Methodological approaches for imputing missing data into monthly flows series

Missing data is one of the main difficulties in working with fluviometric records. Database gaps may result from stations components problems, monitoring interruptions and lack observers. Incomplete series analysis generates uncertain results, negatively impacting water resources management. Thus, proper missing consideration very important to ensure better information quality. This work aims analyze, comparatively, imputation methodologies monthly river-flow time series, considering, as a case study, Doce River, located Southeast Brazil. were simulated 5%, 10%, 15%, 25% 40% proportions following random distribution pattern, ignoring generation mechanisms. Ten used: arithmetic mean, median, simple multiple linear regression, regional weighting, spline Stineman interpolation, Kalman smoothing, maximum likelihood. Their performances compared through bias, root mean square error, absolute percentage determination coefficient concordance index. Results indicate that for 5% data, any methodology imputing can be considered, recommending caution method application. However, proportion increases, it recommended use likelihood when there are support imputation, interpolation Smoothing methods only studied available.
 Keywords: river, data.