Assessing soil organic carbon stocks under land-use change scenarios using random forest models

Abstract

Identifying changes in soil organic carbon stocks (SOCS) is essential for determining appropriate ways to deal with land degradation, for understanding soil and crop management and for gathering useful information for a range of environmental studies. The aim of this study was to predict SOCS and compare under current and potential future land uses. Soil organic carbon (SOC) and bulk density were measured at 137 locations across the Marivan, Kurdistan Province, Iran, and soil SOCS was computed. Auxiliary data including, terrain attributes and Landsat 8 ETM+ data were acquired. Random forest (RF) models were used to relate the SOCS to the auxiliary data. Results suggested that the mean SOCS in the topsoil and subsurface in croplands were lower than in forestland and wetland, although not significantly so. In the area, approximately 18.48% of forestland and 17.39% of wetland has been brought into cultivation. The authors estimate that this has led to a loss of SOCS from forestland topsoil of 22,860 Mg C, and from subsurface of 15,685 Mg C. The SOCS loss from wetland topsoil and subsurface were not as great, at 4193 and 2680 Mg C, respectively, but this was due to the area not being as large.