Quantifying the global cropland footprint of the European Union's non-food bioeconomy

LANDFLOW is a global physical biomass trade accounting model based on data from the UN Food and Agriculture Organization (FAOSTAT 2017). It follows the approach of Kastner et al (2014) and uses detailed and comprehensive agricultural supply and use data (covering production, stock changes, international trade and utilization) measured in physical volumes (i.e. tons) from the FAOSTAT's Commodity Balance Sheets to set up a global tree structure for all commodity flows and tracks embodied cropland along these supply chains. For example, land used to produce soybeans is tracked from harvest via processing to final utilization. In the case of co-production, such as soybean oil and cake, land areas are split and allocated to the derived products in relation to their economic value, i.e. using price allocation.

The method not only covers crops and derived crop products, but also animal products such as milk, meat, fats and hides, among others (table S.1 in the supplementary material is available online at stacks.iop.org/ERL/14/045011/mmedia). Feed balances are estimated for ruminants and monogastrics respectively and available feed crops are allocated according to dietary and energy requirements of the two livestock groups. Once cropland areas are allocated to the two livestock groups, embodied land areas are attributed to multiple derived products (e.g. milk, meat and hides from ruminant livestock) using value shares as described for the case of soybean oil and cake.

The land embodied in products is tracked to final utilization, differentiated into food, seed, waste and other uses. The category of other uses comprises all non-food uses, including, for example, the quantities of vegetable oils used for the production of detergents, polymers and biodiesel, and meat and offal processed into pet food and pharmaceutical products (FAO 2001). In contrast to food use, the category of other uses, however, does not formally describe a final use but rather an industry use. LANDFLOW analysis thus tracks the supply chains of raw materials to the destination of industrial use but cannot track the further trade of highly processed industrial commodities. For instance, once vegetable oils enter the industrial sector to produce detergents, or cotton enters the textile industry, the further trade of detergents or textiles is not recorded in the FAO data.

Therefore, we allocated the results of the LANDLFOW model for the category of other uses, representing the land embodied in agricultural commodities when entering non-food manufacturing industries, to the respective industries of the MRIO model EXIOBASE 3 (Stadler et al 2018). This allowed further tracing upstream flows of non-food biomass commodities from processing industries through the global economy along monetary supply chains to the final consumers. EXIOBASE is an environmentally extended MRIO database ranging from 1995–2011 for 44 countries and five continental rest regions. Its symmetric product-by-product MRIO tables reflect the input structure for the production of 9800 products (200 products per country) and their domestic and bilateral interlinkages. MRIO models, and particularly EXIOBASE, are widely used in footprinting (see, for example, Moran and Wood 2014, Giljum et al 2016, Tukker et al 2016, Tisserant et al 2017, Wiedmann and Lenzen 2018). In this study, the MRIO model was used to complement the limited information on non-food supply chains in the LANDFLOW model, in order to identify the final consumer of crop-based products manufactured in industrial processes.

The decisive step in linking the two models was the mapping of the non-food commodity supply from the LANDFLOW model to the using industries in the EXIOBASE MRIO model. We defined a corresponding EXIOBASE sector for each LANDFLOW commodity, e.g. the EXIOBASE sector 'Products of vegetable oils and fats' corresponds to the LANDFLOW commodity 'vegetable oils'. We then masked the uses of the outputs of this sector in the MRIO entering (domestic and foreign) non-food manufacturing industries, i.e. by removing any uses by the food industry or the service sectors. The resulting correspondence table then delivered the monetary value of the vegetable oil uses by non-food industry (see table S.3 for a summarized representation of the correspondence tables). Based on this information, we derived industry shares and allocate the land inputs proportionally. As a result, we obtained a land use matrix ${\bf{P}},$ with elements ${p}_{ij}$ containing information on the land embodied in each agricultural product i further processed for non-food purposes by manufacturing industry j. For more details see Tramberend et al (2019).

