Climate‐based identification of suitable cropping areas for giant reed and reed canary grass on marginal land in Central and Southern Europe under climate change

Giant reed (GR) and reed canary grass (RCG) have emerged as promising perennial industrial crops for providing sustainable bioenergy from marginal land. However, there is great uncertainty among farmers and researchers about where these crops can be grown in the future due to climate change, which complicates a timely transition to a bioeconomy. Therefore, this study quantifies marginal land and suitable cropping areas for GR and RCG in Europe, as well as their overlap. To derive these areas, the present (1991–2020) and future (2071–2100, RCP8.5) growing degree days, growing season length, annual precipitation, and aridity index were analyzed using the E‐OBS observational dataset and EURO‐CORDEX regional climate simulations. The study concludes that while marginal land will decrease by ~18%, GR and RCG will profit from the changing European climate, increasing by ~24% and ~13%, respectively. Looking at regions of overlap between marginal land and the selected crops, a decrease of ~87% and an increase of ~462% is projected for RCG and GR, respectively. This is due to marginal land shifting southward, benefitting the warm‐season grass GR, while RCG prefers cooler climates.

the latest. Pathways limiting warming to 1.5°C call for large bioenergy cropland areas, ranging from 560,000 to 4,820,000 km 2 globally by 2100 (Pathak et al., 2022).
However, not all land is suitable for the sustainable cultivation of bioenergy crops. Establishing them on arable land could threaten food security in a world where conflicts and political unrest, for example, in Ethiopia or Ukraine already cause food shortages locally and globally. The projected increase in the intensity and frequency of extreme weather events (Seneviratne et al., 2021) will further exacerbate food insecurity by impacting crop production. Crop failures due to extreme weather disasters can already be observed in Afghanistan's food crisis or in Canada, where the 2021 wheat yield dropped by 38.5% compared to 2020 (Statistics Canada, 2022), both due to severe droughts. More locally, drought conditions in northern Italy are reducing the potential yields of winter crops and exacerbating the competition for water between sectors (Toreti et al., 2022). The displacement of food crops for the benefit of bioenergy cropping areas could also lead to rising food prices (Popp et al., 2017), which would be detrimental to food security. Furthermore, the conversion of natural ecosystems could have adverse effects on biodiversity and the provision of valuable ecosystem services (Núñez-Regueiro et al., 2019). The potential of marginal lands to cultivate bioenergy crops as a solution to this "food, energy and environment trilemma" (Tilman et al., 2009) has been widely recognized (Campbell et al., 2008;Csikós & Tóth, 2023;Khanna et al., 2021;Liu et al., 2017;Richards et al., 2014;. By strategically establishing bioenergy cropping systems on marginal lands, land-use conflicts can be reduced. This would also benefit the advancement of the United Nations' Sustainable Development Goals (SDGs) by contributing to SDG 2 on food security, SDG 7 on clean energy, and SDG 13 on climate change mitigation (Bonfante et al., 2017;IPCC, 2018).
European marginal lands have been mapped by Von Cossel, Wagner, et al. (2019) and within the framework of the EU-funded Horizon 2020 project MAGIC (https:// magic -h2020.eu/) for the present. General current and future potential distributions of selected bioenergy crops in Europe have also been studied (Bellarby et al., 2010;Tuck et al., 2006). The current overlap of suitable land for bioenergy crops and marginal land today has been studied by Ciria et al. (2019) for Spain and has also found some attention in Von  for the  This study determines the current and future shares of marginal land suitable for two contrasting bioenergy crops giant reed (GR) (Arundo donax L.) and reed canary grass (RCG) (Phalaris arundinacea L.) (Von Cossel, Wagner, et al., 2019). These are two of the most promising industrial crops for Europe (Alexopoulou, 2018) which do not compete for land with food crops (Rulli et al., 2016). They also share a perennial life cycle, making them superior to annual (bioenergy) crops in many respects: They require less soil tillage, thereby reducing soil erosion (Von Cossel, Wagner, et al., 2019), and have a higher nutrient use efficiency due to their extensive root systems (Fernando et al., 2018;Himken et al., 1997;McCalmont et al., 2017). Where perennials have been established, a faster build-up of carbon in soil compared to annual crops can be observed, due to decreased soil disturbances (Chimento et al., 2016). GR, a warm-season C3 plant, presently grown in Southern Europe, is drought tolerant and therefore suited to more arid regions, while RCG, a coolseason C3 plant, grows in the Northern and Central realms of Europe with short growing seasons and withstands excess soil moisture (Cosentino et al., 2014;Lewandowski et al., 2003;. This allows for an analysis of marginal land's suitability for two crops with inherently different climatic requirements. Other promising perennial crops such as Miscanthus (Miscanthus Andersson) (Winkler et al., 2020) or Sida (Sida hermaphrodita L. var. Rusby) (Jablonowski et al., 2020;Kitczak et al., 2022) were not selected here because their climatic requirements are intermediate between those of GR and RCG. Thus, the selection of GR and RCG is intended to highlight the variance in the degree of the effect of climatic changes on the potential growing areas of perennial crops by showing the north and south boundaries. This is not to say, however, that a close look at Miscanthus, Sida, and other perennial crops would not be useful in the future, especially considering the steady progress in breeding, such as for Miscanthus (Impollonia et al., 2022). Climate change will cause shifts in crop suitability distributions on European marginal lands due to projected rising annual mean temperatures and changing precipitation patterns (Ceglar et al., 2019;Olesen et al., 2011;Peltonen-Sainio et al., 2018). In this regard, the objectives of the present study are to identify changes in (i) marginal land area, (ii) the potential distributions of the two selected bioenergy crops (GR and RCG), and (iii) the overlap between the bioenergy crops and marginal land in Central and Southern Europe under a changing climate.

