Phosphorus adsorption prediction through Decision Tree Algorithm under different topographic conditions in sugarcane fields

• Published in: Catena (Giessen) Vol. 213; p. 106114
• Main Authors: Parra, Jeison Sanchez, de Souza, Zigomar Menezes, Oliveira, Stanly Robson de Medeiros, Farhate, Camila Viana Vieira, Marques, José, Siqueira, Diego
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-06-2022
• Subjects: Classification techniques; Data mining; Macronutrients; Magnetic susceptibility; Soil attributes; Macronutrients; Classification techniques; Soil attributes; Data mining; Magnetic susceptibility
• ISSN: 0341-8162; 1872-6887
• DOI: 10.1016/j.catena.2022.106114

[Display omitted]
•We found patterns in the variation of phosphorus adsorption in the soil through. data mining.•We found that magnetic susceptibility was the most important attribute in the. adsorption of phosphorus in the soil.•We point out the mineralogical importance of the phosphorus adsorption process. in Brazilian soils.•We approach alternative methodologies for the analysis of phosphorus adsorption data in the soil.•We point out different trends in phosphorus adsorption in the soil for concave and convex relief areas.
Phosphorus availability in the soil is essential for plant growth. In Brazil, phosphorous is poorly available in the soil due to its high adsorption in the form of phosphates. This phenomenon requires much studying to assist in the nutritional management of crops. To that end, predicting the fraction of adsorbed phosphorus can be approximated by using attributes that influence soil formation and structure. This study aimed to predict soil phosphorus adsorption based on soil attributes in sugarcane crops with different relief types using data mining techniques. The experiment was carried out in sugarcane agricultural areas, experimental plots with differentiated relief (concave or convex), and identical agricultural practices. The soil was classified as an Alfisol with udic moisture (Udalf) regime and medium to clayey texture. The dataset constituted a matrix of 4580 observations. The analyzed variables corresponded to the chemical, physical, geophysical, and mineralogical attributes in the 0–0.2 m topsoil. Data analysis was carried out based on a decision tree induction model, with an 85% accuracy rate and a high level of agreement between variables. The decision tree recognized magnetic susceptibility as the attribute with the most significant influence on the prediction of soil phosphorus adsorption, validating the relation among adsorption processes and the magnetic properties of oxide minerals characteristic of Brazilian agricultural regions.