Which forest-risk commodities imported to the UK have the highest overseas impacts? A rapid evidence synthesis

3.1 Synthesis of papers from initial screening

The most common commodity and study area examined exclusively by studies was Brazilian soy, followed by Brazilian beef, and Indonesian palm oil (Figure 2). Cocoa was the most commonly studied commodity in Africa, whereas palm oil was the most frequent in Asia and soy in South America. Commodities covered in studies looking at more than one geographical context included soy, palm oil, and rubber.

Figure 2. Country-commodity contexts taken from studies which passed the initial screening in the first sift (n= 152) excluding: duplicates between databases, studies which don’t mention specific contexts, and 33 studies which conducted analyses across various countries and multiple commodities at once.

3.2 Synthesis of selected papers from second screening

In the studies selected after the second screening, a more detailed assessment of commodities covered was conducted. The most common commodity and study area examined exclusively was also Brazilian soy (Figure 3). Other Latin American countries commonly presented in results included Argentina, Bolivia, Colombia, and Paraguay. Indonesia, Malaysia, and Papua New Guinea were also frequently mentioned. After soy, beef, palm oil, and timber were the most common commodities studied exclusively (Figure 3). Other tropical deforestation risk commodities include pulp and paper, leather, rubber, cocoa and coffee.

Figure 3. Forest-risk commodities included in final studies. Leather was incorporated with beef in two of the studies, therefore is counted twice. Likewise, pulp and paper were incorporated in the same category as timber in two studies.

PNG = Papua New Guinea.

Land use data. There were over ten different types of footprint indicators used by the 17 selected studies, including deforestation, ecological, biodiversity, carbon, emissions, and harvest footprints (Molotoks & West, 2021; Appendix B). These were identified by searching for the term ‘footprint’ in each of the selected studies. ‘Land footprint’, ‘land use footprint’ and ‘consumption footprint’ were the most commonly used terminologies. However, in addition to the different indicators representing different aspects of environmental impacts, precise methodology definitions varied between studies and the underlying data between studies also differed. Therefore, similar terminology could represent very different types of environmental impacts. For example, in the global studies synthesised, Hansen et al. (2013) is the most commonly used spatial dataset for representing forest loss, either solely (Pendrill et al., 2019a; Pendrill et al., 2019b) or in combination with national datasets for certain contexts (Persson et al., 2014; Henders et al., 2015). Specific commodity country contexts, however, often use nationally specific remote sensing datasets exclusively for their land cover inputs (zu Ermgassen et al., 2020; Escobar et al., 2020; Godar et al., 2016; Green et al., 2019; Trase, 2018; Walker et al., 2013).

One study uses a combination of spatial and statistical data to estimate relative deforestation risks of each country, using Hansen data on extent of tree loss from the Global Forest Watch data portal in combination with data from the Food and Agricultural Organisation (FAO) on deforestation rates (WWF, 2020). This approach prevents larger countries scoring a higher risk solely based on total land area, as the second dataset accounts for countries losing a large proportion of their small remaining forest cover. Several studies used a combination of different sources of remote sensing spatial data alongside a synthesis of national and international statistical data (Henders et al., 2015; Persson et al., 2014). For studies which did not use spatially explicit methods, data from the Food and Agricultural Organisation Corporate Statistical Database (FAOSTAT) was often used to provide national agricultural land use and production statistics (Niu et al., 2020; Rautner et al., 2013; Sandström et al., 2018). Use of national statistics for the consumption of commodities already defined as ‘forest-risk’ were also sometimes used as a proxy for deforestation impacts (Rautner et al., 2013; Walker et al., 2013; Zhang et al., 2020).

Trade data and methodologies. A variety of trade databases were also used to connect production to imports or consumption activities. For example, the Eora Multiregional Input-Output (MRIO) database (Niu et al., 2020; Zhang et al., 2020), the UN Comtrade database (Rautner et al., 2013; WWF, 2020), and the Global Trade Analysis Project (GTAP) MRIO database (Green et al., 2019). Trade models based on FAOSTAT data were also commonly used (Henders et al., 2015; Persson et al., 2014; Pendrill et al., 2019a; Pendrill et al., 2019b; Sandström et al., 2018; Trase, 2018). Two studies used a combination of two databases: UN Comtrade and (a) FAOSTAT data (Godar et al., 2016), or (b) national statistics on trade of specific commodities (Walker et al., 2013). Furthermore, one study also compares two databases separately, FAOSTAT and EXIOBASE to provide an inter-comparison (Pendrill et al., 2019a).

