Ensemble Model Output Statistics untuk Prakiraan Cuaca Jangka Pendek

Narendra, Ribka Duhita (2017) Ensemble Model Output Statistics untuk Prakiraan Cuaca Jangka Pendek. Masters thesis, Institut Teknologi Sepuluh Nopember.

[img] Text
1315201206-Master_Thesis.pdf
Restricted to Repository staff only

Download (4MB) | Request a copy

Abstract

Operasional prakiraan cuaca jangka pendek saat ini dilakukan dengan menggunakan model Numerical Weather Prediction (NWP). Namun NWP menghasilkan prakiraan yang bias pada dimensi rendah karena NWP diukur dalam skala global sehingga menghasilkan informasi cuaca yang homogen. Sistem prediksi gabungan atau ensemble prediction system dapat mengatasi kelemahan keakuratan dari sistem prediksi tunggal. Prakiraan ensemble merupakan suatu model yang terdiri atas kumpulan dari dua atau lebih model sistem prediksi tunggal yang diverifikasi dalam waktu bersamaan. Meskipun demikian prakiraan ensemble sering bersifat underdispersive. Ensemble Model Output Statistics (EMOS) merupakan metode postprocessing yang mampu mengatasi prakiraan underdispersive dan bias. Metode tersebut didasarkan pada regresi linier berganda, yang menghasilkan prakiraan probabilistik yaitu menggunakan PDF Gaussian untuk variabel cuaca sehingga mendapatkan prakiraan yang terkalibrasi atau reliable. Mean prediktif EMOS adalah rata-rata bobot koreksi bias dari prakiraan ensemble member, dengan koefisien yang dapat diinterpretasikan sebagai kontribusi relatif dari model-model member pada ensemble dan menghasilkan prakiraan deterministik yang kompetitif. Varians prediktif EMOS merupakan fungsi linier dari varians ensemble. Penelitian ini diterapkan untuk komponen cuaca suhu maksimum (TMAKS), suhu minimum (TMIN), kelembaban relatif (RH) pada data NWP di delapan stasiun meteorologi di Jabodetabek tahun 2009-2010. Nilai prakiraan deterministik EMOS memiliki RMSE hingga 87% lebih rendah dibanding RMSE NWP. Untuk prakiraan probabilistik TMAKS di stasiun Tanjung Priok misalnya memiliki CRPS 16% lebih rendah daripada nilai CRPS raw ensemble. Prakiraan EMOS terkalibrasi lebih baik daripada NWP dan raw ensemble. =========================================================================================== Modern-day operational weather forecasts are based on numerical weather prediction (NWP) models. However, NWP produces biased forecasts on a low dimension as it measured on a global scale so that provides weather information that is homogeneous. Ensemble prediction system can overcome the weakness of a single prediction system. Ensemble forecast is a method consisting of a collection of two or more single forecast methods verified at the same time. However, such method systems are often underdispersive and tend to be biased. Ensemble Model Output Statistics (EMOS), an easy to implement postprocessing technique that addresses both forecast bias and underdispersion. The technique is based on multiple linear regression. The EMOS technique produces a probabilistic forecasts that take the form of a Gaussian predictive PDFs for weather variables. The EMOS predictive mean is a bias-corrected weighted average of the ensemble member forecasts, with coefficients that can be interpreted in terms of the relative contributions of the member models to the ensemble, and provides a highly competitive deterministic-style forecast. The EMOS predictive variance is a linear function of the ensemble variance. The EMOS technique was applied to maximum temperature, minimum temperature, and relative humidity over eight meteorogical station of the Jabodetabek area in 2009-2010, using NWP data. The deterministic-style EMOS performed about equally well and were much better than NWP. They had RMSEs for TMAKS until 87% less when compared to the NWP. The probabilistic-style EMOS performed well than raw ensemble. They had CRPS for TMAKS at Tanjung Priok station for example 16% less when compared to the raw ensemble. the EMOS predictive were better calibrated than the raw ensemble or NWP.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Ensemble Model Output Statistics, suhu maksimum, suhu minimum, kelembaban relatif, Numerical Weather Prediction, maximum temperature, minimum temperature, relative humidity
Subjects: H Social Sciences > HA Statistics
Divisions: Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis
Depositing User: Ribka Duhita Narendra
Date Deposited: 12 Oct 2017 08:01
Last Modified: 03 Jan 2018 04:52
URI: https://repository.its.ac.id/id/eprint/45822

Actions (login required)

View Item View Item