Procedures for analysis of data from a hydrologic network usually assume that a collection schedule is fully followed. In practice, however, data are frequently missing (or simply unknown, in regions where measurements were not planned), and must be estimated on the basis of other available observations and information. During the 1980s a simple method to estimate missing data (e.g. on water levels) by use of a spatial approach was developed. Spatial and temporal variations of a measurable state variable (such as hydraulic head) were assumed to be described by a deterministic function, with randomness introduced by the choice of location and timing of observations. The function describes a spatial surface, and allows only for fluctuation parallel to itself along the time axis. Missing (or unknown) data can then be estimated from predicted surfaces. The constraint of parallel fluctuations has now been removed. The proposed method, which is more general, is based on integration of two sources of information: prior estimates and online estimates. It allows for continuous updating of the function as additional information becomes available, and as a consequence, missing (or unknown) data can be more accurately estimated.