At the time of review, there was only one study which conducted a global analysis on a large number of specific commodities which linked embodied deforestation results to UK trade (Pendrill et al., 2019b). Other studies focussed on emissions associated with land use change (Niu et al., 2020; Pendrill et al., 2019a) or only examined a small handful of forest risk commodities (Godar et al., 2016; Henders et al., 2015; Persson et al., 2014; Rautner et al., 2013; WWF, 2020). Sandström et al. (2018) only focuses on certain sectors, with all other studies focusing on a single commodity. Pendrill et al. (2019b) is a recent study that uses a combined approach of utilising modelling based on FAOSTAT data in combination with the latest remote sensing data. Given its global coverage and specific deforestation focus that comprehensively covers all crop production, we present UK specific results from this particular study in section 3.3.

Timeframes. Global studies are often reliant on data which does not reflect present circumstances due to difficulties in accessing up to date data. This is particularly the case for global land use data, with Hansen et al. (2013) being the most commonly used data on forest loss (Henders et al., 2015; Persson et al., 2014; Pendrill et al., 2019b; WWF, 2020) providing annual statistics from 2001- 2019. National land use datasets are also often updated annually e.g., the National Institute for Space Research (INPE) and the Image Processing and Geographic Information System (GIS) Laboratory (LAPIG) for which Godar et al. (2016) uses data from 2015. However, even recent studies base analyses on data which are over a decade old e.g., 2002–2011 (Sandström et al., 2018). For trade data, the majority of studies use FAOSTAT which also provides frequent, often annual, updates on trade statistics. Other sources such as the UN Comtrade database are also frequently updated with studies using various time periods, for example 2011-2018 (WWF, 2020) or 2011 (Rautner et al., 2013; Walker et al., 2013), or used to check consistency of other datasets (Escobar et al., 2020). Studies using MRIO trade models also vary in terms of the time stamp of trade data used, using data from ten years (Green et al., 2019), five years (Pendrill et al., 2019a; Zhang et al., 2020) or four years (Niu et al., 2020) prior to their publication.

Policy recommendations. Despite differences in the underpinning data, methodology and time frame of studies, consistent themes emerge from the reviewed studies around the need for international cooperation for addressing the challenges of promoting sustainable supply chains. It is important to holistically consider the trade-offs between food production and environmental sustainability (Niu et al., 2020) and align monitoring between countries (Efeca, 2019). Monitoring indirect suppliers is also critical for progress towards promoting sustainable supply chains (zu Ermgassen et al., 2020; Trase, 2018; Walker et al., 2013) and applying extensions to cover more stages along the supply chains such as re-exports (Escobar et al., 2020) is also recommended. Broader policy agendas are also recognised as necessary for identifying synergies with policies beyond supply chains e.g., food security, conservation, national economic development and local/traditional knowledge (Godar et al., 2016).

Policy recommendations from studies acknowledge the need for a combination of demand and supply side measures (Henders et al., 2015) and extending supply chain governance beyond biomes of production (zu Ermgassen et al., 2020). A focus on export markets alone will not adequately address deforestation associated with commodities (zu Ermgassen et al., 2020) and bilateral agreements between tropical forest countries are essential for avoiding leakage (Rautner et al., 2013), for example, to countries without REDD+ policies (Green et al., 2019; Henders et al., 2015). Differentiation between emissions from consumption and production related activities could also avoid leakage (Loeff et al., 2018). Emissions are often concentrated in comparatively few trade flows, which suggests effective efforts to reduce deforestation in supply chains should target specific trade relationships and commodities e.g., through use of carbon taxes on food products (Pendrill et al., 2019a). This means increased transparency is also necessary (Trase, 2018). Incentives for facilitating uptake and use of transparency information are needed (Godar et al., 2016), with transparency and corporate disclosure also needing to be incentivised, for example through differential import tariffs and guarantees (Rautner et al., 2013).

