Delineation of site-specific nutrient management zones using PCA and fuzzy C-means clustering in Aroressa district, Sidama region, Ethiopia

Soil fertility variability is a major constraint to agricultural productivity in Ethiopian highlands, yet uniform nutrient management persists due to inadequate characterization of spatial heterogeneity. This study addressed the need for site-specific nutrient management zones (MZs) in Aroressa district, Sidama region. One hundred fifty-four soil samples (77 disturbed, 77 undisturbed) were collected from 0 to 20 cm depth. Analysis included pH, organic carbon (OC), total nitrogen (TN), available phosphorus (av.P), exchangeable cations (K, Ca, Mg), cation exchange capacity (CEC), and bulk density (BD). Geostatistical analysis with semi-variogram modeling and ordinary kriging mapped spatial distribution. Principal component analysis (PCA) reduced dimensionality, and Fuzzy C-Means (FCM) clustering delineated MZs, validated by ANOVA. Soils were strongly to moderately acidic (mean pH 5.19), with critically low TN (0.19%), moderate OC (2.40%), and highly heterogeneous K (CV 62.2%) and av.P (CV 32.8%). Semi-variogram models showed moderate to strong spatial dependence for most nutrients (range: 19.7-317 m). PCA retained three components explaining 71.73% variance. Five optimal MZs were delineated, significantly differentiating all soil properties (p < 0.01). MZ delineation reduced within-zone variability substantially: K (74.2%), CEC (51.0%), OC (49.3%), BD (54.7%), and av.P (40.2%). The PCA-FCM approach effectively captured soil fertility gradients, with Zone 5 showing highest fertility (CEC 35.3 cmol(+)/kg, K 1.33 cmol(+)/kg) and Zone 4 the most acidic (pH 4.66). Results support transitioning from uniform to site-specific nutrient management, with variable-rate K and P fertilization, targeted liming, and zone-specific organic matter restoration strategies.