Abstract: In order to investigate the visible-NIR spectral features of the leaves of a sugarcane variety(ROC 22), the spectral reflectance and the chlorophyll content were measured in the laboratory, and their correlations were analyzed.Prediction models were built eventually.Results showed that the negative correlations with r greater than 0.8 was found in 527-578 nm and 701-731 nm between spectral reflectance and chlorophyll content.And the red edge position(REP) was also found having high correlation with chlorophyll content, with the highest r of 0.8442.In order to explore the most sensitive bands for normalized difference vegetation index(NDVI), 48471 NDVI values were computed with different wavebands for each sample and their correlations with chlorophyll were also analyzed.The distribution maps of NDVI and its correlations in a two-spectral-dimension space both indicated that the red bands had significant influence than the NIR bands.The suggested sensitive red range was 710-735 nm, especially 720-725 nm; and the sensitive NIR range was from 780-850 nm, which had the higher robustness.The chlorophyll predication model with NDVI(725 and 840 nm) in tillering stage had determination coefficients of 0.7386, and was recommended for guiding the subsequent ridging fertilization.