For the assessment of the leaching of crop protection products to groundwater according to EC regulation 1107/2009 a tiered approach has been adopted. A spatially-distributed leaching model is used in tier 3b and tier 4 assessments as described in the EU Guidance. One of the key factors that influences the leaching to groundwater is the soil organic matter content. This content not only depends on factors such as climate and terrain morphology, but also on land use. As the current EFSA organic matter map does not take land use into account, a study was done to investigate whether ISRIC’s SoilGrids product could be used to improve the soil organic matter map for arable land in Europe. Machine learning algorithms were applied using the available data in the SoilGrids database. A 10-fold cross-validation procedure was used to assess the accuracy of the predictions. The best results were obtained using an ensemble model whilst training the models with data on the log-scale and limiting the observations to organic matter contents below 5%, which is common in soils in arable land. As expected, the predictions made by SoilGrids were substantially lower and more accurate than those from the EFSA soil organic matter map.