This is a Submitted Manuscript (pre-print) of the article Corcelli, F., Fiorentino, G., Petit-Boix, A., Rieradevall, J., & Gabarrell, X. (2019). Transforming rooftops into productive urban spaces in the Mediterranean. An LCA comparison of agri-urban production and photovoltaic energy generation. Resources, Conservation and Recycling, 144, 321-336 available online at https://doi.org/10.1016/j.resconrec.2019.01.040 1 Transforming rooftops into productive urban spaces in the Mediterranean. An 2 LCA comparison of agri-urban production and photovoltaic energy generation 3 4 Corcelli, F.a,*, Fiorentino, G.a, Petit-Boix A.b,c, Rieradevall, J.b,d,e, Gabarrell, X. b,d,e 5 a Department of Science and Technology, Parthenope University of Naples, Centro Direzionale – Isola C4, 80143 Naples, 6 Italy 7 b Sostenipra Research Group (2017SGR1683), Institute of Environmental Sciences and Technology (MDM-2015-0552), 8 Z Building, Autonomous University of Barcelona (UAB), Campus UAB, 08193, Bellaterra, Barcelona, Spain 9 c Chair of Societal Transition and Circular Economy, University of Freiburg, Tennenbacher Str. 4, 79106 Freiburg i. Br., 10 Germany 11 d Department of Chemical, Biological and Environmental Engineering, Catalan Biotechnology Reference Network -XRB 12 Autonomous University of Barcelona (UAB), Campus UAB, 08193, Bellaterra, Barcelona, Spain 13 e Department of Chemical, Biological and Environmental Engineering, School of Engineering, Building Q, Autonomous 14 University of Barcelona (UAB), Campus UAB, 08193, Bellaterra, Barcelona, Spain 15 16 Abstract 17 A key strategy towards sustainable urban development is designing cities for increased circular metabolism. The 18 transformation of areas underused, such as urban rooftops, into productive spaces is being increasingly implemented as a 19 result of associated multiple benefits. Rooftop greenhouses (RTGs) are an interesting option for exploiting urban rooftops 20 with direct exposure to sunlight, reducing food miles and creating new agricultural spaces, while building-applied solar 21 photovoltaic (BAPV) panels provide clean energy and reduce greenhouse gas emissions. However, a proper assessment 22 of environmental costs and benefits related to both systems is vital for a successful implementation. By means of life 23 cycle assessment method, this paper aims to compare the environmental performance of different productive uses of 24 rooftops under Mediterranean climatic conditions. The results showed that, in the case of RTG systems, the operation and 25 maintenance phase (i.e.tomato production) has more impacts than the infrastructure and end-of-life phases due to the use 26 of fertilisers (impacts ranging from 21% to 62%). Concerning BAPV systems, the manufacturing phase is dominant in 27 almost all impact categories (impacts ranging from 52% to 93%) due to the electricity used in producing materials for 28 solar panels. The implementation of measures for material and energy efficiency in the assessed systems resulted crucial 29 in lowering the environmental burdens, by avoiding food imports and fossil energy supply. 30 The main finding of this study was that urban planning will have an important role to play in optimizing the circular 31 patterns in highly urbanized areas by integrating these technologies early into the planning process. 32 33 *Corresponding author: fabiana.corcelli@uniparthenope.it; Tel: +390815476666 34 Keyword: cities, LCA, agri-urban, photovoltaic energy, rooftop, circular economy 35 36 37 Acronyms BAPV Building Applied Photovoltaic LCIA Life Cycle Impact Assessment BOS Balance of System MD Metal Depletion CC Climate Change OD Ozone Depletion c-Si crystallin Silicon POF Photochemical Oxidant Formation EoL End-of-Life RTG Rooftop Greenhouse FD Fossil Depletion TA Terrestrial Acidification FE Freshwater Eutrophication TE Terrestrial Ecotoxicity FU Functional Unit UA Urban Agriculture GHG Greenhouse Gas UAB Autonomous University of Barcelona GR Green Roof WD Water Depletion LCA Life Cycle Assessment Wp Peak Watt LCI Life Cycle Inventory WEEE Waste Electrical and Electronic equipment 38 39 40 1. Introduction 41 42 Urban areas play a significant role in addressing the 21st century’s challenges of sustainably realizing climate, energy and 43 economic objectives. Currently, urban environments have proved to be unsustainable, as they heavily rely on imported 44 resources and their environmental footprint exceeds their natural bio-capacity (Doughty and Hammond 2004). Cities host 45 more than 50% of the global population (United Nation, 2014), consuming 60-80% of global primary energy and 46 generating 70% of the world’s total greenhouse gas (GHG) emissions, primarily through the consumption of fossil fuels 47 for energy supply, transportation and food production (UN-Habitat, 2016). To transform modern cities into more 48 sustainable environments, cities need to develop a more circular metabolism where more resources are recycled, reused 49 or produced on-site whilst cleaner forms of energy are produced and consumed (Doughty and Hammond 2004). 50 Additionally, there is a need to optimize land use in over-populated cities where land competition becomes a problem. 51 In this context, sustainable solutions for food, water, energy, and transport of food or waste are needed as integrated 52 components of a city’s climate change adaptation. Sustainable urbanization practices offer many opportunities for 53 optimizing resource use efficiency and developing mitigation measures to deal with such problems, especially through 54 urban planning (e.g., exploitation of unused areas for local resources production, efficient waste management) and design 55 (e.g., green construction, use of recycled materials, upgraded technologies for material/energy recovery) (UN-Habitat, 56 2016). Urban deployment of such strategies is often amalgamated under concepts such as the ‘eco-city’, ‘circular city’ or 57 ‘sustainable city’, and have found wide spread local, regional, and national acceptance (Castán Broto and Bulkeley, 2013; 58 Petit-Boix and Leipold, 2018). In this sense, converting vacant rooftops into productive spaces is a recognised strategy 59 among researchers, city planners and developers (Carter and Keeler 2008). Rooftops have an unprecedented exploitation 60 potential, as they cover up to 32% of cities and built-up areas (Frazer, 2005) and can improve the urban metabolism by 61 producing resources such as energy, greening, food and water (Mahmoud et al., 2014; Specht et al., 2014; Goldstein et 62 al., 2016, Yang and Zou, 2016; Petit-Boix et al., 2018). The transformation of these urban rooftops into productive spaces 63 is becoming standard practice in many cities worldwide (Proksch, 2011). For instance, rooftop greenhouses (RTGs) for 64 food production are gaining popularity in large cities such as New York, Singapore and Montreal (Ackerman et al., 2012; 65 Deng and Quigley, 2012; Haberman et al., 2014). Similarly, as recently noted by Sanyé-Mengual et al. (2015), RTGs are 66 finding deployment in experimental projects also in the Mediterranean context because of the favourable climate 67 conditions. Furthermore, in order to mitigate the environmental impacts of urbanization, researchers worldwide have been 68 recently looking at ways of turning buildings into net energy producers. Solar energy is an infinite and clean resource, 69 and scientists have been assessing systems such as Building-Applied Photovoltaics (BAPV), which consist of generating 70 considerable fractions of urban electricity without the need of dedicating exclusive surface areas for solar photovoltaic 71 (PV) plant installations (Santos and Rüther, 2012). BAPV systems are typically used in retrofits, with off-the-shelf PV 72 panels mounted on a separate metal support structure, superimposed on an existing building’s roof or façades. Among 73 existing applications for BAPV, rooftops are considered the ideal option, since pitched roofs with a proper angle and 74 orientation provide the highest energy harvesting (Heinstein et al., 2013). Food production and energy generation on 75 urban rooftops are also an important source of environmental benefits. Table 1 summarizes the main environmental 76 benefits of using BAPV and RTGs systems. 77 In the recent scientific literature, sustainable urban solutions addressing food supply, on the one hand, and energy supply, 78 on the other, have been increasingly assessed. A range of studies have focused on the role played by urban agriculture 79 (UA) in urban food self-reliance and environmental impact mitigation (Grewal and Grewal, 2012; Haberman et al., 2014; 80 Orsini et al., 2014; Benis and Ferrão, 2016; Wielemaker et al., 2018, among others). Further studies have estimated the 81 potential of renewable energy systems such as PV to fulfil urban energy needs (Hofierka and Kanuk, 2009; Amado and 82 Poggi, 2014; Byrne et al., 2015) and investigated the environmental costs and benefits of this technology (Fthenakis et 83 al., 2009; Peng et al., 2013; Corcelli et al., 2016, 2017; Eskew et al., 2018; among others). Green roofs (GR) are the main 84 rooftop interventions considered, but comparison with food production remains unexplored. For instance, Elzeyadi et al. 85 (2009) investigated the effectiveness of GRs and ‘cool’ roofs in mitigating the Urban Heat Island (UHI) comparing 86 temperatures on a single building and finding that GRs are cooler in both north and south roof orientation. Nonetheless, 87 few experimental works investigated the multi-functionality of rooftops by combining food and energy systems. Hui and 88 Chan (2011) and Nagengast et al. (2013) found that, in places where temperature is normally higher, the benefits of an 89 integrated PV - GR will be more visible. Perez et al. (2012) and Lamnatou and Chemisana (2014, 2015) explored the 90 environmental performance of different roof uses (PV-GR, PV-gravel, GR, standard built-up roof). In all above- 91 mentioned works, the results showed that the technology combining PV - GR offers multiple benefits, in particular in 92 warmer climates and in the long term. 93 Moreover, a wide variety of studies have used the life cycle assessment (LCA) methodology to compare different types 94 of rooftop uses. For example, Saiz et al. (2006) conducted a comparative LCA study between an extensive GR, a white 95 roof and a gravel roof. A similar study was conducted by Kosareo and Ries (2007), which compared extensive and 96 intensive GRs with a conventional roof. Carter and Keeler (2008) conducted two experimental studies in order to examine 97 three environmental local benefits, i.e. stormwater retention, temperature mitigation and habitat creation of GRs compared 98 to impervious surfaces. In all cases, the results verified the advantages of green roofs (e.g. for the energy savings of a 99 building) in comparison with conventional roofs. Sanyé-Mengual et al. (2015) and Sanjuan-Delmás et al.’s (2018) studies 100 on UA in Barcelona (Spain) compared the environmental performance of growing tomatoes in RTGs against conventional 101 supply chains, finding that the former can have lower life-cycle GHG emissions and toxicity impacts. A recent review by 102 Goldstein et al. (2016) found that UA is posited to have numerous advantages over conventional agriculture that will 103 supposedly result in UA's superior environmental performance. 