How much heat can we grow in our cities? Modelling UK urban biofuel production potential

Biofuel provides a globally significant opportunity to reduce fossil fuel dependence; however, its sustainability can only be meaningfully explored for individual cases. It depends on multiple considerations including: life cycle greenhouse gas emissions, air quality impacts, food versus fuel trade‐offs, biodiversity impacts of land use change and socio‐economic impacts of energy transitions. One solution that may address many of these issues is local production of biofuel on non‐agricultural land. Urban areas drive global change, for example, they are responsible for 70% of global energy use, but are largely ignored in their resource production potential; however, underused urban greenspaces could be utilized for biofuel production near the point of consumption. This could avoid food versus fuel land conflicts in agricultural land and long‐distance transport costs, provide ecosystem service benefits to urban dwellers and increase the sustainability and resilience of cities and towns. Here, we use a Geographic Information System to identify urban greenspaces suitable for biofuel production, using exclusion criteria, in 10 UK cities. We then model production potential of three different biofuels: Miscanthus grass, short rotation coppice (SRC) willow and SRC poplar, within the greenspaces identified and extrapolate up to a UK‐scale. We demonstrate that approximately 10% of urban greenspace (3% of built‐up land) is potentially suitable for biofuel production. We estimate the potential of this to meet energy demand through heat generation, electricity and combined heat and power (CHP) operations. Our findings show that, if fully utilized, urban biofuel production could meet nearly a fifth of demand for biomass in CHP systems in the United Kingdom's climate compatible energy scenarios by 2030, with potentially similar implications for other comparable countries and regions.

One approach that ameliorates many of these concerns is to produce biofuel within urban areas, close to where it is to be used. Use of urban land in this way potentially presents a 'win-win-win' scenario by reducing food versus fuel conflicts on high quality agricultural land (Raman et al., 2015), providing a source of fuel close to centres of demand for energy (Renewable Fuels Agency, 2008), and increasing the sustainability and resilience of cities and towns (Norton, Evans, & Warren, 2016). The latter is increasingly important as the global urban population is growing, currently 55% and reaching 68% by 2050 (United Nations, 2018). Consequently, urban areas are drivers of global change and mechanisms to improve their sustainability and resilience are vital, as recognized in UN Sustainable Development Goal 11: Sustainable Cities and Communities. At present, cities account for 37%-49% of global GHG emissions and urban infrastructure for 70% of global energy use (Kekana & ISOCARP, 2019).
Urban biofuel cultivation that supplies local combined heat and power (CHP) systems, if carefully managed to minimize life cycle GHG emissions (comprised of both management/harvesting operations and fuel transport costs between source and destination locations; Bauen et al., 2010;Fajardy & Mac Dowell, 2017;Yang et al., 2017), could be used to reduce city-wide CO 2 emissions. Use of urban greenspace for biofuel production could also deliver additional ecosystem services to urban residents (McHugh, Edmondson, Gaston, Leake, & O'Sullivan, 2015) such as flood protection (O'Sullivan, Holt, Warren, & Evans, 2017), pollution mitigation (Sugiura, Tyrrel, Seymour, & Burgess, 2008) and soil carbon sequestration (Cunniff et al., 2015;Matthews, Grogan, & Matthews, 2002). Although representing poorer and more homogeneous habitat than fully natural or seminatural habitats, biofuel plantations are also likely to support greater floral, invertebrate and avian biodiversity than arable cropland or mown grassland (Rowe, Street, & Taylor, 2009;Sage, Cunningham, & Boatman, 2006).
Urban greenspaces are recognized to provide vital ecosystem services (Grafius, Corstanje, Siriwardena, Plummer, & Harris, 2017;McDonnell & MacGregor-Fors, 2016); however, their ability to support provisioning services is poorly understood and research thus far has focussed on urban agricultural food production (CoDyre, Fraser, & Landman, 2015;Martellozzo et al., 2014). Here, we explore the potential suitability of greenspaces in 10 diverse cities across the United Kingdom to support biofuel production using a Geographic Information System (GIS)-based analysis. We focus on short rotation coppice production (SRC), the planting and regular harvesting of fast-growing tree species and Miscanthus grass. Relative to fossil fuels, SRC and Miscanthus biofuel represent a predictable, carbon-neutral energy source if managed sustainably Wang et al., 2014;Yang et al., 2017). As such, it is recognized as an important component of renewable energy production (European Commission, 2014;Foxon, 2013), and continues to form part of many national and transnational environmental policies and agreements (Alexander, Moran, Rounsevell, Hillier, & Smith, 2014;Matthews et al., 2002;McHugh et al., 2015;Upham & Speakman, 2007). The objective of this research was to estimate the potential of urban biofuel production at a national scale; whereas previous research has focused on rural environments where food versus fuel conflicts will be stronger Hastings et al., 2014) or assessed urban production potential in a single city (McHugh et al., 2015).
In this study, we estimate how much urban greenspace can support biomass production for CHP in UK cities. In so doing we explore whether urban biomass production has a 'sustainable' end market. Areal results for each city were combined with published geographic location-specific yield data on common biofuel crops (SRC willow; SRC poplar and Miscanthus grass) to explore the production potential of each city and, by extrapolation, to urban areas across the entire country. Our research builds directly on that of McHugh et al. (2015), who estimated the production potential in Leicester, United Kingdom; whereas here we consider 10 representative urban areas across the United Kingdom and through this seek to extrapolate our findings to a nationwide estimate. Finally, we investigate the implications of these results in the context of heat and power demand, national sustainability targets and future energy scenarios (FES).

