This study was conducted in 58 producer-field locations in Manitoba from 2003 to 2006 to understand how cropping practices influence Fusarium graminearum inoculum levels on stubble of various crops, including wheat, collected from the soil surface. Colonies per m2 (CN) were determined and converted to base-10 logarithm values (log10CN). Mean log10CN of the sampled field for various crops and groups of crops grown in the 3 years prior to sampling were tested to find significant differences. Average log10CN values were also used to determine significant differences between tillage systems and the effect of number of years. Average log10CN values for zero and minimum tillage systems were not different but were significantly higher than values for conventional tillage. A series of three crop rotation scenarios were tested using weighted log10CN values for crop, tillage, their interaction and their squared terms in step-wise regression models to identify which model was the best predictor of log10CN. This was selected as the cropping practice index (CPI) model and was expressed as: CPI = 1·98423 + 0·55975 (C2 × C1 × T)2 + 0·4390 (C2 × T)2, where C1, C2 and T represent the weighted log10CN values for crops grown 1 and 2 years previously and tillage system, respectively. R2 value for this model was 0·933 (P < 0·0001). The reliability of the CPI model was tested using jack-knife full cross-validation regression. The resulting R2 was 0·899. The CPI model was tested using data collected from seven wheat fields in 2006 (R2 = 0·567). The relationship between CPI and FHB index (R2 = 0·715) was significant.