104 To date, LCA studies have yet to compare alternative uses of building rooftops for food or energy production, accounting 105 for a variety of environmental indicators and using a systematic framework with common assumptions and boundaries 106 for the assessment of both systems. Only Benis et al. (2018) conducted a cost-benefit analysis of the simultaneous 107 production of food and energy in the Mediterranean context. Our study aims to fill this gap by analysing the strategic use 108 of rooftops in urban areas in order to provide a basis to local stakeholders and policy makers for comparing the 109 environmental advantages and disadvantages of implementing these productive uses of rooftops under Mediterranean 110 climatic conditions, such as in the city of Barcelona (Spain). The objective was to answer the following question: “If a 111 given surface of urban roof is available, which is the best option in terms of environmental impacts for solar energy 112 exploitation: food or energy production?” Indeed, Barcelona is endowed with abundance of solar energy, receiving about 113 1,660 kWh/m2/year of solar radiation per year (Perpiña Castillo et al., 2016). When an RTG is placed in the available 114 surface a given amount of food is produced on-site and the conventional production is avoided. Alternatively, if a BAPV 115 system is installed, electricity is produced, but also in this case, it is possible to account for the savings in primary energy, 116 according to the selected electric mix. In particular, this work aims to compare the environmental performances of both 117 pilot rooftop systems located at the Autonomous University of Barcelona Campus (Barcelona, Spain). The life cycle for 118 each system was assessed by means of LCA, with a special focus on those steps and hotspots that present the highest 119 environmental impacts and proposing improvement scenarios for minimizing such impacts. The novelty of this study lies 120 in the comparison of two different rooftop systems by thoroughly assessing the environmental burdens and benefits of 121 both production processes. Real data were used to evaluate the environmental loads of local food and energy production 122 and to quantify the potential benefits deriving from energy and material efficiency measures in order to optimize the 123 environmental performance of both analysed systems. 124 125 126 127 128 129 130 131 132 133 134 Table 1. Summary table of the main environmental benefits of using photovoltaic panels for energy generation and 135 rooftop greenhouses for food production. Sustainable Theme Benefits Rooftop Systems BAPV RTG Energy Produce energya ✓ - Save direct energy consumptiona ✓ - Save indirect energy consumptionb - ✓ Mitigate Urban Heat Islandb - ● Food Increase urban food securityb - ✓ Environmental Benefits Reduce product lossesa,b,c ✓ ✓ Water Prevent aquatic pollution from urban - ✓ runoffb Mitigate storm waterb - ✓ Rainwater harvesting b - ✓ Land Prevent soil erosion d - ✓ Optimize urban space b,c ✓ ✓ Reduce waste through recoverye,f ✓ ✓ Air Reduce GHG emissions g ✓ ✓ Improve air qualityg ✓ ✓ Ecology Enhance biodiversityb - ● Landscape Improve aestheticsh,i ● ● Other Improve rooftop’s performancel ● ✓ Reduce noisel - ✓ 136 137 Legend of symbols: Yes (✓), No (-), Possibly (●) 138 a) Eskew et al., 2018, b) Cerón-Palma et al., 2012, c) Sanyé-Mengual et al., 2015; d) Specht et al., 2013, e) Sanyé-Mengual et al., 2013, 139 f) Sica et al., 2018, g) Goldstein et al., 2016, Sanjuan-Delmás et al., 2018; h) Proksch, 2016, i) Prasad, 2014, l) Proksch, 2011 140 141 2. Materials and methods 142 LCA is the methodological framework used in this paper as defined by ISO standards (ISO 2006 a, b) and ILCD Handbook 143 guidelines (EC 2010, 2011). LCA is one of the main techniques for quantitatively assessing environmental impacts during 144 a product’s life cycle – from raw material extraction through material processing, manufacture, distribution, use, repair 145 and maintenance, and disposal or recycling (from ‘cradle to grave’ or ‘cradle to cradle’, according to a very common 146 definition of LCA). It identifies the most relevant environmental impacts and hotspots and can underpin decision-making 147 strategies for environmental improvements from a life cycle perspective (Baumann and Tillmann, 2004). LCA consists 148 of several interrelated steps: i) goal and scope definition; ii) inventory analysis (LCI); iii) impact assessment (LCIA) and 149 iv) interpretation of results. The same scheme is followed in this paper. 150 151 2.1. Goal and scope definition 152 The goal of this work was to quantify and compare the environmental impacts related to two types of rooftop systems, 153 namely RTG and BAPV in the Metropolitan Area of Barcelona (second largest city in Spain). It is worth clarifying that 154 the investigated roofs are pre-existent, easily accessible and did not undergo any structural modification for implementing 155 both systems. Furthermore, each roof was studied as a ‘single’ system (and not as subsystem of the building) to understand 156 its individual impacts; thereby, the results are not presented in terms of the total building performance. In order to guide 157 decision-making and help select the most suitable system, the functional unit was 1 m2 of flat rooftop using either RTG 158 or BAPV. 159 The system boundaries of the LCA, shown in Fig. 1, include all the life cycle phases, i.e. raw material extraction, 160 manufacturing processes, installation, operation/maintenance and end-of-life (EoL) (dismantling, recycling and final 161 disposal). Therefore, a ‘cradle-to-grave’ approach was adopted. Most of the past studies did not include the EoL of PV 162 technologies, mainly because of the low number of panels that reached their end of useful life and the lack of data 163 (Latunussa et al., 2016). Nevertheless, a comprehensive analysis should consider the contributions of each phase of the 164 life cycle (Fthenakis et al., 2009). During the last years, the recycling processes were investigated and developed and the 165 EoL management of PV is gaining more interest (Xu et al., 2018). Additionally, in Europe, a drive towards responsible 166 EoL management for PV panels has taken form in the Directive on Waste Electrical and Electronic Equipment (WEEE; 167 Directive 2012/19/UE of the European Parliament and the Council), according to which decommissioned PV panels are 168 included as domestic and professional types of WEEE. For this reason, the EoL step of such technology was included in 169 this study as an important step which needs to be investigated. 170 This study is aimed at providing decision-makers with potentially useful recommendations for local resources production 171 planning, without however accounting for large-scale consequences on the background system (e.g., large-scale food and 172 energy sectors, marginal changes of resource costs due to recovery, policy options etc). According to the ILCD Handbook 173 (EC, 2010), this study is centred in the proper accounting of different environmental impacts when comparing systems, 174 hence the attributional LCI modelling principle was chosen for this comparative LCA (so-called situation A). 175 176 177 178 Figure 1. System boundaries and process chain under study. 179 180 2.2. System description 181 Main assumptions and life cycle phases accounted for in the inventory of the RTG and BAPV systems under analysis are 182 presented below. 183 2.2.1. Case study: food production from rooftop greenhouse 184 The ICTA-ICP building houses the headquarters of the Institute of Environmental Science and Technology (ICTA) and 185 the Catalan Institute of Paleontology (ICP). The building is situated in the Autonomous University of Barcelona (UAB) 186 campus, 25 km away from the Barcelona city centre. It can be considered representative of office buildings in cities, 187 because it holds four floors with offices and is similar in terms of size (7,500 m2) and users to other standard buildings 188 (Schloss, 1984). Additionally, its design is based on building-integrated agriculture philosophy, multifunctionality and 189 passive systems that promote energy efficiency (Nadal et al., 2017). The pilot RTG under study, implemented by the 190 Fertilecity project (funded by the Spanish Ministry of Economy and Competitiveness - MINECO), is placed on the 191 building roof and utilises residual heat from the building, CO2 concentrations in this residual air and rainwater collected 192 from the rooftop (Sanyé-Mengual et al., 2015). More specifically, residual heat and CO2 integration are expected to 193 increase crop yields, whilst untreated rainwater is used in the RTG to irrigate the crops and water ornamental plants in 194 the building, reducing the demand for potable water from the conventional distribution network. Despite the potential 195 benefits of the RTG on the building, our study focuses on the greenhouse structure and predicts potential crop outputs 196 but, except for the rainwater collection, does not include an assessment of flow exchanges in the building due to lack of 197 data. 198 199 Characteristics of the crop 200 The RTG has a total area of 122.8 m2 and a crop area of 84.34 m2 (Fig. 2). The crops were beef tomato varieties 201 (Lycopersicon esculentum, Arawak for spring crops and Tomawak for winter crops), grown from February 2015 to July 202 2016. A hydroponic system was used for irrigation to supply a nutrient solution (water plus fertilisers, also called 203 fertigation) to plants located on an inert substrate composed of perlite bags. The system produced 30.1 kg of tomatoes per 204 square metre over 15.5 months, providing a total of 2,540 kg of food and covering the requirements for food of nearly 205 60% of the building. Further technical information about the crop can be found in Sanjuan-Delmás et al. (2018). 206 Infrastructure & Installation. The RTG’s structure mainly consists of steel, aluminium, polycarbonate covers, low-density 207 polyethylene film curtains and concrete block anchors. In addition, the installation is equipped with backup lighting, 208 rainwater harvesting systems, thermal screens, and climate control systems. The installation stage accounts for energy 209 consumption requirements of the machinery used to construct the RTG. According to previous studies, a 50 years lifespan 210 was assumed for the rainwater harvesting system (Vargas-Parra et al., 2013; Sanjuan-Delmás et al., 2015) and the 211 greenhouse structure (Sanye-Mengual et al., 2015), and a 10 years lifespan was assumed for the auxiliary equipment 212 (Hoffman et al., 2007). Transportation of materials from the market to the RTG was also included. The travelled distance 213 was 35 km for fertilisers, pesticides and auxiliary equipment, 60 km for rainwater harvesting construction materials and 214 850 km for substrate bags imported from Almeria (South of Spain). 215 216 Operation & Maintenance. The RTG operation consisted of inputs required for hydroponic cultivation (fertilizers, 217 pesticides, compost, etc.), water and energy needs. In particular, the assessment of fertilisers and pesticides included local 218 emissions to air generated during their application and the treatment of leachates in a wastewater treatment plant. 219 Furthermore, the waste biomass from the crop plants was composted in the greenhouse, thus avoiding transport and 220 landfilling, although emissions generated during the composting process were accounted for. According to Sanjuan- 221 Delmás et al. (2018), a lifespan of 3 and 5 years was assumed for the perlite bags and HDPE materials, respectively. 222 223 End-of-life. For the EoL assessment, the impacts of landfilled materials (substrate) were included. Infrastructure and 224 auxiliary equipment (pumps, rainwater tanks) were assumed to be recycled. A distance of 30 km was assumed from the 225 RTG to the landfill or the recycling facility. 226 227 228 229 230 231 232 233 234 235 236 237 238 Figure 2. Top view of rooftop greenhouse (on the left, source: 2017 © Google LLC) and tomatoes produced (on the right, source: Sanjuan-Delmás et al., 2018). 