| Study area
The 10 study areas this research considered were: the cities of Leicester, Southampton, Nottingham, Newcastle upon Tyne, Bristol, Edinburgh, Swansea/Abertawe, Milton Keynes, Liverpool and Sheffield ( Figure 1). Collectively, these urban areas cover a broad geographic range across the island of Great Britain, as well as representing a wide variety of urban landscape structures and histories in order to act as a representative sample of UK urban areas as a whole.

| Vegetation location using aerial imagery and normalized difference vegetation index
Landmap colour infrared (CIR) aerial imagery (Landmap; Bluesky, 2014) was acquired for all study areas, chosen for its spatial resolution and availability. Due to the nature of the data and acquisition, gaps existed in some areas, usually outer rural or semi-rural regions surrounding the urban areas. Imagery was captured for each study area on different dates, and in some cases, the imagery from individual cities was split over multiple dates (see Table S1). Normalized difference vegetation index (NDVI) was derived for each study area from the imagery using the Image Analysis functionality in ArcGIS (ESRI, 2013). NDVI acts as a commonly used and straightforward measure of relative vegetation density, health and abundance due to biophysical and reflectance properties of living leaves, and is calculated based on the difference in reflectance between visible red wavelengths and near infrared wavelengths: Once calculated, a threshold NDVI value was selected for each study area to distinguish vegetated from non-vegetated areas. The differences in acquisition dates, and thus seasonality of vegetation, between study areas meant that the appropriate NDVI threshold (determined by visual comparison) value varied by study area. Appropriate threshold values were thus selected through visual analysis of the imagery in various land covers of each city (see Table  S1). Note that imagery in Nottingham was split between the east and west halves of the city by date with vegetation differences significant enough to warrant separate NDVI thresholds.

Survey MasterMap
Thresholding of NDVI values was able to separate vegetation from non-vegetation to a large degree; however, image properties and artefacts caused some lingering issues. In Southampton and Edinburgh, slight offsets between image colour bands and seasonal effects due to image timing caused some false vegetation readings in paved or built-up areas. In Newcastle upon Tyne and Swansea, reflectance from shallow offshore areas falsely registered as vegetation. In Sheffield and Swansea, upland areas covered with very sparse vegetation registered as non-vegetated.
In order to correct for these issues, Ordnance Survey MasterMap data (Ordnance Survey (GB), 2017) were used to identify known areas of buildings, structures and water and exclude these from consideration as vegetated surfaces. Polygon features identified either primarily or secondarily as buildings or structures, or primarily as water, were used to remove pixels from the NDVI vegetation layer. This was done on the expectation that the MasterMap data would depict these features reliably and help correct for uncertainties in areas known to be unvegetated, whereas NDVI thresholding is subject to uncertainty due to variations in seasonality, atmospheric properties and sensor characteristics.

