PENERAPAN IMPUTASI LOCF DAN CROSS MEAN DALAM PENGISIAN DATA KOSONG PADA CURAH HUJAN HARIAN ARG

Abstract

The number of installed Automatic Rain Gauges (ARG) today has not been optimally utilized. It is because ARG that works automatically often has missing data due to technical and network problems raising doubts about its accuracy. The data used are ARG rainfall data in 10 minute periods during 2021 and rainfall data from conventional gauge at the same location. The data will be processed until it becomes daily data and will be recovered by missing data entry worked by the Python programming language. Because the ARG data is the longitudinal data type, missing data entry will use LOCF and cross mean imputation. The validity test will compare the recovered ARG data with the conventional gauge data by calculating the MAE, RMSE, and correlation coefficient. The results showed that missing data entry could reduce the percentage of missing from 21.4% to 1.1%. The result of validity tests showed that ARG could produce accurate data determined by a lower error (MAE=0.998mm, RMSE=2.253mm) and a very high correlation (r=0.966). With a higher percentage of data completeness and excellent accuracy, the data usage will become more extensive to provide more benefits, especially for the need of analysis, forecasting, data services, and research.