The consumption footprint of cropland embodied in non-food products ${\bf{F}}$ was then calculated straight-forward by using the environmentally extended demand-driven Leontief model (Miller and Blair 2009) defined by the equation ${\bf{F}}={\bf{E}}\,* \,{\left({\bf{I}}-{\bf{A}}\right)}^{-1}\,* \,{\bf{Y}},$ where ${\left({\bf{I}}-{\bf{A}}\right)}^{-1}$ is the Leontief inverse and ${\bf{Y}}$ is the final demand matrix showing the final demand for each product in each region. The environmental extension matrix ${\bf{E}}$ for the MRIO model was derived by dividing absolute input quantities by the respective output value of each industry: ${\bf{E}}={\bf{P}}\,{\hat{{\bf{x}}}}^{-1}.$

There are some important limitations of the presented data and methods. Even though the data available from FAOSTAT provide full country detail for all UN member states, we run the LANDFLOW model at a more aggregated level (see table S.2). Geographical detail should therefore be improved for assessing region-specific impacts from agricultural production. Some authors even argue that an accurate assessment of impact footprints requires a trade model operating at the subnational level, particularly for big and diverse countries such as Brazil (Flach et al 2016, Godar et al 2016).

Moreover, the model currently does not allow separately reporting of final bio-based products such as biofuels, cosmetics, detergents, lubricants or biopolymers, but rather aggregated product groups such as vegetable oils, covering all products derived thereof.

We downscaled the national results for some major crops to the level of 5 arcminute grid cells (around 10 km × 10 km at the equator) using the spatial distribution of 42 crops provided by the Spatial Production Allocation Model (SPAM) v3.2 (IFPRI and IIASA 2015). In the first step, we aggregated the SPAM maps to three crop groups: (1) maize and sugarcane, (2) oil crops, and (3) fibre crops. We then allocated the EU footprint in each region to the geographically corresponding cells within that region, using the harvested area reported by SPAM to weight the allocation of the EU footprint into the SPAM grid cells. The weight ${\omega }_{i}^{g}$ to allocate a crop group g to a cell i is given by ${\omega }_{i}^{g}={a}_{i}^{g}/{s}_{r}^{g},$ where ${a}_{i}^{g}$ is the harvested area of the crop group g in the grid cell i and ${s}_{r}^{g}$ is the sum of the harvested area of the crop group g for all cells within region r. The weight in a region sums up to one. This approach does not consider subnational differences in the export shares and structure, which obviously biases the results. The downscaled results presented in this article thus should be interpreted as a probability distribution of the EU's footprint, rather than an exact localization. The detailed R codes and data used for this downscaling approach can be found in the previously indicated GitHub repository.

The primary production perspective on the left side of figure 1 shows the land areas used for production of crops and livestock for non-food purposes. The harvested biomass is then further processed by industries, such as the chemical, the rubber or the textile industries. These processing steps may be located in the same country, or may import feedstock from other countries. The processing phase can have many steps. Figure 1 shows the amounts of embodied cropland requirements when the products first enter the processing phase in non-food manufacturing industries. Finally, the end-products are consumed by individuals or governments, or are put on stock for use in the following years. Again, consumers may be located in the country of production or processing, or the final products may be exported to be consumed in other world regions. Note that the aggregated totals of embodied land are identical in all three parts of the Sankey diagram.

Zoom In Zoom Out Reset image size

The EU-28 is a major processor and the biggest consumer region of non-food cropland, but ranks only fifth among the largest crop producing regions. Consequently, the EU is a major net-importer of embodied cropland (figure S.4).

The cropland area within the EU used for non-food purposes increased from 10.4–14.6 Mha between 1995–2010 (table S.4). The latter accounted for about 8% of the global non-food agricultural area in 2010. Oil crops were the most dominant crop type (43%), with rapeseed and sunflower being the most dominant plants. Animal products, such as hides and skins, also played a notable role reaching 31% of total non-food cropland area in the EU in 2010.

