Spatial models for seed yield, wilting, and maturity in furrow-irrigated soybean plots

Abstract

ABSTRACT Field experiments are subjected to spatial variability due to factors such as soil moisture, fertility, pH, and structure, as well as the pressure of diseases and pests. Soybean yields are highly variable across fields. Controlling spatial variability could decrease the risk of erroneous inferences in breeding trials. This study aims at evaluating the spatial variability of furrow-irrigated soybean for seed yield, wilting, and maturity under four different irrigation levels. The field experiment was conducted in four environmzents (location-year combination). A total of 165 soybean lines of similar relative maturity (maturity group 5) along with commercial checks were planted in an augmented strip plot design. Irrigation treatment decisions were triggered using an atmometer based on a threshold at a designated growth stage. Data were analyzed via Analysis of Variance as a linear mixed model using a blocking structure (block model) and spatial covariances using range and column. Two different spatial models were used: exponential and Gaussian. Results showed that the spatial models displayed better data fitting (lower AIC and/or BIC) than the block model in each different irrigation level across different environments and traits. Indeed, genotype ranking for seed yield was different between the block model and the best spatial model, suggesting that spatial adjustment may be necessary for soybean breeding operations under furrow irrigation. Further validation in a breeding yield trial demonstrated similar results of the effectiveness in terms of AIC and/or BIC of the spatial model compared to the block model for soybean seed yield.