Animal production systems provide nutritious food for humans, income and survivability for numerous smallholder farms and transform residues to valuable products. However, animal production is implicated in human health issues (diet-related diseases, zoonosis, antimicrobial resistance) and environmental burdens (ammonia and greenhouse gas emissions, eutrophication of surface waters, biodiversity loss). This paper reviews changes in global animal production and associated nitrogen (N) and phosphorus (P) flows over the past 50 years, during which time total animal production roughly tripled. Cattle still dominate the world in terms of animal biomass, but the number and total production of pigs and poultry have increased faster. Animal production systems are highly diverse and respond to changes in markets. Specialised systems have become more dominant, especially in developed and rapidly developing countries. The annual production of N and P in manure is similar to the amounts of N and P in synthetic fertiliser produced annually, but manure nutrients are often not recycled effectively and used efficiently by plants. Nutrient losses greatly depend on the system, management and regulations. Nitrogen and P use efficiency (NUE and PUE respectively) at the animal level is in the range 5–45%, depending on animal category, feeding and management. NUE of mixed crop-animal systems may range from 5% to 65% depending on NUE at the animal level, and the utilisation of manure nitrogen and new nitrogen inputs. Potentially, values for PUE are higher than those for NUE. Solutions for improving NUE and PUE in animal production are based on a coherent set of activities in the whole chain of ‘feed production–animal production–manure management’. A high efficiency at the system level is achieved through combination of high NUE and PUE at the animal level and effective recycling and utilisation of manure N and P in crop production. Specific regional regulations (low-emission manure storage and application, proper application limits and timing) greatly contribute to high efficiency at a system level.

Abstract. Most scenarios from integrated assessment models (IAMs) that project greenhouse gas emissions include the use of bioenergy as a means to reduce CO2 emissions or even to achieve negative emissions (together with CCS – carbon capture and storage). The potential amount of CO2 that can be removed from the atmosphere depends, among others, on the yields of bioenergy crops, the land available to grow these crops and the efficiency with which CO2 produced by combustion is captured. While bioenergy crop yields can be simulated by models, estimates of the spatial distribution of bioenergy yields under current technology based on a large number of observations are currently lacking. In this study, a random-forest (RF) algorithm is used to upscale a bioenergy yield dataset of 3963 observations covering Miscanthus, switchgrass, eucalypt, poplar and willow using climatic and soil conditions as explanatory variables. The results are global yield maps of five important lignocellulosic bioenergy crops under current technology, climate and atmospheric CO2 conditions at a 0.5∘×0.5∘ spatial resolution. We also provide a combined “best bioenergy crop” yield map by selecting one of the five crop types with the highest yield in each of the grid cells, eucalypt and Miscanthus in most cases. The global median yield of the best crop is 16.3 t DM ha−1 yr−1 (DM – dry matter). High yields mainly occur in the Amazon region and southeastern Asia. We further compare our empirically derived maps with yield maps used in three IAMs and find that the median yields in our maps are > 50 % higher than those in the IAM maps. Our estimates of gridded bioenergy crop yields can be used to provide bioenergy yields for IAMs, to evaluate land surface models or to identify the most suitable lands for future bioenergy crop plantations. The 0.5∘×0.5∘ global maps for yields of different bioenergy crops and the best crop and for the best crop composition generated from this study can be download from https://doi.org/10.5281/zenodo.3274254 (Li, 2019).

Bioenergy crop with carbon capture and storage (BECCS) is a key negative emission technology to meet carbon neutrality. However, the biophysical effects of widespread bioenergy crop cultivation on temperature remain unclear. Here, using a coupled atmosphere-land model with an explicit representation of lignocellulosic bioenergy crops, we find that after 50 years of large-scale bioenergy crop cultivation following plausible scenarios, global air temperature decreases by 0.03~0.08 °C, with strong regional contrasts and interannual variability. Over the cultivated regions, woody crops induce stronger cooling effects than herbaceous crops due to larger evapotranspiration rates and smaller aerodynamic resistance. At the continental scale, air temperature changes are not linearly proportional to the cultivation area. Sensitivity tests show that the temperature change is robust for eucalypt but more uncertain for switchgrass among different cultivation maps. Our study calls for new metrics to take the biophysical effects into account when assessing the climate mitigation capacity of BECCS.