One study gives specific recommendations for UK policy, encouraging mandatory due diligence obligations and legally binding targets for reducing environmental footprints (WWF, 2020), for example, by moving from zero net to zero-absolute deforestation or by introducing annual targets for reducing deforestation (Trase, 2018). Many studies also encourage wider stakeholder engagement (WWF, 2020) and collective action through the establishment of multi-stakeholder partnerships (Green et al., 2019; Walker et al., 2013). Other suggestions include implementation of biome or landscape level mechanisms (Escobar et al., 2020; Green et al., 2019), independent auditing systems (Walker et al., 2013) and improved certification schemes (Rautner et al., 2013). Broader implementation strategies, such as enforcement of public and private sector initiatives and commitments (Rautner et al., 2013), to avoid undue burdens on producers (Trase, 2018) are also recommended. Consumer policy is only mentioned by a few studies with the recommendation of changing consumption patterns (Niu et al., 2020) to reduce consumption of animal products (WWF, 2020), in particular, beef (Sandström et al., 2018).

3.3 Deforestation risk commodities

Data was extracted from the publicly available Pendrill et al., 2020 dataset which is based on the Pendrill et al. (2019b) study. According to this study, the highest risk commodities in terms of deforestation risk for the UK are palm oil, beef, wood products, soy, cocoa, and coffee (Table 1), collectively contributing to over 89% of the UK’s overseas deforestation risk (Pendrill et al., 2020). It is also worth noting that there are other less well-recognised commodities such as sugar and spices represented in the top ten commodities imported by the UK with the highest deforestation risk by this data (Table 1).

Figure 4 shows the country commodity contexts with the highest deforestation risk associated with UK trade. The associated carbon emissions (including peat emissions) from UK imports are also shown (Pendrill et al., 2020). Results illustrate that a high proportion of tropical deforestation associated with UK imports can be attributed to just a few commodities and producer countries. Three contexts alone, Indonesian palm oil, Brazilian beef, and Brazilian soy, are estimated to amount to almost half the total UK deforestation footprint (Figure 4). These same contexts are also identified previously as those that have been most exclusively researched (Figure 2 and Figure 3).

Figure 4. Deforestation risk (hectares/year, (ha/yr), shown in outer ring) and associated carbon emissions (MtCO2/year) from UK imports of agricultural and forestry commodities in 2017 (data from (Pendrill et al., 2020), based on (Pendrill et al., 2019a; Pendrill et al., 2019b)).

For each of the top six commodities (Table 1), the majority of their deforestation risk (over 75%) is concentrated in just three countries ( Figure 5 and Figure 6)

Figure 5. Major countries that the six highest forest-risk commodities (Table 1) are sourced from, by the UK (Pendrill et al., 2020) based on (Pendrill et al., 2019a; Pendrill et al., 2019b), excluding countries within which commodities contribute a small amount to deforestation.

Figure 6. The three countries with the highest deforestation risk (ha/yr) for each of the six top commodities (Pendrill et al., 2020), based on (Pendrill et al., 2019a; Pendrill et al., 2019b)) and all other countries grouped together as ‘Other’.

PNG = Papua New Guinea, DRC= Democratic Republic of the Congo.

It is important to note that other countries also contribute to each commodity’s deforestation risk but are not shown in Figure 5 as they constitute a relatively small proportion to overall risk. This is demonstrated in Figure 6 where the total from all other countries shows a relatively small contribution to each commodities deforestation risk. Furthermore, it is also important to note this study only covers commodities associated with tropical deforestation, hence there is a risk of bias.

Appraisal of land cover datasets employed in reviewed studies

The data from Pendrill et al. (2019b) presented in section 3.3 illustrates the type of information that ‘state of the art’ methods for attributing global deforestation risk (and associated emissions) to traded commodities provide. Although this is not a definitive source, it is one of the most comprehensive, up to date, analyses linking UK trade to specific commodities and their environmental impacts in terms of deforestation risk. However, as with any assessments that attempt to link the impacts of production to international supply chains, there are a few caveats and limitations. The Pendrill et al. (2019b) data only covers tropical and sub-tropical commodities. Furthermore, timber as represented in this dataset only includes timber plantations, a limitation other studies also face (Zhang et al., 2020), as opposed to extraction from primary forest, hence these are potential data gaps. This gap is partially filled by a more recent study which has been published since this review was conducted (Hoang & Kanemoto, 2021). This study also uses Hansen et al. (2013) as well as differentiating between tropical forests from plantations. However, it does not provide commodity-specific deforestation footprints.

