Beneficial land use change: Strategic expansion of new biomass plantations can reduce environmental impacts from EU agriculture

Society faces the double challenge of increasing biomass production to meet the future demands for food, materials and bioenergy, while addressing negative impacts of current (and future) land use. In the discourse, land use change (LUC) has often been considered as negative, referring to impacts of deforestation and expansion of biomass plantations. However, strategic establishment of suitable perennial production systems in agricultural landscapes can mitigate environmental impacts of current crop production, while providing biomass for the bioeconomy. Here, we explore the potential for such “beneficial LUC” in EU28. First, we map and quantify the degree of accumulated soil organic carbon losses, soil loss by wind and water erosion, nitrogen emissions to water, and recurring floods, in ∼81.000 individual landscapes in EU28. We then estimate the effectiveness in mitigating these impacts through establishment of perennial plants, in each landscape. The results indicate that there is a substantial potential for effective impact mitigation. Depending on criteria selection, 10–46% of the land used for annual crop production in EU28 is located in landscapes that could be considered priority areas for beneficial LUC. These areas are scattered all over Europe, but there are notable “hot-spots” where priority areas are concentrated, e.g., large parts of Denmark, western UK, The Po valley in Italy, and the Danube basin. While some policy developments support beneficial LUC, implementation could benefit from attempts to realize synergies between different Sustainable Development Goals, e.g., “Zero hunger”, “Clean water and sanitation”, “Affordable and Clean Energy”, “Climate Action”, and “Life on Land”.


Introduction
The exploitation of fossil fuels has been a powerful driver of global societal development in the twentieth century, resulting in a reduced relative dependency on biomass. One notable example is the complete transformation of the energy systems -from biomass based to fossil based. The food sector has also undergone large changes; while most of our food still comes from agriculture, it is often produced in an intensive manner, relying on fossil fuels and petroleum-based chemicals. This development, especially the invention of synthetic fertilizers, has limited the need for expanding agricultural land, while the global population, and its affluence, has steadily increased. Nevertheless, biomass resources are of major significance for the economy in many countries (FAO, 2014;Alston and Pardey, 2014). As a growing and wealthier global population requires more food, paper, construction wood, and other biomaterials, the demand for land and biomass is expected to increase (Scarlat et al., 2015). This is further accelerated by societal concerns about resource scarcity and impacts associated with the use of non-renewable resources -not the least climate change (Scarlat et al., 2015). Visions of a biobased circular economy have caused countries, organizations, and companies to adopt policies, regulations, and strategies aimed at substituting fossil materials with biomass (D'Amato et al., 2017). Most notably, bioenergy is expected to play a major role in the substitution of fossil energy necessary to meet global climate targets (Clarke et al. (2014);IPCC, 2018).
At the same time, human societies have already put almost half of the world's land surface to their service and have caused extensive land degradation and loss of biodiversity worldwide (Rockström et al., 2009). As we manage landscapes and associated ecosystems for the production of biomass, we often alter their capacity to support other ecosystem services (ES) that are essential for human well-being (Smith et al., 2013). Many ecosystems are currently being degraded or used unsustainably, jeopardizing their capacity to support multiple ES over time (Costanza et al., 2014). The cultivation of annual crops is an important example, as, e.g., nutrient and agrochemical runoff to water bodies, soil carbon losses, and erosion can cause impacts such as eutrophication, climate change, and soil degradation, in the absence of a parallel supply of ES (i.e., nutrient retention, soil carbon sequestration and regulation of mass flows) that can regulate these stressors (Power 2010). Such impacts can be observed in all parts of the world where there is intensive production of annual crops, including Europe (Grizzetti et al., 2012;Panagos et al., 2015;Borrelli et al., 2017;Alfieri et al., 2014;Lugato et al., 2014a).
Implications of an increased biomass supply have therefore been debated for many decades, primarily focusing on bioenergy, with key issues being land use impacts and uncertain climate benefits (Abad et al., 2017;Berndes et al., 2003;Creutzig et al., 2015;Leemans et al., 1996;Slade et al., 2014;Smith et al., 2013). One example is the debate and research activity following the biomass intensive scenario (LESS) in the Second Assessment Report of IPCC. More recently, a similar debate has arisen following IPCC AR5 (Clarke et al., 2014) and IPCC SR1.5 (IPCC, 2018), in which bioenergy with carbon capture and storage is relied upon in most of the considered scenarios where the mean temperature increase is limited to 1.5°C or 2°C above the pre-industrial level. In the discourse, land use change (LUC) has often been considered as negative, referring to environmental and socio-economic impacts of deforestation and expansion of biomass plantations on previously uncultivated land, e.g., habitat loss, greenhouse gas emissions, soil degradation, and water pollution (Searchinger et al., 2008;Kline and Dale 2008;Berndes et al. 2012). In relation to the IPCC AR6 cycle, Smith and Porter (2018) identify key emerging issues to be (i) trade-offs between the use of land for bioenergy production, food and fibre production, and conservation of ecosystem integrity and (ii) the codelivery of bioenergy based climate change mitigation (with or without carbon capture and storage) and the UN Sustainable Development Goals (SDGs). These issues were also prominent in the recently approved Summary for Policymakers of the IPCC report on Climate Change and Land (IPCC 2019).
Society thus faces the double challenge of increasing biomass production to meet the future demands for food, materials and bioenergy, while addressing negative impacts of current (and future) land use. In relation to this, there is a growing body of literature that investigates opportunities for achieving``beneficial LUC'', where a strategic integration of perennial plants (``perennials'') into agricultural landscapes enhances, e.g., landscape diversity, habitat quality, retention of nutrients and sediment, erosion control, climate regulation, pollination, pest and disease control, and flood regulation (Asbjornsen et al., 2014;Berndes et al., 2008;Christen and Dalgaard, 2013;Dauber and Miyake, 2016;Holland et al., 2015;Milner et al., 2016;Styles et al., 2016;Ssegane et al., 2015;Ssegane and Negri, 2016;Zumpf et al., 2017;Cacho et al., 2017). Such LUC can thereby mitigate environmental impacts from intensive agriculture, while maintaining or increasing total productivity. Perennial grasses (e.g. Miscanthus, reed canary grass, switchgrass) as well as woody plants (e.g., short-rotation coppice willow or poplar) can be used for such purposes. There is significant experience of this type of biomass supply systems from both practical field trials and commercial applications (Berndes et al., 2008;2004;Börjesson, 1999a;Börjesson and Berndes, 2006;Christian et al., 1994;Göransson, 1994;Grigal and Berguson, 1998;Gustafsson, 1987;Kort et al., 1998;Perttu and Kowalik, 1997;Rijtema and DeVries, 1994). Implementation of beneficial LUC through such strategic perennialization can support a growing use of bioenergy and other bio-based products while advancing several SDGs, e.g.,``Zero hunger'',`C lean water and sanitation'',``Affordable and Clean Energy'',``Climate Action'', and``Life on Land''.
Most earlier studies of beneficial LUC, as referred to above, are conceptual or adopt a limited geographical scope. Few have investigated the possible extent and spatial distribution at larger scales. This article presents the first attempt to explore the potential for beneficial LUC across EU28, based on high-resolution land use modeling. We identify and quantify: (1) The degree of selected environmental impacts associated with agriculture (soil loss by wind and water erosion, nitrogen emissions to water, accumulated loss of soil organic carbon (SOC), and recurring floods) in ∼81 000 individual landscapes in EU28.
(2) The extent to which strategic introduction of perennials in individual landscapes (from here on referred to as``strategic perennialization'') could mitigate these impacts.
(3) Agricultural areas where strategic perennialization may be particularly beneficial from an environmental point of view.
Finally, we discuss policy implications for realizing beneficial LUC on a larger scale in EU28.

