Establishing Visible–Near Infrared Spectroscopy for the Prediction of Soil Organic Carbon in Arid Croplands of Arequipa, Peru

Foster E.J.; Yeasmin S.; Zuniga L.; Brecheisen Z.; Tomasek A.; Jimenez Castañeda M.E.; Leon-Salas W.D.; Lopa Bolívar J.A.; Schulze D.G.; Villalta Soto M.J.C.; Johnston C.T.; Filley T. · Communications in Soil Science and Plant Analysis (2025) · DOI: 10.1080/00103624.2025.2485416

Resumen

The expansion of irrigated agriculture into deserts can rapidly change soil chemistry, including the soil organic carbon (OC) content, with major impacts on the soil structure and biogeochemical cycling. Direct soil sampling and wet chemistry for frequent monitoring of OC changes can be impractical; a low-cost OC predictive model using Visible Near Infrared (VNIR) spectroscopy was established for arid croplands. To ensure broader model application, soils were sampled under distinct agricultural management types in the southern Arequipa Region of Peru: (1) traditional terraced arid agriculture (with OC contents from 0.38 to 2.31%) and (2) newly developed hyperarid croplands under varying levels of irrigation (with lower OC contents of 0.18–1.37%). Using the two datasets with distinct soil chemistry, soil management representation was increased within this regional spectral library. With the combined dataset, the optimal pre-processing procedure was determined using an international soil standard correction of absorbance, robust baseline correction, multiplicative scatter correction, and Savitzky−Golay smoothing (R2 = 0.81, RMSE = 0.12 (equivalent to 0.09% OC), RPD = 2.5). An equal to or improved predictive power above previous arid VNIR spectroscopic analyses was achieved. These results demonstrate how VNIR predictive modeling can assess rapidly changing soil OC, with implications for arid agricultural management and land development in the region.

Instituciones: Universidad Nacional de San Agustín de Arequipa