Anthropogenic pressures, despite conservation efforts, continue to drive biodiversity loss. Land use is the primary driver, but climate change's impact is predicted to increase substantially, potentially surpassing land use by 2070. Food production, encompassing 38–55% of habitable land, accounts for 21–37% of anthropogenic greenhouse gas (GHG) emissions, potentially adding nearly 1°C to warming by 2100. Creating sustainable food systems is crucial to mitigating both land-driven and climate-driven biodiversity loss. International food trade further complicates the issue, as biodiversity impacts from upstream economic activities are often geographically distant from biodiversity loss locations. Supply-side and demand-side changes are both necessary for reducing food's environmental impact. Understanding the contributions of agricultural land use and GHG emissions to biodiversity loss is essential for developing sustainable international food trade strategies. A land-use-only strategy differs from one incorporating GHG-driven biodiversity loss; for instance, intensive farming uses less land but generates higher GHG emissions from increased fertilizer use. Land use affects local biodiversity, while climate change has global effects. Environmentally extended multi-regional input-output models (EEMRIOs) link downstream environmental impacts to upstream drivers, tracing a product's footprint through its supply chain. While previous studies have assessed biodiversity footprints, they have limitations such as equal threat assumptions across species ranges, exclusion of non-threatened species, or the use of biodiversity threat hotspots. This study aims to improve upon these limitations by using models that capture regional variation in biodiversity sensitivity to both land-use and climate change, employing species richness and rarity-weighted richness as metrics and considering the individual contributions of CO2, CH4, and N2O emissions.

Existing research emphasizes the significant role of the global food system in biodiversity loss, focusing primarily on land-use change. Studies utilizing EEMRIO models have attempted to trace the biodiversity footprints of commodities across global supply chains. Lenzen et al. (2012) conducted an early global biodiversity footprint analysis using the IUCN Red List, but this approach had limitations. Other studies have used biodiversity threat hotspots, bird ranges, or the countryside species-area relationship (cSAR) to estimate biodiversity footprints, but these methods also have limitations regarding spatial resolution, temporal dynamics and weighting of biodiversity facets. Wilting et al. (2017) incorporated both land-use and GHG emissions, but their analysis assumed uniform impacts across regions. Marquardt et al. (2019) highlighted the differences in biodiversity footprints based on local versus regional biodiversity measures. These previous studies have provided valuable insights, but this research builds on them by addressing spatial variations in biodiversity sensitivity and by using multiple and improved biodiversity metrics.

This study employed the EXIOBASE multi-regional input-output model to estimate the biodiversity impacts embedded within the global food system in 2011. The analysis focused on 33 food-related products across 49 regions, considering both production-based and consumption-based footprints. Two biodiversity metrics were used: local species richness and rarity-weighted species richness. Land-driven biodiversity impacts were assessed using agricultural land data from SPAM and EarthStat, while GHG-driven impacts were calculated using EXIOBASE emissions data (CO2, CH4, N2O). Spatial data on biodiversity (species richness and rarity-weighted richness for terrestrial vertebrates) were obtained from various sources and combined with biome maps from The Nature Conservancy. The sensitivity of biodiversity to land use and climate change was estimated using models from the PREDICTS database. Characterisation factors were calculated for both land-driven and GHG-driven biodiversity impacts, accounting for regional variations in biodiversity sensitivity. The impacts of both drivers were estimated in the same units, facilitating direct comparison. Production-based and consumption-based footprints were calculated per km² and per capita, respectively, to facilitate cross-regional comparisons. The study also analyzed the contribution of different GHGs to biodiversity footprints, examining net import/export footprints and the percentage of imported footprints for each region.

