Pemodelan Statistical Downscalling Dengan Projection Pursuit Regression Dan Adaptive Spline Threshold Autoregression Untuk Peramalan Curah Hujan

Santoso, Noviyanti (2014) Pemodelan Statistical Downscalling Dengan Projection Pursuit Regression Dan Adaptive Spline Threshold Autoregression Untuk Peramalan Curah Hujan. Masters thesis, Insititut Teknologi Sepuluh Nopember.

Text 1312201027_Master_Thesis.pdf Download (1MB)

Abstract

Penyediaan data iklim merupakan hal yang penting sebagai upaya untuk mengetahui perubahan iklim di suatu wilayah. Sampai saat ini, general circulation model (GCM) diakui banyak pihak sebagai alat untuk upaya memahami sistem iklim. GCM merupakan alat prediksi utama iklim dan cuaca secara numerik serta sebagai informasi primer untuk menilai pengaruh perubahan iklim. Namun informasi GCM yang dihasilkan masih berskala global (ratusan km), sehingga sulit untuk mendapatkan informasi skala regional. Salah satu upaya untuk mendapatkan informasi skala lokal perlu digunakan statistical downscalling (SD). Metode SD adalah memprediksi curah hujan lokal beresolusi tinggi berdasarkan data General Circulation Model (GCM) berskala global melalui model regresi. Data GCM merupakan data spasial dan temporal yang memungkinkan terjadi korelasi spasial antar grid yang berbeda dalam satu domain dan autokorelasi dalam data deret waktu. Selain itu bentuk fungsi regresi yang tidak diketahui juga menjadi permasalahan tersendiri. Pada p enelitian ini, metode SD yang digunakan adalah Projection Pursuit Regression (PPR) dan Adaptive Spline Threshold Autoregression (ASTAR). Model dengan metode ASTAR digunakan untuk menyusun model error dari hasil model PPR. Model gabungan antara PPR dan ASTAR selanjutnya disebut model hybrid. Oleh karena itu dua pendekatan model yang digunakan, yaitu model hybrid dan non hybrid. Kriteria kebaikan model untuk validasi menggunakan nilai Root Mean Square Error Prediction (RMSEP) dan nilai.
==============================================================================================================================
Climate data supplying is one of an important things as an effort to detect climate change in some certain areas. Till now days, general circulation model (GCM) is admitted as a device to understand the climate system. GCM is a major prediction device numerically for weather and climate also as primary information to verify the effect of climate change. Yet, the result of GCM information is still in global scale (hundred kilometers), so it is difficult to gain information in regional scale. One of the efforts to gain the local scale information is needed the usage of statistical downscaling (SD). SD method is able to predict local fall of rain with high resolution based on GCM global scale’s data through regression model. GCM data is temporal and spatial data which enable a spatial correlation between different grid in one domain and autocorrelation in time line data to happen. Besides, the unknown form of regression function is also becoming a problem. In this research, SD method which is used are Projection Pursuit Regression (PPR) for dimension reduction and Adaptive Spline Threshold Autoregression (ASTAR). Model with ASTAR method is used to arrange error model from PPR model result. Combination model between PPR and ASTAR is known as hybrid model. Hence, there are two approaching model which is used, hybrid and non hybrid model. The criteria of a good model for validation uses Root Mean Square Error Prediction (RMSEP) and

Item Type:	Thesis (Masters)
Additional Information:	RTSt 519.536 San p-1, 2014
Uncontrolled Keywords:	ASTAR, GCM, Statistical Downscaling, PPR, Reduksi dimensi, Dimension reduction
Subjects:	Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression
Divisions:	Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	08 Jan 2024 06:05
Last Modified:	08 Jan 2024 06:06
URI:	http://repository.its.ac.id/id/eprint/105397

Actions (login required)

View Item