Sugarcane is a semi-perennial grass whose cultivation is characterized by an extended harvest season lasting several months leading to very high spatio-temporal variability of the crop development and radiometry. The objective of this paper is to understand this variability in order to propose appropriate spectral indicators for yield forecast. To do this, we used ground observations and Satellite Pour l'Observation de la Terre (SPOT4) and SPOT5 time series acquired monthly over a 2-year period over Reunion Island and Guadeloupe (French West Indies). We showed that variations in the Normalized Difference Vegetation Index (NDVI) of sugarcane at the field scale are the result of the interaction between the sugarcane crop calendar and plant phenology in a given climatic environment. We linked these variations to crop variables measured in the field (leaf area index and leaf colour), and derived simple, appropriate NDVI-based indicators of sugarcane yield components at the field scale (cane yield and sugar content). For biomass forecast, the best correlation (R2 = 0.78) was obtained with images acquired about 2 months before the harvest season, when all the fields are fully developed but before the maturation stage. For sugar content, a polynomial relationship (R2 = 0.75) was observed between the field NDVI acquired during the maturation stage and sugar content in the stalk.