239 240 2.2.2. Case study: energy generation from Building-Applied Photovoltaics 241 Building-Applied Photovoltaics (or ‘BAPV’), in Fig. 3, is a form of on-site electricity generation that can offset the 242 emissions from more environmentally intensive sources of electricity and reduce electricity transmission losses. The 243 BAPV examined has a total nominal power of 50.49 kWp and is placed in UAB campus, on the restaurant and library 244 building’s rooftops. The PV system was implemented within the framework of univERsol, a European project developed 245 between 2002-2004, whose objective was the installation of PV panels in 26 universities, schools, technology centres and 246 city councils in four European Union countries (Spain, France, England, Holland). The total roof area used is 1,600 m2, 247 whereas the roof area covered by PV panels is 380 m2 (UAB’s personnel. Personal communication, 2017). 248 Infrastructure & Installation. The BAPV installation includes 297 single crystalline silicon (c-Si) photovoltaic panels and 249 the balance of system (BOS). Each panel has the following characteristics: 170 Wp, 72 cells, 159 x 79 cm2 dimensions, 250 tilt angle=35°, electrical efficiency=14%, weight=15.4 kg. In particular, the PV cells are enclosed in an aluminium frame 251 and sealed between two plastic sheets in ethylene vinyl acetate (EVA) foil and glued between glass and polyvinyl fluoride 252 (PVF). The system boundary of the research included upstream processes, ranging from silica extraction to the crystalline 253 silicon bar and ingot growth, and midstream processes, which involved cell and panel fabrication as well as aluminium 254 frame and BOS production. 255 Regarding the installation phase, it was modelled by considering the electricity consumption for PV infrastructure 256 installation work. The BOS components included the mounting structure (aluminium and steel), 17 inverters (necessary 257 for transforming the direct current to alternating current and for connecting to the normal electricity grid), copper and 258 plastic materials for cables and contact boxes. Components excluded from the system are the surge protector, 259 pyranometer, digital indicating controller, uninterruptible power supply device, and computer monitoring system. The 260 life expectancy of the PV panels and metal support structures were assumed to be 30 and 60 years, respectively (Peng et 261 al., 2013; Sherwani et al., 2010). Inverters and transformers were considered to last for 20 years, but parts must be replaced 262 every 10 years, according to well-established data from the power industry on transformers and electronic components 263 (Fthenakis and Kim, 2011). 264 The transportation distances were covered by a heavy truck. All the components, except for the PV panels, were assumed 265 to be purchased from factories 100 km away from Barcelona, while the PV panels were purchased from Madrid (UAB’s 266 personnel, Personal communication, 2017). 267 Operation & Maintenance. Usually, PV systems do not show any emission to air or water during operation (Alsema and 268 de Wild-Scholten, 2006; Raugei and Fthenakis, 2010; Tao and Yu, 2014, Eskew et al., 2018). Some panels might be 269 washed by the user on an annual basis. In this study, the use of 20 litres of water per year and square meter for washing 270 the panels was assumed (Frischknecht et al. 1996). Moreover, the inverters were assumed to have a 10-years lifetime, 271 thus requiring to be replaced during the 30-years lifetime of the system. The electricity produced by the BAPV system 272 amounts to 62.089 MWh/yr (UAB’s personnel, Personal communication, 2017). 273 274 End-of-life. Most materials in PV systems are reusable, including aluminium, glass, silicon or copper (IRENA, 2016; Xu 275 et al., 2018). Therefore, a recycling scenario of all recyclable materials was supposed for the EoL phase, assuming a 276 distance of 590 km for the transportation of BAPV system components to the recycling facility (located near Madrid). 277 278 279 280 281 282 283 284 285 286 287 Figure 3. Top view of roof-mounted photovoltaic (on the left, source: 2017 © 288 Google LLC) and PV panels (on the right). 289 290 291 2.3 Data sources and life cycle inventory 292 Inventory data for RTG and BAPV case studies are given in Tables 2 and 3, referred to the selected functional unit (1 m 2 293 of flat rooftop). In order to make possible a comparison between tomatoes production and PV energy production, 294 inventory data for tomato production, which referred to 15.5 months, were averaged over 12 months, taking into account 295 the variability of climate conditions along the whole year for both systems. For the inventory analysis both systems were 296 structured in several stages in order to facilitate the study and interpretation of the results obtained. Regarding the RTG 297 system, specific literature was used as data sources for the LCA. In particular, the inventory for the infrastructure 298 manufacturing, installation, operation/maintenance and transport of waste to the treatment site was deduced from Sanyé- 299 Mengual et al. (2015) and Sanjuan-Delmás et al. (2018). The inventory data about BAPV systems, including the material 300 consumption and environmental emissions involved in the production of solar-grade silicon, wafers, cells, and panels and 301 their EoL were mainly obtained from ecoinvent 3.1 database (Jungbluth et al., 2012) and literature (Corcelli et al., 2016). 302 Additionally, for the installation and operation/maintenance phases foreground data were provided by expert personnel 303 in UAB. Other background data, related to energy use, auxiliary materials and impacts of the waste management (e.g. 304 wastewater treatment, composting, EoL treatments of infrastructure materials) have been derived from the ecoinvent 3.1 305 database (Wernet et al., 2016). It should be pinpointed that the present study is representative of technologies installed in 306 the Spanish territory. As a consequence, this analysis assumes that all the production, installation, operation processes 307 and also the recycling treatments for both roofing systems are developed in Spain, thus the Spanish power mix (2015) 308 was used as a reference. The energy outputs from the BAPV system were considered to be used as alternative to energy 309 produced by means of traditional fossil fuels combustion, in particular, to the electricity from the Spanish grid, where 310 over 60% of electricity come from nuclear and fossil fuels (coal) (Red eléctrica de España, 2015), thus drawing a potential 311 evaluation of environmental benefits. Additionally, the tomatoes production from the RTG system was considered as 312 substitute for tomatoes produced by means of conventional farming in a standard multi-tunnel greenhouse. The avoided 313 cost deriving from the production of tap water for crop irrigation was also considered. Moreover, in both case studies, for 314 crediting recycled materials (metals, glass, plastics, concrete) and tomato biomass composting, the avoided production of 315 primary materials and inorganic fertilizers, respectively, was included. 316 317 Table 2. Life cycle inventory data for tomatoes production in RTG, referred to 1 m2 of flat rooftop (FU) (timeframe 1 318 year). Materials/Energy Unit/ FU Amount Data Sources Infrastructure manufacturing RTG structure: Steel kg 8.36E-01 Sanyé-Mengual et al. (2015) Concrete kg 2.12E-01 // Low Density Polyethylene (LDPE) kg 7.80E-02 // Polycarbonate kg 1.60E-01 // Polyester kg 7.80E-03 // Aluminium kg 7.80E-03 // Auxiliary equipment: Low Density Polyethylene (LDPE) kg 2.30E-02 // Polystyrene kg 2.60E-02 // High Density Polyethylene (HDPE) kg 9.40E-03 // Polyvinylchloride (PVC) kg 4.40E-03 // Steel kg 5.00E-04 // Installation* Electricity kWh 4.00E-04 Sanyé-Mengual et al. (2015) Transport, lorry tkm 3.24E-01 // Transport, transoceanic freight ship tkm 1.61E-01 // Transport, van tkm 2.00E-04 // Operation and Maintenance Substrate: Expanded perlite kg 1.87E+00 Modified from Sanjuan-Delmás et al. (2018) High Density Polyethylene (HDPE) kg 4.88E-02 // Fertilizers: KNO3 kg 3.95E-01 // KPO4H2 kg 2.00E-01 // K2SO4 kg 4.41E-01 // Ca(NO3)2 kg 5.57E-01 // CaCl2 kg 1.82E-01 // Mg(NO3)2 kg 3.72E-01 // Hortrilon/Tradecorp kg 1.49E-02 // Sequestrene kg 1.49E-02 // Pesticides kg 6.89E-02 // Water (rainwater) m3 1.30E+00 // Tap water m3 2.26E-01 // Electricity kWh 4.73E-01 // Local emissions to water: Cl- kg 8.11E-02 // NO3- kg 3.29E-01 // PO43- kg 3.64E-02 // SO42- kg 1.70E-01 // K+ kg 2.06E-01 // Mg2+ kg 2.32E-02 // Ca2+ kg 1.05E-01 // Local emissions to air: NH3 kg 6.40E-03 Estimated from Montero et al. (2009) N20 kg 2.67E-03 // NOx kg 2.13E-02 // Produced tomatoes on-site# kg 2.33E+01 Modified from Sanjuan-Delmás et al. (2018) End-of-life Steel scraps (to recycling)a kg 8.37E-01 Ecoinvent 3.1 database (Wernet al., 2016) Plastics scraps (to recycling)b kg 3.57E-01 // Aluminium scraps (to recycling)a kg 7.80E-03 // Concrete scraps (to recycling)a kg 2.12E-01 // Perlite substrate waste (to landfill) kg 1.87E+00 // Tomatoes biomass waste (to composting)c kg 9.45E+00 // Wastewater treatment m3 1.53E+00 // Transport* to landfill, lorry (perlite tkm 5.62E-02 Modified from Sanjuan-Delmás et al. (2018) substrate) Transport* to recycling facility, lorry tkm 4.24E-02 Sanyé-Mengual et al. (2015) (aluminium, steel, plastics, concrete scraps) 319 320 * Process of transport included vehicle, road manufacture and maintenance, as well as diesel consumption and relative emissions. 321 # For crediting tomatoes production on-site, the avoided environmental burden from conventional production in a standard multi-tunnel greenhouse was assumed. 322 a A substitution ratio of 0.9:1 was assumed for steel, aluminium and concrete scraps, meaning that 1 unit of secondary material replaces 0.9 unit of the corresponding primary 323 material. 324 b A substitution ratio of 1:1 was assumed for plastics scraps, meaning that 1 unit of secondary material replaces 1 unit of the corresponding primary material. 325 c The substitution ratio takes into account the quality of the products from waste in comparison with that of the corresponding avoided product. 0.4:1 means that 1 ton of 326 biomass waste corresponds to 0.4 ton of compost/dry waste (as average value). In the case of composting, 1 ton of compost was assumed to substitute:23 kg of N-fertilizer, 327 9.5 kg of P fertilizer and 9 kg of K-fertilizer (Blengini et al., 2008). 328 329 Table 3. Life cycle inventory data for energy generation from BAPV, referred to 1 m2 of flat rooftop (FU) (timeframe 1 330 year). Materials/Energy Unit/ FU Amount Data Sources Infrastructure manufacturing PV panel components: m2 2.30E-01 Ecoinvent 3.1 database (Jungbluth et al., 2012) Solar cell (c-Si) m2 2.14E-01 // Aluminium alloy kg 6.05E-01 // Polyvinyl fluoride kg 2.54E-02 // Polyethylene terephthalate kg 8.58E-02 // Glass sheet, tempered kg 2.32E+00 // Ethylene vinyl acetate kg 2.30E-01 // Copper kg 2.59E-02 // Nickel kg 3.74E-05 // Soldering flux kg 2.02E-03 // Methanol kg 4.96E-04 // Silicone kg 2.80E-02 // Corrugated board box kg 2.52E-01 // Tap water kg 4.90E+00 // Electricity kWh 1.39E+00 // Installation Balance of System: Steel kg 1.21E+00 Field data supplied by UAB’s engineers Aluminium kg 8.76E-01 // Concrete kg 1.23E+01 // Copper kg 1.14E-01 // Polyvinylchloride (PVC) kg 6.50E-02 // Inverters p 1.06E-02 // Electricity kWh 4.76E-02 // Transport*, van tkm 1.41E+00 // Transport*, lorry tkm 3.37E+00 // Operation and Maintenancea Tap water kg 1.00E+00 Field data supplied by UAB’s engineers Produced electricity on-site from PV# kWh 3.54E+01 // End-of-life Aluminium scrapsb (to recycling) kg 1.74E+00 Modified from Corcelli et al. (2016); Ecoinvent 3.1 database (Wernet al., 2016) Glass scrapsb (to recycling) kg 1.87E+00 // Silicon scrapsc (to recycling) kg 2.25E-01 // Copper scrapsb (to recycling) kg 9.96E-02 // Iron scrapsb (to recycling) kg 5.22E-06 // Steel scrapsb (to recycling) kg 1.21E+00 // Concrete scrapsb (to recycling) kg 1.23E+01 // Plastics scrapsd (to recycling) kg 6.53E-02 // Wastewater treatment kg 1.00E+00 // Transport*, lorry tkm 1.01E+01 Field data supplied by UAB’s engineers 331 * Process of transport included vehicle, road manufacture and maintenance, as well as diesel consumption and relative emissions. 