Key Parameters for Variable-Rate Technology Soybean Prescriptions
- Growers have been collecting numerous spatial data layers over many years for their fields including elevation, soil sampling and soil survey data.
- Many growers also have variable-rate technology on their planters, but are unsure of which data layer(s) to use when creating variable-rate technology seeding prescriptions for soybean.
- Determine key parameters that predict soybean yield in Wisconsin.
|High (left) and low (right) seeding rates on Sept. 9, 2013.
- Conducted on 11 Wisconsin fields in 2013 and 12 fields in 2014.
- High (+30%), standard, and low (-30%) seeding rates were randomly assigned to predetermined strips across each field.
- Standard rates were the growers’ typical rates for each field.
- Soil samples and plant stand counts were taken after emergence at predetermined grid points.
- As-planted and harvest data were collected from growers.
- Random Forest and decision tree analyses were used to determine the most important parameters for predicting yield.
|Locations of Wisconsin variable-rate technology study for 2013 and 2014.
- Soil type was the most important factor for the entire 2013 pooled dataset when used to predict soybean yield (Figure 1).
- Cross-validation procedures found the next 5 most important variables in order were also useful in predicting yield:
- Soil phosphorus (ppm)
- Soil organic matter (%)
- Soil water storage capacity from 0-39 in. depth (in.)
- Elevation (ft)
- Soil pH
- Seeding rate was not an important variable in predicting yield in 2013.
- 2014 data are currently being analyzed.
Research conducted by Ethan Smidt, University of Wisconsin-Madison as a part of the DuPont Pioneer Crop Management Research Awards (CMRA) Program. This program provides funds for agronomic and precision farming studies by university and USDA cooperators throughout North America. The awards extend for up to four years and address crop management information needs of DuPont Pioneer agronomists and customers, and Pioneer sales professionals.
2013 data are based on average of all comparisons made in 11 locations through Jan. 1, 2014. Multi-year and multi-location is a better predictor of future performance. Do not use these or any other data from a limited number of trials as a significant factor in product selection. Product responses are variable and subject to a variety of environmental, disease, and pest pressures. Individual results may vary.