On-the-Go Prediction of Soil pH Using Generalized Additive Models in Mississippi Delta: A Case Study

Soil pH is a critical indicator of soil health and fertility, influencing nutrient availability and crop productivity. Leveraging real-time sensor technology allows for high-resolution data collection across diverse agricultural landscapes, yet effective modeling techniques are required to interpret these complex datasets. This study evaluated applying generalized additive models (GAMs) to predict soil pH in Mississippi using data collected from an on-the-go soil sensor. pH was the dependent variable, and apparent electrical conductivity shallow (ECas) and deep readings (ECad), x and y coordinates, ECa ratio, altitude, curvature, and slope were the independent variables used to develop GAMs capturing the nonlinear relationships between the predictors and soil pH. For the study site, a GAM derived from the x-coordinate and the x- and y-coordinate interaction was best for estimating pH. It achieved an r-squared value of 0.84 and a root mean squared error of 0.08 on the original testing dataset and an r-squared value of 0.87 and a root mean squared error of 0.07 on a bootstrap simulated dataset created from the original testing set. The model effectively exhibited the nonlinear dynamics of soil pH, providing insights into the relative contributions of individual predictors. This approach enhances prediction accuracy and offers interpretability, allowing agronomists to identify critical factors affecting soil pH. The findings support the potential of GAMs as a valuable tool for precision agriculture, facilitating informed decision-making for soil management and crop production.