Treatment on Missing Data of Wind Speed Measurement in Automatic Weather Station Using Multivariate Imputation Chained Equation

Automatic Weather Station (AWS) has an anemometer to measure surface wind speed. In operation, the anemometer has a potential to damage internal components, causing data loss. Imputation is an alternative solution for missing data treatment. This study aims to utilize the spatial relationship of wind speed measurement on multi sites AWS in the imputation model. The imputation of the wind speed dataset uses Multivariate Imputation Chained Equation (MICE) algorithm with the input being the results of anemometer measurements at several surrounded points. The research focuses on the AWS Cisurupan site in the period 1 January 2019 - 31 December 2022. The wind speed dataset has a sampling interval of 10 minutes. MICE algorithm is able to overcome missing data on AWS wind speed measurement using the AWS neighbors input. MICE-based data imputation performance still meets WMO operational requirements for wind speed measurements with a maximum percentage of missing data of 15%. Mean average percentage error (MAPE) value obtained is less than 10%.