There are two usual ways for modeling the realizations of multivariate random fields: Applying kriging individually on each variable or using cokriging, which considers the spatial cross-dependence between the variables.It has been shown that the second way, in general, allows a prediction variance reduction.The use of cokriging may be limited in practice when the number of variables increases because estimating the linear model of coregionalization (the cross-dependence between the variables) becomes complex.This work explores ordinary kriging for functional data based on Andrews curves as an alternative to the classical multivariate approach.Employing a simulation study, we compare the predictor proposed with kriging and cokriging.The methodology is applied to an environmental dataset.