An evaluation of the performance of imputation methods for missing meteorological data in Burkina Faso and Senegal

Addressing data incompleteness issues is crucial for reliable climate studies, especially in regions like Africa that commonly experience data gaps. This study aims to evaluate the performance of five imputation methods (knn, ppca, mice, imputeTS, and missForest) on meteorological data from stations in Burkina Faso and Senegal. The imputed data is compared with ERA5 reanalysis data to validate its accuracy. Temperature, relative humidity, and precipitation observations from the GSOD dataset (1973-2020) were used, creating subsets with missing rates of 5, 10, 20, 30 and 40%. An evaluation was conducted using the Taylor diagram and Kling-Gupta Efficiency (KGE). The results show a good estimation of temperature and relative humidity time series, with missForest performing the best for handling missing values. Precipitation estimation was less accurate, but there was strong agreement between estimated and observed data. ImputeTS was recommended for precipitation. Spatial consistency between imputed data and ERA5 reanalysis products was found. This research improves the quality of meteorological data, provides essential information about climatic characteristics, and serves as a foundation for climate change and weather modeling studies. Key words: Meteorological data, imputation methods, Senegal, Burkina Faso.