Spatial analysis, involving experimental semivariogram evaluation and kriging interpolation, is performed on macroseismic intensity data assumed to represent a regionalized variable. A semivariogram is modeled, showing that data components act at different scale levels. Interpretation of the semivariogram in terms of fractal dimension allows separation of the error component from other scale-dependent components. Use of an objective best spatial-range determination for filtering eliminates the subjective choice that is usually based on data-sampling density, permitting the reconstruction of the smoothed interpolated intensity field. Results are given together with error estimation due to local variability and sampling-density distribution. The method is first applied to synthetic macroseismic data with controlled variable error content and sampling density: the ability to rebuild the original, error-free intensity field is demonstrated. Then macroseismic data from an Italian medium-intensity earthquake are analyzed and spatial intensity attenuation re-evaluated.