Skip to main content
ARS Home » Research » Publications at this Location » Publication #136871
Title: PREDICTING YIELD LOSS IN INDETERMINATE SOYBEAN FROM POD DENSITY USING SIMULATED DAMAGE STUDIES

Author
item Singer, Jeremy
item Malone, Robert - Rob
item Meek, David
item DRAKE, D - RUTGERS UNIVERSITY

Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/27/2003
Publication Date: 3/17/2004
Citation: SINGER, J.W., MALONE, R.W., MEEK, D.W., DRAKE, D. PREDICTING YIELD LOSS IN INDETERMINATE SOYBEAN FROM POD DENSITY USING SIMULATED DAMAGE STUDIES. AGRONOMY JOURNAL. 2004. V. 96. P. 584-589.

Interpretive Summary: Developing relationships between seed yield and pod density can be useful for predicting yield loss in soybean damaged by deer. The objectives of this research were to i) develop a modeling tool using differences between biomass removal treatments and controls for pod density and seed yield to quantify yield loss, and ii) assess the tool using double cross-validation. Model development using linear and polynomial exponential (PE) equations was accomplished using data from 1998-2001 from studies examining different biomass removal treatments, varieties, and row spacings. The polynomial exponential model had a slightly higher coefficient of determination than the linear model. Double cross-validation of both models produced strong relationships with high coefficients of determination and predictive ability. This tool can be used by extension educators and others quantifying wildlife damage to improve the efficiency of measuring deer damage to indeterminate soybean. Farmers can use this information to determine how much yield loss they are suffering from deer depredation and to determine if soybean should be planted in high depredation areas.

Technical Abstract: Developing relationships between seed yield and pod density can be useful for predicting yield loss in soybean [Glycine max (L.) Merr.] damaged by deer (Odocoileus virginianus). The objectives of this research were to i) develop a modeling tool using differences between biomass removal treatments and controls for pod density and seed yield to quantify yield loss, and ii) assess the tool using double cross-validation. Model development using linear and polynomial exponential (PE) equations was accomplished using data from 1998-2001 from studies examining different biomass removal treatments, varieties, and row spacings. The polynomial exponential model had a slightly higher coefficient of determination (R**2 = 0.93) than the linear model (R**2 = 0.92). Double cross-validation of both models produced strong relationships with high coefficients of determination and predictive ability, however, the model performance statistics indicated that the PE model had higher coefficients of determination, lower mean bias error, and more robust slope estimates than the linear model. Depending on the end-user, the simplicity of the linear model should be carefully considered in weighing the benefits of each tool. Nevertheless, these approaches provide robust tools that are not sensitive to moderate abiotic fluctuations, varying cultural practices, and a wide range of temporal biomass removal. Validating the relationship using additional data should be the next step before implementation.