Determine, Predict and Map Soil pH Level by Fiber Optic Sensor

Abstract

Abstract Soil pH is one of the main factors to consider before undertaking any agricultural operation. Methods for measuring soil pH vary, but all traditional methods require time, effort, and expertise. This study aimed to determine, predict, and map the spatial distribution of soil pH based on data taken from 50 sites using the Kriging geostatistical tool in ArcGIS as a first step. In the second step, the Support Vector Machines (SVM) machine learning algorithm was used to predict the soil pH based on the CIE-L*a*b values taken from the optical fiber sensor. The standard deviation of the soil pH values was 0.42, which indicates a more reliable measurement and the data distribution is normal. The Kriging method gave a prediction accuracy of 65% while the SVM algorithm gave an accuracy of 80%. The root mean square error (RMSE) was 0.36, 0.16 and the mean absolute error (MAE) was 0.37, 0.13, respectively, for the two methods. These two methods allow the prediction of soil pH and thus the assessment of soils, allowing for easier and more efficient management decisions and sustaining productivity.