Environmental Impact of Fresh Tomato Production in an Urban Rooftop Greenhouse in a Humid Continental Climate in South Korea †

: In this work, we used life cycle assessment (LCA) to determine the environmental impact of fresh tomato production using a conventional greenhouse (GH) located in a rural area versus a rooftop greenhouse (RTG) located in an urban area in South Korea. The heating and cooling loads were modeled for a period of 12 months using the simulation software TRNSYS. The comparative LCA was then performed for the GH and RTG using these data. It was found that 19% less energy is required for heating an RTG and 38% more energy is used for cooling compared with a GH. Nevertheless, the total energy load reduction for the RTG is 13%. This decreased energy consumption is due to smaller heat losses of the RTG during the colder months. The decreased energy load, combined with the elimination of transportation, storage, and handling losses during the distribution stage, resulted in 43% less global warming potential, 45% less cumulative energy demand and abiotic depletion, 37% less photochemical oxidation and acidiﬁcation, and 27% less eutrophication for the RTG. Further studies with seasonal yield data, energy sources, and integrated energy ﬂows are expected to provide a better understanding of the advantages of urban farming in this region.


Introduction
In view of the sheer expansion of the population and increasing living standards across the planet, our food systems are facing intensification across the entire supply chain [1][2][3]. One of the most intensified food systems is the agricultural sector. This sector is experiencing several changes that are leading to unintended consequences such as forest loss, increase in greenhouse gas emissions, soil degradation, groundwater depletion, and eutrophication, among others [4]. These environmental impacts, linked to the intensified production of food, have been the focus of numerous studies in recent decades. Therefore, a few methodologies have been developed in order to assess the environmental impact that we have on the environment. One of the most successful and widely used in recent years is the life cycle assessment (LCA). A life cycle assessment is defined as a "compilation and evaluation of the inputs, outputs and the potential environmental impacts of a product throughout its lifecycle" by the International Organization for Standardization [5]. An LCA has four stages: (a) definition of goal and scope, (b) inventory analysis, (c) impact assessment, and (d) interpretation of the results obtained from each stage. The original application of the LCA methodology was in the industrial

Methodology Details
Using LCA, it is possible to evaluate and compare the environmental impact of a product or system. In this case, the relative intensity of the energy demand, resources, consumption, and pollutants of the two systems were evaluated. This section outlines the tomato production process, the details of the building energy simulation, and the LCA methodology details.

Fresh Tomato Production Process Details
Fresh tomato production takes place mostly in greenhouses in soilless culture because of the high yield compared with open field production in soil [13]. In this LCA, tomato is produced in greenhouses using different covering materials. For the RTG system, the covering is glass, while double ethylene vinyl acetate (EVA) copolymer sheet is used as the covering material in the GH system. The heating system considered for both systems is a gas boiler system. The modeled production system and its boundary are shown in Figure 1. The production system is divided into infrastructure (greenhouse structure, irrigation, and climate control), production (seedlings, fertilizers, pesticides, water, and transport), energy (heating and cooling), and distribution (packaging, distribution, and losses). Both systems are considered soilless and use rockwool as growing media.
(packaging, distribution, and losses). Both systems are considered soilless and use rockwool as growing media.

Building Energy Simulation: Load Calculation Details.
A building load simulation was performed for both the GH and RTG. The simulation was modeled using TRNSYS [14], which is a transient system simulation program. Both greenhouses are Venlo type with a total area of 1690 m 2 and volume of 7815 m 3 . However, the GH covering is made of EVA copolymer (U-value = 5.62 W m −2 K −1 ), while the covering for the RTG is made of 16 mm double wall tempered glass (U-value = 3.25 W m −2 K −1 ). Another difference is the floor layer; in the case of the GH, the ground floor has a U-value of 0.378 W m −2 K −1 , while the in RTG, the U-value of the floor (or in this case, the adjacent ceiling) is 0.261 W m −2 K −1 . The model contains other data regarding the orientation of the structures, materials, ground temperature, convective coefficients, window frame factor, local weather data, and initial values. Utilizing the typical reference year (TRY), we generated monthly data of 12 specific years and integrated them into the hourly data in this model. The optimal growing temperature for tomatoes ranges from 20 to 25 °C, with a minimum of 5 °C and maximum of 32 °C. Therefore, the following heating temperature program was implemented: during daytime: 22-25 °C, during nighttime: 12 to 15 °C. Cooling was operated as follows: during daytime: 27-30 °C, during nighttime: 24 to 27 °C. Thermal curtains and shading curtains were closed when the global solar radiation was below 5 W m −2 and above 800 W m −2 , respectively. Forced ventilation is activated when the internal temperature is above 26 °C during intense solar radiation (09:00 to 20:00). In the case of the RTG, the simulation does not consider the exchange of energy flows and only considers the initial temperature of the floor at the rooftop. The rated thermal power of the heater was selected as 95 kW. The maximum air flow rate of the blower in the heater was 225 m 3 min −1 , and the boiler efficiency was assumed to be 80%. An energy balance

