Canopy reflectance plays an increasingly important role in crop management and yield prediction at large scale. The relationship of four spectral reflectance indices and cotton biomass, leaf area index (LAI), and crop yield were investigated using three cotton varieties and five N rates in the irrigated low desert in Arizona during 2009 and 2010 growing seasons. Biomass, LAI, and canopy reflectance indices [normalized difference vegetation index (NDVI); simple ratio (SR); near infrared index (NIR); and ratio vegetation index, (RVI)] were determined at different cotton growth stages. The commonly used NDVI and the other three canopy reflectance indices explained over 90% variation in cotton biomass (all R²>0.91) and LAI (R²>0.94). SR, NIR, and RVI all had higher coefficients of determination (R²) compared to NDVI due to the fact that the three indices were not saturated as NDVI at late growth stages. There was not a significant relationship between lint yield and the spectral indices measured at early growth stages. However, the spectral indices determined at peak bloom stage showed significant correlations with lint yield. SR, NIR, and RVI explained 56%, 60%, and 58% of variations in cotton lint yield, respectively, while NDVI only explained 47% of variations in lint yield. This study suggests canopy reflectance indices can be used to predict cotton lint yield at peak bloom and the accuracy of yield prediction can be significantly improved when SR, NIR, and RVI are used.