Dealing with soil organic carbon modeling: some insights from an agro-ecosystem in Northeast Iran

Mapping soil organic carbon (SOC) and its uncertainty are essential for agricultural soil management. The current study was carried out to quantify and map the spatial variability of SOC in an Agro-Ecosystem region (~ 170 km2) in Northeast Iran using Ordinary Kriging (OK), Empirical Bayesian Kriging (EBK), and Inverse Distance Weighting (IDW) techniques. In the study area, a total of 288 soil surface samples (0–20 cm depth) were collected. Results showed that the mean SOC was 0.728% with high variability (CV = 44.78%), and also SOC was found to be deviant from a normal distribution as revealed by Kolmogorov–Smirnov (K-S) test, with positive skewness. Due to the deviation from the normal distribution, and for the purpose of modeling, the data were log-transformed to approximate a normal distribution. The best fit empirical variogram was the Exponential model. Among the various interpolation techniques, estimation with EBK approach was the fairly reliable (RMSE = 0.3156; ME = 0.004), followed by OK (RMSE = 0.3162; ME = 0.003), and IDW (RMSE = 0.3199; ME = 0.005). Moreover, results showed that the spatial analysis of SOC had a strong spatial dependency for this study. Thus, the results revealed that the simple interpolation approach was a fast method and highly suitable for the spatial prediction of SOC in this study.