Precision Water Management in Corn Using ...

The ability to forecast seasonal crop water use estimates is necessary for better resource allocation. Mid-season estimation of entire seasonal crop water requirements can reduce unnecessary water application and spatial water stress, thus increasing crop yield. A crop yield versus seasonal water use model was developed. The crop yield was estimated from late July and early August vegetation aerial image data; bare soil image data; land elevation data; climatic data such as temperature, accumulated growing degree days (AGDD), solar radiation, rainfall; and non-climatic non-imagery data such as irrigation application. All these crop growth-influencing parameters were integrated from the germination date to the aerial image acquisition date. A radial basis functional network (RBFN) was used in the yield prediction model development. The optimized RBFN crop yield prediction model provided an average prediction accuracy of 91% along with a correlation coefficient (r) of 0.65 using the test data. The standard error of prediction (SEP) and Root Mean Square error (RMSE) obtained from the model was only 9.62% and 10.2% of the average actual yield of the test dataset. A linear fit model was created using the spatially predicted corn yields versus the corresponding estimated ET for the crop. An R2 of 0.65 was obtained from the model. Residual analysis was performed as an effective tool to detect outliers because there were a few malfunctioning lysimeters during some years. A performed studentized residual test and Q-test suggested several probable outliers. After the elimination of outliers, the linear fit model between estimated ET and predicted corn yield provided improved R2 of 0.81. Farmers could use the developed water use model to estimate the seasonal water requirement for corn in a midseason cropping period.