Protocols for Sampling and Spatial Prediction in Precision Agriculture: An Integrated Approach to Generate Applied Guidelines

• Grant number: 25/12175-7
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: January 19, 2026
• End date: January 18, 2027
• Field of knowledge: Agronomical Sciences - Agricultural Engineering
• Principal Investigator: Lucas Rios do Amaral
• Grantee: Derlei Dias Melo
• Supervisor: Leonardo Mendes Bastos
• Host Institution: Faculdade de Engenharia Agrícola (FEAGRI). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Institution abroad: University of Georgia, Athens (UGA), United States
• Associated to the scholarship: 24/14044-4 - Detailed pedological mapping and management zones: characteristics, applications, and complementarity between approaches, BP.DR

Abstract

Regular grid sampling followed by univariate interpolation is widely used in soil fertility mapping and variable rate fertilizer recommendation. However, it is known that this approach does not always adequately capture the spatial variability of the soil, which can compromise input use efficiency. Although these sampling methods are commonly adopted, it is widely recognized that each field presents unique variability, requiring a personalized approach - that is, optimized sampling to improve map quality. Additionally, it is necessary to understand the predictive gains when optimization is combined with more sophisticated prediction techniques, such as multivariate approaches - something still little explored in the literature. In this context, the use of orbital remote sensing combined with machine learning emerges as a strategic tool to fill these gaps. This project integrates data from Brazil and the USA with two main objectives: (i) to evaluate whether the combination of sampling techniques based on management zones and micro-variability generates grids that better capture the spatial variability of soil compared to the Latin hypercube and regular grids; (ii) to assess whether multivariate interpolation using covariates employed in sample optimization improves map prediction compared to univariate interpolation. Under the supervision of a leading researcher in fertilizer management for precision agriculture, with expertise in spatial statistics and remote sensing, we expect to generate applicable guidelines for different production contexts. Practically, the resulting protocol is expected to be scalable and adaptable, contributing with practical recommendations that promote rational input use, increased productivity, and reduced environmental impact. (AU)