| Marginal land
Marginal land was defined in terms of the land's unsuitability for growing edible crops. The FAO and CGIAR (2000) definition of marginal land was used in the present study, which describes marginal lands as "having limitations which in aggregate are severe for sustained application of a given use" (here: cultivation of edible crops). The limitations are regional climate or local environmental conditions such as soils and slopes. Subject of this study are the climate induced marginal lands, taking into account temperature and water availability. When marginality thresholds are exceeded, the climatic conditions present a notable "handicap to agriculture, however without making agriculture impossible" (Van Orshoven et al., 2014).

| Climatological data
Pictured in Figure 1 is the orography of the study area. Daily temperature and precipitation time series are required for the analysis (Sections 2.4-2.5). The gridded observational dataset E-OBS v32.1e (Cornes et al., 2018) was used at a resolution of 0.1° for the present period 1991-2020. The data cover the land area. Some regions had to be excluded from the study domain, namely Sicily and Southern Greece, due to low station densities causing missing values in the time series, see areas shaded in gray in Figure 1.
The simulated data for the future period 2071-2100 were taken from the model ensemble of the EURO-CORDEX initiative (Jacob et al., 2020), which provides regional climate projections for Europe at a resolution of 0.11°. We consider 16 EURO-CORDEX simulations (Table 1) which follow RCP8.5, a representative concentration pathway (RCP) that will result in radiative forcing of 8.5 W/m 2 by the end of the 21st century. Selection of the ensemble members was based on the German Meteorological Service (DWD) reference ensemble, with HadGEM simulations being excluded because of their use of the 360-day calendar. The simulations were obtained via the Earth System Grid Federation and the German Climate Computing Center (https://esgf-data.dkrz.de/).

| Bias adjustment
Using the raw climate model output would lead to an inaccurate representation of threshold exceedance (nonexceedance) timing under the climate change scenario (Hempel et al., 2013). Therefore, the projected data were adjusted toward the observed climatology to reduce bias and render an accurate analysis of climate change impacts on energy crop and marginal land distribution possible (Gohar et al., 2017).
The R package downscaleR v3.3.3 (Bedia et al., 2020) from the open source climate4R framework was used to carry out the statistical bias adjustment of the daily variables.
For temperature, the mean and variance of the simulated climate were adjusted toward the corresponding observational mean and variance using the mean and variance adjustment method described in Leung et al. (1999), Räisänen and Räty (2013), Torralba et al. (2017) and Manzanas et al. (2019). This results in the simulated and observed climates having the same mean and standard F I G U R E 1 Orography of the study area. Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset. T A B L E 1 RCM-GCM "matrix," that is, the simulations analyzed in this study.

RCM-GCM
deviation for each grid cell (Leung et al., 1999;Torralba et al., 2017). Precipitation was bias adjusted with local intensity scaling which adjusts the biases in wet-day frequency and wet-day intensity of the simulated climate toward the observed one, thereby covering the major existing precipitation biases in regional climate models. The method applied in this study follows Schmidli et al. (2006) and is also outlined in Themeßl et al. (2011) andWörner et al. (2019).

| Agro-climatic indices
Growing degree days (GDD), growing season length (GSL), and annual precipitation sums (PA) were used to identify suitable land for the cultivation of GR and RCG. The corresponding growth thresholds (Table 2) were taken from Von  and only consider whether the crops could potentially grow, neglecting different yield levels.
For the classification of marginal land, previous work by van Orshoven et al. (2014) was consulted. They used GDD and GSL to determine land surface marginality due to low-temperature constraints, while an aridity index (AI) was used to identify marginal land due to dryness constraints. The marginality thresholds are shown in Table 3. 2.4.1 | Growing degree days GDD describe the amount of effective heat units crops receive in a given time.
In the present study, no upper thresholds were used in the calculations for GDD because they were not readily available from the literature. Russelle et al. (1984) have found that maximum temperatures are essential for crop development, so including them in the calculations might have improved results. The calculations for this index were based on the equation given by the European Climate Assessment & Dataset (ECA&D) Project Team and used by Spinoni et al. (2014) and Wypych et al. (2017) in the following way: where GDD = growing degree days (°C d), T m = daily mean air temperature (°C), T b = threshold base temperature (°C) (Tables 2 and 3), n = number of days over which thermal units are accumulated, i.e., days of the year.