Spatial analysis unit
The spatial analysis unit for the assessment is equivalent to functional elementary catchments (FECs) from the ECRINS database (European Environment Agency, 2012), modified as specified below. FEC is equivalent to sub-watershed. This unit was selected based on the importance of hydrological processes, constrained by a watershed, in determining how nutrient and sediment retention and the control of water and mass flows can be affected by a change in land use. It was also considered an appropriate size for assessing implementation options.
Throughout this article, the analysis units are also referred to as landscapes. While there are varying meanings of the term landscape, it is here defined as an intermediate integration level between the field and the physiographic region (Burel and Baudry, 2003;Turner, 1989), with an extent depending on the spatial range of the biophysical and anthropogenic processes driving the processes under study (Lacoste et al., 2014). The term "landscape-scale" is also commonly used in both scientific studies and policies concerning implementation of measures for mitigating environmental impacts (Englund et al., 2017). A thorough discussion on the use of the terms landscape and landscape scale is provided by Englund et al. (2017).

FECs to landscapes
The following modifications were made to the original FEC dataset. 1 All GIS operations were made using the coordinate reference system ETRS89-LAEA Europe (EPSG:3035).
1 The original dataset included a total of 81,301 FECs in EU28, Norway, and Switzerland. In the construction of the original dataset (European Environment Agency, 2012), a number of FECs were represented by more than one polygon. This had to be resolved since one landscape cannot consist of several polygons. These multipolygon FECs could not be "dissolved" since they in many cases were not located next to each other. Instead, they were split into 1 In the below description, "FEC" refers to features in the original dataset while "landscape" refers to features in the resulting dataset.