332 # For crediting electricity production on-site, the avoided environmental burden from the conventional production of Spanish electricity mix was assumed. 333 a The emissions during the operation phase were considered negligible according to Alsema and de Wild-Scholten (2006) and Peng et al. (2013). 334 b A substitution ratio of 0.9:1 was assumed for aluminium, glass, steel, iron, copper and concrete scraps, meaning that 1 unit of secondary material replaces 0.9 unit of the 335 corresponding primary material. 336 c A substitution ratio of 0.95:1 was assumed for silicon scraps, meaning that 1 unit of secondary material replaces 0.95 unit of the corresponding primary material. 337 d A substitution ratio of 1:1 was assumed for plastics scraps, meaning that 1 unit of secondary material replaces 1 unit of the corresponding primary material. 338 339 340 2.4. Life Cycle Impact Assessment (LCIA) 341 The environmental assessment of the process was modelled by means of LCA software SimaPro v.8.0.5 (Pre-Consultants, 342 2014), integrated with ecoinvent v3.1 database (Wernet et al., 2016). The impact assessment was performed by means of 343 one of the most recent and up-to-date LCA methods, the ReCiPe method (Goedkoop et al., 2009; Vezzoli, 2018). The 344 ReCiPe Midpoint (H) v.1.10 (http://www.lcia-recipe.net/) was chosen, considering that it includes both upstream 345 categories (i.e. referred to depletion of natural resources, such as fossil, metal and water depletion categories) and 346 downstream categories (i.e. referred to impacts generated on natural matrices, such as terrestrial, marine or freshwater 347 acidification) (Frischknecht et al., 2007). In this study, in order to support decision makers by means of a simplified 348 overall assessment across areas of environmental concern, the following impact categories were analyzed: Climate 349 Change (CC, in kg CO2 eq), Ozone Depletion (OD, in kg CFC-11 eq), Terrestrial Acidification (TA, in kg 1,4-DB eq), 350 Freshwater Eutrophication (FE, in kg P eq), Photochemical Oxidation Formation (POF, in kg NMVOC), Terrestrial 351 Ecotoxicity (TE, in kg 1,4-DB eq), Water Depletion (WD, in m3), Metal Depletion (MD, in kg Fe eq) and Fossil Depletion 352 (FD, in kg oil eq). 353 Furthermore, a sensitivity analysis was performed to test the robustness of the results. To this end, alternative scenarios 354 were proposed, based on progressive reductions or substitution of the most sensitive input flows (both energy and material 355 flows) and the effects of these changes on final results were examined. 356 357 3. Results and Discussion 358 As a preliminary approach to the evaluation of the systems performance, the “avoided burdens”, i.e. the credits or benefits 359 resulting from the production of secondary raw materials (recycling), energy, water and biomass recovery, were not 360 included in the assessment, so as not to hide crucial steps still needing improvement (Section 3.1). However, policy 361 makers may also need additional information about potential benefits linked to avoided burdens. As a consequence, a 362 second assessment was performed including a system expansion based on average data (i.e. market mix) for crediting 363 energy and material recovery (Section 3.2). Concerning the avoided costs (observed as the negative values), 364 environmental savings of goods and energy (i.e. metals, glass, biomass, etc) were subtracted from the accounting of the 365 system’s impacts, considering that their production by means of conventional routes for later use in other processes is 366 avoided. Finally, a comparison between RTG and BAPV systems is performed in order to see which system shows the 367 best environmental performance (Section 3.3). 368 3.1. Environmental costs 369 3.1.1. RTG system 370 Table 4 shows the characterized impacts of the RTG system, with reference to the functional unit of 1 m 2 of flat rooftop, 371 broken down into the different life cycle steps with the percentage contribution to the total environmental impacts. The 372 results show that most of the environmental impacts are generated during the Operation & Maintenance phase in almost 373 all analysed impact categories, especially in ozone depletion (OD), freshwater and terrestrial eutrophication (FE, TE), 374 where the contribution from Operation & Maintenance phase ranges from 73% to 80%. In the remaining impact 375 categories, the impacts linked to Operation & Maintenance overcome 58%, except for water depletion (WD) and metal 376 depletion (MD) where the main impacts are generated by EoL (91%) and infrastructure (63%), respectively. 377 Within the Operation & Maintenance of the RTG, the use of fertilizers generates the highest impacts in five out of nine 378 impact categories (from 21% in POF to 62% in TE), whereas the use of pesticides contributes to 47% of the impacts to 379 ozone depletion (OD). Among fertilizers, calcium nitrate resulted the most impactful (data not shown) due to the large 380 amounts of chemicals (such as nitric acid) and energy (heat and electricity) required for its production. Local emissions 381 (namely, emissions from the foreground system) affect only terrestrial acidification (23%), freshwater eutrophication 382 (65%) and photochemical oxidant formation (31%) categories. To a lesser extent, the substrate bags also generate 383 substantial impacts, accounting for 12-19% of climate change (CC), ozone depletion (OD), terrestrial acidification (TA), 384 photochemical oxidant formation (POF) and fossil depletion (FD) and <3% of the other four remaining impact categories. 385 Most of these environmental impacts are generated during the production of the substrate. 386 Infrastructure manufacturing is another environmental hotspot of RTG. The RTG structure generates between 16-63% of 387 the impacts on photochemical oxidant formation (POF), ozone depletion (OD), climate change (CC), fossil depletion 388 (FD) and metal depletion (MD) and 2-10% of the impacts on the remaining categories. In particular, steel is the material 389 that generates the largest environmental load (from 48% in CC to 100% in MD), followed by polycarbonate (2-39%) 390 especially in those categories where thermoplastics tend to have the most significant impact (OD, FD). Concrete only 391 marginally affects the different indicators (<2 %). 392 Concerning the EoL phase, wastewater treatment entails the highest contribution to water depletion (corresponding to 393 90%), while the transport, metals recycling, composting and substrate landfilling steps display minor impacts (always 394 less < 11%) in all analyzed impact categories. 395 396 397 398 399 400 Table 4. Characterized impacts calculated for RTG system, broken down into different process steps, referred to a 401 functional unit of 1 m2 of flat rooftop. CC OD TA FE POF TE WD MD FD 3 kg CO2 eq kg CFC 11 kg SO2 kg P eq kg kg 1,4 - m kg Fe eq kg oil eq eq eq NMVOC DB eq Infrastructure manufacturing RTG structure 3.2E+00 5.1E-07 1.3E-02 1.3E-03 1.1E-02 2.9E-04 3.2E-02 2.5E+00 8.6E-01 [%] 19.5 18.4 10.5 7.2 15.9 6.3 2.1 63.0 25.4 Auxiliary equipment 1.7E-01 2.6E-09 5.4E-04 5.9E-06 6.9E-04 4.1E-06 3.3E-03 2.1E-03 1.1E-01 [%] 1.0 0.1 0.4 0.03 1.0 0.1 0.2 0.1 3.2 Installation Transport 2.9E-02 5.6E-09 1.3E-04 2.2E-06 1.5E-04 1.5E-05 1.0E-04 1.1E-03 1.1E-02 [%] 0.2 0.2 0.1 0.01 0.2 0.3 0.01 0.03 0.3 Electricity 1.6E-04 1.9E-11 9.9E-07 4.1E-08 5.3E-07 5.1E-09 5.5E-07 4.3E-06 3.9E-05 [%] <<1 <<1 <<1 <<1 <<1 <<1 <<1 <<1 <<1 Operation & Maintenance (tomatoes production on-site) Substrate 2.5E+00 3.6E-07 1.5E-02 5.9E-04 8.5E-03 1.4E-04 1.5E-02 5.2E-02 6.5E-01 [%] 14.9 12.8 12.5 3.2 11.8 3.0 1.0 1.3 19.4 Water 9.2E-05 1.0E-11 4.1E-07 5.2E-08 2.8E-07 1.0E-08 2.3E-04 1.4E-05 2.4E-05 [%] <<1 <<1 <<1 <<1 <<1 <<1 0.02 <<1 <<1 Fertilizer 5.6E+00 3.6E-07 3.1E-02 1.6E-03 1.5E-02 2.8E-03 7.5E-02 6.2E-01 9.8E-01 [%] 34.0 13.1 25.6 8.5 21.2 61.7 4.9 15.9 29.1 Pesticides 7.6E-01 1.3E-06 6.9E-03 3.7E-04 3.1E-03 4.8E-04 5.7E-03 8.1E-02 2.8E-01 [%] 4.6 46.7 5.6 2.0 4.4 10.6 0.4 2.1 8.2 Electricity 1.8E-01 2.3E-08 1.2E-03 4.9E-05 6.3E-04 6.0E-06 6.5E-04 5.1E-03 4.7E-02 [%] 1.1 0.8 1.0 0.3 0.9 0.1 0.04 0.1 1.4 Local emissions 8.1E-01 << 1 2.8E-02 1.2E-02 2.2E-02 << 1 << 1 << 1 << 1 [%] 4.9 << 1 23.1 65.5 30.6 << 1 << 1 << 1 << 1 End-of-life Transport 3.7E-02 6.4E-09 1.2E-04 3.3E-06 1.4E-04 1.3E-05 1.1E-04 1.9E-03 1.3E-02 [%] 0.2 0.2 0.1 0.02 0.2 0.3 0.01 0.05 0.4 Recycling metals 6.1E-01 1.0E-07 4.1E-03 7.1E-04 2.4E-03 3.7E-04 1.9E-02 3.3E-01 1.9E-01 [%] 3.7 3.7 3.3 3.9 3.3 8.2 1.3 8.6 5.7 Recycling plastic 1.47E-01 1.98E-08 8.36E-04 5.06E-05 5.08E-04 1.56E-05 1.13E- 2.24E- 4.56E- 03 03 02 [%] 0.9 0.7 0.6 0.3 0.7 0.3 0.1 0.1 0.4 Recycling concrete 1.2E-03 2.2E-10 9.7E-06 4.7E-08 1.7E-05 4.3E-08 2.7E-06 4.5E-05 4.2E-04 [%] 0.01 0.01 0.01 <<1 0.02 <<1 <<1 <<1 0.01 Tomatoes biomass 1.6E+00 2.1E-10 1.4E-02 4.5E-06 2.9E-03 6.7E-07 3.8E-03 5.7E-04 3.2E-04 waste composting [%] 9.8 0.01 11.1 0.02 4.0 0.01 0.3 0.01 0.01 Perlite substrate to 2.0E-02 5.2E-09 1.4E-04 5.7E-06 1.9E-04 1.9E-06 5.2E-04 1.5E-03 1.1E-02 landfill [%] 0.1 0.2 0.1 0.03 0.3 0.04 0.03 0.04 0.3 Wastewater treatment 8.3E-01 8.4E-08 7.1E-03 1.7E-03 4.1E-03 4.1E-04 1.4E+00 3.4E-01 1.8E-01 [%] 5.0 3.0 5.8 9.0 5.7 9.0 89.7 8.7 5.3 TOTAL 1.7E+01 2.8E-06 1.2E-01 1.8E-02 7.2E-02 4.6E-03 1.5E+00 3.9E+00 3.4E+00 [%] 100 100 100 100 100 100 100 100 100 402 403 3.1.2. BAPV system 404 Table 5 shows the characterized impacts of the BAPV system, with reference to the functional unit of 1 m2 of flat rooftop, 405 split over the different life cycle phases with the percentage contribution to the total environmental impacts. The results 406 display that the Manufacturing phase is dominant in almost all impact categories. The portion of total potential 407 environmental impact associated with the Manufacturing phase ranges from 52% in freshwater eutrophication (FE) to 408 93% in terrestrial ecotoxicity (TE), except for a share of 30% in metal depletion (MD). The cell production process 409 especially impacts on ozone depletion (OD) and terrestrial eutrophication (TE) categories, with loads corresponding to 410 41% and 89%, respectively, whilst processes of wafer and panel production contribute to a minor extent (<18%). 411 In contrast, the Installation phase only displays a significant impact on metal depletion (MD), with a share of 56%. In 412 deeper detail, each element of the balance of system (BOS) contributes differently to the Installation stage. Aluminium is 413 the material that shows the largest environmental impacts (33–57%, data not shown) in almost all investigated categories, 414 whilst freshwater eutrophication (FE) and metal depletion (MD) are mostly affected by the inverter (since it contains 415 silver, gold, and zinc). To a lesser extent, the Operation & Maintenance and EoL phases generate impacts lower than 416 11%. 