Building Energy Simulation: Load Calculation Details.
A building load simulation was performed for both the GH and RTG. The simulation was modeled using TRNSYS [14], which is a transient system simulation program. Both greenhouses are Venlo type with a total area of 1690 m 2 and volume of 7815 m 3 . However, the GH covering is made of EVA copolymer (U-value = 5.62 W m −2 K −1 ), while the covering for the RTG is made of 16 mm double wall tempered glass (U-value = 3.25 W m −2 K −1 ). Another difference is the floor layer; in the case of the GH, the ground floor has a U-value of 0.378 W m −2 K −1 , while the in RTG, the U-value of the floor (or in this case, the adjacent ceiling) is 0.261 W m −2 K −1 . The model contains other data regarding the orientation of the structures, materials, ground temperature, convective coefficients, window frame factor, local weather data, and initial values. Utilizing the typical reference year (TRY), we generated monthly data of 12 specific years and integrated them into the hourly data in this model. The optimal growing temperature for tomatoes ranges from 20 to 25 • C, with a minimum of 5 • C and maximum of 32 • C. Therefore, the following heating temperature program was implemented: during daytime: 22-25 • C, during nighttime: 12 to 15 • C. Cooling was operated as follows: during daytime: 27-30 • C, during nighttime: 24 to 27 • C. Thermal curtains and shading curtains were closed when the global solar radiation was below 5 W m −2 and above 800 W m −2 , respectively. Forced ventilation is activated when the internal temperature is above 26 • C during intense solar radiation (09:00 to 20:00). In the case of the RTG, the simulation does not consider the exchange of energy flows and only considers the initial temperature of the floor at the rooftop. The rated thermal power of the heater was selected as 95 kW. The maximum air flow rate of the blower in the heater was 225 m 3 min −1 , and the boiler efficiency was assumed to be 80%. An energy balance based on this model results in both the heating and cooling loads. These results were validated with climate data from an existing conventional greenhouse located  (Figure 2a,b). The details of the simulation and validation of the results can be found in [15]. The building load analysis results are shown in Figures 3 and 4.       . Annual heating (January to April and September to December) and cooling (May to August) load simulation in a conventional greenhouse [7].
Sustainability 2020, 12, x FOR PEER REVIEW 4 of 14 based on this model results in both the heating and cooling loads. These results were validated with climate data from an existing conventional greenhouse located in a rural area of northeast South Korea (37°76′ N, 126°77′ E) and the RTG results with a recently built RTG at the Korea Institute of Machinery and Materials (36°23′ N, 127°21′ E) (Figure 2a,b). The details of the simulation and validation of the results can be found in [15].    The simulation shows that the annual heating load in the case of the RTG is 19% smaller compared with the GH. This difference is driven by the weather conditions in the respective locations, the covering materials, and the floor insulation. However, the cooling load in the RTG is higher by 38% in comparison. Despite the large increment in cooling load, the absolute values of energy requirements are much higher in the heating operation than in the cooling operation. The total energy load of the RTG is 13% lower than that of the GH.

LCA Methodology: Goal, Scope, and Functional Unit
In this study, we followed the international organization for standardization (ISO) guidelines for the implementation of LCA [5,16], which consist of four steps: (1) definition of goal and scope, (2) a life cycle inventory (LCI), (3) a life cycle impact assessment (LCIA), and (4) life cycle interpretation. The goal of this study was to compare the environmental performance of two fresh tomato production systems with similar characteristics in a humid continental climate such as that found in South Korea. The first system is a representative conventional greenhouse (GH) of the Venlo type located in a rural area 315 km away from Seoul in South Korea. The second system is a rooftop greenhouse (RTG) located in an urban area, in this case, in Seoul, South Korea. CML-IA baseline V3.05/World 2000 was the chosen impact assessment methodology for this study. CML is a widely used LCIA method in agriculture [17,18] and a recommended method in the International Reference Life Cycle Data System handbook [19]. To compare the two systems, five impact categories were selected. These impact categories are the most commonly used in agriculture and reflect the general environmental performance of the system [20]. The impact categories are abiotic depletion (AD), global warming potential (GWP), photochemical oxidation (PO), acidification (AP), and eutrophication (EU). In addition, the cumulative energy demand (CED V1.10) was also obtained. The list of impact categories and units is shown in Table 1. The functional unit (FU), related to the inputs and assessed impacts, was defined as the production of 1 kg of fresh tomato retailed for consumption at the local market in the case of the GH. Regarding the RTG, it is considered an idealized scenario in which local consumers arrive at the RTG to directly acquire the produce; this removes the need for extra packaging and transportation.

Details of the Life Cycle Inventory
The data for greenhouse infrastructure, production, and distribution inputs were taken from the ECOINVENT 3.4 database [21]. Energy requirements were obtained from the building energy simulation. The complete inventory list can be found in Appendix A.

Infrastructure Inventory Inputs
The GH infrastructure data correspond to a double-layer EVA copolymer film greenhouse made of galvanized steel. The structure lifetime is considered to be 25 years. This is a commercial greenhouse that includes heating and cooling systems, a fertigation system, and a CO 2 injection system. The structure contains all the emissions from fabrication to dismantling. It also includes the end-of-life activities such as recycling and disposal of materials. The RTG infrastructure data correspond to a structure of galvanized steel and aluminum covered with glass plates. This dataset was modified with actual data of a recently built rooftop greenhouse in order to reflect the extra materials needed to Sustainability 2020, 12, 9029 6 of 13 support this structure such as concrete, steel, and glass. The expected lifetime of this greenhouse is 50 years.

Production Inventory Inputs
For the agricultural inputs, the same dataset was used in both systems and corresponds to an annual production of 48.3 kg m −2 . It also includes fertilization with 0.1025/0/0.0851 kg m −2 mineral nitrogen, phosphorus, potassium (NPK) and pesticide usage of kg m −2 . Seedlings and rockwool as the growing media as well as transportation and waste management were taken into consideration.

Energy Inventory Inputs
The heating demand based on the simulation corresponds to 437 and 367 MJ per kg of tomato per year for the GH and RTG, respectively. A boiler using natural gas was assumed for both the RTG and the GH. Cooling using electrical energy was considered for both cases. The calculated values for energy per kg per year were 7.95 kWh for the GH and 12.88 kWh for the RTG. This electricity consumption was taken from the electricity mix for South Korea of 2014.

Distribution Inventory Inputs
Finally, in the case of the GH, the distribution stage was modeled considering the use of plastic and cardboard packaging. Because of the location of the GH, transportation from the farm to a distribution center and from the distribution center to retail as well as electricity for cooling during storage at the distribution center were considered. In addition, because of the manipulation of the product at the different transportation stages and at the distribution center, we considered a 10% of loss in the case of the GH system as in [22]. As previously stated, in this study, it was assumed that consumers come directly to the RTG to pick up the produce, eliminating the need for extra packaging, transportation, cooling in a distribution center, transportation to the market, and losses due to product handling during all these stages.

Life Cycle Impact Assessment
Using the software SimaPro (ver. 8.5.2), developed by PRé Sustainability B.V., Amersfoort, The Netherlands [23], all the datasets from ECOINVENT and modeled processes were translated into the chosen environmental impacts defined in Section 2.3. Tables 2 and 3 show the results for the GH and RTG, respectively. These results are the absolute values of the impact categories for each of the system processes: infrastructure, production, energy, and distribution.  For easier visualization, the results are plotted as a percentage of the contribution of each process to the impact categories (Figures 5-7). Here, it is possible to see which system process has more influence over the results in the impact categories. Figure 5 shows a direct comparison of the two systems in each of the impact categories.  For easier visualization, the results are plotted as a percentage of the contribution of each process to the impact categories (Figures 5-7). Here, it is possible to see which system process has more influence over the results in the impact categories. Figure 5 shows a direct comparison of the two systems in each of the impact categories.    For easier visualization, the results are plotted as a percentage of the contribution of each process to the impact categories (Figures 5-7). Here, it is possible to see which system process has more influence over the results in the impact categories. Figure 5 shows a direct comparison of the two systems in each of the impact categories.  Contribution to the different impact categories for production of tomato in a conventional greenhouse. Impact categories: AD, abiotic depletion; GWP, global warming potential; PO, photochemical oxidation; AP, acidification potential; EU, eutrophication; CED, cumulative energy demand.

Life Cycle Interpretation
Comparing the total impact in each category of both systems, it can be observed that the RTG has a favorable position in most categories, except in production, where both systems have the same input values. Considering the total value from all of the processes, the RTG has better performance; for example, there is 43% less GWP and 45% less CED compared with the GH. Abiotic depletion also has a high reduction of 45%. Photochemical oxidation and acidification are both reduced by 37%, and eutrophication decreases by 27%. However, energy consumption is not the only reason for this reduction. Regarding infrastructure, the RTG has a higher impact in most categories. For instance, the RTG has 200% higher GWP and PO compared with the GH. The drivers of this large difference are the use of glass in the case of the RTG, as well as the larger amount of steel needed for its construction. The difference is because stiffer supports and steel bars are needed to safely operate the greenhouse on the rooftop of a building. Despite the 50-year lifespan of the RTG, the humidity and weather conditions will probably result in more frequent and difficult maintenance procedures being required than compared with the GH, which has only a 25-year lifespan. Figure 8 shows the GWP comparison per process. From the energy viewpoint, the difference is already substantial between the two systems based on the building energy simulation. The RTG requires 19% less heat compared with the GH, but 38% Contribution to the different impact categories for production of tomato in a rooftop greenhouse. Impact categories: AD, abiotic depletion; GWP, global warming potential; PO, photochemical oxidation; AP, acidification potential; EU, eutrophication; CED, cumulative energy demand.

Life Cycle Interpretation
Comparing the total impact in each category of both systems, it can be observed that the RTG has a favorable position in most categories, except in production, where both systems have the same input values. Considering the total value from all of the processes, the RTG has better performance; for example, there is 43% less GWP and 45% less CED compared with the GH. Abiotic depletion also has a high reduction of 45%. Photochemical oxidation and acidification are both reduced by 37%, and eutrophication decreases by 27%. However, energy consumption is not the only reason for this reduction. Regarding infrastructure, the RTG has a higher impact in most categories. For instance, the RTG has 200% higher GWP and PO compared with the GH. The drivers of this large difference are the use of glass in the case of the RTG, as well as the larger amount of steel needed for its construction. The difference is because stiffer supports and steel bars are needed to safely operate the greenhouse on the rooftop of a building. Despite the 50-year lifespan of the RTG, the humidity and weather conditions will probably result in more frequent and difficult maintenance procedures being required than compared with the GH, which has only a 25-year lifespan. Figure 8 shows the GWP comparison per process.

Life Cycle Interpretation
Comparing the total impact in each category of both systems, it can be observed that the RTG has a favorable position in most categories, except in production, where both systems have the same input values. Considering the total value from all of the processes, the RTG has better performance; for example, there is 43% less GWP and 45% less CED compared with the GH. Abiotic depletion also has a high reduction of 45%. Photochemical oxidation and acidification are both reduced by 37%, and eutrophication decreases by 27%. However, energy consumption is not the only reason for this reduction. Regarding infrastructure, the RTG has a higher impact in most categories. For instance, the RTG has 200% higher GWP and PO compared with the GH. The drivers of this large difference are the use of glass in the case of the RTG, as well as the larger amount of steel needed for its construction. The difference is because stiffer supports and steel bars are needed to safely operate the greenhouse on the rooftop of a building. Despite the 50-year lifespan of the RTG, the humidity and weather conditions will probably result in more frequent and difficult maintenance procedures being required than compared with the GH, which has only a 25-year lifespan. Figure 8 shows the GWP comparison per process. From the energy viewpoint, the difference is already substantial between the two systems based on the building energy simulation. The RTG requires 19% less heat compared with the GH, but 38% From the energy viewpoint, the difference is already substantial between the two systems based on the building energy simulation. The RTG requires 19% less heat compared with the GH, but 38% Sustainability 2020, 12, 9029 9 of 13 more energy for cooling. The total energy is 13% lower for the RTG. This translates into 5% less GWP, 6% less PO, and 7% less CED. The heat losses in the cold months are smaller in the RTG compared with the GH. These favorable environmental conditions could be enhanced in a co-generation scenario or if energy flows are exchanged between the building and the greenhouse.
Distribution represents another area where there is a large reduction in the impacts of the RTG since transportation from the farm to the consumer is no longer needed. In addition, refrigeration and losses occurring during product handling are avoided. Figure 9 shows the contribution of each component of the distribution process to the impact categories. These include packaging, cardboard, loses, transportation, and cooling. Transportation includes the following: farm to warehouse, warehouse to distribution center (DC) and DC to retail. It can be observed that the cardboard utilized for packaging the product represents a major proportion in all categories.
Sustainability 2020, 12, x FOR PEER REVIEW 9 of 14 more energy for cooling. The total energy is 13% lower for the RTG. This translates into 5% less GWP, 6% less PO, and 7% less CED. The heat losses in the cold months are smaller in the RTG compared with the GH. These favorable environmental conditions could be enhanced in a co-generation scenario or if energy flows are exchanged between the building and the greenhouse. Distribution represents another area where there is a large reduction in the impacts of the RTG since transportation from the farm to the consumer is no longer needed. In addition, refrigeration and losses occurring during product handling are avoided. Figure 9 shows the contribution of each component of the distribution process to the impact categories. These include packaging, cardboard, loses, transportation, and cooling. Transportation includes the following: farm to warehouse, warehouse to distribution center (DC) and DC to retail. It can be observed that the cardboard utilized for packaging the product represents a major proportion in all categories.

Further Considerations
Rooftop greenhouses are still in development, and much work is still needed for their integration into the food chain. However, the benefits for large cities are vast in terms of environmental performance. This preliminary study provides a detailed approximation of the environmental impacts of an urban greenhouse focused on the production of fresh tomato. Further studies with seasonal yield data, various heating/cooling systems, and building exchange flow scenarios will help us determine the best possible conditions for urban rooftop farming in the near future.

Conclusions
In this work, it was estimated that 19% less energy is required for heating a rooftop greenhouse (RTG) located in an urban area compared with a conventional greenhouse (GH) located in a suburban area. However, 38% more energy is used for cooling. Nevertheless, the total energy load reduction for the RTG is 13%. This decreased energy consumption is due to smaller heat losses of the RTG during the colder months. The obtained savings represent 5% less GWP for the RTG. All other impact categories were also comparatively reduced. For instance, abiotic depletion decreases by 45%, photochemical oxidation and acidification are both reduced by 37%, and eutrophication decreases by 27%. In addition, other inherent advantages of urban farming are all associated with a large reduction

Further Considerations
Rooftop greenhouses are still in development, and much work is still needed for their integration into the food chain. However, the benefits for large cities are vast in terms of environmental performance. This preliminary study provides a detailed approximation of the environmental impacts of an urban greenhouse focused on the production of fresh tomato. Further studies with seasonal yield data, various heating/cooling systems, and building exchange flow scenarios will help us determine the best possible conditions for urban rooftop farming in the near future.

Conclusions
In this work, it was estimated that 19% less energy is required for heating a rooftop greenhouse (RTG) located in an urban area compared with a conventional greenhouse (GH) located in a suburban area. However, 38% more energy is used for cooling. Nevertheless, the total energy load reduction for the RTG is 13%. This decreased energy consumption is due to smaller heat losses of the RTG during the colder months. The obtained savings represent 5% less GWP for the RTG. All other impact categories were also comparatively reduced. For instance, abiotic depletion decreases by 45%, photochemical oxidation and acidification are both reduced by 37%, and eutrophication decreases by 27%. In addition, other inherent advantages of urban farming are all associated with a large reduction in the environmental impacts. These improvements in the environmental performance are due to a reduction in losses due to product handling, no requirement for transportation from the farm to the consumer, and reduction in packaging. Further analysis using seasonal yield data and scenarios such as different heating systems (co-generation or different fuel sources) as well as the full integration of energy flows between the RTG and the building will be necessary to determine the full range of environmental advantages associated with urban farming.     Urea, as N kg 5.5 × 10 −5 5.5 × 10 −5