Over the past ten years, there has been a fast increase in practicing variable-rate nitrogen (N) application (VRNA), especially for annual field crops.  However, due to the complexity of the soil, plant, and atmosphere interaction, the success of the VRNA depends largely on the soil available N to crops at critical growth stages.  To properly determine how nutrients should be managed in different parts of a field, among different fields, and over different years for the same crop type, a field-scale experiment was carried out on canola in Verner, Ontario, Canada for three consecutive years between 2012 and 2014.  To explore the variability in crop response to fertilizer applications, four N rates (0, 50, 100, and 150 kg N ha^-1) were applied to each crop using the recommended compound fertilize.  The field was divided into 12 to 16 strips, and each strip is 48 m wide and 400 to 800 m long.  The four N rates were randomly assigned to the strips and replicated 3 to 4 times. In-situ field data including plant density count, crop height, phenology, leaf area index (LAI), biomass, chlorophyll content and soil moisture were collected at 10-day interval throughout the entire growing season for all three years.  Soil core samples were collected twice per growing season, once at pre-planting and the other during mid-growing season for soil property analysis, including soil texture and nutrient content. Final yield maps were acquired at harvest using an IH combine with AFS yield monitor system.  Statistical analysis shows that in general crop properties are related to soil nitrate levels.  The correlations between soil nitrite and the canola LAI, fraction of crop cover, biomass, and yield are 0.73, 0.75, 0.71, and 0.38 respectively. However, large variations in crop yield response to N application rate exist among the three years tested. In 2012, a severe drought hit northern Ontario during the mid-season.  As a result, the crop yield did not exhibit strong decency of N rate.  Instead, the higher yield areas within the field occur in areas with better access to soil moisture.  2013 was a normal year with sufficient rainfall, and the yield map shows clear patterns corresponding to the different rates of N treatment zones.  Detailed results from the study will be presented in the paper.