Prediction Of Pm10 Concentrations Using Extreme Value Distributions (Evd) Classical And Bayesian Approaches

Ahmat, Hasfazilah (2016) Prediction Of Pm10 Concentrations Using Extreme Value Distributions (Evd) Classical And Bayesian Approaches. ["eprint_fieldopt_thesis_type_phd" not defined] thesis, Universiti Sains Malaysia.

[img]
Preview
PDF - Submitted Version
Download (593kB) | Preview

Abstract

Keadaan zarahan tinggi (jerebu) secara umumnya dikaitkan dengan kehadiran PM10 atau PM2.5. Ia adalah penting untuk memaklumkan kepada umum terhadap tahap PM10 dan kepentingannya supaya langkah-langkah penyesuaian yang lebih berkesan dapat diambil bagi kalangan umum yang terjejas. Kajian ini dijalankan dengan objektif untuk membandingkan Taburan Nilai Melampau (EVD) menggunakan pendekatan konvensional dan Bayesian dan menggunakan taburan terbaik untuk peramalan kepekatan PM10 pada masa hadapan. Ketika ini, tiada pendekatan Bayesian di dalam kajian kepekatan PM10. Rekod daripada lapan stesen pengawasan di Semenanjung Malaysia telah dipilih untuk tempoh 1 Januari 2000 hingga 31 Disember 2012 selepas analisis awal untuk menilai kewujudan nilai melampau. Taburan dengan pengukuran ralat yang terkecil dan pengukuran kejituan tertinggi di lima stesen pemantauan  Bukit Rambai, Jerantut, Nilai, Pasir Gudang dan Shah Alam adalah taburan menggunakan kaedah Bayesian dengan kebolehjadian GEV dan taburan prior tanpa maklumat menggunakan taburan seragam. Walau bagaimanapun, bagi Klang dan Seberang Jaya taburan EVD GEV disimpulkan sebagai taburan yang terbaik dan EVD dua parameter Weibull adalah taburan terbaik untuk Perai. Pendekatan Bayesian adalah lebih unggul dari kaedah konvensional apabila menggunakan data maksimum harian dan boleh digunakan untuk menilai tahap kepekatan tinggi PM10 untuk penggubal dasar melaksanakan dasar-dasar yang lebih berkesan untuk mewujudkan persekitaran yang lebih bersih. ________________________________________________________________________________________________________________________ High particulate event (haze) is generally associated with presence of PM10 or PM2.5. It is important to make known to public of PM10 level and its importance for more effective adaptation measures among the affected public. This study was conducted with the objectives to compare the best Extreme Value Distributions (EVD) using the conventional and Bayesian approaches and use the best distribution for the prediction of future PM10 exceedances. Currently, there is none on the application of Bayesian approach in the study of PM10 concentrations. Records from eight monitoring stations in the Peninsular Malaysia were selected for the period of 1st January 2000 to 31st December 2012 after preliminary analysis to check for the existance of extreme values. The distribution with the smallest error measures and highest accuracy measures in five of the monitoring stations  Bukit Rambai, Jerantut, Nilai, Pasir Gudang and Shah Alam was the Bayesian GEV likelihood with uniform non-informative prior distribution. However, for Klang and Seberang Jaya the EVD GEV distribution was concluded as the best distribution and EVD two-parameter Weibull was the best distribution for Perai. The Bayesian approach is superior than the conventional method using the daily maximum data and can be used to assess high level of PM10 concentrations for the policy makers to implement effective policies to create cleaner environment.

Item Type: Thesis (["eprint_fieldopt_thesis_type_phd" not defined])
Additional Information: Full text is available at http://irplus.eng.usm.my:8080/ir_plus/institutionalPublicationPublicView.action?institutionalItemId=3165
Subjects: T Technology
T Technology > TA Engineering (General). Civil engineering (General) > TA1-2040 Engineering (General). Civil engineering (General)
Divisions: Kampus Kejuruteraan (Engineering Campus) > Pusat Pengajian Kejuruteraan Awam (School of Civil Engineering) > Thesis
Depositing User: Mr Mohd Jasnizam Mohd Salleh
Date Deposited: 04 Sep 2018 09:06
Last Modified: 04 Sep 2018 09:06
URI: http://eprints.usm.my/id/eprint/41717

Actions (login required)

View Item View Item
Share