Degree of negative environmental impacts
Five environmental impacts that could be mitigated by the introduction of perennials into intensive arable landscapes were included in this assessment (Table 1). Each impact can be attributed to insufficient supply of, or degraded, ES under current agricultural practices. The relationship between ES, environmental impacts, and the spatial indicator used for impact classification is available in Table 1.
Each landscape was classified as having very low, low, medium, high, or very high (i) nutrient emissions to water, (ii) soil loss by water erosion, (iii) soil loss by wind erosion, (iv) recurring floods, and (v) accumulated loss of soil organic carbon (SOC). This classification was made using spatial indicators, as summarized in Table 1 and described below.

Nitrogen emissions to water
Indicated by``annual average diffuse nitrogen emissions to water'', retrieved by running v2 of the Geospatial Regression Equation for European Nutrient losses (GREEN) model (Grizzetti et al., 2012) for the landscape dataset. Diffuse sources include mineral fertilizers, manure applications, atmospheric deposition, crop fixation, and scattered dwellings. For each sub-basin (i.e., landscape), the model considers the total input of diffuse sources and estimates the nutrient fraction retained during the transport from land to surface water.
Thresholds for classification were based on expert (i.e., model developer) recommendations (Table 1).

Soil loss by erosion
Indicated by "annual average soil loss by water erosion on land used for production of annual crops". Annual soil loss was retrieved from a published dataset for the year 2010 with 100 m resolution, available at the Joint Research Centre European Soil Data Centre (ESDAC; https:// 2 This was done as step 4 instead of step 1 due to an initial aim of including these countries in the assessment. O. Englund, et al. Global Environmental Change 60 (2020) 101990 esdac.jrc.ec.europa.eu/) based on the application of a modified version of the Revised Universal Soil Loss Equation (RUSLE) model (RUSLE2015), within which rainfall erosivity, soil erodibility, covermanagement, topography, and support practices were modelled with the most recently available pan-European datasets (Panagos et al., 2015). The degree of soil loss from wind erosion was estimated and classified as described above for water erosion, but using published data with 1000 m resolution (available at ESDAC). The data were derived using a GIS version (RWEQ-GIS) (Borrelli et al., 2017) of the Revised Wind Erosion Equation (RWEQ) model (Fryrear et al., 2000), a tool extensively tested to perform field-based predictions of soil loss due to wind erosion. RWEQ-GIS computes the soil loss potential on a daily basis for each 1000 m cell during the period between January 2001 and December 2010, by combining soil properties and daily data of rainfall, wind speed, evapotranspiration, soil moisture and crop canopy cover.
Total soil loss by erosion was calculated by summing soil loss by water and wind erosion, respectively.
Soil loss by water erosion, wind erosion, and total erosion, respectively, on land classified as annual crop production (see Section 2.4), was then averaged for each landscape. Thresholds for classification were applied based on Panagos et al. (2015; as specified in Table 1.

Recurring floods
Indicated by``share of landscape area subject to 10-year flooding''. Data on 10-year flooding events were retrieved from a published flood hazard dataset with 100 m resolution. The data were derived using a cascading model simulation approach composed of the following steps: (1) Distributed hydrological model setup and calibration; (2) Simulation of a long-term discharge time series and derivation of peak flows with selected return period; (3) Downscaling to 100 m spatial resolution and derivation of design flood hydrographs; and (4) Floodplain hydraulic simulations and merging of output flood depth maps (Alfieri et al., 2014). To indicate the degree to which individual landscapes are prone to recurring floods, the share of the total area in each landscape subject to 10-year flooding events was calculated for each landscape. Thresholds for classification were then applied as specified in Table 1.

Accumulated losses of soil organic carbon
Indicated by``average SOC saturation capacity on land used for production of annual crops''. Data on SOC saturation capacity (expressed as the ratio of current SOC relative to the theoretical maximum potential) were taken from a published dataset with 250 m resolution, available at ESDAC. The data were created using a simulation platform that integrates the CENTURY agroecosystem model (Parton et al., 1988) with several Pan-European spatial and statistical databases (Lugato et al., 2014b) and simulates the changes in SOC over the period 2013-2100 by replacing current land use with alternative management practices (Lugato et al., 2014a). The data used for the purpose of this study represents the conversion of current land use to grassland, as this scenario resulted in the largest positive gain in SOC overall in Europe (Lugato et al., 2014a). For each landscape, the average SOC saturation capacity on land used for annual crop production was then calculated. Thresholds were defined using geostatistical properties to define five equal intervals between the minimum and maximum aggregated average SOC saturation capacity values, based on expert (i.e. model developer) recommendations (Table 1).
For this impact, high (10% higher threshold values) and low (10% lower threshold values) estimates were also defined, to enable a sensitivity test (see Section 2.7).

