The questions of how to account for upstream land requirements is highly relevant for assigning responsibility for global land use. Two approaches, physical trade flow analysis and multi-regional input–output analysis have been frequently used for land accounting of international trade leading to diametrically opposed results for countries such as China. In this study, we investigate and explain the differences by comparing the estimates of cropland embodied in international trade for China from studies using physical trade flows (PTF) and multi-regional input–output (MRIO) analysis and provide a step-wise calculation to explain the gap between estimates from these different approaches and their interpretation. Our results show that the gap between PTF and MRIO is largely due to the system boundary selection and truncation errors from the boundary cut-off. These two approaches should be used for different research purposes. If focusing on the flows of a particular product, in particular primary products such as rice, wheat or other grains, among countries the higher level of detail of physical flow model is more suited. Whereas when accounting for the total embodied land in trade and consumption-based land use by recipient countries to analyze drivers of land use, MRIO is more suitable for tracking entire global supply chains.