Estimación e imputación de datos faltantes mediante métodos de interpolación espacial para precipitación mensual acumulada en el departamento de Antioquia durante el periodo 2014-2018

Abstract

Missing values are common in the databases we work on a daily basis, knowing what to do with those missing data is essential, sometimes the immediate solution is to remove the records and lose information that can be of great value, the purpose is to take advantage of the information that is available in the greatest way, for this objective, there are several techniques that allow the efficient imputation of the missing values. This paper seeks to contrast methods of spatial interpolation in the imputation of missing values of accumulated monthly precipitation in the department of Antioquia during the 2014-2018 period, with data provided by the Institute of Hydrology, Meteorology and Environmental Studies (IDEAM), to achieve this goal, weighted inverse distance (IDW), thin plate spline (SPT) and ordinary kriging are used as evaluated methods through the root of the mean square error (RMSE). Ordinary Kriging was effective when you have more than 10% of the missing values, the weighted inverse distance method was the one that yielded the best results when you have 5% of the missing values. The results obtained were applied to the data corresponding to the year 2018