1. UKM Learning and Research Repository
2. PHD Thesis / Tesis PHD
3. Faculty of Science and Technology / Fakulti Sains dan Teknologi

Title:	Pemodelan siri masa kepekatan bahan pencemar udara O3 PM10 dan jerebu menerusi pendekatan kalut
Authors:	Nor Zila Abd Hamid (P52173)
Supervisor:	Mohd. Salmi Md. Noorani, Prof. Dr.
Keywords:	Bahan pencemar udara Jerebu Pendekatan kalut Air - Pollution - Mathematical models
Issue Date:	29-Jan-2015
Description:	Kajian ini adalah aplikasi pendekatan kalut ke atas pemodelan siri masa bahan pencemar udara ozon (O3), zarah terampai (PM10) dan jerebu yang dicerap mengikut jam di stesen-stesen asas dan metropolitan di Malaysia. Pemodelan kalut melibatkan dua peringkat iaitu (i) analisa dinamik siri masa dan (ii) pembinaan model peramalan. Peringkat (i) melalui kaedah Cao, kaedah msongsang dan plot ruang fasa menunjukkan kehadiran dinamik kalut dalam setiap siri masa. Peringkat (ii) melibatkan dua langkah iaitu (a) pembinaan semula ruang fasa dan (b) proses peramalan. Untuk (a), dua parameter diperlukan iaitu masa tunda  dan matra pembenaman m . Parameter  ditentukan melalui penetapan  1, kaedah purata maklumat bersama dan kaedah baharu  songsang. Parameter m dikira melalui kaedah Cao dan kaedah msongsang. Langkah (b) dijalankan melalui kaedah penghampiran purata setempat (kpps), kaedah penghampiran linear setempat (kpls) dan kaedah penambahbaikan penghampiran linear setempat (kppls). Peramalan satujam ke hadapan mendapati kaedah  songsang membantu meningkatkan prestasi peramalan. Prestasi model kppls adalah mengatasi model-model kpps dan kpls. Model kalut O3 boleh meramal sehingga dua hari ke hadapan manakala bagi model kalut PM10, peramalan semakin merosot bermula dua jam ke hadapan. Peramalan pelbagailangkah memperlihatkan penurunan prestasi model dengan pertambahan bilangan langkah peramalan. Model peramalan kalut pemetaan logistik adalah cemerlang dengan nilai pekali kolerasi satu. Tiada model peramalan siri masa bahan pencemar udara dengan pekali kolerasi satu kerana siri masa ini mengandungi gangguan hingar. Mengecilkan skop siri masa mengikut musim adalah berbaloi memandangkan prestasi model peramalan PM10 ketika jerebu adalah lebih baik berbanding tempoh biasa. Prestasi model kalut adalah lebih cemerlang berbanding model tradisional regresi linear. Pendekatan kalut telah berjaya menganalisa dan meramal siri masa O3, PM10 dan jerebu. Dengan hasil-hasil yang diperoleh, diharapkan kajian ini mampu memudahkan pengurusan pencemaran udara oleh pihak-pihak berkepentingan dan menjadi petunjuk baharu dalam arena pemodelan siri masa bahan pencemar udara di Malaysia.,This study is an application of the chaotic approach in modeling the time series of air pollutants namely ozone (O3), particulate matter (PM10) and haze observed hourly at the Malaysian background and metropolitan stations. Chaotic modeling involves two stages, namely (i) dynamic analysis of time series and (ii) development of the prediction model. Stage (i) through Cao method, m inverse method and phase space plot discovered the presence of chaotic dynamics in each time series. Stage (ii) involves two steps, namely (a) reconstruction of phase space and (b) prediction process. For (a), two parameters are needed, the delay time  and embedding dimension m . Parameter  is determined by the setting of  1, the average mutual information method and a new method called the   inverse method. Parameter m is calculated by the Cao and m inverse methods. Step (b) is performed by means of the local average approximation method (kpps), local linear approximation method (kpls) and the improved local linear approximation method (kppls). The one-hour ahead prediction models found that the   inverse method helped to improve the prediction performance. The performance of kppls model overcomes both kpps and kpls models. Chaotic model of O3 can predict up to two days ahead while for PM10 model, the prediction performance declined starting from two hour ahead prediction. The multistep ahead prediction models show the decrease of the prediction performance with increasing number of prediction steps. Prediction model for the chaotic logistic map is excellent where the value of correlation coefficient is one. However, there is no prediction model of air pollutants time series with a correlation coefficient equal to one as the time series contains noise disturbance. Narrowing the scope of the time series according to its season was worth it as the performance of the prediction model of PM10 during haze episode is much better than the normal period. Prediction performance through chaotic model is better than the traditional model of linear regression. Chaotic approach has successfully analyzed and predicted the time series of O3, PM10 and haze. With the promising results, it is hoped that this study will facilitate the management of air pollution by the stakeholders and it becomes the new guidance in the realm of the air pollutants time series modeling in Malaysia.,PhD
Pages:	238
Call Number:	TD883.1.N646 2015 tesis
Publisher:	UKM, Bangi
Appears in Collections:	Faculty of Science and Technology / Fakulti Sains dan Teknologi

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.