Mitigation potential of strategic perennialization
Perennialization in the form of wind breaks can increase yields for annual crops on land protected from wind, due to reduced crop damages (e.g., plant blasting, coverage of plants, uncovered roots and seeds), while also avoiding losses of organic matter and fine soil particles that can lead to decreased soil fertility. To be effective, windbreak cultivations need to be several meters high, hence preferably based on woody crops. For example, 50-meter wide willow plantations located 100 m apart can provide continuous sheltering in areas exposed to wind erosion and on sensitive soils, if half of the plantation width is harvested at a time (Börjesson, 1999a).
Perennial cultivations can be used as riparian buffer strips and filter zones reducing nutrient (and other agrochemical) emissions from arable land. Plantations designed and managed similarly as for windbreaks can be located along open waterways to continuously capture nutrients (Berndes et al., 2008;Styles et al., 2016;Ferrarini et al., 2017). Riparian buffer zones may consist of perennial grass cultivations and/or short-rotation woody plantations. Field trials have shown that N removal rates between herbaceous and woody crops, and between planted and spontaneous crops, are comparable (Ferrarini et al., 2017). A 20 m buffer with SRC and/or grass has been suggested to have 100% nitrate removal effectiveness (Ferrarini et al., 2017). However, several different designs have been suggested in the literature, from 50 m with SRC willow (Styles et al., 2016) to 5 m with grass (Ferrarini et al., 2017). On arable land with covered drainage systems, nutrient-rich drainage water can be collected in storage ponds and used for irrigation. Besides efficient nutrient retention and water purification, the irrigation can improve yield levels and reduce the need for commercial fertilizers (Börjesson and Berndes, 2006). Vegetation zones, or strips of perennial crop cultivations, can also be located in areas sensitive to rill erosion, particularly on fields with clayey and silty soils in hilly areas (Börjesson, 1999a). Prevention of water erosion requires continuous soil cover, which can make perennial grass cultivations preferable to short-rotation woody plantations. Similar types of vegetation zones can also be used for flood prevention (Berndes et al., 2008). Besides the onsite benefits of reduced soil losses, there are also offsite benefits, such as reduced sediment loading in reservoirs and irrigation channels, as well as reduced deterioration in the quality of river water due to the suspended load that accompanies flood waters formed mostly by runoff.
Independently of the type of perennial cultivation, replacement of annual crops with perennial crops normally leads to increased soil carbon sequestration (Whitaker et al., 2018). This is due to a combination of an increased input of organic matter to the soil and reduced soil tillage, leading to decreased decomposition of soil organic matter by microorganisms. Thus, this benefit will normally be provided in all situations where annual crops are replaced (Berndes et al., 2008). The extent may however vary geographically, due to local and regional climate conditions as well as the historical land use, e.g., the intensity in previous cultivation of annual crops (Berndes et al., 2012). This is also illustrated in the concept of SOC saturation capacity (Lugato et al., 2014a;2014b), used as indicator for accumulated SOC losses in this study.

Annual crop dominance
The introduction of perennial crops for mitigating environmental impacts can only be effective in landscapes dominated by the production of annual crops, which has caused the environmental impacts by degrading the regulating ES supply. To estimate the effectiveness of perennialization, the annual crop dominance, i.e., the share of land in each landscape used for the production of annual crops compared with the total vegetated area, was calculated for each landscape.
The share of annual crops in each landscape was calculated using the CORINE 2012 100 m LULC dataset (Copernicus Land Monitoring Service, 2018). The CORINE raster was first reclassified from 47 to four land use classes,``annual crops'',``other agriculture'',``other vegetation'' and``unvegetated'' ( Table 2). The number of 100 m cells was then calculated for each of the four land use classes within each landscape O. Englund, et al. Global Environmental Change 60 (2020) 101990 unit. Finally, the share of annual crops of all vegetation was calculated in each landscape (annual crops / (annual crops + other agriculture + other vegetation)). Thresholds for annual crop dominance classes were defined based on univariate statistics, as specified in Table 3. The distribution was skewed (mean: 0.33, median: 0.27, skewness: 0.62) so quantiles were used to define reasonable thresholds. Note that landscapes without annual crops were excluded in the computation of quantiles but still (naturally) classified as very low annual crop production dominance. This class therefore has significantly more observations than other classes.

Mitigation effectiveness of strategic perennialization
The annual crop dominance and the estimated degree of the five environmental impacts were combined to define four levels of expected effectiveness of perennialization, as illustrated in Table 4. This level was calculated for each environmental impact in each landscape.