| Calculation of urban areas suitable for biofuel production
Based on research described in McHugh et al. (2015) and adapted from Aylott et al. (2010), multiple exclusion criteria were applied in GIS to remove areas from analysis that were deemed unsuitable for biofuel production and operation due to legality, desirability or public usage rights. These include various protected lands (e.g. Sites of Special Scientific Interest-SSSI, national and local nature reserves and World Heritage Sites), common land (land with shared usage rights), public rights of way (land with public access rights), recipients of other government schemes such as countryside stewardship agreements (which encourage environmental improvements to rural lands) or areas otherwise set aside for uses incompatible with biofuel cultivation (Natural England, 2013; see Table S2). Current government guidance and grants for SRC and other biofuel operations in the United Kingdom focus on agricultural lands, but our research assumes that this guidance could be readily adapted for use in urban settings.
Lands listed under the Agricultural Land Classification (ALC) system as Grades 1-3 arable land were also excluded, as these represent lands with high-quality agricultural potential, in order to avoid conflicts between biofuel plantations and arable land that could be better utilized in food crop production. The remaining grades (4, 5, non-agricultural and urban) were theorized to be only marginally less efficient for biofuel production given recent findings about urban soil quality, compaction and carbon content being similar to or more favourable than agricultural lands (Edmondson, Davies, Gaston, & Leake, 2014;Edmondson, Davies, McCormack, Gaston, & Leake, 2011), while less desirable for food crop cultivation. Although the food/fuel conflict may seem an unnecessary consideration in urban areas at present, growing concerns over global food security and interest in urban agriculture supported the exclusion of high-grade land that may be better used for food production, even in urban areas (Edmondson et al., 2019). Subsequent refinement of spatial layers did not further consider ALC grade, as regional/climatic differences were found by Aylott et al. (2010) to explain a greater degree of variation in SRC yield across the United Kingdom than
ALC grade/soil quality, and were thus treated as a higher priority. Aylott et al. (2010) considered both region and ALC grade in England-wide estimates of SRC potential, finding a mean variation in yield of 2.9 oven dry tonnes (ODT) ha −1 year −1 across geographic regions, compared with a mean yield variation of 1.2 ODT ha −1 year −1 across ALC grades. Given this finding, we chose to use figures from Hastings et al. (2014) for yield estimations (discussed below) for their more complete regional coverage (inclusive of Scotland and Wales rather than only England), more recent figures and inclusion of multiple crop types. Although we did not explicitly consider climate impacts on crop yields, these data nonetheless enabled a representative coverage across Great Britain and region-specific yield figures, which were assumed to correlate with any noteworthy climatic effects. Selection by attribute query was used to isolate the key exclusion features in each data set, which were then removed from the analysis area. The raster vegetation layer was clipped to those areas deemed suitable for biofuel plantation placement. Some areas initially intended for exclusion were not removed from analysis due to unavailability of appropriate spatial data on those areas (see Table S3), but their impact on our urban study sites was assumed to be minimal. Any existing woodlands and trees not already excluded under the above criteria were identified using Environment Agency LIDAR. All features with a height greater than 2 m were removed from analysis.
The resulting raster data sets, having been created from NDVI vegetation thresholding and the exclusion of local and national designated lands, private gardens and trees, were converted to vector. Enclosed holes in potential SRC area polygons smaller than 100 m 2 were removed as these appeared to be small artefacts from NDVI calculation and were assumed in reality to be valid areas belonging to their surrounding sites. Buffers were calculated to restrict potential biofuel sites to be greater than 10 m from urban residential private gardens (rural lands outside of built-up urban areas, even if privately owned, were still available for analysis if not excluded based on other criteria). A 3 m buffer around the edge of all potential sites was removed to allow vehicle access before production potential was calculated. Finally, the area of all resulting green patches was calculated, and patches smaller than 0.5 ha in size were removed to ensure that sites under consideration would be of an adequate size to make operations practical (McHugh et al., 2015).
Once calculated, the area potentially suitable for biofuel production in each study city was compared to each city's total administrative boundary and total built-up extent (i.e. developed urban landscape but still containing small greenspaces potentially suitable for biofuel cultivation; ONS, 2011; Rowland et al., 2017). This enabled comparisons between the study cities and provided a consistent basis from which to extrapolate mean biofuel potential to urban administrative and built-up areas across the entire United Kingdom, correcting for variation in the shape and urban makeup of administrative boundaries, with the caveat that urban landscape structure is highly variable and direct comparisons are difficult (e.g. these suitability proportions were not normally distributed in our analysis). Similarly, the total area of urban greenspace within each city's administrative boundary was calculated based on the NDVI thresholding described previously, and compared to total biofuel-suitable area to produce the proportion of greenspace in each city that may be suitable for biofuel production.
We conducted a validation process where 500 of the resulting study sites were randomly selected (stratified to 50 from each study city, representing 18.7% of the total 2,680 sites) and each of these compared with high-resolution Google Earth imagery to assess the accuracy of our exclusion process. This validation process found an agreement rate with the sites our analysis had deemed potentially suitable for biofuel production of 78%. The discrepancies were driven by two factors: (a) urban development between acquisition of the CIR imagery used in our analysis and the more current 2018 Google Earth imagery, and (b) some schools and private sports fields being incorrectly classified in the OS MasterMap data we used. Additionally, some airport grassland areas were included as potential sites (clearly unsuitable, although our exclusion process had not explicitly ruled them out-nevertheless, these spaces account for a relatively small area in only some of our study cities; see Table S4 for validation rates and information by city). Despite the discrepancies, the research here is not intended to represent an exhaustive multi-criteria site suitability analysis, but rather the consideration of a maximal edge case where we look to exclude obviously unsuitable areas while still allowing for the conversion of spaces such as mown grass in parks. Unconvertible areas such as cemeteries and airport grasslands fall within what we consider to be the allowable margin of error in our results.