The use of different forest datasets will also result in different outcomes depending on the methodology used. Pendrill et al. (2019b), like many studies, uses Hansen et al. (2013) which is one of the main sources for tree loss data. The other main source of deforestation data comes from the Global Forest Resources Assessment, which uses statistics from the UN Food and Agricultural Organisation (FAO). A detailed study comparing these methods found estimates of deforestation rates lower using Hansen data due to the different definitions of deforestation applied (McNicol et al., 2018). Hansen data is based on tree cover, defined as over five metres tall, whilst FAO estimates are based on land use classifications and extrapolations of rates of change, which are a net change figure. The way Hansen defines tree cover as all vegetation over five metres in height gives higher estimates of tree cover (McNicol et al., 2018). However, its sensitivity for capturing smaller canopy disturbances can ultimately lead to an underestimation of deforestation rates driven by small-scale disturbances, particularly on local-regional scales (Milodowski et al., 2017).

There are therefore likely to be benefits of using nationally specific deforestation datasets as opposed to applying global datasets to national scales. This is particularly important for countries with large areas of forest cover and low levels of deforestation where small disturbances make up a large proportion of forest loss (Galiatsatos et al., 2020). Many of the studies that focus specifically on one country-commodity context use national datasets as these datasets may provide more granular and ultimately more accurate representations of forest cover. Although the Hansen dataset is 30m resolution, it is rescaled from a coarser 250m resolution which limits its applicability for detecting afforestation or degradation changes (Milodowski et al., 2017). Differences in definitions between global and national datasets as to what constitutes ‘forest’ can also result in differences, as found when PRODES data showed systematically lower rates of deforestation than Hansen, despite use of the same satellite data and mapping methods (Milodowski et al., 2017). However, use of national datasets inevitably makes inter-comparisons between different countries or regions more difficult. Using Hansen data on a national scale and suitably calibrating for percent tree cover using national datasets could provide a potential solution to harmonisation. Given that different thresholds will be necessary for providing the closest correspondence to different national forest cover statistics, this could prevent over or under estimation (Sannier et al., 2015; Galiatsatos et al., 2020). However, there is a need for more inter-comparisons of national and global datasets for assessing accuracy and increasing confidence in the data, with provision of confidence limits for data users being particularly important.

Appraisal of different trade databases/approaches to modelling

FAOSTAT data was used in the majority of studies, which is unsurprising given its position as the most detailed global dataset on agricultural commodity production and trade. However, the methodologies and criteria for data collation within FAOSTAT may vary in quality depending on the national reporting guidelines and capabilities. This is reflected in, for example, the mismatch between export and import reports which, when compiled into footprinting studies, require methodological choices to harmonise which may impact results. Any attempts to correct bilateral trade statistics for the ‘re-export’ of materials (as per Pendrill et al., 2019b) will also necessitate a modelling-based approach whose assumptions may vary with impact on the distribution of production impacts between production and consumption.

Several studies extend beyond the use of bilateral trade data and use multi-regional input-output approaches (MRIOs) to provide a ‘final consumption’ account. Of these, GTAP (Green et al., 2019), Eora (Niu et al., 2020; Zhang et al., 2020) and EXIOBASE (Pendrill et al., 2019a) MRIOs were used in the studies which were included after the second screening in our review. Choosing to apply an MRIO in combination with (Green et al., 2019; Niu et al., 2020; Zhang et al., 2020) or as an alternative to (Pendrill et al., 2019b) FAOSTAT can have a large impact on results. This is driven by a combination of the different perspective (i.e., true consumption vs trade) adopted by the inclusion of an MRIO method, and by differences in the underpinning data and the methods via which they are applied. As an example, a recently published (but currently un-peer reviewed) interim report compiling the UK deforestation footprint (Croft et al., 2021) used the same deforestation dataset as Pendrill et al. (2019a). The combination with a hybridised MRIO approach (similar to that applied in (Croft et al., 2018)), estimated ~20,000ha for UK total deforestation risk (Croft et al., 2021) compared to ~16,000ha for the Pendrill et al., 2019a account presented above (Figure 4). Such a different reflects the fact that a full consumption footprint encompasses indirect utilisation of commodities (e.g., in more complex products) which are not included to the same degree when only bilateral trade information is considered.

