Declustering Peaks Over Threshold Pada Data Curah Hujan Ekstrim Dependen Di Sentra Produksi Padi Jawa Timur

Malika, Rosna (2014) Declustering Peaks Over Threshold Pada Data Curah Hujan Ekstrim Dependen Di Sentra Produksi Padi Jawa Timur. Other thesis, Insititut Teknologi Sepuluh Nopember.

Text 1310100073_Undergraduate_Thesis.pdf Download (4MB)

Abstract

Iklim ekstrim merupakan fenomena langka, namun dapat memicu dampak merugikan salah satunya bagi sektor pertanian. Curah hujan yang ekstrim perlu diidentifikasi untuk mengurangi kerugian pertanian. Extreme Value Theory (EVT) merupakan salah satu metode statistika yang digunakan untuk mempelajari perilaku nilai-nilai ekstrim. Salah satu yang menjadi penting dalam model extreme value adalah return level. Return level merupakan nilai maksimum yang diharapkan akan dilampaui satu kali dalam jangka waktu tertentu. Sebelum menentukan nilai return level, terlebih dahulu harus diketahui nilai parameter distribusi EVT yang dipilih, dengan syarat urutan data harus independen. Namun pada kenyataannya, seringkali ditemukan data ekstrim yang dependen, akibatnya akan diperoleh nilai return level yang tidak valid. Salah satu cara untuk mengatasi hal tersebut dengan metode declustering. Penelitian ini menganalisis data curah hujan ekstrim dependen dengan pendekatan Peaks Over Threshold (POT) serta mengatasi dependensi data dengan declustering. Hasil penelitian menunjukkan bahwa metode declustering berhasil mengatasi dependensi data, namun tidak menghasilkan return level yang lebih baik daripada non declustering. Hal ini ditunjukkan nilai RMSE return level declustering yang sama dengan non declustering. Oleh karena itu, pada analisis selanjutnya sebaiknya menggunakan metode pemodelan parameter Generalized Pareto Distribution.
===============================================================================================================================
Extreme climate is a rare phenomenon that lead bad impact for the agricultural sector. Extreme rainfall event need to be identified to minimize losses due to the incident. Extreme Value Theory (EVT) is one of the statistical methods that used to study the behavior of extreme value. It is usually more convenient to interpret extreme value models in terms of return levels. To determine the value of return level, must be known parameter values of distribution first. Each of extreme value models derived so far has been obtained through mathematical arguments that assume an underlying process consisting of a sequence of independent random variables. However, in the real case this condition is often violated, then its reducing the validity of conclusions. This research was conducted in five districts as the center of paddy production in East Java. In this study the method of approach to the Peaks Over Threshold (POT) is used to determine the return level, and its also used declustering method to solve the problem of dependent data. The results showed that the order of the data become more independent when declustering method is applied. But the RMSE of return level by using declustering method have same value as RMSE of return level with non declustering method. Furthermore, it will be better if the data analyze with approach to modelling of Generalized Pareto Distribution parameter.

Item Type:	Thesis (Other)
Additional Information:	RSSt 519.5 Mal d-1, 2014
Uncontrolled Keywords:	Curah Hujan, Extreme Value Theory (EVT), Peaks Over Threshold, Generalized Pareto Distribution, Declustering, Return Level.
Subjects:	Q Science > QC Physics > QC871 Meteorology--Observations.
Divisions:	Faculty of Mathematics and Science > Statistics > 49201-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	03 Jan 2024 06:36
Last Modified:	03 Jan 2024 06:36
URI:	http://repository.its.ac.id/id/eprint/105348

Actions (login required)

View Item