| Calculation of potential biofuel yield
Calculations of total available biofuel-compatible land area within each urban area were combined with figures on the potential mean yield per area in ODT per hectare per year in each region. We considered willow (Salix spp.) and poplar (Populus spp.) trees as well as Miscanthus grass-three of the most commonly used biofuel crops in the United Kingdom. Yield figures were taken from modelled results in , specific to each region of Great Britain to account for climatic differences expected to affect yield (see Table  S5). This enabled the estimation of the total potential biofuel yield per study city.
These yield estimates were then combined with information on per-building heating demand and the number of dwellings present in each study city to calculate a broad estimate of each city's residential heating demand, and the proportion of it that could potentially be met by locally urban-grown biofuels. Figures from Forest Research United Kingdom (Forest Research, 2017) were used after McHugh et al. (2015) to define typical heating demands of urban structures (e.g. residential homes, district heating schemes and municipal buildings). According to these data, a typical domestic house/dwelling requires an average of 3.99 ODT of wood chips per year, a district heating scheme requires an average of 119 ODT/year and a municipal building requires an average of 203 ODT/year. Note that these values are reported by Forest Research as tonnes of wood chips with 30% moisture content/year (5.7 tonnes/year for a typical domestic house, 170 for a typical district heating scheme and 290 for a municipal building); McHugh et al. converted these to ODT/ year and it is McHugh's converted figures that we use here.

| Potential urban biofuel cultivation area
The 10 UK urban areas we studied were variable in their underlying urban structure, which translated to variations in the amount and location of greenspace potentially available for biofuel production (Table 1). Cities such as Leicester ( Figure  2a), Southampton ( Figure S1) and Nottingham ( Figure S2) represent typical cases where urban built-up extent closely matches the urban administrative boundary, so potential biofuel sites generally represent underused greenspaces within the urban matrix (3% of Leicester's total administrative area was deemed potentially suitable, 2% for Southampton and 6% for Nottingham).
Other urban areas such as Swansea included rural lands that were outside of urban built-up areas but within the city's administrative jurisdiction, and thus may be suitable for biofuel production according to the criteria used here. In the case of Swansea, this produced high absolute and proportional values of greenspace land deemed suitable for biofuel production (Figure 2b; 8% of total administrative area), although uncertainties in the suitability of uplands for biofuel production may lead these figures to overestimate suitable land area. By contrast, in other locations, the impact of these extraurban lands was tempered by conflicting exclusion criteria; much of the extra-urban land in Milton Keynes is ALC grade 1-3 and was thus excluded to avoid conflict with high-grade food production lands ( Figure S6; 3% of administrative area deemed suitable), and much of Sheffield's extra-urban land is part of the Peak District National Park and was thus excluded due to the presence of protected and sensitive habitats ( Figure  S8; 2% of administrative area deemed suitable).

