As a grass (Poaceae), sugarcane needs supplemental mineral nitrogen (N) to achieve high yields on commercial production areas. In Brazil, N recommendations for sugarcane ratoons are based on expected yield and the results of N response trials, as soil N analyses are not a suitable basis for decisions on optimum N fertilizer rates under tropical conditions. Since the vegetative parts in sugarcane are harvested, yield components such as the number of stalks and stalk height are directly correlated with crop biomass, which, at early growth stages, can be determined by a vehicle mounted optical crop canopy sensor. With the aim to investigate the relationship between the vegetation index (VI) obtained from early season crop canopy sensing and final yield, a study on three commercial sugarcane fields located in the state of São Paulo, Brazil was conducted between 2010 and 2013. The fields included in the investigation ranged in size between 10 and 16 hectares and represented typical soils for sugarcane production, with soil textures ranging from sandy to clayey, where sugarcane of different ages (1^st, 2^nd, and 3^rd ratoon crops) was grown. The harvest of the cane occurred in September/October (end of the dry season). After the previous harvest, all fields were soil sampled (0 - 0.25 m depth) on a 0.5 ha regular grid for chemical and physical soil attributes. After sprouting, during the early season, fields were scanned with an optical crop canopy sensor (N-Sensor® ALS, Yara International ASA). On one field, these measurements were repeated three times (at approximately 0.2, 0.4, and 0.6 m stalk height) and on the two other fields just once, at 0.4 m of stalk height. After maturation, fields were mechanically harvest (no burning) with a harvester that was equipped with a yield monitor system, logging data points every two seconds. The yield data was filtered to eliminate errors and noise. Using a GIS software, buffer zones with a diameter of 20 m were created around the georeferenced soil sampling points. Average values for the measured sensor VIs and yields were calculated for the data points located within a certain buffer zone and related to each other and the respective soil properties. Finally, all factors were correlated in a matrix. From all the sampled parameters, optical sensor VI was the only one with stable good correlation with yield on all three study fields. At a stalk height of approx. 0.4 m on average in the field, correlation coefficients (r) for this relationship ranged between 0.5 and 0.6. The optical canopy sensor seems to be a valuable tool to predict in-field variability of yields. As the expected yield is the predominant factor for decisions on optimum N fertilizer supply in sugarcane production systems, this gives the opportunity for a crop sensor based variable rate nitrogen fertilizer application, aiming for improved nitrogen use efficiency (NUE) in this crop.