|Drummond, Scott - UNIV OF MO|
|Birrell, Stuart - UNIV OF MO|
Submitted to: American Society of Agri Engineers Special Meetings and Conferences Papers
Publication Type: Abstract Only
Publication Acceptance Date: June 22, 1995
Publication Date: N/A
Technical Abstract: Site specific crop management aims to improve production efficiency by adjusting crop treatments, especially fertilizer and chemical application, to local conditions within the field. The full benefit of site specific crop management will not be realized until the critical factors that cause variations in yield can be related to a measured parameter or identified as sbeing caused by unknown parameters. A study was initiated to explore analysis and correlation methods that would assist in the detection of these critical factors on a point by point basis. Methods investigated included standard linear correlation, multiple linear regression, stepwise multiple linear regression, partial least squares regression, projection pursuit regression, and back propagation neural network analysis. The data set used for analysis consisted of two years of yield data and nine soil properties kriged to a 10-m cell size across a field area of approximately 25 ha. Standard statistical techniques such as linear correlation and multiple linear regression were unable to adequately model the spatial variability. More complex linear methods such as stepwise linear regression and partial least squares regression also failed to separate factors satisfactorily. Complex nonlinear, non-parametric methods such as projection pursuit regression and back propagation neural network analysis provided much more accurate yield predictions (r**2=0.74 and r**2=0.67, respectively) and allowed more complex relationships to be modeled.