International trade of food and feed has facilitated the specialization and agglomeration of agricultural production systems in many countries. Confined animals in specialized production systems are increasingly supplied with soybean and maize, imported from other countries. This has increased animal productivity but has also contributed to spatially decoupled crop and animal production systems. We analyzed the changes in the trade of soybean and maize at the global level in the period 1961-2011, and related these to the changes in livestock density and nutrient balances in the whole food system for 11 selected countries. Export of soybean and maize remained dominated by few countries (mainly USA, Argentina and Brazil) during the period 1961-2011, while the number of importing countries increased. Increases in the import of maize and soybean are positively related with changes in livestock density and N and P balances of national food systems. Imported soybean accounted for 12-36% of the calculated N balance at country level, and imported maize for 0-26%. There were large differences between importing countries; increases in the N surplus ranged from 75 to 306 kg N/ha and in the P surplus from 2 to 49 kg P/ha when the mean livestock density increased 1 LU/ha. This variation is related to differences in nutrient management regulations and to spatial variations in livestock density within countries. Our study contributes to the understanding of the complex relationships between the international trade of animal feed, livestock density and environmental impacts associated with N and P balances.

Different types of forests and edgesAs shown in Extended Data Fig. 8, forest and edge pixels were classified into different types in our study.When calculating edge effects across Africa, forest pixels (30 m) 1 were firstly separated into moist (M) and dry (D) forests using the MODIS land cover map (500 m) (MCD12Q1, version 6) 2 to avoid the fitting biases induced by the biomass gradient between these two forest types. Moist forests comprise Evergreen Broadleaf Forests, Deciduous Broadleaf Forests and Mixed Forests, while dry forests comprise Savannas and Woody Savannas. Next, to quantify the fire impacts on edge effects, we defined forest pixels with fire-related edges as forests influenced by fire. Forests having suffered fire, but without fire edges, were not included in our study (mainly natural fires, e.g. ignited by lighting). We overlaid the forest edges on the FireCCI burned area data set (250 m) 3 , and the edge pixels that had experienced at least one fire event during 2004-2009 were set as fire edges, while the other edge pixels were set as non-fire edges. Forest pixels in the same grid cell are separated according to their nearest edge types (fire: MF/DF, non-fire: MN/DN), and then the edge effect curves are fitted separately. Values of β were averaged over fire and non-fire edges separately in the same 0.25° grid cell, and Δβ (fire minus non-fire β) was used as indicator of the indirect fire impact on edge effects.Forest pixels with fire edges were further separated, according to whether fire intruded into forests, to calculate fire distance (fire intrusion: MF_F/DF_F, no fire intrusion: MF_N/DF_N). Fire distance was set as the distance to edge (d) for forest pixels with fire intrusion (MF_F/DF_F), and set to 0 for those pixels without fire intrusion (MF_N/DF_N). Then the median of the fire distances of forest pixels with fire edges was used for each 0.25° grid cell (MF/DF). Therefore, the fire distance also potentially reflects the area of forests with fire intrusion.2 Supplementary Discussion 1: Comparison with previous studies Methods of previous edge effect studiesForest edge effects have been studied using field studies, fragmentation experiments, remote sensing data and models. Field studies generally focus on specific variables affected by edge effects such as temperature and carbon storage 4,5 , but are not able to isolate edge effects from the impacts of long-term climate change and surrounding environment change 6 . Long-term fragmentation paired-experiments can quantify the temporal dynamics of edge effects 6,7 . These experiments are performed over forest fragments of different shapes and sizes which are maintained by artificial disturbances, compared to control and replication areas to exclude other factors (e.g. climate change, surrounding environment). The development of remote sensing products has made it possible to estimate large-scale AGB heterogeneities 8 . Briant et al. 9 pioneered the use of such products to calculate the carbon deficit due to forest edges in eastern Amazonia. C...