417 418 Table 5. Characterized impacts calculated for BAPV system, broken down into different process steps, referred to a 419 functional unit of 1 m2 of flat rooftop. CC OD TA FE POF TE WD MD FD kg CO2 eq kg kg SO2 eq kg P eq kg kg 1,4 - m3 kg Fe eq kg oil eq CFC 11 eq NMVOC DB eq Infrastructure manufacturing Solar grade 2.7E+01 2.9E-06 1.5E-01 9.1E-03 7.9E-02 1.6E-03 1.9E+00 5.5E-01 6.9E+00 silicon [%] 35.6 17.2 32.3 21.4 24.3 1.4 71.2 2.0 33.5 Single-Si wafer 7.9E+00 6.5E-07 4.1E-02 2.9E-03 2.4E-02 6.9E-04 1.1E-01 1.3E+00 2.4E+00 [%] 10.6 3.9 9.1 6.7 7.5 0.6 4.1 4.9 11.7 Single-Si PV cell 5.3E+00 6.8E-06 3.1E-02 4.5E-03 6.1E-02 1.1E-01 2.1E-01 3.1E+00 1.1E+00 [%] 7.2 40.6 6.9 10.6 18.8 89.2 7.4 11.3 5.4 Single-Si PV 1.1E+01 1.8E-06 8.2E-02 5.6E-03 5.2E-02 1.7E-03 1.3E-01 3.1E+00 3.5E+00 panel [%] 14.5 10.8 18.2 13.3 15.9 1.4 4.5 11.5 16.7 Installation Steel 2.6E+00 1.5E-07 1.2E-02 1.9E-03 1.1E-02 3.5E-04 4.1E-02 3.6E+00 5.7E-01 [%] 3.5 0.9 2.6 4.4 3.4 0.3 1.5 13.2 2.7 Aluminium 5.7E+00 1.6E-06 3.9E-02 3.2E-03 3.1E-02 1.2E-03 6.5E-02 1.8E+00 1.9E+00 [%] 7.7 9.5 8.8 7.4 9.6 1.0 2.4 6.5 9.1 Concrete 2.2E+00 1.2E-07 6.5E-03 3.0E-04 6.5E-03 1.6E-04 2.1E-02 1.8E-01 3.1E-01 [%] 3.0 0.7 1.5 0.7 2.0 0.1 0.8 0.6 1.5 Copper 2.2E-01 2.6E-08 6.6E-03 2.7E-03 2.8E-03 2.3E-04 1.1E-02 5.0E+00 6.2E-02 [%] 0.3 0.2 1.5 6.3 0.9 0.2 0.4 18.3 0.3 PVC 1.3E-01 9.8E-10 3.6E-04 4.5E-06 6.7E-04 5.9E-06 1.2E-02 4.9E-04 7.1E-02 [%] 0.2 0.01 0.1 0.01 0.2 0.01 0.4 <<1 0.3 Transport 2.4E+00 4.2E-07 9.1E-03 1.9E-04 1.2E-02 5.5E-04 7.1E-03 1.2E-01 8.3E-01 [%] 3.2 2.5 2.0 0.5 3.6 0.5 0.3 0.4 4.0 Electricity 1.9E-01 2.3E-09 1.2E-04 4.9E-06 6.4E-05 6.1E-07 6.6E-05 5.1E-04 4.7E-03 [%] 0.3 0.01 0.03 0.01 0.02 <<1 <<1 <<1 0.23 Inverter 2.2E+00 4.5E-07 2.9E-02 7.5E-03 1.4E-02 1.2E-03 3.6E-02 4.7E+00 5.9E-01 [%] 3.0 2.7 6.6 17.6 4.1 1.0 1.3 17.2 2.8 Operation & Maintenance (electricity generation on-site) Tap water 2.4E+00 4.5E-07 1.7E-02 1.8E-03 1.0E-02 3.7E-03 8.6E-02 8.6E-01 6.5E-01 [%] 3.2 2.7 3.7 4.3 3.0 3.1 3.1 3.1 3.0 End-of-life Transport 2.2E+00 3.9E-07 7.1E-03 1.7E-04 8.5E-03 4.7E-04 6.5E-03 9.9E-02 7.8E-01 [%] 3.0 2.4 1.6 0.4 2.6 0.4 0.2 0.4 3.7 Recycling metals 3.01E+00 8.63E-07 1.77E-02 2.60E-03 1.02E-02 9.28E-04 6.71E-02 2.91E+00 8.19E-01 [%] 4.0 5.2 4.0 6.1 3.1 0.8 2.4 10.6 3.9 Recycling glass 6.4E-01 1.1E-07 4.0E-03 7.8E-05 1.8E-03 5.8E-05 1.5E-03 9.9E-03 2.2E-01 [%] 0.9 0.7 0.9 0.2 0.6 0.1 0.2 0.04 1.0 Recycling 7.1E-02 1.3E-08 5.6E-04 2.7E-06 9.6E-04 2.5E-06 1.6E-04 2.6E-03 2.4E-02 concrete [%] 0.1 0.1 0.1 0.01 0.3 <<1 <<1 0.01 0.1 Recycling plastics 4.7E-02 6.3E-09 2.7E-04 1.6E-05 1.6E-04 4.9E-06 3.6E-04 7.2E-04 1.5E-02 [%] 0.1 0.04 0.7 0.04 0.1 <<1 0.01 <<1 0.1 Energy recovery 3.4E-02 4.9E-09 2.1E-04 9.4E-06 1.2E-04 1.2E-06 1.4E-04 1.0E-03 8.6E-03 [%] 0.1 0.03 0.1 0.02 0.04 <<1 <<1 <<1 0.04 TOTAL 7.4E+01 1.7E-05 4.5E-01 4.2E-02 3.3E-01 1.2E-01 2.8E+00 2.7E+01 2.1E+01 [%] 100 100 100 100 100 100 100 100 100 420 421 3.2. Environmental benefits of circular use of resources 422 From a circular economy standpoint, the implementation of measures for energy and material efficiency in the assessed 423 systems leads to environmental benefits that can be quantified by accounting for the avoided impacts of conventional 424 production of electricity, tomatoes, tap water, fertilizers and virgin metals. The characterized results of the assessment for 425 the RTG and BAPV systems, referred to the selected functional unit (1 m2 of flat rooftop in a timeframe of 1 year), are 426 shown in Table 6 and 7, respectively. 427 In the case of RTG systems (Table 6), the environmental benefits – i.e. negative values (in bold)– deriving from material 428 efficiency are much higher than the environmental loads attributable to the structure, local emissions during operation 429 and wastewater treatment in six out of nine impact categories. In particular, the most relevant benefits are achieved thanks 430 to the tomato production on-site (e.g., 25 kg CO2 eq/m2 are saved in climate change category), whilst a smaller benefit is 431 provided by the avoided production of fertilizers (N, P, K) thanks to biomass waste composting. The use of rainwater to 432 supply water requirements of the crop provides constrained benefits apart from the water depletion (WD) category, 433 amounting to –1.59 m3 of water/m2 of rooftop. Metal recycling (aluminium and steel) shows a valuable contribution only 434 in the metal depletion (MD) category, with a saving of 0.89 kg Fe eq/m2 of rooftop. 435 The percentage contribution to the total environmental impacts of each step in the RTG is shown in Fig. 4. 436 437 Table 6. Characterized impacts calculated for the RTG system (broken down into different process steps), referred to a 438 functional unit of 1 m2 of flat rooftop. Negative values (in bold) correspond to avoided impacts thanks to energy and 439 material recovery. 440 441 442 CC OD TA FE POF TE WD MD FD 3 kg CO2 eq kg CFC 11 kg SO2 eq kg P eq kg kg 1,4 - m kg Fe eq kg oil eq eq NMVOC DB eq Infrastructure 3.4E+00 5.2E-07 1.3E-02 1.3E-03 1.2E-02 2.9E-04 3.5E-02 2.5E+00 9.6E-01 manufacturing Installation 2.9E-02 5.6E-09 1.3E-04 2.2E-06 1.5E-04 1.5E-05 1.0E-04 1.1E-03 1.1E-02 Operation & -2.5E+01 6.4E-07 -8.5E-02 3.2E-03 -1.2E-01 -8.4E-03 1.1E-02 1.9E-01 -5.6E+00 Maintenance End-of-life Transport 3.7E-02 6.4E-09 1.2E-04 3.3E-06 1.4E-04 1.3E-05 1.1E-04 1.9E-03 1.3E-02 Recycling metals -1.2E+00 -2.7E-08 -3.8E-03 -3.4E-04 -4.9E-03 2.5E-04 -5.4E-03 -8.9E-01 -1.6E-01 Recycling plastic 1.5E-01 1.9E-08 8.4E-04 5.1E-05 5.1E-04 1.6E-05 1.1E-03 2.2E-03 4.6E-02 Recycling concrete 1.2E-03 2.2E-10 9.7E-06 4.7E-08 1.7E-05 4.3E-08 2.7E-06 4.5E-05 4.2E-04 Tomatoes biomass 5.3E-01 -6.7E-08 8.1E-03 -2.1E-04 2.3E-04 -3.7E-04 -1.8E-02 -8.4E-02 -1.7E-01 waste composting Perlite substrate to 2.0E-02 5.2E-09 1.4E-04 5.7E-06 1.9E-04 1.9E-06 5.2E-04 1.5E-03 1.1E-02 landfill Wastewater 7.3E-01 7.4E-08 6.7E-03 1.6E-03 3.8E-03 4.0E-04 -1.6E+00 3.3E-01 1.5E-01 treatment 443 444 CC [kg CO2 eq] Infrastructure manufacturing OD [kg CFC-11 eq] Installation Operation & Maintenance (avoided TA [kg SO2 eq] conventional tomatoes production) Transport FE [kg P eq] Recycling metals (aluminium and POF [kg NMVOC] steel) Recycling plastics TE [kg 1,4-DB eq] Recycling concrete WD [m3] Tomatoes biomass waste composting MD [kg Fe eq] Perlite substrate to landfill FD [kg oil eq] Wastewater treatment (water recovery) -100% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100% Avoided impacts Impacts End-of-life 445 446 Figure 4. Percentage contribution of each phase to the overall environmental impacts of RTG system, referred to a 447 functional unit of 1 m2 of flat rooftop. Results include avoided impacts (negative values) due to recovery of energy and 448 material flows. 449 450 In the case of the BAPV system (Table 7), overall negative scores (in bold) can be observed for almost all the 451 environmental impact categories. This means that, except for TE, WD and MD categories, the environmental benefits 452 from electricity production on-site from renewable source (Operation & Maintenance phase) (e.g., 475 kg CO2 eq/m2 are 453 saved in climate change category) prevail on the environmental loads produced during the Manufacturing, Installation 454 and EoL phases. Metal recycling (aluminium, silicon, steel, copper) plays a minor role in lowering the environmental 455 loads as well, with a relevant share on metal depletion (MD) category only. 456 The percentage contribution to the total environmental impacts of each step in the BAPV is shown in Fig. 5. 457 458 Table 7. Characterized impacts calculated for the BAPV system (broken down into different process steps), referred to a 459 functional unit of 1 m2 of flat rooftop. Negative values correspond to avoided impacts thanks to energy and material 460 recovery. CC OD TA FE POF TE WD MD FD kg CO2 eq kg CFC 11 kg SO2 eq kg P eq kg kg 1,4 - m3 kg Fe eq kg oil eq eq NMVOC DB eq Infrastructure 5.1E+01 1.2E-05 2.9E-01 3.5E+01 2.2E-01 1.1E-01 2.4E+00 8.1E+00 1.4E+01 manufacturing Installation 1.5E+01 2.7E-06 1.0E-01 2.8E+01 7.7E-02 3.7E-03 1.9E-01 1.5E+01 4.3E+00 Operation & -4.8E+02 -5.7E-05 -3.0E+00 -1.0E+02 -1.7E+00 -1.3E-02 -1.7E+00 -9.4E+00 -1.2E+02 Maintenance End-of-life Transport 2.2E+00 3.9E-07 7.1E-03 4.3E-01 8.5E-03 4.7E-04 6.5E-03 9.9E-02 7.8E-01 Recycling metals -1.4E+01 -2.5E-06 -9.2E-02 -6.7E+00 -7.5E-02 -1.9E-03 -1.2E-01 -6.5E+00 -4.0E+00 Recycling glass -9.7E-01 -1.1E-07 -8.2E-03 -3.7E-01 -4.9E-03 -3.6E-04 -9.9E-03 -3.9E-02 -2.9E-01 Recycling concrete -2.2E+00 -1.0E-07 -5.9E-03 -4.5E-01 -5.5E-03 -1.6E-04 -2.1E-02 -1.7E-01 -2.8E-01 Recycling plastics -2.1E-01 2.5E-09 -5.7E-04 -5.2E-03 -8.3E-04 -6.2E-07 -3.6E-03 -4.1E-03 -1.7E-01 Energy recovery -6.1E-01 -3.3E-08 -1.7E-03 -4.6E-02 -6.7E-04 -3.3E-05 -1.3E-04 -6.8E-03 -2.1E-01 Solar grade silicon CC [kg CO2 eq] Single-Si wafer OD [kg CFC-11 eq] Single-Si PV cell TA [kg SO2 eq] Single- Si PV panel Installation FE [kg P eq] Operation & Maintenance (avoided POF [kg NMVOC] conventional electricity generation) Transport TE [kg 1,4-DB eq] Recycling metals (aluminium, silicon, steel, iron, copper) Recycling glass WD [m3] Recycling concrete MD [kg Fe eq] Recycling plastics FD [kg oil eq] Energy recovery -100% -80% -60% -40% -20% 0% 20% 40% 60% 80% 100% Infrastructure manufacturing Avoided impacts Impacts End-of-life 461 462 463 Figure 5. Percentage contribution of each phase to the overall environmental impacts of BAPV system, referred to a 464 functional unit of 1 m2 of flat rooftop. Results include avoided impacts (negative values) due to recovery of energy and 465 material flows. 466 3.3. Comparison RTG – BAPV systems 467 Table 8 compares the total characterized results of RTG and BAPV systems, with reference to the selected functional 468 unit. Almost all values are negative, meaning that both systems turn out to be favourable (i.e. they contribute to decreasing 469 the impacts) thanks to the production of resources on-site. Material and water recycling provide a relevant environmental 470 benefit for both roofing systems, thus confirming the efficiency of circular economy patterns. Such advantages are 471 negligible if compared with the avoided emissions produced by the substitution of fossil energy in the case of BAPV and 472 the traditional tomato supply chain in the case of RTG. 473 Except for TE, WD and MD, BAPV system shows the highest avoided impacts in comparison with RTG: for instance, 474 the impacts generated by BAPV on climate change (CC) and fossil depletion (FD) categories, corresponding to -4.3E+02 475 kg CO2 eq/m2 and -1.1+02 kg oil eq/m2, respectively (versus -2.2E+01 and -4.7E+00 in the RTG system), are around 20 476 times lower. Conversely, in the case of TE and WD, BAPV delivers environmental loads while RTG generates benefits 477 (i.e. negative impacts). Moreover, in MD, the impacts due to BAPV are higher than RTG. 478 479 Table 8. Characterized impacts calculated for RTG and BAPV systems, referred to a functional unit of 1 m 2 flat rooftop. 480 Negative values correspond to avoided impacts thanks to energy and materials efficiency. Impact category Unit/FU RTG BAPV CC kg CO2 eq -2.2E+01 -4.3E+02 OD kg CFC-11 eq 1.2E-06 -4.4E-05 TA kg SO2 eq -5.9E-02 -2.7E+00 FE kg P eq 5.7E-03 -4.9E+01 POF kg NMVOC -1.0E-01 -1.4E+00 TE kg 1,4-DB eq -7.8E-03 9.9E-02 WD m3 -1.6E+00 8.1E-01 MD kg Fe eq 2.0E+00 7.6E+00 FD kg oil eq -4.7E+00 -1.1E+02 481 482 When seeking to identify the most environmentally friendly way of using 1 m 2 of flat rooftop, the competition between 483 rooftop greenhouse farming and electricity production by means of PV technology seems to be an unsolved question. 484 Modern urban systems often import food, energy, water and other resources to fulfil essential needs, which results in the 485 emission of harmful greenhouse gases (Grewal and Grewal, 2012). In this study, throughout the production of local 486 resources, RTG and PV systems enhance the practical realization of the circular economy in cities, which might increase 487 the efficiency of the system. The choice of the Mediterranean context is very appropriate, since it is considered one of the 488 world’s best locations for solar energy use with a large exploitation potential (Girard et al., 2016). Both agri-urban and 489 PV energy systems positively affect the current metabolism of Mediterranean cities and increase their multifunctionality, 490 implementing a new symbiotic model between urban and natural systems from a circular perspective. In particular, RTG 491 systems can be easily realized in Mediterranean cities without additional heating because of their warm climate and their 492 high level of solar radiation, in contrast to northern Europe where additional heat is required thus resulting in an increased 493 energy demand. The implementation of RTG systems may also represent a means to reduce the food losses during the 494 transport and retail stages and to eliminate transport requirements, namely the main contributor to GHG emissions (Smith 495 et al., 2005). Additionally, organic waste from RTGs can be reused as fertilizer, resulting in less waste collection costs 496 for the city and reduced expenditures on synthetic fertilizers. Likewise, recirculating rainwater for food production (here, 497 tomatoes) reduces the demand for water and the costs of stormwater management. According to existing literature, among 498 the benefits coming from local food production, the satisfaction of basic food needs can be promoted only in cities located 499 in developing countries or neighbourhoods with limited food retail. In developed countries, additional social benefits 500 prevail, such as learning and education facilities for children and adult city-dwellers or bridging the gap between 501 consumers and producers (Specht et al., 2014). 