Machine learning for the textural classification of soils based on the particle size distribution
Year: 2025
Type: article
Abstract: Soil texture, defined as the relative proportion of sand, silt, and clay, is a fundamental property that influences water retention, fertility, and soil classification. The use of machine learning algorithms to classify soil textures in Cuban sugarcane fields can contribute to sustainable agricultural management and accurate digital mapping. The aim of this study was to use machine learning algorithms for soil textural classification based on particle-size distribution. A total of 109 soil samples were collected at different depths within the solum (A + B horizons) from various soil types in sugarcane fields belonging to the “Cristino Naranjo” Agroindustrial Sugar Company, located in the municipality of Cacocum, Holguín Province, Cuba. Soil particle size was determined by mechanical analysis using the pipette method. A mechanical or textural analysis of soils was performed to classify the samples according to their clay, silt, and sand content into four textural classes: clay, clay loam, silty loam, and loam. Orange Data Mining software version 3.36.2 was used for machine learning. Unsupervised classification methods were applied, and six models were evaluated: Decision Tree, Logistic Regression, Support Vector Machine, k-Nearest Neighbors, Naive Bayes, and Random Forest. Random Forest proved to be the most accurate, stable, and reliable algorithm for textural classification based on particle size distribution, establishing reproducible qualitative thresholds.
Source: Agronomía Colombiana
Institution: University of Holguín
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
SDGs: Life on land, Zero hunger
Open Access status: diamond
OpenAlex ID: W7156255803