Hazhiah, Indria Tsani (2016) Spatial Extreme Value With Bayesian Hierarchical Model (Studi Kasus: Pemodelan Curah Hujan Ekstrem Di Kabupaten Ngawi). Masters thesis, Institut Technology Sepuluh Nopember.
Preview |
Text
1314201204-Master_Thesis.pdf - Accepted Version Download (3MB) | Preview |
Abstract
Ekstrem Curah hujan ekstrem merupakan fenomena alam yang langka sehingga
tergolong salah satu iklim ekstrem yang dapat memicu bencana alam seperti
banjir dan longsor. Bencana tersebut menjadi penyebab keresahan masyarakat dan
menimbulkan berbagai penyakit pada masyarakat di wilayah tersebut sehingga
perlu adanya identifikasi curah hujan ekstrem untuk mengurangi dampak negatif
yang ditimbulkan. Extreme value theory (EVT) adalah salah satu metode statistika
yang digunakan untuk mempelajari perilaku-perilaku nilai-nilai ekstrem. Salah
satu yang dihitung dalam EVT adalah return level yaitu suatu level ( kuantil
dengan peluang 1/T) kejadian ekstrem yang diharapkan terlampaui dalam periode
T. Sebelum menentukan nilai return level, terlebih dahulu harus diketahui
parameter distribusi EVT yang dipilih dengan menggunakan data yang banyak.
Namun kenyataannya peristiwa curah hujan ekstrem jarang terjadi sehingga data
pengamatan jumlahnya terbatas yang digunakan dalam penaksiran parameter.
Salah satu metode yang diharapkan dapat mengatasi jumlah data pengamatan
yang terbatas adalah Bayesian. Keunggulan Bayesian salah satunya adalah dapat
melakukan updating informasi terhadap likelihood melalui distribusi prior karena
informasi pada likelihood terbatas apabila data pengamatan sedikit. Selain itu,
informasi geografi dan klimatologi yang tidak bisa diakomodasi secara penuh
dalam model diharapkan dapat ditangkap oleh struktur spasial pada Bayesian
Hierarchical Model (BHM). Penelitian ini menganalisis data curah hujan ekstrem
di Kabupaten Ngawi, Jawa Timur dengan pendekatan BHM pada Generalized
Pareto Distribution (GPD) yang merupakan distribusi asimtotis dari Peak Over
Threshold (POT). Pada penelitian ini dengan menggunakan BHM menghasilkan
return level yang lebih baik dilihat dari SE yang terkecil dibandingkan dengan
hanya menggunakan POT.
============================================================================================================
Extreme precipitation is a rare natural phenomenon that classified as one of
extreme climates. It could lead to natural disasters such as floods and landslides
which cause public unrest and inflict various diseases in the region thus need to
identify the extreme precipitation in order to reduce the negative impact of the
causes it. Extreme value theory (EVT) is one of statistical models that used to
learn the behaviors of extreme values. The one of the counted in the extreme value
theory is return level. Return level is maximum value the future periods that
informs time scales incident of the next extreme precipitation else. The first that
should know is parameter of distribution estimation in extreme value theory that
be chosen by using many data before be determine the value of return level. The
actually extreme precipitation events are rare thus we just have few amount of
observational data that used in the valuation of parameters. One of the models that
hope coping limited amount of observational data is Bayesian. One of the
Bayesian advantage could be updating of information on the likelihood through
the information on the prior distribution because the limited information obtained
from the likelihood if only using a small data. Coley (2007) states that the extreme
values at different locations are influenced by factors of geography and
climatology. Factors Geography and climatology could not be accommodated
fully in the model of these deficiencies was captured by the spatial structure of
Bayesian hierarchical models. This paper analyzed data extreme precipitation in
Ngawi regency of East Java with the approach of the Bayesian hierarchical model
(BHM) on Generalized Pareto Distribution which a asimtotis distribution of peak
over thershold (POT) methods. The results of this research showed that BHM be
able cope a limited amount of observational data and Factors Geography and
climatology that could not be accommodated fully in the model. This is shown in
the result data using BHM that has Square Error (SE) of return level smaller than
without using the BHM. BHM approach applied on generalized Pareto
distribution which is asimtotis of POT.
| Item Type: | Thesis (Masters) |
|---|---|
| Additional Information: | RTSt 519.542 Haz s |
| Uncontrolled Keywords: | Curah Hujan, EVT, Peaks Over Threshold, Generalized Pareto Distribution, Return Level, Bayesian hierarchical model |
| Subjects: | Q Science > QA Mathematics > QA279.5 Bayesian statistical decision theory. |
| Divisions: | Faculty of Mathematics and Science > Statistics > 49101-(S2) Master Thesis |
| Depositing User: | Mr. Tondo Indra Nyata |
| Date Deposited: | 22 Jan 2020 08:01 |
| Last Modified: | 29 Apr 2024 00:59 |
| URI: | http://repository.its.ac.id/id/eprint/72902 |
Actions (login required)
 |
View Item |