| Growing season length
The growing season can be defined as the period of the year when temperatures allow for plant growth to potentially take place (Carter, 1998;Linderholm et al., 2008).
The start, end, and length of the growing season can be defined in different ways, depending on the region of interest and its climate. The present study looks at the GSL in Central and Southern Europe to determine the marginality of Europe's land surface and its suitability for the selected crops. Therefore, an approach based on the ECA&D Project Team's definition of the GSL was used (Project Team ECA&D, 2013), as outlined in the following: • GSL is defined as the number of days between the first occurrence of at least six consecutive days with T ij > 5°C (1) GDD i T A B L E 2 Giant reed's and reed canary grass' minimum agro-climatic requirements (adopted from Von . • and the first occurrence of at least six consecutive days with T ij < 5°C within the last 6 months of the year with T ij = mean temperature on day of period in °C.

| Annual precipitation sum
Since the crops in this study are assumed to be rainfed, precipitation plays a vital role in determining an area's suitability for their cultivation. Because the perennial crops stay on the field year-round, PA is calculated instead of seasonal precipitation sums. For marginal land, a cropindependent index for dryness was used (Section 2.4.4).

| Aridity index
After Pereira et al. (2009), aridity is defined as "a natural permanent imbalance in the water availability consisting in low average annual precipitation, with high spatial and temporal variability, resulting in overall low moisture and low carrying capacity of the ecosystems." Arid regions can therefore be defined as marginal for crop cultivation. However, dryness does not only depend on annual precipitation. Temperature also plays a role in determining the degree of aridity as it influences the rate of evaporation. The AI is therefore calculated by dividing the annual mean precipitation by the annual mean potential evapotranspiration PET ′ a , which is adjusted for variable durations of daylight for different latitudes and length of the months during the course of a year. The method follows the World Atlas of Desertification by the United Nations Environment Programme (UNEP, 1997): For calculating monthly unadjusted potential evapotranspiration (PET, in mm), a modified version of Thornthwaite's method (Thornthwaite, 1948;Van Der Schrier et al., 2011;Willmott et al., 1985) was used following UNEP's World Atlas of Desertification (UNEP, 1997). Using the more sophisticated Penman's method (Penman, 1948) instead would require meteorological measurements (e.g., wind velocity) that are not readily available from the E-OBS dataset for the study area.
where is the monthly mean temperature in °C, and Finally, the results for PET were adjusted in the following way: where ℎ is the duration of daylight in hours and is the length of the month in days. To arrive at PET ′ a , the 12 monthly PET′ values are added up for each year.

| Crop suitability and marginality maps
Following the methods and equations given in the previous sections, time series for the agro-climatic indices GDD, GSL, PA, and AI were calculated from the biasadjusted daily data for temperature and precipitation. Subsequently, the 30-year mean values were determined for the present period (meanE-OBS) and each future simulation included in Table 1. The mean values from each simulation were averaged to give an ensemble mean (ensRCP8.5) for each index. meanE-OBS and ensRCP8.5 were then analyzed separately for each crop and marginal land by comparing them against GR and RCG suitability and marginality thresholds (Tables 2 and 3).
The production of crop suitability maps follows Tuck et al. (2006). Hereby a grid cell on the map is classified as suitable (i.e., coded as 1), if the 30-year mean of an index lies at or above the corresponding threshold for a crop. It follows that for GDD < 1843°C d, no suitability for GR is identified and the grid cell is coded as 0. Once the masks had been calculated for each index, they were overlayed on a map of the study area for each crop and both time periods. Only areas where all three indices are equal to 1 for a crop were classified as suitable for its cultivation. Consequently, areas where at least one index does not meet the threshold (i.e., was coded as 0) are identified as unsuitable.
The marginality of land surfaces was determined using a probabilistic approach following van Orshoven et al. (2014). It employs annual time series of the indices instead of the climatological means. The publication suggests using a probability of exceedance/non-exceedance of 80%/20% for each index. Hereby land is classified as marginal if the threshold of an index is not surpassed in at least 7 of 30 years. This method accounts for interannual variability of the climate. For GDD, this means that grid cells where the annual GDD sum is equal to or under 1500°C d are coded as 1 to designate marginal conditions, while the remaining grid cells are coded as 0. The time series is then summed up over 30 years. Now, grid cells that show values of 7 or higher (i.e., if the threshold is not surpassed in at least 7 of 30 years) are again coded as 1 to indicate marginal conditions due to severe low temperature. Thus results a mask for marginal land based on exceedance or non-exceedance of the GDD threshold for marginality ( Table 3). The same method is repeated for the remaining indices.
The three marginality masks (one for each index in Table 3) were then overlaid on a map for the observational period and each of the simulations. For a region to be classified as marginal overall, it suffices for one of the three indices to be less than or equal to the marginality threshold (i.e., to be coded as 1). In this case, the area is again coded as 1 to designate marginal land, while the remaining area is coded as 0 to designate land that is favorable for the cultivation of food crops. This is in accordance with van Orshoven et al. (2014), who combined the diagnostic criteria for identifying marginal land according to the law of the minimum (i.e., Liebig's law). In this way, overall marginality masks are produced from the overlaid indices. In a last step, en-sRCP8.5 was calculated from the individual simulations for the future scenario.