Priority areas for strategic perennialization
Priority areas for beneficial LUC are conceptually referred to as landscape units where the environmental effects of perennialization are estimated to be particularly beneficial. In the modeling framework, priority areas are defined as landscapes where 1 one environmental impact could be mitigated with very high effectiveness, or 2 multiple impacts could be mitigated with either high or very high effectiveness To identify the latter, the number of impacts for which perennialization was classified as having a high and very high expected effectiveness, respectively, were identified (see Section 2.5) and counted for each landscape.

Sensitivity analysis
Accumulated SOC losses had a very high influence in the identification of priority areas (see Results). To test how sensitive the identification of priority areas is to variations in threshold definitions, a high (thresholds increased with 10%) and low (thresholds decreased with 10%) estimate of accumulated SOC losses (Table 1) were used in the identification of priority areas.

Effectiveness of strategic perennialization
The extent to which the assessed environmental impacts can be mitigated by perennialization depends on the degree of environmental impact in the landscape, and the dominance of annual crops relative to other vegetation. As summarized in Table 5 (see Table S1 and S2 for more information) and detailed below, the results indicate that there is a substantial potential for effective mitigation regarding all the assessed impacts.
The production of annual crops is an important determinant for accumulated loss in SOC. For other impacts, the spatial correlation is weaker, indicating that there are additional important biophysical factors influencing the degree of soil erosion, nitrogen emissions to water, and recurring floods. For example, nitrogen emissions to water can be very high in areas with high precipitation and/or intensive livestock production, even if the land is largely covered by perennials (see, e.g., Ireland in Fig. 1). The same can be seen for soil loss by water erosion which can be high in mountainous areas or on land with steep slopes, regardless of the land use. Soil loss by wind erosion, the least severe impact overall, is largely driven by wind exposure, hence mainly limited to coastal areas or higher altitudes, but also by structural deficits and topsoil texture. It can be observed that where several contributing parameters co-exist, the degree of environmental impact is particularly high.

Nitrogen emissions to water
Nitrogen emissions to water is classified as high to very high in 9% of all landscapes in EU28, containing 11% of the total area under annual crop production ( Table 5). The majority of these landscapes are located in north-western Europe; most notably in Ireland, Western UK, Denmark, and the Netherlands (Fig. 1).
Mitigation of nitrogen emissions to water by strategic perennialization could be achieved with high or very high effectiveness in 4.4% of all landscapes, containing 12% of the total area under annual crop production ( Table 5). As for the impact, the mitigation effectiveness is significant mainly in north-western Europe; primarily in large parts of the UK and Denmark, as well as parts of the Netherlands and Belgium, northern France, western Germany, the Po Valley in Italy and in the western parts of the Danube basin (Fig. 1).

Soil loss by erosion
Soil loss by water erosion is classified as high to very high in 12% of all landscapes in EU28, containing 12% of the total area under annual crop production ( Table 5). The majority of these landscapes are located in southern Europe; most notably in large parts of Italy and parts of Spain, Romania, Slovakia, and southern Poland (Fig. 2).
Soil loss by wind erosion is a lesser concern, in general; classified as high to very high in 0.4% of all landscapes in EU28, containing 1% of the total area under annual crop production ( Table 5). The majority of these landscapes are located in western UK, Denmark, the Netherlands and eastern Bulgaria (Fig. 2).
Total loss by wind and water erosion combined is classified as high  to very high in 14% of all landscapes in EU28, containing 15% of the total area under annual crop production ( Table 5). Mitigation of soil loss by either wind or water erosion by strategic perennialization could be achieved with high or very high effectiveness in just over 8% of all landscapes, containing about a quarter of the total area under annual crop production ( Table 5). The mitigation effectiveness is significant in areas scattered all over Europe, but most notably in Eastern UK, Denmark, Spain, Italy, Romania, Bulgaria, and southern Poland (Fig. 3).

Table 4
Expected effectiveness of perennialization in mitigating negative environmental impacts by enhancing corresponding ecosystem services. Colours indicate marginal (blue), low (purple), medium (light red), high, (orange), and very high (yellow) expected effectiveness. Colours are identical as in Figs. 1-5.

Table 5
Degree of environmental impacts and mitigation effectiveness of strategic perennialization in European landscapes. More information is available in O. Englund, et al. Global Environmental Change 60 (2020)

Recurring floods
Recurring floods is classified as high to very high in 11% of all landscapes in EU28, containing 14% of the total area under annual crop production (Table 5). These landscapes are primarily located along major rivers, such as the Danube, Po, Elbe, Oder, Vistula, and Rhône (Fig. 4).
Mitigation of recurring floods by strategic perennialization could be achieved with high or very high effectiveness in 6% of all landscapes, containing 16% of the total area under annual crop production ( Table 5). The mitigation effectiveness is significant mainly in the Po Valley in Italy and along the Danube basin, but also in areas around other rivers throughout Europe (Fig. 4).

