Imputation of Missing Rainfall Data Using Revised Normal Ratio Method

Accurate estimation values are very necessary in imputation of missing rainfall data to provide precise information on meteorological characteristics. Therefore, a modification of the old normal ratio method by adapting trimmed mean and geometric median is proposed to produce more accurate estimation results. The performances for each of the methods were assessed at six different levels of missingness and three levels of outliers. The results indicated that the modified methods improved the estimates based on several performance criteria. The modified normal ratio based on geometric median was found to be the best method in imputing missing rainfall values at all levels of missingness and outliers. The verification results indicated that the dataset imputed by this method successfully fits to a lognormal distribution. Generally, the proposed methods are highly recommended to be applied as an alternative method in imputation of missing values, particularly in the existence of outliers in the dataset.