| Changes in temperature and precipitation
The agro-climatic indices GDD, GSL, and AI depend on temperature, while AI and PA also depend on precipitation. Under RCP8.5, annual mean temperatures and annual precipitation sums over Europe are expected to change drastically (Jacob et al., 2014).
The present study finds increases in annual precipitation in Western Europe and the British Isles by up to 10% ( Figure 2) and in the Baltic States by up to 15%. A band of minor to no change in precipitation spans over Southern France, the Balkan peninsula, and Italy. The models predict a decrease in annual precipitation in the southernmost regions of the Mediterranean (as far as missing values allow for an observation to be made). Spain's Mediterranean coastline, in particular, will see a decrease in precipitation by over −15%.
Results indicate an increase in annual mean temperature by over 4°C in the High Alps ( Figure 3). The British Isles are expected to exhibit the smallest changes in temperature while Southern and Eastern Europe will experience greater warming. Under the RCP8.5 scenario, marginal land area is expected to shrink in the northern regions of the study area and the British Isles (Figure 4b). The mountainous regions of Central and Eastern Europe will also see a decline in marginal land, except in the High Alps, where it persists. Especially noticeable is the predicted increase in marginal land in Spain, where it is expected to cover more than half of the country's land area by the end of the century. In Romania, the marginal land area will span the country's entire Black Sea coast and reach into the inlands, where marginal conditions are bounded by the Carpathians. The Apulia region of Italy (the "spur" and "heel" of Italy's boot), the southern tip of Sardinia, and large parts of mainland Greece are also expected to exhibit marginal conditions under RCP8.5. F I G U R E 2 Average change in annual mean precipitation under ensRCP8.5 for 2071-2100 as compared to meanE-OBS for 1991-2020. Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.

| Distribution of marginal land
The distribution of marginal land is based on lowtemperature and dryness constraints, that is, GDD, GSL, and/or AI, which are shown in the following.
The marginality threshold for GDD lies at 1500°C d (Table 3). For the observational period 1991-2020, GDD values range from less than 1000°C d in the northernmost parts of the study region to 5000°C d in Southern Spain (Figure 5a). F I G U R E 3 Average change in annual mean temperature under ensRCP8.5 for 2071-2100 as compared to meanE-OBS for 1991-2020. Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.  A latitudinal gradient is visible, meaning GDD increases from north to south. The main European mountain ranges-the Alps, the Carpathian Mountains, and the Pyrenees-show a decrease in GDD with increasing altitude. Presently, marginal land due to limiting GDD (i.e., areas with values <1500°C d) is located in Europe's mountainous regions like the Alps and the Carpathians, as well as the northern regions of the study area, that is, Scotland, Northern Ireland, and Estonia.
Under RCP8.5 (Figure 5b), the latitudinal gradient of GDD for marginal land is intensified. Values will range from less than 1000°C d in the northwestern study area to a maximum of around 6000°C d in the south of Spain and Eastern Sicily. The altitudinal gradient will still be pronounced but is going to shift toward higher values of GDD. Coppola et al. (2021) also projected increasing GDD values across Europe under RCP8.5, using a larger ensemble of EURO-CORDEX simulations. This means that marginal land in terms of GDD will decrease by the end of the century.
GSL follows a continental gradient (Figure 6a), along which GSL values decrease from oceanic to more continental areas. This manifests itself in a West-East gradient (toward greater continentality). Furthermore, the data show an altitudinal gradient, with values for GSL decreasing with increasing altitude, for example, as evident in the Alpine region. Marginal conditions due to limiting GSL (i.e., areas with GSL values <180 days, Table 3) can presently be found mainly in the Alps, the Carpathians and southern regions of Norway and Sweden, as well as the Baltic States.
In the future scenario ( Figure 6b) the continental gradient for GSL is expected to remain. There will be a shift toward higher values for GSL in regions where its values had not already amounted to the maximum of >360 days in the observational period ( Figure 6a). The altitudinal gradient will also persist and shift toward higher values for GSL. By the end of the century, marginal land due to under-exceedance of the GSL threshold will decrease. It will persist only in the High Alps and Southern Norway.
Most of Europe exhibits an AI of over 0.65 during the time period 1991-2020 (Figure 7a), except for Spain, where values reach a minimum of 0.2-0.5 at the coastlines bordering the Mediterranean Sea and in the inland plateau region. The plains of Hungary (0.5 < AI ≤ 0.65) and the Balkan region's Black Sea coast (0.5 < AI ≤ 0.65 and 0.2 < AI ≤ 0.5) also exhibit a low AI.
By the end of the century (Figure 7b), gains in arid areas and intensification of aridity (i.e., a decrease in the AI) are expected mainly in areas where the AI is already below 0.65 in the reference period (Figure 7a). Gains in new areas with an AI ≤ 0.65 will be located at the coast of Southern France and the Po Plain in Northern Italy.
Presently, most Western Europe, the British Isles, and Eastern Europe show values for PET below 750 mm/a (Figure 8a). Europe's mountainous regions can be easily distinguished on the map because of their low PET values as compared to the surrounding area. Mediterranean Europe exhibits PET values exceeding 800 mm/a in many places, especially in Southern Spain. Overall, values range from around 350 mm/a in the High Alps to a maximum of 1020 mm/a in Southern Spain.
Under RCP8.5, values for PET are projected to increase across the entire study area (Figure 8b). PET values will range from 409 mm/a in the High Alps to a maximum of 1230 mm/a in Southern Spain. F I G U R E 6 Mean annual growing season length for (a) 1991-2020 (meanE-OBS) and (b) 2071-2100 (ensRCP8.5). Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.