502 In the case of PV installations, flat rooftops are generally acknowledged to be good spots for a solar energy system, given 503 that solar panels can be adjusted to the correct angle and the most appropriate orientation (Specht et al., 2014). Many 504 advantages of PV applications on rooftop can be obtained: they reduce dependence on fossil fuel use for electricity 505 generation, reduce the energy losses associated with transmission and distribution and do not require land for installation. 506 Nevertheless, several environmental constraints have been highlighted in association with both retrofitting uses of 507 rooftops. In the case of RTG, some concerns were expressed within the social dimension, since consumers may be 508 reluctant to use soil-less growing techniques. Furthermore, health risks might possibly derive from air pollution or 509 irrigation with improperly treated wastewater (Specht et al., 2014), although Ercilla-Montserrat et al. (2018) concluded 510 that the heavy metal content in the air of Barcelona is not a source of contamination for urban crops including high traffic 511 areas. On the other hand, the environmental impacts of PV electricity are mainly determined by the upstream emissions 512 associated with the mining and purification of raw materials and by the emissions associated with the electricity needed 513 for the manufacturing of panels (Meijer et al., 2003; Fthenakis and Kim, 2011; Chatzisideris et al., 2016). In particular, 514 within Si-based panels life cycle, the energy requirement for the purification step from the metallurgical-grade silicon 515 that is extracted from quartz to solar-grade silicon by either a silane (SiH4) or trichlorosilane (SiHCl3)-based process 516 resulted to be the main criticality (Fthenakis et al., 2008). 517 For the sake of clarity, it should be noted that the outcomes of both analysed systems are highly site-specific, depending 518 on yields which vary according to local conditions, such as solar radiation that determines the efficiency of the system. 519 Moreover, the magnitude of generated environmental loads and of attained benefits is susceptible to other factors, such 520 as the uncertainties in inventory data, the definition of system boundaries, impact assessment methods and the modelled 521 sources of electricity and heat. Although desirable, a direct comparison of the results achieved in this study with previous 522 LCA literature is hardly possible: even if the FU selected in these studies is most often 1 m 2 of rooftop, food and PV 523 energy production technologies can widely vary (PV in combination with GR or gravel, GR, building applied or building 524 integrated PV, etc). For instance, a comparison with Benis et al. (2018) would seem thinkable. However, the type and 525 lifespan of PV roofing technology, as well as the assessment of operation and EoL phases are different if compared to 526 systems analysed in this study. Therefore, generalizations are hardly possible. 527 528 3.4. Sensitivity 529 The assessment carried out in this study elucidated both pros and cons of RTGs and BAPV systems separately. In 530 particular, in accordance with previous studies, the use of fertilizers resulted to be a very impacting input flow for RTGs 531 (Sanyé-Mengual et al., 2015; Sanjuan-Delmás et al., 2018), whereas the electricity used in the production of solar panels 532 was the main hotspot in the BAPV systems (Pacca et al., 2006; Bekkelund, 2013; Eskew et al., 2018). In order to check 533 the robustness of LCA results and their sensitivity to changes in the input flows included in the study, a sensitivity analysis 534 was performed by assuming a reduction or substitution of the inputs correlated with the highest environmental loads (i.e. 535 electricity and chemicals). Therefore, two independent sensitivity analyses have been performed for the investigated RTG 536 and BAPV systems. 537 Regarding the RTG system, the first step of the sensitivity analysis consisted in a careful sensitivity check related to the 538 variability of major chemical inputs in the Operation & Maintenance phase. Indeed, as described in section 3.1.1, 539 fertilizers were the most important flow (from 5% to 62%) for all the considered impact categories, especially due to the 540 use of nitrogen fertilizers. Taking this hotspot into consideration, a sensitivity analysis was carried out on the basis of 541 three alternative scenarios for tomatoes grown in greenhouses, compared with the original tomato cultivation analysed in 542 this study (S0, base scenario): 543 1) S1– reduction (−10%) of the amounts of each fertilizer input (Torrellas et al., 2012); 544 2) S2– reduction (–20%) of the amounts of each fertilizer input (Torrellas et al., 2012); 545 3) S3 – horn meal (an organic fertilizer) was assumed to substitute calcium nitrate at a ratio 1:1, due to the relatively high 546 content of N. 547 These scenarios were developed assuming that the reduction or change of used fertilizers does not affect the crop yield, 548 thanks to more efficient technological application per area (e.g. amount of fertiliser per area of crop) rather than marginal 549 consumption per amount of production. 550 As shown in Figure 6, the decreased use of fertilizers generates a negligible decrease in the impacts on the investigated 551 categories of the Operation & Maintenance phase (WD category is not shown since it is not affected by this hotspot at 552 all). When the use of fertilizers is reduced by 20% of the original amount (S2 versus S0), impacts decrease in a range of 553 2-15% approximately, while a smaller decrease of impacts in the range 1-9% is achieved when fertilizer use decreases by 554 10% (S1 versus S0). Additional reductions of impacts could be reached by means of further optimized fertilization 555 techniques, such as drip irrigation and fertigation, that are commonly recognized to increase fertilizing efficiency (Worrell 556 et al., 1995; Kennedy et al., 2013; Solis et al., 2013) or by improving the efficiency of the background processes for 557 industrial fertilizer production. 558 100% S0 (Base scenario) 80% S1 (10% Reduction of fertilizers) 60% S2 (20% Reduction of fertilizers) 40% S3 (- Calcium nitrate + Horn meal 20% (organic fertiliser)) 0% S4 (RTG structure: - primary steel/aluminium + secondary steel/aluminium; - PCB + PET) S5 (S3 + S4 ) 559 560 Figure 6. Sensitivity analysis for changes to fertilizers and infrastructure materials input flows, referred to RTG system 561 (PCB= polycarbonate, PET= polyethylene terephthalate). 562 Since the RTG structure manufacturing was noted as a potential limitation to the implementation of RTGs due to the high 563 environmental impact (see Section 3.1.1.), the second step of the sensitivity analysis was focused on the infrastructure 564 materials. Thus, scenario S4 was designed considering both the substitution of virgin metals (aluminium and steel) with 565 secondary (i.e. recycled) metals and the substitution of polycarbonate (PC) with polyethylene terephthalate (PET) (with 566 a better environmental profile according to Franklin Associates, 2010). 567 The S4 scenario analysis showed that almost all impacts categories were sensitive to the material substitutions. A 568 reduction of impacts can be observed with respect to the reference system (S0), ranging from 5% in freshwater 569 eutrophication (FE) to 60% in metal depletion (MD). The terrestrial eutrophication (TE) is the only category that remained 570 unchanged. 571 Finally, scenarios S3 and S4 were combined in the scenario S5. The results indicate that using organic fertilizer for 572 cultivation and recycled materials for RTG infrastructure may improve the sustainability of supplying locally produced 573 food. Indeed, the adoption of both strategies at the same time significantly reduces the impacts in all investigated 574 categories, especially in climate change (CC) (32% reduction) and metal depletion (MD) (66%). 575 Concerning the BAPV system, one of the main aims of renewables is to contain the greenhouse effect. Therefore, a 576 sensitivity analysis was carried out only for the key impact category, i.e. climate change (CC). As observed in Section 577 3.1.2., during the life cycle of PV, emissions to the environment mainly occur from using electricity when producing 578 materials for solar panels (manufacturing phase). These emissions are strongly linked to the electricity mix used. Since 579 the world’s PV market is mainly dominated by China, USA, Germany and Italy (Solar Power Europe, 2018), a production 580 chain of crystalline silicon-based panels in these countries was modelled and compared with the base scenario (Table 4). 581 In addition, a comparison with the PV production chain in Sweden – leader country on renewable energy among the EU 582 Member States (Eurostat, 2018) – was included, in order to see how the use of an electricity mix with a high rate of 583 renewable sources affects the results. The life cycle inventory data of silicon production and crystalline silicon ingots, 584 wafers, cells and panels were extrapolated from ecoinvent database and modelled by changing the Spanish electricity mix 585 (base scenario) with the electricity mixes of the other six countries (referred to 2015) and by taking into account the 586 transport of the manufactured panels to Barcelona, by transoceanic freight ships in the case of China, USA and Japan, 587 and by truck in the case of European countries (Germany, Italy and Sweden). 588 Figure 7 shows the results of the comparison between the different electricity mixes applied during the manufacturing 589 phase and variation of transport distances (the Operation & Maintenance and EoL phases of the base scenario are not 590 shown in the figure since they remain unchanged). As shown in Figure 7, the c-Si panels produced in China cause about 591 40% higher GHG emissions compared with Spanish panels. The highest emissions result from the electricity mix used in 592 the Chinese industry with a high share of fossil fuel power (72%). If the panel production in Japan, USA, Germany and 593 Italy is considered, it is possible to observe GHG emissions about 15.8%, 13.6%, 14.3% and 12.9%, respectively, higher 594 compared to panels produced in Spain. In contrast, the use of electricity from renewable energy sources (e.g. in the case 595 of Sweden the share of renewable power is 63%) in the production chain reduces the GHG emissions by about 14% (- 7 596 kg CO2-eq per m2 of rooftop). The contribution from the transport step does not exceed 3.4%, except for the case of 597 Sweden (9%) due to the longer distance travelled by truck. 80 PV Manufacturing Transport 73.29 59.77 60.55 59.04 59.44 60 52.31 kg CO2 eq/ FU 44.94 40 20 0 SWEDEN SPAIN (Base ITALY GERMANY USA JAPAN CHINA scenario) 598 599 Figure 7. Sensitivity analysis for changes to electricity-mix and transport input flows, referred to BAPV system (only 600 manufacturing phase is included). 