All MRIOs have their own strengths and weaknesses, which can dictate their choice in different applications. However, deforestation footprints and associated emissions will vary according to the MRIO used due to differences in geographic and temporal resolution, time lags, country and sectoral representation, and methods used to harmonise the statistics they contain (Giljum et al., 2019). There are limited comparisons between MRIOs to understand how differences affects results, therefore methodological development and alignment between global MRIOs is an important research priority. There is also a need for more detailed data on specific locations of risk, especially for countries with high deforestation rates (Pendrill et al., 2019a), as international supply chain assessment will lack specific supply chain detail. There are approaches which attempt to give this detail; for example, the Transparency for Sustainable Economies (Trase) initiative (http://www.trase.earth). This approach is used in several of the studies in this review (zu Ermgassen et al., 2020; Escobar et al., 2020; Trase, 2018) and uses publicly available data to link imports of consumer countries to subnational places of production. More detailed, sub-national assessments are not possible internationally at this time, however, and efforts to improve data availability and target other interventions for risk-mitigation first requires an understanding of the likely hotspots (Figure 5, Figure 6).

Context of results in UK policy

The hotspots i.e., the main deforestation risk commodities analysed within papers comprising this synthesis, and the results presented from Pendrill et al., 2019a are a good match with current UK policy-linked dialogue. The commodities mentioned in the UK’s due diligence consultation (DEFRA, 2020), for example, include beef and leather, cocoa, palm oil, rubber, and soy. The main commodities illustrated here that have the highest deforestation risk yet are not included in the DEFRA consultation list are wood products and coffee (Table 1). Timber and pulp and paper are mentioned briefly in introductory materials, despite not appearing in the list of included commodities for questions set within the due diligence consultation, but it is worth noting there are already due diligence requirements on timber imports (OPSS, 2014). However, these highest risk commodities are not considered equally across the reviewed literature, with coffee and cocoa having significantly less dedicated studies, for example (Figure 2 and Figure 3). Whilst not mentioned in the UK due diligence consultation, certain spices such as nutmeg, pepper, and sugar are also listed as being in the top ten forest risk commodities imported by the UK (Table 1) with major producer countries being Indonesia, Vietnam, and Belize respectively (Pendrill et al., 2020). This is despite having a higher estimated deforestation risk than some commodities included in the consultation, for example rubber. No studies in our rapid review specifically covered these emerging commodities as deforestation risk commodities linked to the UK (Figure 2 and Figure 3). There is therefore a potential gap related to the inclusion of less well-known commodities, shown in recent research to have a deforestation risk associated with UK imports, in terms of their scientific and policy coverage.

This point about the balance of coverage in the literature and policy documentation and identified forest-risk commodities within emerging recent analysis is an important consideration. There may be commodities which are less extensively studied for reasons of data availability or scientific capacity which have a relatively high deforestation footprint. Furthermore, the inclusion of certain commodities but not others in policy materials could bias efforts towards certain areas at the expense of others. Brazilian soy for example was the most commonly studied commodity in our review (Figure 2), likely in part due to data availability and a recognised historical link between its production and impacts on critical biomes such as the Amazon rainforest (Fearnside, 2001). There is no doubt that Brazilian soy is a critical deforestation-risk commodity, and the presence of lesser-studied commodities in deforestation footprints should not distract from attention on this and the landscapes where it is a threat. However, additional attention on other commodities that are less commonly studied is warranted, particularly to avoid these being associated with deforestation frontiers in future (Table 1).

It is particularly important to keep commodities listed as being a deforestation risk in legislation under review as, should reducing deforestation associated with one commodity be successful, there is a risk of deforestation being displaced to others via leakage effects (Meyfroidt et al., 2013). Quite a few of the studies mention the issue of leakage and the importance of avoiding impacts of commodity production simply moving to a different area, which is only likely to be minimized by wider adoption and enforcement of zero-deforestation commitments (Garrett et al., 2019). The impacts of climate change may also contribute to changing crop yields and shifting areas of cultivation in future. Although negative impacts of climate change on major crops are well recognised (Zhao et al., 2017), particularly in high and low latitudes, uncertainty remains for the impact of climate change on crop yields in mid-latitude regions (Rosenzweig et al., 2014). Climate change impacts are likely to be most severe in the tropics where the majority of global commodity-driven deforestation is also concentrated (Curtis et al., 2018). Uncertainties in future projections are also more severe for certain agricultural commodities such as soy, which have more concentrated production areas, increasing sensitivity to regional differences in model projections. Furthermore, depending on the model and scenario used, impacts of future cropland expansion can vary considerably (Molotoks et al., 2020). Predicting future hotspots of commodity-driven deforestation is therefore difficult and subject to variations in model structures and assumptions, underpinning a necessity for policy, which can respond to ‘early warning’, signs of deforestation occurring in new areas should these emerge in response to changing climate.