| Distribution of GR
Presently, GR can potentially be grown in large parts of Europe ( Figure 9a). Unsuitable regions are located namely in the Alps and the lower mountain ranges of continental Europe. The Mediterranean coast of Spain and the country's inner plateau also exhibit an unfavorable climate for GR. On mainland Europe, conditions for GR become unsuitable around the 55° N latitude.
Under the RCP8.5 scenario (Figure 9b), an overall increase in the area suitable for GR is projected as compared to the present period (Figure 9a). This increase manifests itself especially in the northern parts of the study area and Central and Eastern Europe. In the Alps, the suitable area will advance into higher regions while the lower mountain ranges in Europe, like the Pyrenees, the Black Forest, or the Carpathians, become almost entirely suitable for GR. The scenario also predicts a considerable increase in the British Isles.
Underlying GR's potential distribution in Figure 9 are the agro-climatic indices PA, GDD, and GSL. The crop requires a minimum PA of 400 mm (Table 2). In the present, values for precipitation range from <200 on Spain's Mediterranean coastline to >2100 mm/a in the Alps and other mountain ranges as well as the western coasts of the British Isles (Figure 10a). Accordingly, GR's requirements are fulfilled in most of Europe, except Southern and inland Spain, and Ukraine's Black Sea coast.
Under RCP8.5, areas with less than 400 mm a year will remain essentially unchanged (Figure 10b). A further decrease in PA can be made out in Southern Spain.
Presently, the crop cannot be grown above 55° N in terms of GDD ( Figure 5), as it requires a minimum of 1843°C d ( Table 2). The Alps, the Carpathians, and Europe's lower mountain ranges in Germany, Spain and the Baltic Peninsula are also unsuitable. The figure shows that the potentially suitable area for GR in terms of GDD will increase by the end of the century under RCP8.5 compared to the observational period to cover most of the study area except Southern Norway, Scotland's Grampian Mountains, and the higher altitudes of the Alps, where GDD remains limiting.
According to Figure 6, GR, needing at least 210 days GSL (Table 2), can currently potentially be grown anywhere in Europe in terms of GSL, except for mountainous regions and the Baltic states. Under RCP8.5, GR's potential F I G U R E 7 Mean annual aridity index for (a) 1991-2020 (meanE-OBS) and (b) 2071-2100 (ensRCP8.5). Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.
distribution is expected to increase, with only the highest altitudes of the Alps and the Carpathians remaining unsuitable because of a limiting GSL.

| Distribution of RCG
Most of Eastern and Southern Europe, as well as the Balkan Peninsula show limited suitability for RCG's cultivation under present-day climatic conditions (Figure 11a). The East of England, the Alps, Italy's Apulia, and Germany's Brandenburg region also stand out as unsuitable areas for RCG.
In the northern regions of the study area, the RCP8.5 scenario projects an increase in suitable area for RCG, with the East of England and the southernmost regions of Norway and Sweden becoming potential growing areas ( Figure 11b). The High Alps will become almost entirely suitable for RCG. The suitable area also advances into Eastern Europe. Apart from these increases, however, "blind spots" where conditions remain unsuitable for RCG are projected to persist in Poland, parts of the Czech Republic, and large parts of Hungary. In Germany, the Palatinate and parts of Brandenburg and Saxony-Anhalt are to remain unsuitable climatologically. Southern Europe is also expected to remain largely unsuitable, except for Northern and North-Western Italy.
The underlying indices for RCG's potential distributions in the present and future time periods are analyzed in the following.
The crop needs at least 2000°C d (Table 2) in annual GDD. Presently, RCG-specific GDD values range from F I G U R E 8 Mean annual potential evapotranspiration for (a) 1991-2020 (meanE-OBS) and (b) 2071-2100 (ensRCP8.5). Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.

