Skip to main content
ARS Home » Midwest Area » Morris, Minnesota » Soil Management Research » Research » Publications at this Location » Publication #203179
Title: Structural and Fractal Dimensions are Reliable Determinants of Grain Yield in Soybean

Author
item Jaradat, Abdullah
item SUREK, DERYA - MIN. OF AGRIC., TURKEY
item Archer, David

Submitted to: Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 1/25/2007
Publication Date: 2/28/2007
Citation: Jaradat, A.A., Surek, D., Archer, D.W. 2007. Structural and Fractal Dimensions are Reliable Determinants of Grain Yield in Soybean. In: Proceedings of PMA06: The Second International Symposium on Plant Growth Modeling, Simulation, Visualization and Applications, November 13-17, 2006, Beijing, China. p. 153-158.

Interpretive Summary:

Technical Abstract: Reliable models are needed to describe plants with complex geometric structures, quantify the impact of management strategies on the plant’s geometric distribution in space and time, and predict yield as a function of fractal dimension. We measured growth and development variables on single soybean [Glycin max (L.) Merr.] plants under five management strategies in the upper Midwestern USA. Measurements derived from digital imagery of stems and leaves were subjected to multivariate and neural network analyses to identify interrelationships among plant variables and the impact of management strategies on these variables. Plants grown under different management strategies differed significantly in their geometric structures. We separated the plants into their proper categories with 75 to 100% correct classification, based mainly on differences in their fractal dimension (Do), midday differential canopy temperature (dT), and canopy light interception [Log(I/Io)]. A multilayer perception neural network with back propagation identified plant dry weight, volume, circularity (ratio of minor to major axes) and perimeter, in decreasing order, as reliable predictors (R2=0.76) of Do. The fractal dimension was the most important predictor in a generalized regression neural network, followed by plant dry weight, volume and circularity, in decreasing order, in predicting grain yield m-2 (R2=0.79). A conventional system with moldboard tillage created the most ideal microenvironment for single soybean plants to develop complex geometric structures with significantly larger Do (1.477) values and grain yield (11.2 g plant-1) as compared to plants grown under organic system with strip tillage (Do =1.358, and grain yield = 2.32 g plant-1). Knowledge of how plants respond to single and multiple management strategies will help agronomists develop better predictive models and will help farmers refine management practices to optimize yield.