The study's key findings include: 1. Land-driven biodiversity footprints show a different picture from simple land-area footprints; important information is missed if land area is used solely as a proxy. Differences are driven by species richness variations and biome-specific sensitivities to land use. Regions with the greatest land-area footprints were Rest of World (ROW) Africa, China, and RoW Asia & Pacific. However, RoW Africa, Brazil, and RoW Central & South America showed the highest land-driven species richness footprints. RoW CS America had the highest rarity-weighted richness footprint. 2. India and China had the highest GHG-driven biodiversity loss, followed by RoW Africa. However, RoW Africa had a much lower GHG-driven footprint relative to its land-driven footprint. Several regions (China, India, RoW Asia) showed concerning ratios of land-driven to GHG-driven biodiversity loss, indicating that direct emissions from a single year’s food production cause biodiversity loss equivalent to 2% or more of the loss caused by historic land use. 3. Some regions were net importers of land-driven biodiversity loss but net exporters of GHG-driven biodiversity loss (e.g., India), and vice versa (e.g., Indonesia). RoW Asia & Pacific, RoW CS America, Australia, and Mexico were net exporters of both, while China, the United States, Russia, and RoW Middle East were net importers. 4. Production-based footprints varied considerably across food-related groups and regions. Animal-derived products generally had much higher footprints than plant-derived ones, although Asia & Pacific was an exception due to rice and cereal production. The 'Other Food' category often had high GHG-driven footprints. 5. Taiwan showed the highest per-area land-related production footprints, while Brazil showed higher per-area land-driven species richness footprints despite a lower land-use footprint. 6. Australia had extremely high land-driven consumption-based footprints per capita, driven by animal-based products, while Luxembourg's were driven by plant-derived products. 7. Methane emissions constituted 70% of the total GHG-driven biodiversity footprint of all food-related products. 8. The United States, United Kingdom, Germany, Russia, Japan, China, and RoW Middle East were net importers of both land-driven and GHG-driven biodiversity loss. 9. 10%, 15%, and 8% of land-driven species richness, land-driven rarity-weighted richness, and GHG-driven richness footprints, respectively, were embedded in trade between world regions. A greater percentage of plant-derived footprints were traded compared to animal-derived ones.

This study's findings emphasize the importance of considering both land use and GHG emissions when assessing the biodiversity impacts of the global food system. The use of multiple biodiversity metrics reveals different facets of biodiversity loss and highlights the disproportionate impacts on species-rich regions and those with a higher proportion of narrow-ranged species. The significant contribution of methane emissions underscores the need for strategies to reduce methane production, particularly from livestock. The results also highlight the role of international trade in shifting biodiversity impacts across regions, suggesting a need for policies that promote sustainable food production and consumption patterns. Differences in findings compared to previous studies can be attributed to methodological variations such as biodiversity metrics used, spatial and temporal resolution, and the consideration of land conversion emissions. The consistency in the metric used here allows for a more direct comparison of land-driven and GHG-driven biodiversity impacts. The high proportion of imported footprints in several regions underscores the need for policies that address the embedded biodiversity impacts of imports. The findings support policies that encourage dietary shifts towards plant-based diets and the reduction of emissions from the food system. The study contributes to the growing body of evidence highlighting the urgent need for sustainable changes in food production, consumption, and trade to mitigate biodiversity loss.

This study demonstrates that using land area alone to assess biodiversity impacts of food production is insufficient. Multiple biodiversity metrics (species richness and rarity-weighted richness) and a consideration of both land-use change and GHG emissions provide a more comprehensive picture. Methane emissions are a major driver of GHG-driven biodiversity loss, highlighting the urgent need to reduce them. International trade significantly influences the distribution of biodiversity impacts, calling for policies that promote sustainable practices throughout the entire food supply chain. Future research could focus on further disaggregating data, explicitly modelling agricultural intensification's impact, and integrating more sophisticated biodiversity metrics. The presented methods are relatively simple to use, allowing for integration into policy-making and environmental impact assessments.

The study uses data from 2011, potentially underestimating current impacts due to agricultural land expansion. Sectoral and regional aggregation in the EXIOBASE model may underestimate the volume of trade and embodied impacts. The study does not explicitly model the impacts of agricultural intensification or indirect biodiversity loss through ecosystem function loss (e.g., soil impoverishment). While the PREDICTS database captures agricultural intensity gradients and ecosystem degradation implicitly, there are data limitations, particularly for climate impacts which currently only include terrestrial vertebrates. There's also high variability in pasture maps stemming from different definitions and datasets used; however, this study uses one of the best available datasets.