F I G U R E 9
Mean potential distribution of giant reed based on the crop's minimum climatic requirements; (a) 1991-2020 (meanE-OBS), (b) 2071-2100 (ensRCP8.5). Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.
<1000°C d in the High Alps to >6500°C d in Southern Spain (Figure 12). With 0°C RCG has a lower T b than GR for calculation of GDD, but like for GR, the values follow a latitudinal gradient and decrease with latitude ( Figures 5 and 12). Presently, RCG can be grown almost anywhere in the study area in terms of GDD, except the High Alps and Southern Norway, where GDD restrict the crop's potential distribution. By the end of the century (Figure 12b), values for GDD will increase, ranging from around 1500°C d in the High Alps to around 8000°C d in Southern Spain and Sicily. Consequently, the potential distribution of RCG is expected to increase, and most of the study area will become suitable (with the exception of the highest peaks of the Alps) for the crop in terms of its GDD requirements.
RCG requires a minimum of 111 days GSL (Table 2). GSL restricts RCG's potential distribution only in the High Alps in the reference period ( Figure 6). Under RCP8.5, GSL will no longer limit RCG's growth anywhere in Europe.
At 600 mm/a, RCG's PA requirements are higher than those of GR (Table 2). For the observational period, most of Spain, Southern Portugal, and Italy's Apulia region are unsuitable for RCG due to too little precipitation (Figure 10a). In Central Europe, too dry conditions limit RCG's growth in the Palatinate and Brandenburg regions of Germany and large parts of the Czech Republic as well as Poland. For Eastern Europe, unsuitable areas are mainly located in Hungary, Romania, and along the Black Sea coast. Eastern England also exhibits too little rainfall for RCG. Under RCP8.5 (Figure 10b), the Mediterranean regions' unsuitability for RCG remains unchanged. In Central and Eastern Europe, however, more land will become suitable for the crop's cultivation. Suitable land is expected to advance in Brandenburg, Poland, Hungary, and Romania, while Eastern England will become fully suitable for RCG in terms of PA.

| Summary of results
In the following, the absolute surface areas and relative changes in ensRCP85 compared to meanE-OBS are summarized (Tables 4 and 5), before the spatial distribution of overlapping areas is examined in Sections 3.5.1 and 3.5.2.

| Overlap between GR and marginal land
Presently, regions where marginality and suitability for GR overlap are sparse. They are located in Southern Spain and Greece (Figure 13a).
Under RCP8.5 (Figure 13b), there will be a gain in areas both marginal and suitable for GR in Southern Spain and the Algarve region of Portugal. The southerly lowlands of Bulgaria, as well as the coastal regions of Romania, will also show an overlap. Other regions of overlap include the Apulia region in Italy, the southern tip of Sardinia, the Macedonia region in Greece, and the central valley of the river Vardar in North Macedonia.

| Overlap between RCG and marginal land
Areas of overlap in the present (Figure 13a) are located mainly in the northern regions of the study area, that is, in Scotland, the Baltic states, and the southernmost tips of the Scandinavian Peninsula. Regions with marginal climate and suitability for RCG can also be found in the Alps, the Carpathians, and the mountains of Slovakia and the Czech Republic. In Southern Europe, areas of overlap are scarce and can only be found in mountain ranges, like the Pyrenees and some of the higher peaks of the Balkan region.
By the end of the century, regions of overlap between marginal land and RCG are expected to decrease remarkably under RCP8.5 (Figure 13b). They will persist in the High Alps, the Grampian Mountains of Scotland, and Southern Norway.

| Distribution of marginal land
The increase in aridity is mainly responsible for the projected gains in marginal land area (Figures 4 and 7) throughout Southern Europe by the end of the century. The AI is calculated from the quotient of PA and annual PET. While negative changes in PA (Figure 2) can only account for a small part of this increase (e.g., along Spain's Mediterranean Sea coast), most of the gains in marginal land in this area can be ascribed to an increase in annual mean temperature under RCP8.5 (Figure 3). In inland Spain and Northern Greece, the temperature increases considerably (>4°C). This temperature rise, in turn, causes annual PET values to increase (Figure 8), which are dependent on monthly mean temperatures. The higher the values for PET (i.e., the denominator for calculation of AI), the smaller is the resulting AI, which designates an increase in aridity. Greve and Seneviratne (2015) also projected an increase in aridity in Southern Europe under RCP8.5 for the time period 2080-2100.
Low-temperature constraints (i.e., GDD and GSL) cannot be held responsible for increasing marginality in Southern Europe because GDD and GSL increase with increasing annual mean temperature. GDD and GSL are already above the marginality threshold under present climatic conditions (Figures 5 and 6) and, since annual mean temperatures are increasing under RCP8.5 (Figure 3), they will remain so in the future. This is because no upper temperature threshold was implemented for calculations of GDD, which might have led to an even more substantial projected increase in marginal land for this region. Values In terms of meanE-OBS total marginal land area (km 2 ). c In terms of the corresponding meanE-OBS mean surface area (km 2 ) for each row.
for GSL are already above 350 days under present climatic conditions in Southern Europe, so no further major increase can be seen under RCP8.5, as GSL cannot exceed 365 days (366 days for leap years).
The projected decrease in marginal land (Figure 4) on the British Isles and throughout the Baltic region, as well as at the higher altitudes of the study area, can be attributed to decreasing low-temperature constraints (i.e., non-limiting GDD and GSL, Figures 5 and 6), as none of these regions exhibit constraints due to dryness (i.e., arid conditions) at present (Figure 7). This means that marginal land will shrink in the aforementioned areas due to increasing annual mean temperatures under global warming following RCP8.5 (Figure 3). Because aridity plays no role in the present, the projected positive changes in precipitation will not affect marginality there.
Overall, marginal land due to climatic constraints is expected to decrease by −17.43% under RCP8.5 (Table 4). Unlike Von , uncultivatable land (e.g., woodlands and urban areas) was not excluded from the mapping of marginal land. This means that marginal land, rather than marginal agricultural land, is mapped and may result in marginal land suitable for the cropping of the selected bioenergy crops being overestimated.

