Hyper-spectral remote sensing is one of the effective means for prediction of soil nutrients. Taking Wangjiagou small watershed of Three Gorges Reservoir Area as researching zone, based on the soil physicochemical properties, reflective spectrum analysis and measurement, were built predictive models for total phosphorous and total phosphorus concentrations in purple soil. Meanwhile, 33 soil samples from paddy soil were used to validate the prediction models for soil nutrients in purple soil. Results show that the total correlation coefficients between their predicted values and measured values of total nitrogen and total phosphorus concentration in purple soil are 0.672 and 0.498, respectively. Correlation coefficients obtained from predictive model of purple soil nutrients validated by paddy soil samples are 0.550 and 0.124. Therefore, it is reasonable to use hyper-spectrum method to prediction total nitrogen concentration. But prediction accuracy for total phosphorus concentration in purple soil is relatively poor. The prediction model of soil nutrients in purple soil is not suitable for paddy soil.