Abstract: In this study, in order to detect isofenphos-methyl pesticide residues in watermelon fastly, the amount of isofenphosmethyl in watermelon was determined by using surface-enhanced raman spectroscopy ( SERS) and a rapid prediction model was built by chemometrics methods. Our study showed that the characteristic Raman peak of isofenphos-methyl was located in1043 cm^-1, and its intensity changes with the concentration of isofenphos-methyl between 0 ⁵ μg/g.Then the model was established by using principal component analysis ( PCA) , the detection limit were analysed to be 0.01 μg/g. The prediction model was created by using partial least squares ( PLS) after the original Raman spectra had been preprocessed.The performance of the model was tested, the correlation coefficient of calibration and validation were 0.9960 and 0.9952, respectively. And the root mean square error of calibration ( RMSEC) and validation ( RMSECV) were 0.163 and 0.183, which indicated that the model was reliable.This study proved that SERS method was capable of providing a simpler and more sensitive way to identify and detect isofenphos-methyl pesticide residues in watermelon.