The EU also has a significant processing industry with around a quarter of the required raw materials and related land use being imported from other world regions. In particular, vegetable oils for biofuel, polymer and detergent production were imported from Indonesia and other Asian countries. In 2010, the EU's processing industry required 19.8 Mha of cropland. Most of the processing output served consumption within the EU itself. In addition, processed products were imported from all other world regions, including China (4.4 Mha; primarily embodied in oleochemical products), Rest of Asia-Pacific (3 Mha; vegetable oils and rubber) and the USA (1.6 Mha; primarily maize and ethanol).

The EU was the largest consuming region in absolute terms with 28.2 Mha in 2010 followed by China (27.7 Mha). In relation to population, Australia leads the ranking (1199 m²/capita) followed by the USA (828 m²/capita), Canada (807 m²/capita), the EU (562 m²/capita) and Brazil (468 m²/capita). In comparison, the average non-food cropland demand in India was only 75 m²/capita (see figure S.2 and table S.6). From 1995 to 2010, the overall cropland footprint of the EU's consumption of non-food products increased by 23% from around 23–28 Mha, after reaching a peak in the year 2007 with 31.5 Mha (see figure S.3).

While the vast majority (86%) of cropland embodied in the EU's food consumption in 2010 stemmed from the EU itself (Fischer et al 2017), for the case of non-food products only 35% (9.9 Mha) were based on domestic land resources (table S.5). The remaining 65% of the cropland (18.3 Mha) was imported from outside the EU-28 (figure 2). Large amounts of embodied land (7.3 Mha) were also imported to serve manufacturing processes in the EU.

Zoom In Zoom Out Reset image size

With 2.7 Mha of embodied land, China was a major supplying country for the EU, accounting for almost 10% of the EU's non-food cropland footprint, mainly in the form of oil crops, maize, and fibre crops, or products derived therefrom (figure 2 and table S.5). Indonesia, with 2 Mha, also provided large areas, largely related to palm and coconut oil. The group Rest of Asia-Pacific, including Malaysia, Bangladesh, the Philippines and Thailand, among others, supplied Europe particularly with vegetable oils, rubber, fibre crops and non-food alcohol. Northern America also played an important role as an exporter of maize for industrial uses (e.g. in the form of starch or ethanol).

In 2010, more than one third of the EU's cropland footprint for non-food products was related to vegetable oils and oil crops, which are mainly consumed in the form of biofuels, detergents, lubricants and polymers (FNR 2014). This is more than double the embodied land of this category in 1995. Increasing consumption of vegetable oils was a main determinant for the overall growth of the EU non-food cropland footprint.

Figure 3 provides a probability distribution of the EU's footprint over a 5 arcminute grid for selected crops: (a) maize and sugarcane, which together represent more than 90% of the global ethanol feedstock and in addition are used for material purposes e.g. in the production of adhesives or bioplastics; (b) oil crops, which is the biggest crop category in the EU's non-food cropland footprint; and (c) fibre crops, mainly represented by cotton used in the textile industry.

Zoom In Zoom Out Reset image size

Spatially explicit footprint maps allow identifying regional hotspots, such as the maize plantations in the Great Plains of the US, sugarcane in south-central Brazil, or cotton in the big river basins of Pakistan. Consistent spatially explicit supply chain and footprint assessments are essential to fully capture the spatiotemporal heterogeneity of biomass production and related impacts, such as deforestation, biodiversity loss or water scarcity, which differ greatly between production regions.

Another noticeable aspect is the change in composition of the EU non-food cropland footprint between 1995–2010 (figure S.1). While in 1995, crop products contributed 63% to the overall land footprint of the EU bioeconomy, this share increased to 80% in 2010. This includes increasing quantities of cereals, non-food alcohol (mainly from maize and sugarcane) and vegetable oils for fuel and material use. In contrast, the cropland area related to the consumption of animal products, such as hides and skins, showed a declining trend.

Our results emphasize that particular attention should be given to the non-food sector, as it is the main driver of growing biomass demand, in recent years particularly due to increasing vegetable oil demand for fuel use. The EU's high external non-food land footprint indicates that a big part of the environmental impacts related with the EU's consumption occur in other world regions. Our findings show that the EU increasingly sources non-food biomass feedstocks from tropical regions, which have been identified as hotspots of both deforestation and biodiversity loss (Sodhi et al 2004, Koh and Wilcove 2008).

