Remote Sensing Modeling of Mycotoxins

The infection and establishment of mycotoxigenic fungi (e.g. Aspergillus flavus, Fusarium verticillioides, etc.) on crops and their subsequent production of mycotoxins are exacerbated by plant stress, which is influenced by the environment (e.g. drought stress at flowering). Satellite-derived remote sensing datasets offer information on the spatio-temporal variation in key environmental factors and can model the distribution of mycotoxins across a landscape. Such models could be valuable to surveillance of mycotoxins and could help inform preemptive actions to avert public health crises caused by mycotoxin exposure.

The Nelson Lab is interested in creating these types of models in smallholder farming systems, which are largely unregulated and can be sites of extremely high mycotoxin levels. A barrier to this modeling is accurate ground-truthed mycotoxin data; mycotoxin levels are highly variable across scales (over landscapes, within a field, and even in a single ear of maize).

To address this issue, members of the Nelson Lab have conducted multiple surveys of local grain mills in Kenya and Tanzania. Smallholder farmers frequent their local grain mills to process maize for household consumption or local markets. Using remote sensing data of environmental factors, we have modeled the highly local variation in maize aflatoxins and fumonisins and identified spatio-temporally explicit pre-harvest growing conditions that are predictive of mycotoxin contamination. Key factors that influence mycotoxin variation include normalized difference vegetation index (NDVI), which is an indicator of plant stress across a growing season, and soil measurements such as organic carbon.

For more information see Smith et al. (2016).

Top right: Preparing maize for grinding at a local mill in Kongwa District, Tanzania.

Bottom right: Example of a remote sensing-based model predicting aflatoxin levels across Kongwa District, Tanzania.