A study was conducted using maize samples collected from different agroecological zones of Kenya (n = 471) and Tanzania (n = 100) during the 2013 maize harvest season to estimate a relationship between aflatoxin B1 concentration and occurrence with weather conditions during the growing season. The toxins were analysed by the ultra-high-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS) method. Aflatoxin B1 incidence ranged between 0–100% of samples in different regions with an average value of 29.4% and aflatoxin concentrations of up to 6075 µg/kg recorded in one sample. Several regression techniques were explored. Random forests achieved the highest overall accuracy of 80%, while the accuracy of a logistic regression model was 65%. Low rainfall occurring during the early stage of the maize plant maturing combined with high temperatures leading up to full maturity provide warning signs of aflatoxin contamination. Risk maps for the two countries for the 2013 season were generated using both random forests and logistic regression models.
Bibliographical noteFunding Information:
Funding: This research was funded by the Australian Department of Foreign Affairs grant number 57685.
Acknowledgments: This study was sponsored by Australian Government through an Australian Awards scholarship of Benigni Temba and the Australian Department of Foreign Affairs and Trade (DFAT) funding of CAAREA Project in Kenya as part of the DFAT-Commonwealth Scientific and Industrial Research Organisation (CSIRO) Africa Food Security Initiative (Grant 57685). Very sadly, George Frederick (Frikkie) Liebenberg passed away on 10 March 2017 under tragic circumstances. Frikkie was involved in this project from its inception, working out ways to make the field surveying practical, and handling the downstream data management. He and his team captured the field data digitally, ensuring that it passed stringent quality-assurance checks. He is survived by his two children, a daughter, Nienke, a son, Nieko, three brothers and two sisters. Frikkie will be sadly missed by his colleagues in this project. We would also like to acknowledge the contribution of Jason Beddow (University of Minnesota) who died suddenly and unexpectedly during the course of the research that led to this paper. Jason was a great colleague and friend and he will be sorely missed.
© 2021 by the authors. Licensee MDPI, Basel, Switzerland.
PubMed: MeSH publication types
- Journal Article