Spatial approach to estimation of missing data

Abstract

Frequently a hydrologic network exists and the analysis of the collected data is based on the assumption that the schedule is fully followed. In practice this is very seldom the case as many factors are involved in the process of data collection, and therefore misleading conclusions may be arrived at. To obviate this the missing data must be estimated on the basis of the observations in the present and in the past. A special but frequently encountered case is addressed: at each site a sufficiently long record of observations exists and the phenomenon is characterized by a spatial configuration that moves parallel to itself in time. For such a case an extremely simple method is proposed which enables one to estimate the missing data and the standard error of the estimate. Should this error be too large spatial correlations may be employed to improve the estimates. The method is illustrated on the case of groundwater network in the coastal aquifer of Israel.