This study evaluated the TBET model for improving P indices in southern states. TBET predictions were compared to measured P loss data from Arkansas, Georgia, and North Carolina. The uncalibrated model performed poorly but predictions improved with calibration, especially for runoff. Calibrated TBET and the NC P Index had similar accuracy for annual P loss predictions. While TBET could potentially improve P indices, calibrating it for each state would be time-consuming. Overall, TBET shows promise for predicting P losses but needs work to better estimate sediment and dissolved P.
6. Study Sites
Site
Field size Duration Crop
Annual
rainfall
Annual
runoff
Annual
erosion
Soil test P
ha yrs cm cm Mg ha-1 mg kg-1
AR 0.4 3 Pasture 98 – 137 0.3 – 19 0.007 – 0.2 81 – 183
GA 0.72 – 0.79 4 Pasture 101 – 118 5 – 33 < 1 14 – 142
NC 0.04 3
Corn with
wheat
cover
80 – 157 3 – 83 0.05 – 8 41 – 121
7. Calibration and Validation Periods
Site # Fields Years Site-years
Uncalibrated,
Calibration
Washington Co., AR 7 2009, 2010 14
Putnam Co., GA 6 1995, 1998 12
Henderson Co., NC 5 2011, 2012 10
Validation
Washington Co., AR 7 2011 7
Putnam Co., GA 6 1996, 1997 12
Henderson Co., NC 5 2013 5
8. Model process
• Single year simulations run on a daily time‐step
(1/1/YYYY – 12/31/YYYY)
• 2 years of warm‐up
Initialize soil‐moisture profile and nutrient pools
• Model evaluation
Nonparametric regression Kendall‐Theil Robust Line (KTR) method
Nash‐Sutcliffe Efficiency (NSE) primary criteria
NSE >0.30 for daily time step considered “successful”
Developed prediction intervals to measure uncertainty
10. Overview of Results
• Event‐based predictions for AR, GA, and NC
• Uncalibrated model performed poorly
• Predictions improved with calibration
• Runoff predictions were successful (NSE ≥ 0.3) in 2 of 3 states during calibration and in
all 3 states during validation
• Erosion predictions were still poor
• Dissolved P predictions under‐estimated losses and could be due to lack of manure pool
• Total P predictions better but a matter of getting the right answer for the wrong reasons
in AR and GA
• Over‐predicting sediment loss and under‐predicting dissolved P loss resulted in reasonably
good predictions of total P loss
• Annual predictions for NC
• Calibrated TBET and P Index had about the same accuracy
• Will show the results for event‐based validation runs and annual
predictions