Historical increases in agricultural production were achieved predominantly by large increases in agricultural productivity. Intensification of crop and livestock production also plays a key role in future projections of agricultural land use. Here, we assess and discuss projections of crop yields by global agricultural land-use and integrated assessment models. To evaluate these crop yield projections, we compare them to empirical data on attainable yields by employing a linear and plateauing continuation of observed attainable yield trends. While keeping in mind the uncertainties of attainable yields projections and not considering future climate change impacts, we find that, on average for all cereals on the global level, global projected yields by 2050 remain below the attainable yields. This is also true for future pathways with high technological progress and mitigation efforts, indicating that projected yield increases are not overly optimistic, even under systemic transformations. On a regional scale, we find that for developing regions, specifically for sub-Saharan Africa, projected yields stay well below attainable yields, indicating that the large yield gaps which could be closed through improved crop management, may also persist in the future. In OECD countries, in contrast, current yields are already close to attainable yields, and the projections approach or, for some models, even exceed attainable yields by 2050. This observation parallels research suggesting that future progress in attainable yields in developed regions will mainly have to be achieved through new crop varieties or genetic improvements. The models included in this study vary widely in their implementation of yield progress, which are often split into endogenous (crop management) improvements and exogenous (technological) trends. More detail and transparency are needed in these important elements of global yields and land use projections, and this paper discusses possibilities of better aligning agronomic understanding of yield gaps and yield potentials with modelling approaches.

历史上农业生产的增长主要是通过农业生产率的大幅度提高而实现的。作物和牲畜生产的集约化在未来农业用地预测中也起着关键作用。在这里，我们通过全球农业土地利用和综合评估模型评估和讨论作物产量的预测。为了评估这些作物的产量预测，我们通过使用观测到的可达到的产量趋势的线性和平稳延续，将它们与可达到的产量的经验数据进行比较。在牢记可实现的单产预测的不确定性并且不考虑未来气候变化的影响的同时，我们发现，就全球而言，所有谷物的平均水平到2050年的全球预计单产仍低于可实现的单产。对于具有较高技术进步和减缓努力的未来途径，也是如此，这表明即使在系统转型的情况下，预计的增产也不是过分乐观。在区域范围内，我们发现，对于发展中地区，特别是撒哈拉以南非洲地区，预计单产仍远低于可达到的单产，这表明可以通过改善作物管理来弥合巨大的单产差距，将来可能还会持续。相比之下，在经合组织国家，当前的单产已经接近可达到的单产，而预测方法或某些模型甚至在2050年之前甚至超过了可达到的单产。这一观察结果与研究相似，表明发达地区可达到的单产的未来进展将主要是通过新的农作物品种或遗传改良来实现。本研究中包括的模型在实现产量进步方面差异很大，通常分为内源性（作物管理）改进和外源性（技术）趋势。在全球产量和土地利用预测的这些重要要素中，需要更多的细节和透明度，并且本文讨论了通过建模方法更好地使对产量差距和产量潜力的农艺学理解相一致的可能性。