While the production-based approach measures territorial land use, the consumption perspective brings in the global socio-economic dynamics. Literature indicates that the EU's consumption-based cropland use is already beyond a globally equitable limit (Bringezu et al 2012, O'Brien et al 2015, O'Neill 2015, Tukker et al 2016, Häyhä et al 2018). Anthropogenic land modification, in particular deforestation, has already transgressed the planetary boundary for land system change, causing increasing pressure on climate and biodiversity (Steffen et al 2015, Campbell et al 2017). Many global energy and land use scenarios envision that the systemic change towards a bio-based economy will be more heavily reliant on terrestrial ecosystems and land resources (e.g. Lotze-Campen et al 2010; Popp et al 2014, Schipfer et al 2017, Di Fulvio et al 2019). The expanding bioeconomy will then add to the already high land demand for food supply, resulting in growing pressure on planetary boundaries. This relates closely to issues of global justice when it comes to a fair distribution of biophysical resources (Häyhä et al 2016).

Assessments of social and environmental impacts related to the consumption of bio-based commodities are usually focussing on certain products or regions. Only few studies conducted comprehensive consumption-based assessments of certain impacts with global coverage of all traded products. The model approach presented in this article facilitates the analysis of impacts from a consumption perspective. Potential environmental impacts to be studied include, for example, increased water scarcity (Mekonnen and Hoekstra 2016) and nutrient pollution (Zhang et al 2014), but also potential negative climate impacts, in particular due to deforestation in tropical regions (Achard et al 2014; Lawrence and Vandecar 2015), driven by a growing demand for raw materials for the bioeconomy (Sheppard et al 2011). Social impacts may arise due to the dislocation of vulnerable socio-demographic groups in developing countries, such as subsistence farmers with unclear land access rights (McMichael 2012), and the commodification of land and food crops (Birch et al 2010).

There is a need to analyse pathways for reducing negative impacts of the bioeconomy, for example by optimizing feedstock composition or sourcing from world regions with favourable social and environmental production conditions, including the partial substitution of globally sourced biomass by local or regionally produced alternatives (Kpdonou and Barbier 2012, Priefer et al 2017). However, as responsible consumers pull out of producer regions with questionable impacts, voids will eventually be filled by others, if incentives prevail.

At the current level of the model's geographical aggregation, most countries and world regions are net-exporters of biomass for non-food use and related land areas between the steps of primary production and processing, implying that a part of the involved manufacturing processes (and related value added) does not take place in the producer country of the raw material. For example, in 2010, Brazil produced crops destined for non-food uses on around 11.7 Mha. However, Brazilian industries only processed crops equivalent to around 9.2 Mha. This means that products equivalent to an area of around 2.5 Mha were exported to processing industries in other countries and regions. This pattern is even more pronounced in Indonesia, where the domestic industry processed only around half of the primary products produced within Indonesia (7.8 Mha compared to 14 Mha). Indonesia is a major exporter of palm oil and other non-food products, most notably to the EU and the region 'Rest of Asia-Pacific'. These results have implications for ongoing debates about the economic benefits of developing and emerging economies engaging in global value chains (GVCs). Studies have illustrated that participation of these countries in GVCs can have positive economic impacts, e.g. through dissemination and uptake of new technologies, but results are particularly positive when combined with an upgrading of exports (UNCTAD 2013). The adoption of bioeconomy strategies in an increasing number of countries, including import-dependent regions, such as the EU, offers new options for value creation in developing countries (Dietz et al 2018). However, the key challenge will be to ensure that value addition through processing will take place in the countries of production (Virchow et al 2016). The results illustrated above suggest that—from the perspective of biomass producer countries—there is still significant room for increasing domestic upgrading of biomass exports and develop a biomass export portfolio oriented towards higher value added products.

