[EN] We adapt the universal kriging method of prediction to contexts with uncertainty and imprecise data. Using fuzzy set theory, we develop a fuzzy universal kriging that we use to estimate quality classes for a quartz deposit. The method models the trend in the data by linearly combining functions that depend on the coordinates for each point, which are fuzzy numbers. The spatial dependence between the data is determined by means of non-fuzzy covariograms. The proposed method provides fuzzy numbers and non-fuzzy variances at unsampled locations. Finally, the estimated fuzzy numbers are translated back into quartz qualities and graphically represented. From the grade model obtained by using fuzzy kriging it has been possible to perform selective mining at the deposit depending on market demand of the four products exploited