Chapter 34 - Digital soil mapping of organic carbon at two depths in loess hilly region of Northern Iran

Abstract

This chapter was conducted to investigate the spatial variability of soil organic carbon (SOC) content using a digital soil mapping (DSM) framework in the semiarid region of northern Iran. To do this, a total of 252 soil samples were collected from two soil depths (0–20 and 20–40cm) in 126 sampling sites determined by the conditioned Latin Hypercube Sampling (cLHS) method for an area of 5390ha in a part of the Iranian Loess Plateau (ILP), Golestan Province, Iran. To predict SOC, a random forest (RF) model was applied with twenty-eight environmental covariates, including soil characteristics, topographic attributes, and remotely sensed data. Findings depicted that SOC ranged from 0.08 to 2.40% with a mean of 0.68% in 0–20cm and from 0.01 to 1.60% with a mean of 0.49% in 20–40cm. The 10-fold cross-validation was applied to assess the performance of the RF model by “root mean square error (RMSE)” and the “coefficient of determination (R2).” RMSE and R2 values differed between 0.36 and 0.38 for the first depth and 0.30 and 0.16 for the second depth. For both soil depths, slope aspect, silt, clay, elevation, and remote sensing indices, including carbonate (CI), grain size (GSI), coloration (CI), and saturation (SI) were the most important variables controlling the spatial variation of SOC. Remote sensing indices are very capable of predicting SOC in this region with complex topography. It was concluded that the predicted SOC maps at both soil depths are consistent with the realities that are highly consistent with expert knowledge in the field of soil sciences. However, a more accurate prediction of SOC in deep soil requires additional relevance. It recommended that topographic attributes and remote sensing indices are useful ancillary data for quick and cost-effective spatial prediction of SOC content to improve management practices and planning in hilly regions.