In a geophysical survey, one of the main challenges is to estimate the physical parameter using limited geophysical field data with noise. Geophysical datasets are measured with sparse sampling in a survey. However, the limited data constrain the geophysical interpretation. Traditionally, the field data has been interpolated using mathematical algorithm. In many cases, the estimated field data uncertainties are required to determine which earth models are consistent with the observations. A model-based data-estimation method can provide precise information for imaging and interpretation. The approach used in this paper is based on a stochastic partial differential equation, and it is employed to predict the geophysical data. With this statistical model-based approach, the sparse sample from a survey is used to estimate the underlying spatial surface, and it is assumed that the predicted geophysical data have the same probability density function as the observed data. Furthermore, this method can return the uncertainties of the prediction. Both the synthetic data and the gold mineral exploration field data cases illustrate that this approach leads to better results than traditional methods.