Predicting Yield Potential, 2007
Can Yield Potential (similar to “yield goals”) be Predicted MID-SEASON Can Yield Potential (similar to “yield goals”) be Predicted MID-SEASON? Better than a preplant N decision?
= Winter Wheat NDVI at F5 INSEY Days from planting to sensing, GDD>0 Winter Wheat Units: biomass, kg/ha/day, where GDD>0
Predicting Yield Potential in Corn NDVI, V8 to V10 = INSEY Days from planting to sensing CORN
Long-Term Winter Wheat Grain Yields, Lahoma, OK
Response to Fertilizer N, Long-Term Winter Wheat Experiment, Lahoma, OK “After the FACT” N Rate required for “MAX Yields” Ranged from 0 to 140 lbs N/ac
Can RI be Predicted in Wheat?.... YES
Can RI Be Predicted in Corn?... YES Mullen Agronomy Journal 95:347-351 (2003) Winter Wheat
Improved Prediction of Yield Potential SuperPete to the Rescue
The mechanics of how N rates are computed are really very simple RI-NFOA YPN=YP0 * RI YPN YPN YP0 YPMAX RI=1.5 Grain yield RI=2.0 INSEY (NDVI/days from planting to sensing) Nf = (YP0*RI) – YP0))/Ef The mechanics of how N rates are computed are really very simple Yield potential is predicted without N The yield achievable with added N is #1 times the RI Grain N uptake for #2 minus #1 = Predicted Additional N Need Fertilizer Rate = #3/ efficiency factor (usually 0.5 to 0.7)
INSEY works, but needs to be more robust Problems: Extremely early season prediction of yield can be overestimated (Feekes 4, wheat) (V6, corn) Inability to reliably predict yield potential at early stages of growth should be accompanied by more risk averse prediction models (small slope)
NDVI and days from planting to sensing where GDD>0 interact with one another Model includes > 2800 observations (1996 to present)
Response Mean 2.427489 Root MSE 1.383095 R-Square 0.5857 Coefficient of Variation 56.9764 Type I Sum Regression DF of Squares R-Square F Value Pr > F Linear 2 2148.878927 0.1731 561.67 <.0001 Quadratic 2 5084.614676 0.4095 1329.00 <.0001 Crossproduct 1 37.934739 0.0031 19.83 <.0001 Total Model 5 7271.428342 0.5857 760.23 <.0001 Sum of Residual DF Squares Mean Square Total Error 2689 5143.930740 1.912953 Standard from Coded Parameter DF Estimate Error t Value Pr > |t| Data Intercept 1 6.841693 0.488513 14.01 <.0001 10.500821 days 1 0.017736 0.007928 2.24 0.0254 -1.395237 ndvi 1 -23.847124 0.819606 -29.10 <.0001 30.358037 days*days 1 -0.000065116 0.000036672 -1.78 0.0759 -0.200572 ndvi*days 1 -0.027534 0.006183 -4.45 <.0001 -1.478405 ndvi*ndvi 1 27.065163 0.524480 51.60 <.0001 25.332527
47-66 days, GDD>0
67-95 days, GDD>0
96-119 days, GDD>0
120-139 days, GDD>0
146-158 days, GDD>0
Combined RI = (NDVI-N Rich Strip/NDVI-Farmer Practice) CoefA = (0.323123*Gdd2 - 77.8* Gdd + 5406) CoefB = -0.0003469*Gdd2 + 0.08159*Gdd - 2.73372 YP0 = (CoefA * exp(CoefB * NDVI-FP)) If ((NDVI-N Rich Strip/NDVI-FP)< 1.72) RI = (NDVI-N Rich Strip/NDVI-FP)*1.69 - 0.7 If (RI<1) RI=1 YPN = YP0*RI; NRate = ((YPN-YP0)*0.0239/0.6)
Variable Rate Technology Treat Temporal and Spatial Variability Returns are higher but require larger investment
Just remember boys, you can always trust SuperPete!