Fuzzy c means clustering was used to define soil nutrient management zones. Remote sensing (RS) data, soil sampling data, and a combination of both were tested to identify which data source was the best for partitioning optimum zones, using a geographical information system and various statistical techniques. The study area was a region of large scale drip irrigated cotton cultivation in China. For all three data sources, the area was portioned into three zones. With the aim to confirm the resulting zones, the coefficient of variation of the nutrient index was calculated for the RS data, soil data, and combination of both types of data. There was no significant difference among the results calculated using the three data types. The least spatial variation in soil nutrient content was found within the same management zones, with larger variation between zones. The highest degree of conformity (91.36%) with zones derived using actual cotton production data was found for the management zones defined using the combination of RS and soil data. Using soil nutrient data alone, the degree of conformity was lower, at 84.40%. The lowest conformity (75.46%) was found for the zones based on the RS data alone (using the normalized difference vegetation index). The method proposed here, using fuzzy c means clustering and a combination of RS and soil sampling data, can be useful in determining zones for optimal fertilizer application and resource management in cotton systems.