Infilling of missing values of hydrometeorological time series in distributed relational databases

Infilling of missing values and extension of records are related procedures, performed either to facilitate other procedures or to improve the statistical characteristics of a time series. In this thesis, some of the most important existing methods are examined. These are: mean value, reciprocal distance, normal ratio, linear regression, maintenance-of-variance extension (MOVE), use of an AR(1) and PAR(1) model, and use of a multivariate stohastic model. Most of these have been implemented in a computer program in order to serve the needs of the national databank of hydrological and meteorological data, named "Hydroscope", which is based on a distributed relational database. Some significant elements of the implementation are described. Finally, an application on daily rainfall time series, together with a brief evaluation of the methods, is presented.