Logistic regression modeling of cropping systems to predict fumonisin contamination in maize

Abstract

The aims of this research were to monitor the presence of fumonisins in maize crops in northern Italy over a 6 year period, to study the role of the cropping system on fumonisin levels, and to contribute to the development of a predictive system for fumonisin contamination. In the 6 year period from 2002 to 2007, 438 maize samples were collected in five regions, supported by agronomic data, and analyzed for fumonisin content. Fumonisin was detected in almost all of the grain samples, but 2007 was less and 2005 more contaminated compared to the other years. Preceding crop, maturity class of hybrids, nitrogen fertilization, sowing and harvest week, and grain moisture significantly affected the level of contamination. The logistic regression developed explained around 60% of variability with major roles for longitude, maturity class, and growing weeks. The function can be used to quantify the effect of these factors in a predictive system.