| Urban biofuel production potential
Total potential biofuel production in each city was calculated based on region-specific yield estimates for each crop combined with the suitable area in that city (Figure 3). In all cases except Bristol, SRC poplar generated the highest predicted yields, whereas in Bristol, Miscanthus grass was predicted to be the most productive crop. Swansea showed the greatest total production potential of the study cities (three times that of Edinburgh, the next highest) in terms of total ODT/ year, or six times that of Milton Keynes when compared on a per-dwelling basis. However, in per-hectare comparisons, Swansea (1.13 ODT ha −1 year −1 ) was similar to Liverpool (0.93 ODT ha −1 year −1 ), with Edinburgh (0.62 ODT ha −1 year −1 ) and Nottingham (0.61 ODT ha −1 year −1 ) considerably higher than the remaining six urban areas. Collectively, the results highlight the importance that urban form can have on biofuel production potential. Although not explicitly studied here, this can include considerations such as urban history, built-up versus administrative area, fragmentation of land parcels, road layout and relative proportions of different land covers.
In order to extrapolate areal suitability findings to the entire United Kingdom, the proportion of potential biofuel production land was calculated in each study city relative to the total administrative area and total built-up area contained within the administrative boundary of that city (Table 1). On average, 3.3 ± 0.5% SE of our study cities' T A B L E 1 Total potential biofuel production area in each study site, compared proportionally to total administrative area, total built-up land and total urban greenspace in that site. Proportions of suitable total administrative areas, built-up lands and urban greenspace were all not significantly different from normal distribution according to Shapiro-Wilk's test built-up area and 4.0 ± 0.7% SE of their administrative area were potentially suitable for supporting biofuel cultivation. Proportional calculations of biofuel potential relative to total available greenspace show that a mean of 9.9 ± 1.6% SE of the greenspace in UK urban areas may be suitable for consideration for biofuel production. Given that Great Britain contains a total of 1.64 million ha of built-up land (ONS, 2011 for England/Wales; Rowland et al., 2017 urban/suburban land cover for Scotland), our average figure of 3.3% for built-up area suitability potential suggests that a mean of 53,800 ha (±7,400 ha SE, and a mean potentially as low as 42,000 ha based on the 78% accuracy rating found in our mapping validation) of this land may be suitable for biofuel production. By applying yield estimates for the most productive crops in each region from Hastings et al. (2014), this represents a maximum nationwide biofuel production potential in urban built-up lands of 624,800 ODT/ year. Although not explored here, reconsideration of legal and social frameworks defining some current exclusion criteria, such as high agricultural land grade and woodland grant schemes, could expand these figures further (Natural England, 2013). These yield estimates represent a maximal potential biofuel supply, which were next considered in the context of heating demand. Figures from Forest Research UK (2017) were used after McHugh et al. (2015) to predict the heating potential of the most productive locally sourced biofuel crop in each city (Table 2). Additionally, comparing the recorded number of dwellings in each city (for England sites: Ministry of Housing, Communities, & Local Government, 2016; Swansea: StatsWales, 2016; Edinburgh: National Records of Scotland, 2017) to the estimated number of dwellings that could be heated this way enabled the calculation of the proportion of urban domestic heating demand that could potentially be met in each case. When expanded to the total built-up area of Great Britain, the maximal potential yield represents the ability to heat 156,591 dwellings, 5,250 district heating schemes or 3,078 municipal buildings. The Forest Research figures do not account for details such as transmission losses of heat and only report figures based on generalized approximate sizes of dwellings/schemes/buildings used for their calculations, but this still provides us with a broad national estimate of heating potential.

| Potential of urban biofuel to meet national sustainability targets
The factors contributing to estimates of each city's heating demand, and the potential of indigenous biofuel to meet it, account for a considerable degree of variation between cities; but they suggest that, on average, locally sourced urban biofuel is capable of meeting 2% of a city's residential heating demand (Table 2). Swansea stands out by being able to supply an order of magnitude more heat from local biofuel production than other study cities, due to a combination of its extensive biofuel-suitable extra-urban land and relatively modest demand by number of dwellings. Most other cities fall between zero and two percent, perhaps reflecting the more realistic or common case. However, the estimation of total potential to meet demand underplays the role urban biomass can have in achieving climate change targets. It is unlikely, for example, that the retrofit of biomass heating to average urban dwellings will be a realistic proposition due the range of end-user problems that can occur, concerns about pellet supply and pricing and maintenance requirements (Thomson & Liddell, 2015). Additionally, there are several technologies competing to become the low carbon option of choice for UK households to move space heating requirements away from natural gas combi boilers (which currently dominate provision in the United Kingdom), including air source heat pumps and hydrogen boilers with associated gas network retrofit (Hanna, Parrish, & Gross, 2016). As such, the most F I G U R E 3 (a) Total, (b) per-hectare of total administrative area and (c) per-dwelling estimated biofuel production (ODT/year) by city based on suitable area and predicted yield for various biofuel crops (see Table S4). ODT, oven dry tonnes; SRC, short rotation coppice likely destination for urban biomass is as a primary or mixed fuel for CHP generators feeding urban heat networks. Heat networks with CHP are seen as a critical pillar in UK energy system decarbonization, and are substantially present in those scenarios which meet the United Kingdom's climate change commitments (National Grid, 2018). With the potential yield data developed here, we can estimate the extent to which indigenous urban biomass can satisfy the fuel demand for these networks, thus ameliorating reliance on international supply chains and their associated emissions.
National Grid in its role as UK system operator produces annual FES, which in their most recent incarnation describes four potential energy system futures to 2050; two of which, 'Community Renewables' and 'Two Degrees', meet the United Kingdom's climate change commitments (National Grid, 2018). Both scenarios incorporate a rapid expansion of urban heat networks, and 'Community Renewables' predict up to 2 GW of installed decentralized biomass CHP capacity by 2040. National Grid does not compute the land area requirements for this, nor specify biomass sources such as SRC. However, using the Digest of UK Energy Statistics (DUKES) 2018 (BEIS, 2018) we can derive a ratio of installed electrical CHP capacity to heat output for both conventional CHP and Organic Rankine Cycle (ORC) installations. DUKES 2018 suggests that recent UK Government policy for distributed renewable heat has led to installation of a substantial number of ORC CHP engines, so our calculations here incorporate this technology (Table 3). From these calculations, we find that within the urban boundary of UK major cities (comprising 53,846 ha), we could grow up to 18% of the biomass fuel for CHP systems needed to meet the demands of a climate compatible energy scenario by 2030. As stated previously, the data underlying our analysis do not account for some relevant details such as heat loss in transmission which would lower the overall efficiency of operations; nevertheless, the work presented here displays the broad potential of urban biofuel production at a national scale, and producing them as near to the point of use as possible (i.e. within the same city) where they would then be used to produce heat and power for local consumers would seek to maximize efficiency across the life cycle.

