Heavy metal pollution in soil has severe impact on the human health and global environment. There is an urgent need for cost-effective devices capable of recognizing and detecting heavy metallic matters in soils. A broadband photoacoustic spectrometric (PAS) system was established for the detection of heavy metal contaminants in soil. The heavy metal element lead (Pb) was selected as a preferred detection target. The near infrared photoacoustic spectra of contaminated soil samples with various concentrations of Pb were collected and the spectroscopic relationship between the soil absorption peaks and Pb concentrations was analyzed. Four advanced spectral preprocessing methods were explored to improve the robustness of the prediction model. Based on the maximal correlation coefficient and minimal root mean square error criteria of prediction, a two-layer feed-forward network with a continuum removing preprocessing method with a correlation coefficient as 0.96 was tested as the most appropriate method for Pb concentration prediction. This fact verifies that the broadband PAS evaluation methodology, as a nondestructive testing method, can be potentially used as an alternative quantification detection method of heavy metal contaminants in soil without the involvement of complicated sample pretreatment. © 2009-2012 IEEE.