Overview of policy recommendations

Recommendations for policy drawn from the synthesised literature include a need for international collaboration between producer and consumer countries, multi-stakeholder engagement, and increased transparency of supply chains to address deforestation footprints. For the country-commodity contexts with the highest deforestation risks, Brazilian soy and beef, and Indonesian palm oil (Figure 6), forest and commodity-based exports are often particularly important for economic development. Hence there is a strong trade-off between development and environmental sustainability which makes working closely with these countries to support forest-risk commodity producers particularly important (Bager et al., 2021). There is also often a trade-off between political feasibility and impact on reducing deforestation footprints. Policy options which are more feasible to implement tend to have a weaker theory of change and, in light of voluntary commitments being relatively ineffective (Garrett et al., 2019), multiple policy levers are likely to be required.

Voluntary certification programmes have been the focus of previous initiatives to reduce deforestation, however, these schemes only cover a small share of the global market (Garrett et al., 2019). They also place the burden for proving compliance on producers and changing policies or inconsistent enforcement can affect the ability of companies to implement commitments. Adoption of zero-deforestation certifications is very low for soy (Garrett et al., 2019) and certified production only covers a small proportion of conventional production of major forest-risk commodities (van der Ven et al., 2018). Furthermore, there has been a lack of success in halting land use change patterns through certification schemes, partly due to lack of uptake, but also due to regulatory loopholes and lack of enforcement (van der Ven et al., 2018). There is also a major gap in commitments around beef production, with indirect suppliers not considered, hence there has been very little impact on reducing deforestation in this sector (Alix-Garcia & Gibbs, 2017)

Other supply-side commitments have been more successful. For example, the soy moratorium has been evaluated by recent reports to confirm their effectiveness due to a combination of multi-stakeholder action enforced through strict accountability (Gibbs et al., 2015). However, studies included in this rapid review have stressed that a combined approach of both demand and supply side policies will be essential for effective reduction in commodity-driven deforestation. Policies to prevent deforestation therefore also need measures that target consumers, hence demand-side measures which target lifestyle changes can also encourage deforestation-free production along supply chains (Henders et al., 2018). Reducing meat consumption, in particular beef, can effectively reduce deforestation and associated emissions (Sandström et al., 2018). By reducing demand for beef, it is possible to reduce emissions related to feed production, particularly soy, which is often embedded within animal products consumed in the EU. As per the recommendations of the UK’s Global Resource Initiative, policy makers should therefore carefully consider the potential to introduce demand-linked measures (such as the promotion of lifestyle changes) in addition to the measures focusing on supply chain actors placing forest-risk commodities on the market (GRI, 2020).

Provision of economic incentives is one option for shifting the demand for forest-risk commodity consumption, however carbon taxation of food for example has shown to be relatively ineffective at affecting consumption (Bager et al., 2021). This approach is more likely to be effective if levied on supply-chain actors where economic incentives are stronger for shifting consumption patterns. However, for this to be effective in reducing deforestation risk, emissions should be differentiated relative to contributions to deforestation. Economic incentives are also mentioned in studies included in this review as likely being necessary to promote increased supply chain transparency (Godar et al., 2016; Trase, 2018). Furthermore, there have recently been calls from 22 major companies, supported by the Sustainable Trade Initiative UK, for provision of additional measures, including reporting guidelines, sector specific requirements, and incentive-based approaches alongside due diligence legislation (IDH, 2020). This suggests that, despite mandatory due diligence for supply chain actors being effective in reducing imported deforestation and relatively implementable (Bager et al., 2021), the UK due diligence does not go far enough. Therefore, strengthening this legislation and applying other measures are likely to be required to overcome feasibility barriers that due diligence alone faces, such as jurisdictional approaches and wider stakeholder engagement (von Essen & Lambin, 2021).

Underlying data

Zenodo. Which forest-risk commodities imported to the UK have the highest overseas impacts? A rapid evidence synthesis systematic review. DOI 10.5281/zenodo.5227100

This project contains the following underlying data:

Appendix A: Original review protocol

Appendix B: Data extraction table.

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).