Quantitative Spatial Validation of Management Zones by Integrating Multivariate Analysis, Geostatistics and Fuzzy Clustering in the Brazilian Cerrado

作者：Gustavo Luciano Pereira de Castro, Logam Miguel Costa Pires, D. Caetano, Sharrine Omari Domingues de Oliveira Marra, P. S. Batista, Lucas Vinícius De Souza Cangussú, A. M. de Andrade, Marinaldo Loures‐Ferreira · 发表于：Soil use and management · 年份：2026 · DOI：10.1111/sum.70260

The delineation of management zones is widely used in precision agriculture to improve input‐use efficiency; however, their quantitative validation remains a methodological challenge. This study evaluated management zone delineation in an irrigated field in the Brazilian Cerrado, covering 121.69 ha and based on 66 georeferenced soil samples collected on a regular grid, using an integrated framework combining multivariate analysis, geostatistics and fuzzy clustering. Soil chemical attributes were sampled at the 0.0–0.20 m layer, and Principal Component Analysis (PCA) was applied to summarize the main fertility gradients controlling spatial variability. The retained components explained 80.39% of the total variance, with the first component representing a base‐related fertility gradient. The spatial structure of soil attributes and the composite PCA fertility score was modelled using ordinary kriging (OK), revealing predominantly moderate spatial dependence. Management zones were delineated using the Fuzzy K‐Means algorithm applied to (i) selected soil attributes and (ii) the composite PCA fertility score. The optimal number of zones was defined using the Fuzzy Performance Index and the Normalized Classification Entropy. Agreement between management zones and soil attribute maps was assessed using Cohen's Kappa, Kendall's Tau‐b and overall accuracy after block aggregation to reduce spatial autocorrelation. PCA‐based zoning achieved higher agreement for calcium and soil organic ...