Land use is at the core of various sustainable development goals. Long-term climate foresight studies have structured their recent analyses around five socio-economic pathways (SSPs), with consistent storylines of future macroeconomic and societal developments; however, model quantification of these scenarios shows substantial heterogeneity in land-use projections. Here we build on a recently developed sensitivity approach to identify how future land use depends on six distinct socio-economic drivers (population, wealth, consumption preferences, agricultural productivity, land-use regulation, and trade) and their interactions. Spread across models arises mostly from diverging sensitivities to long-term drivers and from various representations of land-use regulation and trade, calling for reconciliation efforts and more empirical research. Most influential determinants for future cropland and pasture extent are population and agricultural efficiency. Furthermore, land-use regulation and consumption changes can play a key role in reducing both land use and food-security risks, and need to be central elements in sustainable development strategies.


Scenarios of land use and land cover play an important role in exploring future developments and policy options for climate change, biodiversity, food security, ecosystem services, and sustainable development. The exploration of future global land use in international assessments started in the 1990s (SRES)1 and gained increasing importance in environmental outlooks and assessments2,3. During the last decade, the number of studies and models on global land-use projections has increased tremendously, reflecting concerns about land scarcity, climate change impacts or bioenergy threatening food security4,5,6,7, loss of natural areas and biodiversity, and sustainable development in general8,9. However, despite the central role of land use in future environmental change, the modeling and systematic model comparison of global land-use projections is still in its infancy10,11 and the uncertainty in results is large12,13,14,15. Land use results from multiple interactions between regionally specific demand and supply systems, numerous feedback processes, and smaller-scale factors such as land-use regulation and land ownership, and therefore projecting land-use change is highly complex. As an example, although agricultural production has increased by about 60% during the last 40 years, global cropland area has increased by only 5% as a result of increased agricultural productivity (intensification)16. Therefore, the intricate interplay between demand and production, and agricultural intensification is a core determinant of future land use. To account for the range of possible developments in these drivers, long-term projections are often dealt with following a scenario approach, where fundamental uncertainties in socio-economic factors are explicitly varied along so-called storylines to explore contrasting futures. However, the range of model outcomes for such scenarios tends to be large and results may depend equally on model characteristics as on storylines and assumptions15. Given the importance of land-use projections for informing policy makers in the areas of climate change, food security, and biodiversity protection, it is important to investigate the spread across storylines and model results, to better understand the possible evolution of land use and the food system.

Here we use the scenario framework of the Shared Socio-economic Pathways (SSPs)17 and their implementation by integrated assessment and agricultural models14, to explore how long-term drivers determine projections of land use and food availability. We follow a sensitivity methodology recently applied to projections of CO2 emissions of the SSP scenarios18 to assess the contribution of each driver to the scenario outcomes. This allows us to explain model spread and results and, more importantly, to identify the key determinants of future land use, their relative importance, and interactions. Although single model studies have identified factors that determine future land use and support a transition to a sustainable land and food system, this study allows a systematic comparison across factors and models, and identifies the interaction between factors. Furthermore, the discussion highlights gaps and shortcomings in the current modeling of global land use to guide future research priorities.

The SSP scenario framework consists of five contrasting storylines19, with SSP2 representing a baseline development with continuation of current trends, a sustainability scenario (SSP1), a regional rivalry scenario (SSP3), a fragmentation scenario (SSP4), and a fossil fuel scenario (SSP5). Here we focus on the socio-economic developments of SSP1, SSP2, and SSP3 as their land-use implications were initially explored in most detail and by the largest number of models12,14. Results related to SSP4 and SSP5 are however also available. The SSP scenarios play a crucial role in the ongoing assessment and reports by The Intergovernmental Panel on Climate Change (IPCC)20, in the agricultural modeling community AGMIP6,21,22 and in the Inter-Sectoral Impact Model Intercomparison Project ISIMIP23, and also in assessments of biodiversity within IPBES24 and probably a range of future assessments of sustainable development. The SSP narratives represent contrasting global developments with respect to population growth, economic development, technological change, consumption preferences, environmental protection, and international cooperation (see Methods and Supplementary Information). Although projections for  population and GDP were harmonized in quantitative terms within the SSP process25,26, all other storyline elements were translated into scenario drivers by each team separately, because differences in model structure hampered quantitative harmonization.

Our analysis covers all five models that participated in the initial quantification of land use in the SSPs14, namely AIM27, GCAM28, GLOBIOM29, IMAGE-MAGNET30,31, and MAgPIE32, plus an additional model, IMPACT33, frequently used in agricultural assessments3,34.

To determine the sensitivity of scenario outcomes to the scenario drivers, we rely on a sensitivity analysis protocol, which allows to identify the interaction between drivers35 and which was previously applied to the analysis of CO2 emission trajectories as well18. We distinguish six groups of scenario drivers: population growth (POP), economic growth in gross domestic product per capita (GDPpc), land-use regulation (LUR), agricultural productivity growth (PRD), consumption preferences (CON), and trade development  (TRD). Using a matrix of scenario experiments, the deviation of other SSPs from the SSP2 baseline is attributed to these six driver groups, showing their individual, final, and interaction effect (Supplementary Fig. 1). The individual effect describes how scenario results change if a single factor of SSP2 baseline is replaced by the one used in SSP1 (or SSP3) and the final effect shows how results in the full SSP1 (or SSP3) scenario change if this single factor is changed back to the SSP2 default setting; the interaction effect is computed as the difference between individual and final effect, and shows how other factors interact with this single factor, i.e., reduce the individual effect to the final effect (see Methods). In order to analyze how sensitivity results depend on the quantity of a driver, we derive from the experiments the size of the drivers and relate them to the change in outputs.