This paper was focused on analyzing the feasibility of predicting nitrogen contents in humid and dry soils with Near Infrared Spectroscopy (NIRS). The spectral characteristics of paddy soils in southern China were studied and the effects of different soil moisture contents on soil spectrum were analyzed. With moisture increasing, spectral absorbance hoisted and humid soil showed higher absorbance than dry one by adopting the pretreatment methods of spectrum, the calibration models were established with PLSR and SMLR. The result showed that total nitrogen in humid and dry soil samples both could be predicted by using near infrared spectroscopy, particularly by using the model established by PLSR. From external validation, Rv2 and RMSEP of dry soil and humid soil were 0.89 and 0.91, 6.92% and 5.92%, respectively. It may provide technical support for measuring soil total nitrogen content in the field.