601 602 4. Conclusion 603 Sustainable urban planning is essential to effectively meet the needs of a growing population and respond to changes in 604 consumption patterns without exhausting our planet's finite resources. In this sense, the transformation of urban underused 605 rooftops into productive spaces can improve the urban metabolism by producing or collecting locally resources such as 606 energy, greening, food or water. The innovative contribution provided by this study is the comparison of two different 607 rooftop systems for resources production (i.e. food and energy) with the aim of supplying additional helpful elements for 608 promoting the circular economy at the urban scale. The adoption of the avoided burden approach allows to highlight the 609 environmental benefits deriving from the implementation of energy and material efficiency measures, especially in warm 610 Mediterranean climates, and from the concomitant avoided costs of conventional production of tomatoes and electricity 611 (with savings of 22 kg CO2 eq and 425 kg CO2 eq for RTG and BAPV respectively). The BAPV system is more 612 environmentally sound in all impact categories, except for TE, WD and MD, the latter being the only category that is not 613 advantaged by neither RTGs nor BAPVs. Even if the solar PV technology impacts produced during the material 614 manufacturing phase are high (from 52% to 93%), they are copiously balanced by the avoided impacts associated to 615 energy output during the operation phase. Nevertheless, similar advantages are achieved by means of rooftop farming. 616 The options suggested in the sensitivity analysis may provide some useful tools for improved environmental 617 performances, but cannot definitely overcome the limits of each system. 618 In the future, potentially improved performances may be gained by means of synergy effects achieved when combining 619 both food and energy production (Marucci and Cappuccini, 2016; Trypanagnostopoulos et al., 2017; Loik et al., 2017). 620 Preliminary tests have shown that PV panels work more efficiently over a green roof that cools down the cells through 621 evapotranspiration (Köhler et al., 2007). Meanwhile, the panels shade the plants, thus reducing sun exposure and 622 favouring heat-sensitive crops. Therefore, sustainable rooftop planning for urban buildings might lead to multifunctional 623 uses of the same roof throughout the integration or the coexistence of both investigated systems. 624 In conclusion, in a future prospect, attention to additional energy and material efficiency and material selection, in favour 625 of the more environmentally sustainable choice, should also remain a main point of investigation. For instance, in the 626 case of RTGs, both the greenhouse structure and the rainwater harvesting system can be reduced in size, and in the design 627 phase the amount as well as the quality of materials used for their manufacture can be optimised. Likewise, a decrease in 628 the electricity demand for manufacturing BAPV systems together with a well-designed recovery treatment would 629 contribute to a more sustainable supply chain. Additionally, economic and social assessments should be performed in 630 further research in order to provide a more nuanced contextualisation (either at country or city level) of RTG and BAPV 631 systems within these sectors. 632 633 Conflicts of interest 634 The authors declare no conflict of interests. 635 636 Acknowledgements 637 The authors Corcelli and Fiorentino gratefully acknowledge the financial support received from the European Union’s 638 Horizon 2020 research and innovation program under grant agreement No. 649342 (EUFORIE). G. Fiorentino also 639 acknowledges the research grant received from Parthenope University, project DSTE332 - Material and Energy Efficiency 640 in Chemical Processes for Industry and Environment. A. Petit-Boix thanks the Spanish Ministry of Education, Culture 641 and Sports for the grant awarded (FPU13/01273) while employed at UAB, and the German Federal Ministry of Education 642 and Research for funding the research group “Circulus - Opportunities and challenges of transition to a sustainable circular 643 bio-economy”, grant agreement No. 031B0018. 644 645 646 References 647 Ackerman, K., Plunz, R., Katz, R., Dahlgren, E., and Culligan, P., 2012. Potential for urban agriculture New York City. 648 (New York). 649 Alsema, E., de Wild-Scholten, M., 2005. Environmental impact of crystalline silicon photovoltaic module production. In: 650 Material Research Society Fall Meeting, Symposium G: Life Cycle Analysis Tools for ‘‘Green” Materials and Process 651 Selection, Boston, MA. 652 Amado, M. and Poggi, F., 2014. Solar urban planning: A parametric approach. Energy Procedia, 48, 1539–1548. 653 http://dx.doi.org/10.1016/j.egypro.2014.02.174. 654 Baumann, H. and Tillman, A.M., 2004. The hitch hiker's guide to LCA: an orientation in life cycle assessment 655 methodology and application. Studentlitteratur, Lund, Sweden 656 Benis, K. and Ferrao, P., 2016. Potential mitigation of the environmental impacts of food systems through Urban and 657 Peri-Urban Agriculture (UPA) – A Life Cycle Assessment approach. Journal of Cleaner Production, 140, 784–795. 658 http://dx.doi.org/10.1016/j. jclepro.2016.05.176. 659 Benis, K., Turan, I., Reinhart, C., Ferrão, P., 2018. Putting rooftops to use – A Cost-Benefit Analysis of food production 660 vs. energy generation under Mediterranean climates. Cities 78, 166–179. doi:10.1016/j.cities.2018.02.011 661 Bekkelund K., 2013 A Comparative Life Cycle Assessment of PV Solar Systems, Norwegian University of Science and 662 Technology. https://daim.idi.ntnu.no/masteroppgaver/010/10240/masteroppgave.pdf. 663 Blengini, G.A., Genon, G., Fantoni, M., 2008. LCA del sistema integrato di gestione dei rifiuti nella provincia di Torino 664 (Research programme financed by Servizio Pianificazione Sviluppo Sostenibile e Ciclo Integrato di rifiuti della 665 Provincia di Torino). Politecnico di Torino, Turin, Italy. 666 Broto V. C. and Bulkeley H., 2013. A survey of urban climate change experiments in 100 cities. Global Environmental 667 Change: human and policy dimensions, 23, pp. 92-102. 668 Byrne, J., Taminiau, J., Kurdgelashvili, L. and Nam, K., 2015. A review of the solar city concept and methods to assess 669 rooftop solar electric potential, with an illustrative to the city of Seoul. Renewable and Sustainable Energy Reviews, 670 41, 830–844. http://dx.doi.org/10.1016/j.rser.2014.08.023. 671 Carter, T. and Keeler A.., 2008. Life-Cycle Cost-Benefit Analysis of Extensive Vegetated Roof Systems. Journal of 672 Environmental Management 87: 350–63. doi:10.1016/j.jenvman.2007.01.024. 673 Cerón-Palma, I., Sanyé-Mengual, E., Oliver-Solà, J., Montero, J.I., Rieradevall, J., 2012. Barriers and Opportunities 674 Regarding the Implementation of Rooftop Eco.Greenhouses (RTEG) in Mediterranean Cities of Europe. J. Urban 675 Technol. 19, 87–103. doi:10.1080/10630732.2012.717685 676 Corcelli, F., Ripa, M., Leccisi, E., Cigolotti, V., Fiandra, V., Graditi, G., Sannino, L., Tammaro, M., Ulgiati, S., 2016. 677 Sustainable urban electricity supply chain - Indicators of material recovery and energy savings from crystalline silicon 678 photovoltaic panels end-of-life. Ecol. Indic. doi:10.1016/j.ecolind.2016.03.028 679 Corcelli, F., Ripa, M., Ulgiati, S., 2017. End-of-life treatment of crystalline silicon photovoltaic panels. An emergy-based 680 case study. J. Clean. Prod. 161. doi:10.1016/j.jclepro.2017.05.031 681 Doughty, M.R.C. and Hammond, G.P., 2004. Sustainability and the Built Environment at and beyond the City Scale. 682 Building and Environment 39 (10): 1223–33. doi:10.1016/j.buildenv.2004.03.008. 683 Deng, Y., Li, Z., and Quigley, J. M., 2012. Economic returns to energy-efficient investments in the housing market: 684 Evidence from Singapore. Regional Science and Urban Economics, 42(3), 506–515. 685 http://dx.doi.org/10.1016/j.regsciurbeco.2011.04.004. 686 Ercilla-Montserrat, M., Muñoz, P., Montero, J.I., Gabarrell, X., Rieradevall, J., 2018. A study on air quality and heavy 687 metals content of urban food produced in a Mediterranean city (Barcelona). J. Clean. Prod. 195, 385–395. 688 doi:10.1016/j.jclepro.2018.05.183 689 Eskew, J., Ratledge, M., Wallace, M., Gheewala, S.H., Rakkwamsuk, P., 2018. An environmental Life Cycle Assessment 690 of rooftop solar in Bangkok, Thailand. Renew. Energy 123, 781–792. doi:10.1016/j.renene.2018.02.045 691 Eurostat, 2018. Share of energy from renewable sources. Available on-line at: http://ec.europa.eu/eurostat/statistics- 692 explained/index.php/Renewable_energy_statistics#Renewable_energy_produced_in_the_EU_increased_by_two_third 693 s_in_2006-2016 (accessed June 2018). 694 European Commission (EC) - Joint Research Center - Institute of Environment and Sustainability, 2010. International 695 Reference Life Cycle Data System (ILCD) Handbook - General Guide for Life Cycle Assessment - Detailed Guidance. 696 First edition March 2010. EUR 24708 EN. Publications Office of the European Union, Luxembourg, LU. 697 European Commission (EC) - Joint Research Center - Institute of Environment and Sustainability, 2011. Supporting 698 Environmentally Sound Decisions for Waste Management – A technical guide to Life Cycle Thinking (LCT) and Life 699 Cycle Assessment (LCA) for waste experts and LCA practitioners. EUR 24916 EN. Publications Office of the European 700 Union, Luxembourg, LU. 701 European Photovoltaic Industry Association (EPIA), 2012. Solar generation 6. Solar photovoltaic electricity empowering 702 the world. 2011. Available at: https://www.greenpeace.org/archive- 703 international/Global/international/publications/climate/2011/Final%20SolarGeneration%20VI%20full%20report%20l 704 r.pdf (accessed March 2018) 705 Fthenakis, V.M., Kim, H.C., Alsema, E., 2008. Emissions from Photovoltaic Life Cycles Emissions from Photovoltaic 706 Life Cycles, vol. 42, pp. 2168e2174. https://doi.org/10.1021/es071763q. 707 Fthenakis, V.M., Kim, H.C., Held, M., Raugei, M., Krones, J., 2009. Update of PV energy payback times and life-cycle 708 greenhouse emissions. 24th Eur. Photovolt. Sol. Energy Conf. 4412. doi:10.4229/24thEUPVSEC2009-6DO.10.5 709 Fthenakis, V.M., Kim, H.C., 2011. Photovoltaics: Life-cycle analyses. Sol. Energy 85, 1609–1628. 710 doi:10.1016/j.solener.2009.10.002 711 Frischknecht, R., Editors, N.J., Althaus, H., Bauer, C., Doka, G., Dones, R., Hischier, R., Hellweg, S., Kollner, T., 712 Loerincik, Y., Margni, M., 2007. Implementation of life cycle impact assessment methods. Am. Midl. Nat. 150, 1e151. 713 Frazer, L. 2005. Paving Paradise. Environmental Health Perspectives. 113: 457-462 714 Franklin Associates, 2010. Final Report—Life Cycle inventory of 100% Postconsumer HDPE and PET recycled from 715 postconsumer containers and packaging. Prepared for The Plastics Division of the American Chemistry Council Inc., 716 the Association of Postconsumer Plastic Recyclers (APR), the National Association for PET Container Resources 717 (NAPCOR) and the PET Resin Association (PETRA). 718 Girard, A., Gago, E.J., Ordonez, J., Muneer, T., 2016. Spain’s energy outlook: a review of potential and energy export, 719 Renew. Energy 86, 703e715. 720 Goedkoop, M., Heijungs, R., Huijbregts, M., Schryver, A., De Struijs, J., Zelm, R.V., 2009. Report I: Characterisation. 721 ReCiPe A life cycle impact Assess method which comprises Harmon. Categ. Indic. Midpoint Endpoint Lev. 132. 722 Goldstein, B., Hauschild, M., Fernández, J., Birkved, M., 2016. Testing the environmental performance of urban 723 agriculture as a food supply in northern climates. J. Clean. Prod. 135, 984–994. doi:10.1016/j.jclepro.2016.07.004 724 Grewal, S. S., and Grewal, P. S., 2012. Can cities become self-reliant in food? Cities, 29(1), 1–11. 725 http://dx.doi.org/10.1016/j.cities.2011.06.003. 726 Haberman, D., Gillies, L., Canter, A., Rinner, V., Pancrazi, L., and Martellozzo, F., 2014. The potential of urban 727 agriculture in Montréal: A quantitative assessment. ISPRS International Journal of Geo-Information, 3, 1101–1117. 