| Management considerations, cobenefits and trade-offs
The availability of greenspace for biofuel production in any given city, as a proportion of either its administrative or builtup area, reflects characteristics of urban form and landscape structure that also have implications for ecosystem service provision. Since biofuel suitability was restricted in part to T A B L E 2 Estimated annual heating potential of locally sourced urban biofuel based on the most productive crop at each site (Table S4) and expected heating demand (3.99 ODT/year per house/dwelling, 119 ODT/year per district heating scheme and 203 ODT/year per municipal building, Forest Research, 2017;McHugh et al., 2015), number of residential dwellings present in each study site (England: Ministry of Housing, Communities, & Local Government, 2016;Swansea: StatsWales, 2016;Edinburgh: National Records of Scotland, 2017)  relatively large patches, low proportional suitability relative to the available greenspace in a given city may reflect higher degrees of greenspace fragmentation. Conversely, cities with a relatively high proportion of their greenspace suitable for biofuel production (e.g. Nottingham, Bristol, Swansea and Liverpool) possess more and larger greenspace patches with associated environmental benefits (Beninde, Veith, & Hochkirch, 2015;Grafius, Corstanje, & Harris, 2018;Saunders, Hobbs, & Margules, 1991). If managed effectively, biofuel production operations may support peripheral ecosystem service benefits such as biodiversity (O'Sullivan et al., 2017;Rowe et al., 2009;Sage et al., 2006), mitigation of air and soil pollution (Rowe et al., 2009;Sugiura et al., 2008), mitigation of flood risk (O'Sullivan et al., 2017), sequestration of carbon in the soil (Matthews et al., 2002), noise suppression from major roads and railways and aesthetic benefits (Rowe et al., 2009). These benefits must be balanced against potential disservices such as negative aesthetics (Dockerty, Appleton, & Lovett, 2012), noise and access conflicts from maintenance and harvesting operations, perceived increase in crime risk (Chiesura, 2004;Dobbs, Escobedo, & Zipperer, 2011), and land conflict with any alternate preferred uses of a given site. Additional factors, such as albedo, may remain relatively unchanged in lands converting from mown grass to SRC (De Groote et al., 2015;Markvart & Castaner, 2003;Schmidt-Walter, Richter, Herbst, Schuldt, & Lamersdorf, 2014). The use of marginal lands for energy crop production may help to avoid land conflicts while providing ecosystem services and benefitting remediation efforts (Blanco-Canqui, 2016). The most appropriate crop type will depend on local considerations; regional differences in climate will determine which crops will maximize production, but again this will be balanced against other considerations. For example, Miscanthus or an SRC crop may maximize raw fuel production but be deemed undesirable in some locations, whereas an alternate approach such as pollarding of trees (Read, 2006;Smith, Pearce, & Wolfe, 2012) or short rotation forestry operations may produce only 15%-25% as much oven dry weight of fuel material (Forest Research, 2018; based on Forest Research figures for expected yields from forestry residues after thinning operations, specific figures for pollarding yields were unavailable) but be deemed more aesthetically desirable in some frequently used public lands such as parks. In some cases, it may also be possible to include secondary sources of fuel from other operations, such as cuttings from tree pruning. One example in Ecuador found that lechero trees (Euphorbia laurifolia L.), although perhaps exhibiting faster growth rates than most urban UK trees, produced an average of 9.95 kg of wood and leaf pruning waste per tree (Velázquez-Martí, Gaibor-Cházvez, Niño-Ruiz, & Narbona-Sahuquillo, 2018).
There are also myriad economic and logistical factors involved in the consideration of biofuel production, T A B L E 3 Percentage of national biomass CHP fuel demand that can be met using biofuel grown within urban boundaries particularly in urban areas with their probable smaller site areas compared to rural areas, which remain largely unexplored, but addressing such complexities can be increasingly aided with the use of modelling tools (Bauen et al., 2010;Foster, 1993). Although economies of scale suggest that only larger sites might prove economically viable (Fiala & Bacenetti, 2012), a component-based understanding of the relative costs incurred by fuel transport, travel and harvest method and timing may find alternative solutions (Pecenka & Hoffmann, 2015): for example, an operation structured around frequent harvesting of relatively small plots would enable the use of lower performance equipment and provide a more continuous fuel supply, all of which could prove optimal for urban biofuel production (Sims & Venturi, 2004).
Finding an appropriate balance between production output, economic and logistical practicality, co-benefits, residential preferences, rural versus urban land value and other considerations must be studied further and addressed on a case-by-case basis by individual local interests when deciding whether and where to locate urban biofuel operations. However, having a clear and broad understanding of the production potential, as we explore here, will strengthen this decision-making process.

