Assessment of two data sources to delineate productive environments in agricultural fields of south-central Entre Ríos and northwestern Buenos Aires

Abstract

Delineating within-field productive environments is a key strategy in precision agriculture to optimize input use and enhance production efficiency. This study compared the effectiveness of two data sources—yield maps and NDVI satellite imagery—for identifying productive environments in four agricultural fields located in south-central Entre Ríos and northwestern Buenos Aires, Argentina. An unsupervised classification using the k-means algorithm (k = 3) was applied to normalized datasets, and pixel-level congruence between sources was assessed. Yield data exhibited greater variability than NDVI, reflecting cumulative growing conditions more comprehensively. While both sources significantly differentiated productive environments, NDVI tended to overrate pixels by assigning them to higher categories than yield, possibly due to saturation during high-biomass stages and its dependence on specific acquisition dates. Congruence between classifications varied by region and field: it was higher in Entre Ríos (80.00  % in field 1-ER and 59.41  % in field 3-ER) and lower in Buenos Aires (48.70  % in field 6-BA and 43.99  % in field 28-BA). This disparity was attributed to greater edaphic and topographic heterogeneity in Entre Ríos, associated with water erosion processes that structure spatial variability and enhance the alignment of NDVI with yield patterns. It is concluded that freely available satellite imagery constitutes a complementary and valid tool for delineating productive environments when historical yield maps are unavailable or when existing maps present quality limitations. Future research should assess the impact of increasing the number of NDVI observations per field, defining variability thresholds to guide site-specific management decisions, and exploring alternative vegetation indices that overcome NDVI limitations.