728 http://dx.doi.org/10.3390/ ijgi3031101. 729 Haberman, D., Gillies, L., Canter, A., Rinner, V., Pancrazi, L., and Martellozzo, F., 2014. The potential of urban 730 agriculture in Montréal: A quantitative assessment. ISPRS International Journal of Geo-Information, 3, 1101–1117. 731 http://dx.doi.org/10.3390/ ijgi3031101. 732 Heinstein, P., Ballif, C. and Perret-Aebi, L.-E., 2013. Building Integrated Photovoltaics (BIPV): Review, potentials, 733 barriers and myths. Green. http://dx.doi.org/10.1515/ green-2013-0020. 734 Hofierka, J. and Kanuk, J., 2009. Assessment of photovoltaic potential in urban areas using open-source solar radiation 735 tools. 34, 2206–2214. http://dx.doi.org/10.1016/j.renene. 2009.02.021. 736 Hoffman, G.J., Evans, R.G., Jensen, M.E., et al., 2007. Design and Operation of Farm Irrigation Systems, second ed. 737 American Society of Agricultural Engineers 738 Hui, S.C.M., Chan, S.C., 2011. Integration of Green Roof and Solar Photovoltaic Systems. Joint Symposium 2011: 739 Integrated Building Design in the New Era of Sustainability. Nov. 22, Hong Kong. 740 IRENA (International Renewable Energy Agency) and IEA-PVPS (International Energy Agency), 2016. End-of-Life 741 Management: Solar Photovoltaic Panels, International Renewable. Available online at: 742 http://www.irena.org/publications/2016/Jun/End-of-life-management-Solar-Photovoltaic-Panels (accessed February 743 2018) 744 ISO (International Organization for Standardization), 2006a. Environmental management — life cycle assessment — 745 principles and framework. Standard ISO 14040. Geneva, Switzerland. 746 ISO (International Organization for Standardization), 2006b. Environmental management — life cycle assessment — 747 requirements and guidelines. Standard ISO 14044. Geneva, Switzerland. 748 Jungbluth, N., Stucki, M., Flury, K., Frischknecht, R., Büsser, S., 2012. Life cycle Inventories of Photovoltaics. ESU- 749 Services Ltd. 750 Kennedy, T.L., Suddick, E.C., Six, J., 2013. Reduced nitrous oxide emissions and increased yields in California tomato 751 cropping systems under drip irrigation and fertigation. Agri. Ecosys. Environ. 170, 16-27. 752 Köhler, M., Wiartalla, W. and Feige, R., 2007. Interaction between PV-systems and extensive green roofs, In Proceedings 753 of the Fifth Annual Greening Rooftops for Sustainable Communities Conference, Awards and Trade Show, 754 Minneapolis, April 29-May 1, 2007. 755 Kosareo, L. and Ries, R., 2007. Comparative environmental life cycle assessment of green roofs. Building and 756 Environment, 42, 2606–2613. doi:10.1016/j.buildenv.2006.06.019 757 Loik, M. E., Carter, S. A., Alers, G., Wade, C. E., Shugar, D., Corrado, C., Jokerst, D., & Kitayama, C., 2017. Wavelength- 758 Selective Solar Photovoltaic Systems: Powering Greenhouses for Plant Growth at the Food-Energy-Water Nexus. 759 Earth’s Future, 5, 1044–1053, https://doi.org/10.1002/2016EF000531 760 Mahmoud, W.H., Elagib, N.A., Gaese, H., Heinrich, J., 2014. Rainfall conditions and rainwater harvesting potential in 761 the urban area of Khartoum. Resour. Conserv. Recycl. 91, 89–99. doi:10.1016/j.resconrec.2014.07.014 762 Marucci, A., Cappuccini, A., 2016. Dynamic photovoltaic greenhouse: Energy efficiency in clear sky conditions. Appl. 763 Energy 170, 362–376. doi:10.1016/j.apenergy.2016.02.138 764 Meijer, A., Huijbregts, M.A.J., Schermer, J.J., Reijnders, L., 2003. Life-cycle assessment of photovoltaic modules: 765 comparison of mc-Si, InGaP and InGaP/mc-Si solar modules. Prog. Photovoltaics Res. Appl. 11, 275e287. 766 https://doi.org/10.1002/pip.489. 767 Montero, J.I., Anton, A., Torrellas, M., et al., 2009. EUPHOROS deliverable 5. Report on environmental and economic 768 profile of present greenhouse production systems in Europe. In: European Comssion FP7 RDT Project Euphoros 769 (Reducing the Need for External Inputs in High Value Protected Horticultural and Ornament). 770 Nadal, A., Llorach-Massana, P., Cuerva, E., Lopez-Capel, E., Montero, J.I., Josa, A., Rieradevall, J., Royapoor, M., 2017. 771 Building-integrated rooftop greenhouses: An energy and environmental assessment in the mediterranean context. Appl. 772 Energy 187, 338–351. doi:10.1016/j.apenergy.2016.11.051 773 Orsini, F., Gasperi, D., Marchetti, L., Piovene, C., Draghetti, S., Ramazzotti, S., Gianquinto, G., 2014. Exploring the 774 production capacity of rooftop gardens (RTGs) in urban agriculture: The potential impact on food and nutrition security, 775 biodiversity and other ecosystem services in the city of Bologna. Food Security, 6(6), 781–792. 776 http://dx.doi.org/10.1007/s12571-014-0389-6. 777 Pacca S., Sivaraman D., Keoleian G.A, 2006. Life Cycle Assessment of the 33 kW Photovoltaic System on the Dana 778 Building at the University of Michigan: Thin Film Laminates, Multi-crystalline Modules, and Balance of System 779 Components, Center for Sustainable Systems, Report No. CSS05e09, University of Michigan. 780 Peng J, Lu L, Yang H. Review on life cycle assessment of energy payback and greenhouse gas emission of solar 781 photovoltaic systems. Renewable and Sustainable Energy Reviews 2013;19:255–74. 782 Perez, M.J.R., Wight, N.T., Fthenakis, V.M., Ho, C., 2012. Green-Roof Integrated PV Canopies – an Empirical Study 783 and Teaching Tool for Low Income Students in the South Bronx. May 13-18, Colorado. 784 Perpiña Castillo, C., Batista e Silva, F., Lavalle, C., 2016. An assessment of the regional potential for solar power 785 generation in EU-28. Energy Policy 88, 86–99. doi:10.1016/j.enpol.2015.10.004 786 Petit-Boix, A., Leipold, S., 2018. Circular economy in cities: Reviewing how environmental research aligns with local 787 practices. J. Clean. Prod. 195, 1270–1281. doi:10.1016/j.jclepro.2018.05.281 788 Petit-Boix, A., Devkota, J., Phillips, R., Vargas-Parra, M.V., Josa, A., Gabarrell, X., Rieradevall, J., Apul, D., 2018. Life 789 cycle and hydrologic modeling of rainwater harvesting in urban neighborhoods: Implications of urban form and water 790 demand patterns in the US and Spain. Sci. Total Environ. 621, 434–443. doi:10.1016/j.scitotenv.2017.11.206 791 Prasad, D., 2014. Snow, M. Designing with Solar Power: A Source Book for Building Integrated Photovoltaics (BiPV); 792 Routledge: Abingdon, UK 793 Proksch, G., 2011. Urban Rooftops as Productive Resources. Rooftop Farming versus Conventional Green Roofs. ARCC 794 Considering Res. Reflecting upon Curr. themes Archit. Res. 497–509. 795 Proksch G., 2016. Creating Urban Agricultural Systems: An Integrated Approach to Design Published by Taylor Francis 796 Ltd, United Kingdom 797 Raugei, M., Fthenakis, V., 2010. Cadmium flows and emissions from CdTe PV: Future expectations. Energy Policy 38, 798 5223–5228. doi:10.1016/j.enpol.2010.05.007 799 RED Eléctrica de España, Informe del Sistema Eléctrico Español, 2015. Documento resumen con nivel de accesibilidad 800 AA, Madrid, Spain. 801 Saiz, S., Kennedy, C., Bass, B. and Pressnail, K., 2006. Comparative Life Cycle Assessment of Standard and Green Roofs. 802 Environmental Science and Technology, 40(13), 4312–4316. 803 Santos, Í.P. Dos, Rüther, R., 2012. The potential of building-integrated (BIPV) and building-applied photovoltaics 804 (BAPV) in single-family, urban residences at low latitudes in Brazil. Energy Build. 50, 290–297. 805 doi:10.1016/j.enbuild.2012.03.052 806 Sanjuan-Delmás, D., Llorach-Massana, P., Nadal, A., Ercilla-Montserrat, M., Muñoz, P., Montero, J.I., Josa, A., 807 Gabarrell, X., Rieradevall, J., 2018. Environmental assessment of an integrated rooftop greenhouse for food production 808 in cities. J. Clean. Prod. 177, 326–337. doi:10.1016/j.jclepro.2017.12.147 809 Sanyé-Mengual, E., Cerón-Palma, I., Oliver-Solá, J., Montero, J.I., Rieradevall, J., 2013. Environmental analysis of the 810 logistics of agricultural products from roof top greenhouses in mediterranean urban areas. J. Sci. Food Agric. 93, 100– 811 109. doi:10.1002/jsfa.5736 812 Sanyé-Mengual, E., Oliver-Solà, J., Montero, J.I., Rieradevall, J., 2015. An environmental and economic life cycle 813 assessment of rooftop greenhouse (RTG) implementation in Barcelona, Spain. Assessing new forms of urban agriculture 814 from the greenhouse structure to the final product level. Int. J. Life Cycle Assess. 20, 350–366. doi:10.1007/s11367- 815 014-0836-9 816 Schloss, N., 1984. Technical note: use of employment data to estimate office demand, Monthly Labor Review, Vol. 107 817 No. 12, pp. 40-44. 818 Sica, D., Malandrino, O., Supino, S., Testa, M., Lucchetti, M.C., 2018. Management of end-of-life photovoltaic panels 819 as a step towards a circular economy. Renew. Sustain. Energy Rev. 82, 2934–2945. doi:10.1016/j.rser.2017.10.039 820 Sherwani, F., Usmani, J., Varun, 2010. Life cycle assessment of solar PV based electricity generation systems: A review. 821 Renew. Sustain. Energy Rev. 14, 540–544. doi:10.1016/j.rser.2009.08.003 822 Smith, A., Watkiss, P., Tweddle, G., McKinnon, A., Browne, M., Hunt, A., et al., 2005. The validity of food miles as an 823 indicator of sustainable development. Oxon, UK: Defra. ED50254, -103. 824 Solar Power Europe, 2018. Global Market Outlook for Solar Power / 2018-2022. Available on-line at: 825 http://www.solarpowereurope.org/global-market-outlook-2018-2022/ (accessed June 2018). 826 Solis, A., Vidal, I., Paulino, L., Johnson, B.L., Berti, M.T., 2013. Camelina seed yield response to nitrogen, sulfur, and 827 phosphorus fertilizer in South Central Chile. Ind. Crop. Prod. 44, 132-138. 828 Specht, K., Siebert, R., Hartmann, I., Freisinger, U.B., Sawicka, M., Werner, A., Thomaier, S., Henckel, D., Walk, H., 829 Dierich, A., 2014. Urban agriculture of the future: An overview of sustainability aspects of food production in and on 830 buildings. Agric. Human Values 31, 33–51. doi:10.1007/s10460-013-9448-4 831 Tao, J., Yu, S.R., 2014. Review on feasible recycling pathways and technologies of solar photovoltaic modules. Sol. 832 Energy Mater. Sol. Cells 141, 108–124. 833 Torrellas, M., Antón, A., López, J.C., Baeza, E.J., Montero, J.I., 2012. LCA of a tomato crop in a multi-tunnel greenhouse 834 in Almeria. Int J Life Cycle Assess 17, 863–875. doi:10.1007/s11367-012-0409-8 835 Trypanagnostopoulos, G., Kavga, A., Souliotis, Tripanagnostopoulos, Y., 2017. Greenhouse performance results for 836 roof installed photovoltaics. Renew. Energy 111, 724–731. doi:10.1016/j.renene.2017.04.066 837 United Nations, Department of Economic and Social Affairs, Population Division, 2014. World Urbanization Prospects: 838 The 2014 Revision, Highlights (ST/ESA/SER.A/352) Available on-line at: 839 https://www.compassion.com/multimedia/world-urbanization-prospects.pdf (accessed June 2018). 840 UN-Habitat, 2016. World Cities Report 2016: Urbanization and Development – Emerging Futures. Available on-line at: 841 https://unhabitat.org/books/world-cities-report/ (accessed July 2018). 842 Vargas-Parra, M.V., Villalba, G., Gabarrell, X., 2013. Applying exergy analysis to rainwater harvesting systems to assess 843 resource. Resour. Conserv. Recycl. 72,50-59 844 Wernet, G., Bauer, C., Steubing, B., Reinhard, J., Moreno-Ruiz, E., Weidema, B., 2016. The ecoinvent database version 845 3 (part I): overview and methodology. Int. J. Life Cycle Assess. 21, 1218-1230. 846 Wielemaker, R.C., Weijma, J., Zeeman, G., 2018. Harvest to harvest: Recovering nutrients with New Sanitation systems 847 for reuse in Urban Agriculture. Resour. Conserv. Recycl. 128, 426–437. doi:10.1016/j.resconrec.2016.09.015 848 Worrell, E., Meuleman, B., Blok, K., 1995. Energy savings by efficient application of fertilizer. Resour. Conserv. Recycl. 849 13, 233-250. 850 Xu, Y., Li, J., Tan, Q., Peters, A.L., Yang, C., 2018. Global status of recycling waste solar panels: A review. Waste 851 Manag. 75, 450–458. doi:10.1016/j.wasman.2018.01.036 852 Yang, R.J., Zou, P.X.W., 2016. Building integrated photovoltaics (BIPV): Costs, benefits, risks, barriers and improvement 853 strategy. Int. J. Constr. Manag. 16, 39–53. doi:10.1080/15623599.2015.1117709