CONDITION NUMBERS AS CRITERIA FOR EVALUATION OF ATMOSPHERIC AEROSOL MEASUREMENT TECHNIQUES

The steps required to compute the condition numbers and a description of appropriate scaling are discussed. The method is illustrated with three examples: the selection of indicator elements in source-receptor models, the relevance of condition number to classifier type and inversion algorithms, and in the location of experiments in size distribution measurements. From the magnitude of the condition number it is possible to choose a priori likely candidates for indicator elements and to discriminate among algorithms used to invert penetration measurements for the size distribution.