| Distribution of GR
The potential distribution of GR is projected to increase throughout most of Europe (Figure 9). The positive change in the northern parts of the study area, the Baltic states, the British Isles, and the mountainous regions can be attributed to the rise in annual mean temperature (Figure 3), which removes the present-day lowtemperature constraints, that is, GDD and GSL, for GR in these regions (Figures 5 and 6). Changes in precipitation play no role in the increase in suitable land for GR in these areas since annual precipitation sums are not a limiting factor at present and are not expected to become limiting under RCP8.5 ( Figure 10).
There will be virtually no change in suitable land area for GR in Southern Europe, as far as missing values allow for observations to be made. Areas in Southern and inland Spain that are currently not suitable for GR will remain so in the future under RCP8.5 and vice versa (Figure 9). F I G U R E 1 3 Mean overlap between marginal land, giant reed, and reed canary grass for (a) 1991-2020 (meanE-OBS) and (b) 2071-2100 (ensRCP8.5). Areas shaded in gray indicate regions that were excluded from the analysis due to high shares of missing values in the E-OBS dataset.
The projected increase in annual mean temperature under global warming following RCP8.5 will not affect GR's potential distribution in these areas, as the crop's requirements for GDD and GSL already exceed the limiting threshold value under present-day climatic conditions. As a result of decreasing annual precipitation sums in this region (Figure 2), current limiting rainfall conditions will persist in the future.
Overall, suitable land area for GR is expected to increase by +24.16% by the end of the century under RCP8.5 (Table 4). While this sounds promising, it needs to be examined in the following whether this area overlaps with marginal land to prevent competition with food production.

| Distribution of RCG
The observational data and the simulations indicate the crop's preference for cooler climates and its relatively high precipitation requirements, as Southern and Southeastern Europe show unsuitability for the cool-season grass ( Figure 11). The meanE-OBS  findings are in keeping with Tuck et al. (2006), who displayed the climate suitability map for RCG based on observed data from 1961 to 1990. Only the suitability status for Spain, Hungary, and Romania differs somewhat from the present study. The slightly elevated share of suitable area in these regions in the study conducted by Tuck et al. (2006) could be ascribed to the earlier observational period that was used when global warming was not yet as advanced as in the observational period of the present study, and the use of the old emissions scenarios from the Special Report on Emissions Scenarios (IPCC, 2000).
The projected increase in suitable land for RCG in Eastern England, Central, and Eastern Europe, as well as the Alps and the northern regions of the study area, also corresponds with Tuck et al. (2006), who reported similar findings between 45 and 60° N by the 2080s. The present study shows that these changes are due to an increase in annual precipitation sums in these areas (Figure 2). The rise in annual mean temperatures (Figure 3) will not lead to a notable increase in the crop's distribution. Its unsuitability for Southern Europe will persist and only a slight increase in the northern parts and the mountainous regions of the study domain can be noted in Figure 11 due to decreasing low-temperature constraints.
Most of Southern Europe presently falls below RCG's minimum precipitation threshold. Because no notable positive change in annual precipitation sums is expected for Southern Europe (Figure 2), the area will remain unsuitable for RCG under RCP8.5. In contrast, Tuck et al. (2006) projected a 16%-30% decrease in suitable land area for RCG between 35 and 44° N by the 2080s relative to 1990.
Overall, the potential distribution of RCG is expected to increase by 13.30% by the end of the century because of global warming following RCP8.5 (Table 4).

