Estimation and mapping of soil fertility index in arid agricultural environments of the Tambo Valley using regression kriging

Efficient soil fertility management in arid environments requires a clear understanding of the spatial variability of soil properties and their relationship with vegetation vigor. This study presents an integrated geospatial–edaphic framework that combines multivariate analysis and spatial modeling to quantify and map soil fertility in arid agricultural landscapes. We developed soil fertility maps for the Tambo Valley (Arequipa, Peru) by integrating edaphic and geospatial indicators through regression kriging. A total of 491 soil samples were analyzed for 22 physicochemical variables—including macro- and micronutrients, pH, texture, and bulk density—complemented with NDVI and geomorphological factors. Spearman’s correlation analysis showed positive associations between NDVI and the availability of P, Cu, and Co (r = 0.37–0.54), and negative correlations with pH and sand content (r = –0.33 and –0.31). Principal component analysis (PCA) identified fertility gradients driven by phosphorus availability, alkalinity, and micronutrient content (PC1 = 48.6%; PC2 = 11.9%). A weighted soil fertility index (SFIw) derived from the PCA was classified into low (≤0.26), medium (0.27–0.50), and high (>0.50) categories, based on data-driven tertiles of the index distribution. Regression kriging of the SFIw achieved robust spatial prediction (R² = 0.68; RMSE = 0.11), ensuring reliable mapping of fertility patterns. The highest SFIw values were found in Cocachacra and Deán Valdivia districts, linked to fertile fluvial–alluvial soils, whereas Mejía and Mollendo exhibited low indices associated with sandy and alkaline conditions. Based on these spatial patterns, three management zones were delineated: (1) high-fertility areas requiring balanced nutrient replacement, (2) medium-fertility areas needing phosphorus regulation, and (3) low-fertility areas requiring soil amendments and pH correction. The resulting maps revealed that 86.7% of the agricultural area has low or medium fertility, demonstrating the potential of PCA-weighted regression kriging as a scalable tool for precision nutrient management and sustainable intensification in arid regions.