For the prediction model: Z(s) = U(s) + e(s), in which Z denotes the predicted value, U denotes the global trend, e denotes random, autocorrelated error(autocorrelation + nugget), s indicates the location or point; if the trend U(s) is a linear function of the point coordinates, e.g. U(s) = b0+b1X+b2Y+b3X2+b4Y2+b5XY(a second-order polynomial trend), and this trend vary and the coefficients(b) are unknown, this model is called universal kriging.