| Overlap between marginal
land and the selected bioenergy crops  found that presently, 21% and 53% of marginal land across Europe (EU-28) is suitable for the cultivation of GR and RCG, respectively. At 8% overlap between GR and marginal land, the present study found a lower value for the observational period ( Table 5). The reason for this may be that while Von  studied the whole EU-28, the present study excludes some of Europe's Mediterranean region, which holds all of the overlapping area between marginal land and GR at present (Figure 13a), from the study area. For RCG, however, areas of overlap are presently estimated at 55% (Table 5), therefore only differing by 2 percentage points from the findings by Von .
By the end of the century, areas of overlap between GR and marginal land will almost quintuple in size under RCP8.5 (Table 5). In contrast, areas of overlap between RCG and marginal land will see a decrease of almost 90% (Table 5). For GR, this increase is due to marginal land gaining in area in Southern and Southeastern Europe (Figure 4), regions of which the crop is well suited to climatically ( Figure 9). Since RCG considered individually will overall gain in suitable area under the climate change scenario (Figure 11), the negative change in regions of overlap can be attributed to the decline of marginal land in the northern parts of the study domain ( Figure 4). The crop's suitability will remain virtually unchanged in the Southern regions of Europe (Figure 11), whereas marginal land will gain in area there (Figure 4). While this benefits GR, RCG will not contribute to overlapping land area below 45° N ( Figure 13).
It is evident that the majority of land available for growing the selected crops is located outside marginal lands for both the observational and the future time periods (Tables 4 and 5). As the demand for bioenergy crops will continue to grow in the future due to the switch to a low-carbon economy (Popp et al., 2021), this could indicate increasing competition between areas suitable for bioenergy crops and other land uses, like food production, protected areas, and woodlands (Cronin et al., 2020). In addition, the marginal area available for the cultivation of GR and RCG may be overestimated because woodlands, urban, and mountain areas were not filtered out. In these regions, establishing bioenergy cropping systems is not possible because of the steepness of the slopes or because the land is used for other purposes. Forests, for instance, should not be converted for use as bioenergy cropping systems since this would entail more greenhouse gas emissions than the subsequent carbon sequestration by GR in the soil could make up for (Ale et al., 2019).
It is also noticeable that Spain's marginal inland plateau shows large areas of no overlap for both crops and time slices (Figure 13). Neither of the selected crops is suited to the region's climate because it is too arid (Figure 7). However, this could open up the potential for the cultivation of other, more drought tolerant bioenergy crops. Tuck et al. (2006) found good suitability at present and a potential for increased suitability by the end of the century for cardoon (Cynara cardunculus L.), sorghum (Sorghum bicolor), and prickly pear (Opuntia fiscus-indic) between 35 and 44° N. Meanwhile, Von  mentioned pencil tree (Euphorbia thirucalli L.), prickly pear, and agave (Agave tequilana F.A.C. Weber), all dependent on the crassulacean acid metabolism photosynthetic pathway, as suitable for cultivation on drought affected sites. Whether the aforementioned crops' potential distribution areas will actually overlap with future marginal lands has yet to be examined. At least for cardoon  found good suitability in the present for cultivation on Mediterranean marginal agricultural lands, specifying an overlap of ~170,000 km 2 .

| CONCLUSIONS AND OUTLOOK
In summary, it can be concluded that European marginal land will decrease by nearly 20% by the end of the century under RCP8.5, while the potential distributions of GR and RCG are expected to increase by ~25 and ~15%, respectively. Although marginal land will shrink due to decreasing low-temperature constraints under global warming, the rising annual mean temperatures will benefit the selected crops. Moreover, the overlapping area between marginal land and RCG will decrease, while it will increase for GR and marginal land. In the case of RCG, this is because rising annual mean temperatures have opposite effects on the studied entities, decreasing marginal land and shifting it southward, and benefiting the bioenergy crop in the northern parts of the study domain. Consequently, for the cold-season grass RCG, a decrease of about ~90% in its overlapping area with marginal land is projected by the end of the simulated time period. This is due to a discrepancy in the agro-climatic thresholds, with most marginal land being found in Mediterranean Europe and RCG preferring cooler climates. For the warm-season grass GR, this effect is reversed. The crop will experience a notable increase in overlap with marginal land of over ~450%, as it is able to thrive in warmer and drier climates and can thus take advantage of some of the marginal land in Southern Europe.
The overlapping area between marginal land and either of the two selected bioenergy crops (Table 5) would not suffice to meet the required land area for compliance with the IPCC's mitigation pathways as specified in Section 1. This hints at the enormous land surface required for landbased mitigation measures like BECSS and highlights the importance of using marginal land so as not to threaten food security further.
The climate-based suitability maps for GR and RCG merely show areas of potential suitability, without taking into account their productivity and yield quality. Using crop models or field experiments, the profitability of growing GR or RCG on potentially suitable marginal sites could be elaborated upon (Cappelli et al., 2021;Meehan et al., 2017;Nassi o Di Nasso et al., 2013). It could also be useful to map the influence of different input intensities (e.g., irrigation or nitrogen fertilization) on site-specific yields and quality performance of GR and RCG in future studies, as was done by Ramirez-Almeyda et al. (2017).
Unlike  and Elbersen et al. (2020), the mapping of marginal land was not limited to arable areas only. This means that marginal land area in the present study contains steep slopes across Europe's mountain ranges, urban or environmentally protected areas, among other land surfaces which, even though they may be subject to marginal climatic conditions, are not suitable for agricultural use and the cultivation of bioenergy crops. While the present results serve as a starting point for verifying Europe's potential for sustainably growing bioenergy crops on marginal land, the results could be elaborated upon in the future by using the CORINE Land Cover inventory to mask the data for agricultural areas only.
The methodology outlined in the present study could be applied to examine supplementary promising bioenergy crops and their suitability for growth on marginal land under climate change. The method could also be transferred to identify marginal land and its suitability for bioenergy crops on other continents, although in climates outside the temperate zone, marginality thresholds may need to be adjusted to reflect the thresholds of the most common food crops in those regions. For example, in Canada, Liu et al. (2017) have analyzed marginal land's economic potential for bioenergy crops for the present. There is a noticeable gap in research regarding the identification of marginal land under scenarios of climate change and mapping its suitability for bioenergy crops.
The framework provided by the present paper can be used to examine European marginal land's suitability for other promising bioenergy crops, such as miscanthus, sida, sorghum, or camelina (Camelina sativa L.) under future climatic conditions. Tapping Europe's full potential in this regard can further foster its growing bioeconomy, while simultaneously reducing land competition and contributing to the achievement of climate targets.