| Comparison with previous research
We have explored the potential of biofuels in UK urban areas at a deliberately broad scale, considering 10 different study areas in order to generate a nationally representative picture that spans many UK regions and urban landscape types. Few published studies exist that offer comparable results; however, McHugh et al. (2015) estimated urban biofuel potential for the city of Leicester, United Kingdom, albeit based on different SRC yield estimates found in . This research found 297 ha of potentially suitable land within the urban boundary when using comparable exclusion criteria to our study, capable of producing an average of 3,052 ODT of wood chip per year. Our analysis here, by comparison, estimated 238 ha producing 1,639 ODT/year for SRC willow or 2,347 ODT/year for SRC poplar. Based on heating demand figures from Forestry Research, United Kingdom (Forest Research, 2017;McHugh et al., 2015), for residential dwellings, district heating schemes and municipal buildings, McHugh et al. (2015) calculated that local urban SRC in Leicester could supply heat for 1,566 houses, 52 district heating schemes or 30 municipal buildings. Our research produced more modest results, finding that urban biofuel production in Leicester could potentially supply heating for 588 domestic houses, 19 district heating schemes or 11 municipal buildings. The differences between these findings may stem in part from methodological differences in suitability mapping, although the criteria were largely the same between studies; as such, the greatest portion of this difference is believed to originate from the different sources of predicted yield values, for which much uncertainty remains given the limited degree to which urban biofuel potential has been investigated thus far. Another consideration for future research is that climate change may alter the suitability and optimal selection of crops in different regions over time (Bellarby, Wattenbach, Tuck, Glendining, & Smith, 2010). Understanding the productivity of urban biofuel operations is in its infancy and data poor, so current uncertainties are large but will decrease with further research and real-world implementation.

| Local policy implications
In addition to national targets and scenarios, numerous local authorities in the United Kingdom express independent aspirations towards increasing urban resilience and sustainability through the greater uptake of locally sourced biofuels, often in concert with CHP systems and/ or district heating schemes.  , 2012;Minshull et al., 2015;Newcastle City Council, 2010), or at least the costs of importing them externally (Nottingham City Council, 2010), and in Southampton, a proposed 100 MW biomass plant was rejected largely over concerns about the import of biofuel and the associated whole-system detriment to residents and the environment (Vinson, 2011). These examples strengthen the case for greater consideration of urban production of biofuels and their associated use in local district heating and/or energy production schemes, as such operations could support many stated aspirations of local authority strategies while avoiding the identified environmental and cost disadvantages of importing biofuel from distant or foreign suppliers.