The mismatch between domestic production on the one hand and industry demand for crops for material and energy uses on the other hand will likely grow in the future. The industry perspective can be expected to further gain importance, considering the fact that the share of agriculture on the value added of food supply chains is decreasing while the share of processing industries continues growing, as documented by the European Commission (2009). The economic (and environmental) benefits and costs of a global bioeconomy transformation will therefore likely be geographically unevenly distributed as countries have largely varying competitive advantages for the production and processing of bio-based materials.

Besides socio-ecological considerations, the vulnerability of export crop production to climate change in some major supplying countries (Vörösmarty et al 2005, McGregor et al 2016) also puts highly import-dependent economies at risk of supply constraints.

Given the far-reaching global implications of an expanding European bioeconomy, robust methods and indicators need to be developed and applied, to comprehensively assess Europe's resource use as well as the related environmental and social impacts.

This paper contributes to advancing land footprint accounting and demonstrates a hybrid approach integrating the biophysical accounting method with the EXIOBASE MRIO model. As discussed extensively in the earlier literature (Vringer et al 2010, Liang and Zhang 2013, Schoer et al 2013, Weinzettel et al 2014, Bruckner et al 2015), a hybrid footprint model allows to increase product and country detail, and (partially) avoids the assumption of unique sector prices. At the same time, the model keeps a comprehensive coverage of the entire economy including all manufacturing industries and service sectors, and considers non-market commodity flows. To exploit the full potential of hybrid methods, the highest possible level of country and commodity detail provided by FAO statistics should be used. Adding more spatial and product detail will be an important task for future modelling, as yields and environmental impacts may differ largely within product and country groups, thus introducing an avoidable aggregation error.

Moreover, there is still significant room and need to expand the presented method by including other biomass commodities of key importance (e.g. timber and forest areas). Furthermore, current statistics from the FAO and EXIOBASE do not allow to explicitly separate bioenergy (e.g. biodiesel and ethanol) from biomaterial uses (e.g. detergents, adhesives, polymers). Industry data could help refine the model for addressing more detailed research questions.

Alternative accounting approaches based on economy-wide material flow analysis (ew-MFA) can reach far greater level of product detail than the present study. O'Brien et al (2015), for example, calculate the land footprint of the EU accounting for a list of 991 commodities, including both food and non-food products. The ew-MFA method basically accounts for imports and exports of all commodities and, in the case of the land footprint, converts them into land equivalents, i.e. the area required for their production. For this conversion, data from Life Cycle Assessment studies and process analyses are used to derive land use coefficients in hectares per ton of product. While being the most detailed method in terms of products, the regional resolution of ew-MFA studies is very limited, as it is not possible to specify the country of origin of the raw materials, consequently not being able to consider differences in yields or local environmental impacts.

Finally, cropland footprints are only a part of a much larger puzzle that involves the quantification and equitable sharing of the costs and benefits associated with the production and consumption of biomass-based commodities. Footprinting methods thus need to be downscaled from national to local levels to account for regional differences and dynamics in the socio-environmental conditions that determine biomass production and its impacts in producer regions (Godar et al 2015, Flach et al 2016, Godar et al 2016, Kanemoto et al 2016, Moran and Kanemoto 2016, 2017).

Our results clearly indicate a growing demand for non-food bio-based products. This means that cropland demand is increasingly driven by other than traditional food value chains, including more complex or completely new value chains that emerge in response to new biomass applications (Philp et al 2013). Moreover, biomass production may gradually shift from traditional sources in the Americas and South East Asia to new agricultural frontiers with lower governance capacities in Africa (Gasparri et al 2016). Hence, better information and transparency about the socio-economic and environmental benefits and costs associated with globally traded biomass will become key to inform the increasing number of value-chain based governance initiatives (Gardner et al 2018). Key governance challenges include substitution effects between value chains with heterogeneous levels of regulation or regulatory enforcement that can lead to environmentally costly indirect land use change (Arima et al 2011). Hybrid footprinting approaches with high spatial and temporal resolution can help to address this challenge by serving as early warning systems, when biomass sourcing patterns shift to regions or value chains that exhibit severe governance gaps.