Accumulated loss of soil organic carbon
Accumulated losses of SOC is classified as high to very high in about a third of all landscapes in EU28, containing 70% of the total area under annual crop production (Table 5). These landscapes are scattered all over Europe, having a strong spatial correlation with the production of annual crops (Fig. 5).
Mitigation of SOC losses by strategic perennialization could be achieved with high or very high effectiveness in 17% of all landscapes, containing almost two thirds of the total area under annual crop production (Table 5). The mitigation effectiveness is significant in areas all over Europe; primarily in eastern UK, northern France, and large parts of Denmark, Italy, Spain, Germany, Poland, Lithuania, Czech Republic, Hungary, Romania and Bulgaria (Fig. 5).

Priority areas for strategic perennialization
The majority of annual crops cultivated in EU is located in landscapes where strategic perennialization can help mitigating different environmental impacts, in different ways and to different extents. Areas   O. Englund, et al. Global Environmental Change 60 (2020) 101990 where perennialization can be particularly beneficial, from an environmental perspective, are here identified as Priority areas for beneficial LUC. A total of 1764 landscapes, harboring 9% of total annual crop production in EU, can be considered priority areas, due to expected mitigation of a single environmental impact by perennialization with very high effectiveness (Table 6, see also Table S3 for more information).
Priority areas could also be defined as landscapes where multiple impacts can be mitigated with either high or very high effectiveness. Depending on the required number of impacts to be mitigated, such priority areas contain 1% (for four mitigated impacts), 9% (at least three impacts), or 37% (at least two impacts) of total annual crop production in EU, respectively (Table 6). Combined, these two types of priority areas cover 15-60 million hectares, harboring 10-46% of total annual crop production in EU. These areas are scattered all over Europe, but there are notable "hotspots" where priority areas are concentrated. This can be seen in, e.g., large parts of Denmark, western UK, The Po valley in Italy, and the Danube basin, but also in northern France, and several regions in, e.g., Spain, Germany, and Italy (Fig. 6).

Sensitivity analysis
Both the high and the low estimates of accumulated SOC losses had substantial effects on impact classification (Table S4). Effectiveness classification was less affected, which was expected as this is also influenced by the annual crop dominance (Table S4). There were, however, notable relative differences compared with the main analysis on the number of landscapes where perennialization was classified as having a very high mitigation effectiveness (Table S4). This can also be seen in the identification of priority areas defined as landscapes with very high expected mitigation effectiveness of a single environmental impact, where the low estimate resulted in a 25% decrease, and the high estimate resulted in a 51% increase, respectively, in the number of landscapes (Table 7). In total, the number of priority areas decreased with 13% in the low estimate and increased with 16% in the high estimate (Table 7).
Spatial patterns of impact and effectiveness classification, respectively (Fig. S1), as well as of priority areas (Figs. 6-7), in both the high and the low estimate are comparable with the main analysis. Priority "hot spots" are therefore very similar (Figs. 6-7).

Discussion
While the results indicate that large areas under annual crops could be subject to strategic perennialization, only parts of these areas would need to be converted to perennial systems. The area that need to be converted for achieving successful impact mitigation basically depends on the type and degree of the impact and what management system is implemented, which in turn can be influenced also by other factors, such as practicality in terms of planting and harvesting (determining, e.g., size of plantations) and local preferences concerning the landscape aesthetics (determining, e.g., selecting woody or herbaceous crops). It also depends on how to interpret "successful" impact mitigation. For example, to completely restore accumulated losses of SOC throughout a landscape, the entire cropland area in this landscape need to be converted to, e.g., grassland and maintained as such for a long period of time. If this is not desirable, a smaller share of the cropland area could instead be converted to enhance SOC at the landscape scale. Furthermore, the area of riparian buffers needed to mitigate N emissions to water depends on the width of the strip (5-50 m; see Section 2.3), as well as the total length of rivers in the landscape. It is thus difficult at this point to provide estimates of areas needed for strategic perennialization, and their corresponding impact mitigation effectiveness and biomass production. Preliminary calculations for riparian buffers however indicate that it could suffice to convert about 1% of the total cropland area in EU, to establish 20 m wide buffer strips in all landscapes where the effectiveness of strategic perennialization for mitigating nitrogen emissions to water is classified as high or very high.
While regional and national assessments can indicate areas where strategic perennialization could be environmentally beneficial, the actual effects of introducing perennials in agricultural landscapes depend on crop selection, management system, location in the landscape, and  O. Englund, et al. Global Environmental Change 60 (2020) 101990 Table 6 The total number of landscapes and areas under annual crops where strategic perennialization can mitigate different numbers of environmental impacts, with a high and/or very high effectiveness. Numbers in the coloured rows can be linked to identically coloured areas in Fig 6. See Table S3 for more information.

