Numerous of studies have confirrmed that the kriging algorithm is most successful in the geospatial data prediction (Goovaerts, 1997). Firstly, kriging allows a great flexibility to control the characteristics of the prediction surface by the modification of variogram model. Secondly, it weighs the information locally, and the influence of a support point on the target value is controlled by the spatial correlation. In addition, kriging has the advantage to provide the expected mean square error of prediction in comparison with non-stochastic methods (Stein et al., 2002). In this study, the universal kriging (UK) algorithm is made use of to predict the wet delay. In UK, it is customary to decompose the predicted value into a deterministic component, the so-called trend, which models the large-scale variation, and a stochastic component, modeling the smooth small-scale fluctuations and the irregular part of the variation. The general model of UK is given as follows: