In Eritrea, conventional agriculture is the main stay for most of the population for livelihood where crop productivity is very low, and food insecurity and poverty are common. Though initiatives are being taken to promote the agricultural sector in the country, soil resources are not addressed. Soil studies are very limited, and the country lacks digital soil mapping. Thus, the study aims to develop robust soil organic carbon (SOC) prediction model/s for the western lowland soils where most of the agricultural activities of the country are carried out. We employed MLR, PLS, Cubist, RF, GB and XGB algorisms, and regressed multiple soil, climatic, Landsat 8 (L8) bands and spectral indices against SOC (n=178) through machine learning. The SOC modelling was done in 3 steps with 25, 14, and 06 independent environmental variables to identify the main SOC driver variables. Models performances were evaluated using the RMSE, R2, and RPD metrics. The SOC content in the study area was low with an average of 0.44%, which needs effective soil carbon improvement planning. The accuracies of all the tested models were good enough in all the three steps. The PLS model with 14 input variables gave the highest accuracy (RMSE = 0.1128%, R2 = 0.8268, RPD = 2.4393), and RF model with 06 input variables recorded the lowest (0.1435%, 0.7032, 1.9173). MLR and XGB models improved but GB model worsened with dimensionality reduction. PLS, Cubist, and RF models gave better results with 14 input variables. According to the RPD category, the PLS, XGB, and Cubist models were very good, and RF was good in all the three steps. MLR improved from good to very good but GB deteriorated from very good to good. Rainfall was the most important variable for SOC spatial variability prediction in the study area. Temperature, Green and SWIR2 Landsat 8 bands, NDSI, BR2, Sand, and MSAVI2 also had good capacity to predict the SOC spatial variability. We conclude that all the developed models have good predictive accuracies to be employed in short-mid-long-term planning and monitoring of soil fertility and productivity improvements, ecosystem restoration, and climate-change mitigation action. The study, being the first of its kind in the country, has laid the foundation for digital soil mapping (DSM) and management in the country, and more detailed SOC modelling studies are advised with more soil samples in the study area and other parts of the country for better results.