Exploratory functional extreme data analysis: An application to rainfall data in Peninsular Malaysia

• Published in: AIP Conference Proceedings
• Main Author: Adnan N.I.M.; Adam M.B.; Shafie N.A.; Babura B.I.; Hassan R.
• Format: Conference paper
• Language: English
• Published: American Institute of Physics Inc. 2023
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85166763433&doi=10.1063%2f5.0117189&partnerID=40&md5=d777a299fdd54024a22bf2aca1c84d89

Due to the geographical factor, Malaysia is divided into two parts which are the Peninsular Malaysia and West Borneo. Frequently, the highest number of rainfalls occur per year in Peninsular Malaysia is between November and December, while in the West Borneo is between December and February. Due to this situation, some regions in Malaysia receive heavy rainfalls and floods every year. This study aims to explore the characteristics of an extreme rainfalls data in Malaysia by using a new functional data analysis method and to observe the performance of three types of extreme data as a result of the new functional data analysis method. Functional data analysis is one of the techniques to transform a discrete or continuous data into a function. The monthly maxima and r-largest data approaches of extreme value theory were used in selecting the extreme data. This study used a data of 30 years period (1987-2016) collected from five stations of five different states in peninsular Malaysia. Functional descriptive statistics was performed to explore the behaviors and characteristics of the extreme rainfall data. Hence, this finding concludes that most of the stations recorded the extreme rainfalls amount in the early and end of each year. The latest extreme rainfalls amount was slightly increasing over the past years. Based on the functional extreme rainfall data, the extreme events such as very heavy and heavy rainfalls level can be identified specifically at particular rainfalls amount level and time by observing the pattern of each data. Therefore, functional data analysis is suggested as one of the techniques to summarize the data with additional pattern information instead of the classical method in conventional statistics. © 2023 Author(s).