Dynamic geospatial modeling of mycotoxin contamination of corn in Illinois: unveiling critical factors and predictive insights with machine learning

Abstract

Mycotoxin contamination of corn is a pervasive problem that negatively impacts human and animal health and causes economic losses to the agricultural industry worldwide. Historical aflatoxin (AFL) and fumonisin (FUM) mycotoxin contamination data of corn, daily weather data, satellite data, dynamic geospatial soil properties, and land usage parameters were modeled to identify factors significantly contributing to the outbreaks of mycotoxin contamination of corn grown in Illinois (IL), AFL >20 ppb, and FUM >5 ppm. Two methods were used: a gradient boosting machine (GBM) and a neural network (NN). Both the GBM and NN models were dynamic at a state-county geospatial level because they used GPS coordinates of the counties linked to soil properties. GBM identified temperature and precipitation prior to sowing as significant influential factors contributing to high AFL and FUM contamination. AFL-GBM showed that a higher aflatoxin risk index (ARI) in January, March, July, and November led to higher AFL contamination in the southern regions of IL. Higher values of corn-specific normalized difference vegetation index (NDVI) in July led to lower AFL contamination in Central and Southern IL, while higher wheat-specific NDVI values in February led to higher AFL. FUM-GBM showed that temperature in July and October, precipitation in February, and NDVI values in March are positively correlated with high contamination throughout IL. Furthermore, the dynamic geospatial models showed that soil characteristics were correlated with AFL and FUM contamination. Greater calcium carbonate content in soil was negatively correlated with AFL contamination, which was noticeable in Southern IL. Greater soil moisture and available water-holding capacity throughout Southern IL were positively correlated with high FUM contamination. The higher clay percentage in the northeastern areas of IL negatively correlated with FUM contamination. NN models showed high class-specific performance for 1-year predictive validation for AFL (73%) and FUM (85%), highlighting their accuracy for annual mycotoxin prediction. Our models revealed that soil, NDVI, year-specific weekly average precipitation, and temperature were the most important factors that correlated with mycotoxin contamination. These findings serve as reliable guidelines for future modeling efforts to identify novel data inputs for the prediction of AFL and FUM outbreaks and potential farm-level management practices.

Keywords: Aspergillus; Fusarium; aflatoxin; fumonisin; gradient boosting; machine learning; neural network; soil.

Figure 1

Pair-wise correlation analysis of the significant input variables influencing GBM and summary of the GBM model using multinomial mycotoxin outcome. (A) Correlation analysis of the features used for AFL modeling; (B) top 20 influential input features and their relative influence over the model in the prediction of AFL; (C) correlation analysis of the features used for FUM modeling; (D) top 20 influential input features and their relative influence over the model in the prediction of FUM. The correlation is depicted from positive (blue) to negative (red), with blank squares representing non-significant p-values of correlation between variables. For the correlation analysis, the p-value cutoff was 0.05, and the confidence level was 0.95. The blue hue in GBM plots represents the relative influence of the input variables, with light blue high and dark blue low influence levels. The green line is the testing set multinomial deviance loss, the black line is the training set multinomial deviance loss, and the blue dotted vertical line is the number of iterations used.

Figure 2

ARI relationship with AFL contamination levels and their geospatial distribution in IL from 2003 to 2021. (A) Average ARI in week 4 (January); (B) average ARI in week 11 (March); (C) average ARI in week 27 (July); (D) average ARI 45 (November). Box–Whiskers plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile) and third (75th percentile) quantiles distribution; For AFL, high is >20 ppb and low ≤20 ppb. The violin plot is shaded in red and depicts the density distribution of the soil property in low and high levels of mycotoxin contamination; and the gray dots depict each data point. Maps are shaded in red in relation to the ARI value for a specific week in each year; and the y-axis is latitude and the x-axis is longitude.

Figure 3

Crop-specific NDVI relationship with AFL contamination levels and their spatial distribution in Illinois from 2003 to 2021. (A) Corn NDVI in week 30 (July); (B) wheat NDVI in week 5 (February). Box–Whisker plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile), and third (75th percentile) quantiles distribution. For AFL, high is >20 ppb and low ≤20 ppb. The violin plot is shaded in red and depicts the density distribution of the soil property in low and high levels of mycotoxin contamination. The gray dots depict each data point. Maps show the average NDVI values for corn (green) and wheat (yellow) for specific weeks each year.

Figure 4

Calcium carbonate (CaCO₃ Kg/m²) relationship with (A) AFL contamination levels from 2003 to 2021 and its (B) spatial distribution in IL. Box–Whisker plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile), and third (75th percentile) quantiles distribution. For AFL, high is >20 ppb and low ≤20 ppb. The violin plot is shaded in red and depicts the density distribution of calcium carbonate in low and elevated levels of mycotoxin contamination; and the gray dots depict each data point. Maps show the weight percentage of calcium carbonate values. For color legend, a non-linear ramp was used.

Figure 5

Temperature, precipitation, and veg. index relationships with FUM contamination levels and their geospatial distribution in IL from 2003 to 2021. (A) Average temperature in week 31, July, (B) average temperature in week 44, October, (C) average precipitation in week 7, February, (D) average veg. index in week 11, March. Box–Whisker plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile) and third (75th percentile) quantiles distribution; For FUM, high is >5 ppm, and low ≤5 ppm. The violin plot is shaded in blue and depicts the density distribution of the temperature, precipitation, or veg. index in low and high levels of mycotoxin contamination; and the gray dots depict each data point. Maps of geography are shaded in relation to temperature (red), precipitation (blue), or vegetation index (green) values for specific weeks each year, the y-axis is latitude and the x-axis is longitude.

Figure 6

Crop-specific NDVI relationship with FUM contamination levels and their geospatial distribution in IL from 2003 to 2021. (A) Corn NDVI in week 5, February, (B) corn NDVI in week 19, May, (C) corn NDVI in week 42, October, (D) wheat NDVI in week 26, June. Box–Whisker plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile) and third (75th percentile) quantiles distribution; For FUM, high is >5 ppm, and low ≤5 ppm. The violin plot is shaded in blue and depicts the density distribution of the corn or wheat NDVI in low and high levels of mycotoxin contamination; and gray dots depict each data point. Maps of geography are shaded in relation to corn NDVI (green) or wheat NDVI (yellow) values for specific weeks in each year.

Figure 7

Soil properties relationship with FUM contamination levels and their geospatial distribution in IL. (A) soil moisture (m³ H2O/m³ soil), (B) available water-holding capacity (cm). Box–Whisker plot depicts the maximum (25th – 1.5 * interquartile range “IQR”) and minimum [75th percentile + 1.5 * interquartile range (IQR)], and the Box–Whisker plot depicts median, first (25th percentile) and third (75th percentile) quantiles distribution; (C) percentage of clay content from 0 to 5 cm below surface, (D) percentage of clay content from 25 to 50 cm below surface; For FUM, high is >5 ppm, and low ≤5 ppm. The violin plot is shaded in blue and depicts the density distribution of the soil properties in low and elevated levels of mycotoxin contamination; and the gray dots depict each data point. Maps of geographical shaded red in relation to soil properties values, scale is in miles. For color, legend of soil moisture, a linear ramp was used, and for available water-holding capacity, and clay content, and a non-linear ramp was used.