Fig. 6.
Priority areas for beneficial LUC through strategic perennialization. In case a landscape appears in both the orange "very high" category and any of the blue "high to very high" categories (cf. Table 6), the latter is prioritized for visualization.
biotic and abiotic landscape characteristics. To fully understand-quantitatively as well as spatially-the effects of perennialization, high-resolution spatially explicit analysis within individual landscapes is required (Englund et al., 2017). One important characteristic to include in such assessments can be sub-landscape or sub-field variations in cropland productivity, as demonstrated by Ssegane et al. (2015). Targeting strategic perennialization to such land could provide environmental benefits with minimal impact on of total agricultural production. This could also inform farmers about alternative management systems for land with limited productivity, which can provide environmental benefits and possibly also increase total productivity at the farm level.

Model evaluation
The classification of environmental impacts and dominance of annual crops was supported by high-quality spatial models and datasets, calibrated and validated using empirical data, as summarized below. Further information is available in the original articles.
• Nitrogen emissions to water was estimated using the GREEN model, run specifically for the landscape dataset in this study. The model is calibrated using about 1400 measurement points of surface water quality and quantity, between 1985-2005. The Nash-Sutcliffe (1970) coefficient of efficiency of calibration was 92%, with yearly efficiencies ranging from 76% to 97%. The comparison between measured and estimated loads did not show any significant systematic or temporal deviations. For the 63 stations where complete time series were available, the correlation between the trends in measurements and in model estimates (computed as the slope of the linear interpolation) was 84%, indicating that the model is capturing rather well the observed temporal trends (Grizzetti et al., 2012).
• Soil loss by water erosion was estimated using a 100 m pan-European dataset derived from the RUSLE2015 model. The mean loss rates and spatial patterns are very close to national data reported in the EIONET-SOIL database for Germany, the Netherlands, Bulgaria, Poland and Denmark. It was found to be the most suitable modelling approach for estimating soil loss at the European scale, in terms of validation, usability, replicability, transparency, and parameterisation (Panagos et al., 2015).
• Soil loss by wind erosion was estimated using a 1000 m pan-European dataset from the GIS-RWEQ model. A cross-validation of the model showed that the predicted soil loss rates were generally in agreement with wind erosion sites reported in literature; 85 of 90 reported locations (94.4%) were classified by the model as being susceptible to erosion. Thereof, 23.3% of the literature sites fell into areas modelled as high erosion areas, whereas 48.9% fell into areas where slight to moderate erosion was predicted. The remaining 22.2% literature sites fell into areas classified as being very low to low erosive (Borrelli et al., 2017).
• Recurring floods was estimated using the first quantitative pan-European flood hazard assessment, representing the largest application of its kind at 100 m resolution. The map was evaluated against national/regional maps for three areas: the state of Saxony in Germany, the Thames, and the Severn River basin in the United Kingdom. Overall, the overlap between the pan-European and the national/regional maps ranges between 59% and 79%, depending on the region and the aggregation scale considered (Alfieri et al., 2014).
• Accumulated SOC losses was estimated using a 250 m pan-European dataset created using a simulation platform that integrates the CENTURY agroecosystem model with several pan-European spatial and statistical databases. Simulation values were validated against two independent empirical datasets, LUCAS and EIONET-SOIL. Simulated values showed a good agreement with measured values in Table 7 Priority areas calculated as described in Section 3.2 using a high and low estimate of accumulated SOC losses, respectively. O. Englund, et al. Global Environmental Change 60 (2020) 101990 all the aggregated land uses considered, with no particular bias (Lugato et al., 2014b).
• Dominance of annual crops was estimated using the 100 m CLC2012 land use dataset, produced within the CORINE Land Cover programme coordinated by the European Environment Agency. CORINE provides consistent information on land cover and land cover changes across Europe. CLC products are based on the photointerpretation of satellite images by the national teams of the participating countries. The resulting national land cover inventories are further integrated into a seamless land cover map of Europe (Copernicus Land Monitoring service, 2018).
The results are however sensitive to the threshold values used for the classification of negative impacts and annual crop dominance, as illustrated in Section 3.3. Impacts were classified based on advice from the model developers and providers of indicator datasets, except for the case of recurring floods, where the classification was based on arbitrarily defined thresholds. Thresholds for annual crop dominance classes were also arbitrarily defined based on univariate statistics. While results for individual landscapes are sensitive to threshold definitions, spatial patterns are generally not. The results presented here are therefore considered particularly useful for indicating relative differences between areas, and for identifying locations where perennialization can be particularly interesting from an environmental point of view. Such locations could later be subject of more detailed assessments, as discussed below.

