Abstract: A method for rapid and nondestructive detection of pesticide residues on spinach surface by using surface enhanced raman scattering ( SERS) was built in this paper. SERS of spinach samples with and without deltamethrin were pretreated with some methods such as first derivation, Norris derivation and Savitsky-Golay convolution derivation. Discriminant analysis and distance match method were used to build qualitative model for determination of spinach samples with and without deltamethrin, and the highest accuracy rate of prediction was proved to be 100%. Then several quantitative models of those samples contained deltamethrin pesticide were built with partial least squares ( PLS) .The results showed that the best model was built with the differential spectrum. The correlation coefficient of calibration ( Rc) and the correlation coefficient of validation ( Rp) of the best model were 0.9908, 0.9552 respectively.The root mean square error RMSEC and the root mean square error RMSECV were 0.0802, 0.23 respectively. All results indicated that this method can be used for a rapid and nondestructive detection of deltamethrin on spinach, as well as other agricultural products.