Developing and Testing a Soil Property Database for Forensic Applications in Southern California.

The research sought to develop and test a forensic database of surface soil variability within previously mapped geologic and soil units in southern California. This type of database could be used to link suspects to crime scenes or determine source locations of soil sample evidence. Variability was evaluated using (i) color, (ii) magnetic susceptibility, and (iii) particle-size distribution. Soil properties were analyzed for their ability to discriminate source areas using stepwise discriminant analysis. The percent correct predictions for geologic unit groups ranged from 30% to 100%. A blind study experiment matched four of the 18 samples to their unit of origin with the first choice by stepwise discriminant analysis, and eight were matched as second and third choices. The probability of selecting the appropriate unit of origin increased by 54% over random chance and eliminated as much as 99% of the field area as a potential search location.