Associations between field characteristics and soybean plant performance using canonical correlation analysis

The main objective of this study is to find associations between site characteristics (topographic, and soil physical and chemical properties) and soybean [Glycine max (L.) Merr.] plant performance (e.g. yield, canopy development) occurring at a field scale. The study took place in an Illinois production field in the 2000 and 2001 seasons. These associations were studied with canonical correlation analysis (CCA) followed by a spatial analysis of the resulting canonical variables with semivariography. The CCA discovered several significant associations between site characteristics. The first pair of canonical variables had a correlation coefficient of 0.76. The site characteristics most consistently correlated with the first pair of canonical variables were organic matter (OM) (r = 0.64 and 0.51 for the 2000 and 2001 seasons, respectively), pH (r = 0.39 and 0.51 for the 2000 and 2001 seasons, respectively), and deep electrical conductivity (EC_D) (r = 0.53 and 0.49 for the 2000 and 2001 seasons, respectively). Site variables soil phosphorous (P) and soil potassium (K) were inconsistently correlated with the site characteristics canonical variable. These results indicate that site variables related to soil water retention are more consistently associated with soybean performance than site variables related to soil fertility. The plant performance characteristic most correlated with the soybean performance canonical variable were NDVI_N (r = 0.76 and 0.72 for the 2000 and 2001 seasons, respectively), SPAD (r = 0.70 and 0.47 for the 2000 and 2001 seasons, respectively), and yield (r= 0.44 and 0.58 for the 2000 and 2001 seasons, respectively). The variables NDVI_N, yield, EC_D are obtained with sensors and thus they can be easily used at a production field scale. The common spatial structures in pairs of the canonical variables confirm the relationship between site properties and soybean performance, proving their potential in the demarcation of uniform areas within production fields. This approach can be used to explore soil plant relationships in other field studies.