Kingdom and beyond
The United Kingdom has recently committed to a goal of net zero carbon emissions by 2050 (Prime Minister's Office, Clark, & May, 2019), as one of many countries around the world seeking to reduce its GHG emissions through an increased focus on renewable energy, and with fast-growing biofuel resources in particular as one way to work towards this goal. Research in the Northeast United States, for example, modelled the potential yield of different energy crops for that region, finding short rotation willow to be the most favourable albeit dependent on various market conditions and incentives (Brandt et al., 2018).
Research in China has explored the balances and trade-offs between bioenergy crops, food crops and local water table depletion with a goal of informing governmental policy on the optimal management of these resources (Yang, Chen, Pacenka, Steenhuis, & Sui, 2019). Biofuel research in Germany alternately has explored GHG abatement in the transportation sector given its wide-ranging impact, finding that a prioritized focus on land passenger transport is key (Millinger, Meisel, & Thrän, 2019). In all such cases, production of biofuel crops in urban spaces could potentially contribute. From a policy standpoint, nations have historically acted in accordance with their own unique situations, such as the United States' focus on the economics of maize corn-based ethanol (Dutta, 2019), and France and Germany's adoption of biofuel policies in response to agricultural surpluses in the early 20th century and later the 1970s oil crisis. In many cases, national policy remains driven more by economic self-interest and energy security than global environmental concerns (Oliveira, McKay, & Plank, 2017); however, multinational agreements may offer a more holistic perspective that better recognizes the advantages of small-scale urban biofuel operations. Examples of such agreements include the European Commission's 2030 Climate & Energy Framework (European Commission, 2014), which sets a binding target for European nations of generating at least 27% of their total energy from renewable sources by 2030, and the Conference of the Parties (COP 21) Paris Agreement of the United Nations Framework Convention on Climate Change, in which 174 countries have set a goal of limiting global average warming due to anthropogenic GHG emissions to 'well below 2°C' compared to pre-industrial levels (United Nations Framework Convention on Climate Change (UNFCCC), 2016). Such international agreements commonly account for local differences in resources and feasibility by not dictating explicitly how the signatory nations achieve these goals, and in many cases, fast-growing biofuel crops for local consumption in heat and/or power generation may represent a valuable resource; being relatively straightforward to implement at local levels, but potentially scaling up to have noteworthy regional and global impacts on GHG reduction.

| CONCLUSIONS
Our findings demonstrate that a typical UK urban area holds a great deal of largely unconsidered potential to provision its residents with sustainable biofuel for electrical power and/or heat generation. Optimal decisions pertaining to biofuel species used, type of energy produced, supply chain details and location of operations must be determined on a case-by-case basis. The estimates calculated here represent a maximal edge case which is unlikely to be desirable in practice; however, more land could become classified as suitable if some restriction criteria were loosened. Such decisions will necessarily address conflicting preferences and values from local residents and users, consider entire supply chains and life cycles of operations (Strohbach, Arnold, & Haase, 2012) and balance trade-offs and synergies between different ecosystem services and disservices that may be provided by biofuel operations beyond their primary goal of fuel production. However, the potential is clear and compelling, and modelling tools are increasingly becoming available to address the inherent challenges (Tallis et al., 2013). Urban areas in the United Kingdom have the capability to supply a significant proportion of their own energy demand using low-carbon or carbon-neutral methods, produced near the point of demand to avoid costs incurred in long-distance fuel transport, and without land conflict with rural food crop production. The advantages of such practices are in line with the expressed goals of numerous local authorities as well as national policies and international agreements. Overall, our findings suggest that urban biofuel production can make a significant contribution to the volume of biomass called for in local, national and potentially international sustainability scenarios. Furthermore, the largely unconsidered nature of this renewable fuel source makes this potential contribution particularly striking. Future research should seek to further investigate the feasibility of such operations at various scales and in different locations, as well as according to different criteria that were not explicitly studied here. A more exact consideration of the effects of site fragmentation and size on economies of scale, trade-offs between urban and rural lands that consider potential land value as well as transport costs and the impacts of local climate on biofuel crop yields are perhaps the most vital issues for immediate consideration, particularly in the context of a sensitivity analysis capable of comparing the relative importance of these factors.
We have shown that ca. 3% of the built-up area of a typical UK city, 4% of its total administrative area and nearly 10% of its total urban greenspace are potentially suitable for biofuel production. Scaled up, this urban biofuel potential may be capable of meeting approximately 18% of the United | 129

GRAFIUS et Al.
Kingdom's biomass CHP heating fuel demand in a climate compatible scenario by 2030 as outlined by the National Grid's Future Energy Scenarios. Urban biofuel operations face multiple challenges that must be addressed at local scales and on a case-by-case basis in order to ensure they are viable and sustainable, but if such challenges can be met urban biofuel production may represent an opportunity to make cities and energy systems more sustainable and resilient while providing numerous peripheral benefits to urban residents and nature.