Qona`ah, Niswatul (2018) Spatial Bayesia Model Averaging untuk Prakiraan Cuaca Jangka Pendek Terkalibrasi. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
Thesis_Niswatul Qona'ah.pdf - Accepted Version Download (8MB) | Preview |
Abstract
Hasil prakiraan cuaca menggunakan ensemble seringkali masih bersifat underdispersive dan tidak terkalibrasi. Bayesian Model Averaging (BMA) adalah metode postprocessing statistik untuk mengkalibrasi prakiraan ensemble dan membuat interval prediktif yang lebih reliabel. BMA menghasilkan ramalan terkalibrasi, tetapi, tidak mempertimbangkan korelasi spasial. Tidak seperti BMA, Geostatistical Output Perturbation (GOP) mempertimbangkan korelasi spasial di antara beberapa lokasi secara serentak. Metode ini memiliki parameter spasial yang memodifikasi output prakiraan untuk menangkap informasi spasial. Namun, GOP hanya berlaku untuk perkiraan deterministik tunggal. Spasial Bayesian Model Averaging (Spatial BMA) adalah metode yang menggabungkan BMA dan GOP. Metode ini memperluas BMA untuk menghasilkan perkiraan probabilistik yang dikalibrasi untuk beberapa lokasi secara bersamaan. Anggota ensemble Spatial BMA diperoleh dengan memakaikan ramalan cuaca dengan hasil simulasi eror spasial, dalam proporsi yang sesuai dengan bobot BMA. Suhu menjadi fokus penelitian ini karena cenderung memiliki korelasi yang relatif kuat dengan unsur-unsur lainnya. Analisis yang diterapkan untuk mengkalibrasi ramalan suhu di delapan lokasi meteorologi di Jabodetabek, Indonesia adalah BMA Spasial. Awalnya, Partial Least Square (PLS), Principal Component Regression (PCR), dan metode Regression Ridge menjadi anggota ensemble BMA. Prakiraan suhu BMA simultan digunakan untuk mendapatkan simulasi kesalahan yang secara spasial berkorelasi yang memodifikasi prakiraan suhu. Untuk masa pelatihan selama 30 hari, prakiraan suhu Spatial BMA di 8 stasiun meteorologi tergolong cukup baik. Metode ini mampu mengkalibrasi ramalan suhu lebih baik daripada raw ensemble yang nilai coverage-nya mendekati standar 50% tetapi tidak lebih baik daripada menggunakan BMA asli. Berdasarkan beberapa indikator ramalan cuaca, seperti RMSE dan CRPS, dalam hal ini BMA asli lebih baik dari Spatial BMA dalam hal akurasi dan presisi. Namun, jika dibandingkan dengan GOP, prakiraan suhu udara menggunakan Spatial BMA menunjukkan akurasi dan presisi yang lebih tinggi.
==================================================================================================
The output of weather forecasts using ensemble are often underdispersive and uncalibrated. Bayesian Model Averaging (BMA) is a statistical postprocessing method to calibrate the ensemble forecasts and create more reliable predictive interval. BMA producing the calibrated forecast, but, does not consider spatial correlation. Unlike BMA, Geostatistical Output Perturbation (GOP) consider spatial correlation among several locations altogether. It has spatial parameters that modifying the forecast output to capture spatial information. But, GOP only applies to single deterministic forecasts. Spatial Bayesian Model Averaging (Spatial BMA) is a method which combines BMA and GOP. It extends BMA to produce calibrated probabilistic forecasts for several location simultaneously. The members of the spatial BMA ensemble are obtained by dressing the weather forecasts with simulated spatially correlated error, in proportions that correspond to the BMA weights. Temperature becomes the focus of this study because it tends to have a relatively strong correlation with the other elements. Analysis applied to calibrate the temperature forecast at eight meteorological sites in Jakarta, Indonesia is Spatial BMA. Initially, Partial Least Square (PLS), Principal Component Regression (PCR), and Regression Ridge methods become ensemble members of BMA. Temperature forecasts of BMA are used to obtain simulated spatially correlated error that modifying temperature forecasts. For training period over 30 days, Spatial BMA temperature forecasts at 8 meteorological stations belongs to good enough. It is able to calibrate the temperature forecast better than raw ensemble which coverage comes closer to the standard 50% but it is not better than using original BMA. Based on some indicators of weather forecast, such as RMSE and CRPS, in this case original BMA is better than Spatial BMA in terms of accuracy and precision.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | BMA, Ensemble, GOP, Spatial BMA |
| Subjects: | G Geography. Anthropology. Recreation > GA Mathematical geography. Cartography Q Science > QA Mathematics Q Science > QA Mathematics > QA276 Mathematical statistics. Time-series analysis. Failure time data analysis. Survival analysis (Biometry) Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression |
| Depositing User: | Niswatul Qona'ah |
| Date Deposited: | 05 Aug 2021 00:05 |
| Last Modified: | 05 Aug 2021 00:05 |
| URI: | http://repository.its.ac.id/id/eprint/58292 |
Actions (login required)
 |
View Item |