrasi memiliki nilai yang dekat dengan data observasi
====================================================================================================================================
The development of rainfall forecasting methods is still ongoing to obtain more accurate results. There are several methods for predicting rainfall, from single models, hybrid models to combinations of several prediction models. Combining the prediction results from several models into a value is a form of ensemble prediction. In general, ensemble prediction results are underdispersive or overdispersive so calibration is required. Bayesian Model Averaging (BMA) with a Normal distribution is often used to calibrate ensemble predictions. However, the characteristics of extreme observations cannot be approximated by a normal distribution. One distribution that can capture extreme rainfall observations is the Gamma and Generalized Extreme Value (GEV) type II distribution, namely the Frechet distribution. This research aims to develop a calibration method that involves extreme distributions, namely Modified BMA-Gamma and BMA-Frechet. Theoretical studies were carried out to estimate the Modified BMA-Gamma and BMA-Frechet parameters. Implementation of Modified BMA-Gamma and BMA-Frechet is used to calibrate ensemble predictions of extreme rainfall for each rainfall pattern in Indonesia. The research results show that the performance of modified BMA-Gamma and BMA-Frechet can calibrate extreme rainfall data well, as indicated by the small bias values. This means that the calibration results have values that are close to the observation data.
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Bayesian Model Averaging, Frechet, Gamma, Curah Hujan Ekstrem, Kalibrasi; Bayesian Model Averaging, Frechet, Gamma, Extreme, Calibration, Maximum Likelihood Estimation |
| Subjects: | Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) |
| Divisions: | Faculty of Science and Data Analytics (SCIENTICS) > Statistics > 49001-(S3) PhD Thesis |
| Depositing User: | Defi Yusti Faidah |
| Date Deposited: | 06 Feb 2024 07:51 |
| Last Modified: | 06 Feb 2024 08:00 |
| URI: | http://repository.its.ac.id/id/eprint/106318 |
Actions (login required)
 |
View Item |