Prediction of topsoil organic carbon based on vegetation indices derived from Landsat-8 OLI images in semiarid rangelands of Semnan province, Iran.

Organic carbon is one of the foremost basic soil quality indices, affecting almost all the soil’s physical, chemical, and biological properties. Digital soil mapping (DSM) of soil organic carbon (SOC) is crucial in building different environmental models. Regarding the high correlation of vegetation spectral indices (VSI) with SOC, in numerous studies, the VSI has been used as one of the appropriate information layers to predict the SOC. Due to the diversity of VIS, a question arises as to which of them is better for prediction modeling of SOC. This study aimed to compare the different VSI correlations with SOC in the semiarid rangelands of Aseran in the Semnan province of Iran. To this end, One hundred forty-five topsoil (0–20 cm) samples were collected from the studied area with a stratified random method to measure the SOC. This study used 17 independent variables, including enhanced vegetation index (EVI), normalized difference vegetation index (NDVI), green normalized difference vegetation index (GNDVI), infrared percentage vegetation index (IPVI), soil-adjusted vegetation index (SAVI), optimized soil-adjusted vegetation index (OSAVI), modified simple ratio index (MSR), modified soil-adjusted vegetation index (MSVI-1, MSVI-2, MSVI-3), normalized difference moisture index (NDMI), renormalized difference vegetation index (RDVI), simple ratio (SR), stress-related vegetation index (STVI-1, STVI-3, STVI-4), and tasseled-cap’s greenness index (TCT-GI). The relationship between the studied SOC and the VSI was investigated using the Pearson correlation coefficient. Then, a stepwise linear regression analysis using 80% of total SOC location data (145 samples) between SOC (dependent variable) and VSI (independent variables) that significantly correlated with SOC was conducted. For validating the model, 20% of total SOC location data (145 samples) were used. The statistical comparison of the results was carried out in terms of root mean square error (RMSE) and mean error (ME). Results showed that the correlation between STVI-1 and SOC was higher than other studied VSI. The determination coefficient (R²), RMSE, and ME of the prediction model were 0.33, − 0.06, and 0.15, respectively. Using the STVI-1 vegetation index along with other independent variables affecting SOC in areas with climatic conditions similar to the studied area where the soil spectral effect is dominant can increase the accuracy of the SOC spatial prediction model. [ABSTRACT FROM AUTHOR]