Model extension and adaptation
The model presented here for estimating effectiveness of strategic perennialization can be further developed to, e.g. include additional environmental impact categories, and adapted to better suit, e.g., application at other geographical scales. As discussed above, additional analytical work is needed to provide estimates of areas for strategic perennialization, and corresponding impact mitigation effectiveness and perennial biomass production.
Applying the model at larger scales would be challenging, as sufficiently reliable data at high resolution is often lacking. It would require combining many different datasets and accepting large uncertainties.
Applying the model at national scale can be done using national or regional datasets with higher precision, and thus produce more reliable results. Such results can also be more easily evaluated as the national context can be fully considered with the involvement of relevant stakeholders.

Policy considerations
Policies and regulations put in place to establish a societal transition towards the Paris targets will likely lead to an increased biomass demand for bioenergy and other bio-based products. Yet, despite that knowledge and practical experience from field trials and commercial applications have existed for several decades, perennialization activities of the type described in this study rarely takes place in EU. Studies commonly find significant socioeconomic values (Berndes et al., 2008;Börjesson, 1999b;Börjesson and Berndes, 2006), but the incentives for farmers to achieve such beneficial LUC have not been sufficiently strong. The Common Agricultural Policy (CAP) of the EU has historically not provided direct support for perennial plantations producing biomass feedstock for, e.g., energy purposes. Inadequate knowledge support, low biomass prices, and market uncertainty are other reasons behind slow development for production systems with perennial grasses and woody crops (Dimitriou et al., 2018;Dimitriou et al., 2011).
The effectiveness in promoting beneficial LUC may increase if policies and regulations seek synergies between climate change mitigation, energy security, and other societal goals, e.g., related to SDGs. Recent policy development is favorable in some areas. For example, the CAP currently requires that all arable areas exceeding 15 ha must set aside 5% of the area for``ecologically beneficial elements'' (Ecological Focus Areas, EFAs). The main purpose of EFAs is to enhance biodiversity, but also to provide other environmental benefits. EFAs can be in the form of, e.g., fallow land, terraces, landscape features, buffer strips, agroforestry, strips along forest edges, short rotation coppice with no use of fertilizers and/or plant protection products, catch crops, and nitrogen-fixing crops (European Parliament and the Council, 2013). The biomass produced on these areas is allowed to be used as feedstock for various purposes, including bioenergy. This may act as a driver for increased perennialization in agricultural landscapes, hence beneficial LUC.
Localization of EFAs in the landscape will be determined by biotic O. Englund, et al. Global Environmental Change 60 (2020) 101990 and abiotic landscape characteristics as well as stakeholder preferences.
In some cases, EFAs may provide the highest environmental benefits by being scattered across the landscape, while in other cases it may be more beneficial to connect EFAs to provide green infrastructure, which would also simplify potential biomass harvesting. The approach presented in this article can be further developed to provide more detailed information on how to localize EFAs to meet different objectives in individual landscapes. Such information can facilitate landscape design processes where landowners, local decision makers, and other relevant stakeholders jointly develop strategies for beneficial LUC that reflect local conditions and preferences (Busch, 2017). If the achievement of beneficial LUC causes losses in the production of agriculture commodities, the production of the same commodities will need to increase elsewhere, unless changes in demand and efficiency improvements along supply chains can fully buffer the losses. Effects of such indirect LUC (iLUC) need to be considered in relation to any measure that aim to reduce land use impacts, e.g., changes from conventional to organic agriculture, restrictions of fertilizer use to protect water, or lower stocking densities in animal agriculture.
In response to concerns that iLUC will cause large negative effects, various approaches to identify so-called low iLUC risk options have been developed (Peters et al., 2016). Options for achieving beneficial LUC through perennialization can provide opportunities to reduce land use impacts while achieving high biomass yields. The biomass can then be refined to multiple products, including biofuels and animal feed, hence substituting conventional (cultivated) feed and reducing grazing requirements (Egeskog et al., 2011;Larsen et al., 2017;Manevski et al., 2017Manevski et al., , 2018Solati et al., 2018;Sparovek et al., 2007). Such options can help maintain or increase agricultural production in a region while limiting environmental impacts, or reduce imports of agricultural commodities that are associated with negative impacts where they are produced. In other cases, when reduced food commodity production will be compensated by increased production elsewhere, this need not imply adverse environmental impacts; outcomes critically depend on the context where production increases, including governance of land use.
Beneficial LUC need not be premised on the requirement that the production of agriculture commodities in a region is not reduced. However, it remains important to consider possible iLUC impacts when evaluating how options for achieving beneficial LUC contribute to set policy objectives, such as GHG emissions reduction. These issues are further addressed in subsequent ongoing studies that quantifybiomass supply potentials and GHG mitigation associated with strategies for achieving beneficial LUC in EU.