Abstract
This paper gives a general two-term approximation for the tail probability of the maxima of a class of differentiable Gaussian random fields and illustrates its potential statistical applications.