This project highlights one component of a broader portfolio of center-funded research initiatives focused on developing novel approaches to complex scientific challenges.
Center Update: Improving Bermudagrass Putting Green Nitrogen Fertilization Management by Understanding Growth Rate
Submitted by: Qiyu Zhou (Assistant Professor, Department of Crop and Soil Sciences)
Objectives
This three-year project investigated precision nitrogen (N) fertilization strategies for bermudagrass putting greens in North Carolina's transition zone. The goals were to quantify growth dynamics, understand nutrient needs, and develop data-driven tools to optimize fertilization while maintaining playing surface quality.
Key Findings:
1. Machine Learning Growth Prediction Model
An XGBoost model was developed using field data from North Ridge Country Club and Lake Wheeler Turfgrass Research Field (Raleigh, NC). Using inputs such as weather variables, soil moisture, NDRE, and N application history, the model achieved R² = 0.79. A site-specific version trained on ~60 clipping samples per course reached R² = 0.72 across 18 greens (Figure 1), which makes it practical for real-world deployment.
Figure 1. Site-specific model performance across 18 greens at North Ridge Country Club using a reduced input set of weather variables, NDRE, soil moisture, and historical N fertilizer application
2. Near-Infrared (NIR) Spectroscopy for Tissue N Prediction
A portable NIR model using a NeoSpectra mhandheld scanner was developed to rapidly estimate turfgrass tissue N content without laboratory analysis (R² = 0.66) (Figure 2). While promising at standard mowing heights, the NIR-based fertilization strategy produced unacceptable turf quality at low mowing heights, requiring further threshold calibration before operational use.
3. Comparison of Decision Support Strategies
Five N and plant growth regulator (PGR) decision strategies were evaluated on 'G12' ultradwarf bermudagrass at two mowing heights (0.085" and
0.125"). Key results:
- Threshold-based ML model: Reduced N use by 40–66% vs. experience-based fixed rate while maintaining acceptable turf quality at both mowing heights. The most effective strategy overall.
- PACE Turf Growth Potential (temperature-driven): Reduced N by 23–44%; accessible for facilities without sensor infrastructure.
- NDRE sensor-based: Reduced N by 13–25%; practical intermediate option requiring no model training.
- NIR spectroscopy: Acceptable at higher mowing heights (50% N reduction) but needs refinement at lower heights.
Figure 2. Relationship between laboratory-measured and NIR-predicted N in fresh bermudagrass leaves using the SG5 model. SEP = Standard error of prediction
