Management zones for irrigated and rainfed grain crops based on data layer integration

• Ano de defesa: 2025
• Autor(a) principal: Sterle, Luiz Gustavo de Goes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Agricultura de precisão; Análise de agrupamento; Cluster analysis; Estatística multivariada; Geoestatística; Geostatistics; Multivariate statistics; Precision agriculture; Spatio-temporal variability; Variabilidade espaço-temporal
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11152/tde-18092025-145520/

The precision agriculture (PA) cycle involves the analysis of fields spatial variability and is characterized by the steps of data collection, information manement, site-specific intervention, and performance evaluation. Management Zones (MZs) represent an approach that simplifies the spatial resolution of within-field variability management. They support agricultural decision- making by identifying areas where crop yield may be limited and require differentiated management practices. This study evaluated the effectiveness of two methods for delineating MZs (i.e., fuzzy c-means (FCM) and hierarchical) in an 54.6 ha irrigated and 7.9 ha rainfed areas of comercial sorghum and soybean field in Brazil. The historical records of yield maps, apparent soil electrical conductivity (ECa) at two depths and terrain atributes were used to delineate MZs by using multivariate estatistical analyses (PCA and cluster) and the variance reduction of the analised atributes in the field were used as reference data to evaluate the delineation methods. Results showed distinct spatial variability patterns between irrigated and rainfed systems. Rainfed areas exhibited higher variability in yield (CV 912%) and soil properties (ECa CV 2027%) compared to irrigated fields (yield CV <7%, ECa CV ~5%). Geostatistical analysis shows moderate to strong spatial dependence for both ECa and yield data. PCA revealed that elevation and subsoil ECa dominated irrigated field variability (48% PC1), while topography and surface ECa were key in rainfed systems (66% PC1). FCM clustering produced more homogeneous MZs, particularly in irrigated fields (soybean CV <6%), requiring fewer zones for effective variance reduction. Hierarchical clustering captured extreme values better but needed more zones (5 MZs for 54% variance reduction). In rainfed areas, hierarchical clustering improved internal homogeneity, though FCM remained viable for operational simplicity. Yield determinants also differed: irrigation minimized topographic effects, while rainfed yields were strongly influenced by slope (sorghum r = -0.40) and low-lying moisture (soybean r = -0.34). The study concludes that MZ delineation should align with production system characteristicsadjusting water application rates (VRI) by MZs for irrigated fields and moisture conservation for rainfed systems. FCM offers practical advantages for routine implementation, while hierarchical clustering provides higher precision where economically justified.