Abstract: A rapid method for the determination of total nitrogen in feces of large-scale dairy farms was established combining near infrared diffuse reflectance spectroscopy with CARS-PLS. Near infrared diffuse reflectance spectra of 111 feces samples were collected from the whole process of wastewater treatment. Modeling variables were selected by using interval partial least squares（iPLS）,synergy intervalpartial least squares（siPLS）and competitive adaptive weighted sampling（CARS）. For full-wavelength modeling, the correlation coefficient（R）and root mean square error of prediction（RMSEP）were 0.928 and 0.161 3%, respectively. For iPLS modeling, the R and RMSEP were 0.926 and 0.151 4%, respectively. For siPLS modeling, the R and RMSEP were 0.928 and 0.149 1%, respectively. For CARS model, the R and RMSEP were 0.981 and 0.084 1%, respectively.The results showed that the prediction accuracy of the model couldbe improved by variable selection, and the prediction error could be reduced and the prediction accuracy could be improved by using CARS method. This study provides a theoretical basis for the development of field convenient near infrared spectrometer.

Key words: near infrared reflectance spectroscopy, dairy cow manure, total nitrogen（TN）, partial least squares, interval partial least square, synergy interval partial least square, CARS