Determination of the best single imputation algorithm for missing rainfall data treatment

• Main Authors: Gamil Abdulraqeb Abdullah Saeed, Chuan, Zun Liang, Roslinazairimah Zakaria, Wan Nur Syahidah Wan Yusoff, Mohd Zuki Salleh
• Format: Article
• Language: English
• Published: Pusat Pengajian Sains Matematik, Fakulti Sains dan Teknologi, Universiti Kebangsaan Malaysia 2016
• Online Access: http://journalarticle.ukm.my/10205/1/jqma-12-1_2-paper7.pdf; http://journalarticle.ukm.my/10205/; http://www.ukm.my/jqma/jqma12_1_2a.html
• Institution: Universiti Kebangsaan Malaysia

The presence of missing rainfall data is inevitable due to error of recording, meteorological extremes and malfunction of instruments. Consequently, a competent imputation algorithm for missing data treatment algorithm is very much needed. There are several such efficient algorithms which have been introduced in earlier studies. However, the limitations of current algorithms are they are highly dependent on the information and homogeneity of adjoining rainfall stations. Therefore, this study is intended to introduce several single imputation algorithms for missing data treatment, which believed to be more competent in treating missing daily rainfall data without the need to depend on the information of adjoining rainfall stations. The proposed algorithms use descriptive measures of the data, including arithmetric means, geometric means, harmonic means, medians and midranges. These algorithms are tested on hourly rainfall data records from six selected rainfall stations located in the Kuantan River Basin. Based on the analysis, the proposed singular imputation algorithms, which treated missing data by geometric means, harmonic means and medians are more superior compared to the other imputation algorithms, irrespective of missing rates and rainfall stations.