This document is concerned with developing linear regression techniques for interpolation of air pollutant concentrations over an area and, using these techniques in a computer program for determining the optimum location of air pollution observing stations. The general interpolation problem is surveyed and the advantages of using linear regression formulas as interpolation formulas are discussed. The case of observations containing errors of observation or effects of limited range of influence is emphasized. Since the use of linear regression methods depends on knowledge of the two-point correlation function for pollutant concentration measures, the construction of correlation coefficients from synthetic data is taken up. Attention is given to the estimation of residual variances or the effects of limited range of influence, using Factor Analysis. In extending these methods to a continuous formulation in integral equation form, the lack of accuracy in the integral equation solution is more important than the statistical significance of the data unless the residual variances are removed. If this is done, then the tests for accuracy and statistical significance are reconciled. If the user carefully handles the residual variances in constructing program input, difficulties encountered in code development are avoidable.

Prepared by Kaman Sciences Corporation, Colorado Springs, Colo., under contract no. 68-02-0699. Includes appendices. Includes bibliographical references (pages 194-197).