Based on the monitored data of soil PH and measured VIS-NIR reflectance on given spots, the relationship between measured reflectance and soil PH was analyzed. Besides original field-measured spectrum (R), several spectral indices were also calculated: first derivative reflectance spectrum (FDR), inverse-log spectrum (lg(1/R)) and band depth (BD). Multivariate linear regression models were built to evaluate soil alkalinization level based on these four spectral indices and the model accuracy of PH fitting was discussed with validated sample group. The results showed that there is a significant positive correlation between soil PH and original reflectance. The accuracy of the model based on original spectrum(R) is the best with a value of R2 as high as 0.873. Thus original spectrum(R) had potential ability of rapid and exact estimation of changes in the alkalinization soil. The model can help to further the analysis of the ability of detecting alkalinization with image reflectance because of the original spectrum (R) measured directly from field .The accuracy of inverse-log spectrum predicting model was slightly lower than the accuracy of original reflectance predicting model, so inverse-log spectrum calculating was of less help to improve the predicting efficiency. The R2 of first derivative reflectance spectrum (FDR) and band depth (BD) were 0.728 and 0.648, which were not ideal for the prediction of alkalinization.