To study effect of soil moisture on soil reflection spectrum, Black soil, as a single soil class, was selected as subject for the study. Soil samples were collected and prepared to have different water contents. And then, hyperspectral reflectances of the samples was obtained in lab for analysis, statistics and quantitative description. On such a basis, a reflection spectrum-based model for prediction of soil moisture in black soil was established. Results show that with more water added to a certain threshold (300 g kg^-1), the phenomenon of oversaturation would appear with soil reflectance. However, the problem could be solved with its derivative of log-transformed reflectance (DDW), of which the response to soil moisture displayed a 3-phased variation, thus resulting in nonlinear variation of DDW at 1 870 nm. It is, therefore, essential to use piecewise functions to rapidly and accurately predict soil water content with spectral data.