A quick gap-filling of missing hydrometeorological data

Data-gaps are ubiquitous in hydrometeorological time series and filling these values remains still a challenge. Since datasets without missing values may be a prerequisite in performing many analyses, a quick and efficient gap-filling methodology is required. In this study the problem of filling sporadic, single-value gaps using time-adjacent observations from the same location is investigated. The applicability of a local average (i.e., based on few neighboring in time observations) is examined and its advantages over the sample average (i.e., using the whole dataset) are illustrated. The analysis reveals that a quick and very efficient (i.e., minimum mean squared estimation error) gap-filling is achieved by combining a strictly local average (i.e., using one observation before and one after the missing value) with the sample mean.