Life cycle assessment of greenhouse tomatoes for the Swedish market

The food supply chain is responsible for a large share of the anthropogenic contribution to global warming, as well as being a major contributor to several other impact categories such as acidification and eutrophication. Therefore, it is necessary to find ways of limiting the impact from food production and the food supply chain. Many crops are not adapted to growing in regions with cold climate, which creates the need to either import them or to use production methods such as greenhouses to artificially create good conditions for the crops. Sweden is currently reliant on imports for many different crops, including tomatoes where most of the consumption is covered by import from the Netherlands. This study uses life cycle assessment to analyze the potential environmental impact of Swedish tomato consumption, by comparing several year-round domestic production scenarios with scenarios representing import from the Netherlands. This is done by using a green-house simulation software to simulate a theoretical greenhouse placed in both countries, and then using the simulation results in combination with data from the database EcoInvent to perform a life cycle assessment. The results showed that Swedish domestic production has the potential to decrease the environmental impact of tomatoes consumed in Sweden, when compared to import from the Netherlands. There were a couple of combinations of production scenarios and impact categories where the Dutch production performed better, but the Swedish production scenarios performed better in general. The results also clearly showed that scenarios using LED lighting systems consistently had a lower impact than similar production scenarios using high-pressure sodium lighting systems. The choice of energy sources was identified as a crucial factor when it comes to the environmental impact of the studied systems.


Introduction
The anthropogenic emissions of carbon dioxide need to decrease if humanity is going to reach the goal of limiting the global warming to below 1.5 • C (IPCC, 2018).Research has shown that the food supply chain is currently responsible for 26 % of the anthropogenic contribution to climate change (Poore and Nemecek, 2018), while also being a major contributor to other environmental impact categories such as acidification and eutrophication (Poore and Nemecek, 2018).This creates a need to find ways of limiting the impact from food production and the food supply chain.
Looking at the production of crops, much food for human consumption is produced by open cultivation where the crops are cultivated in open fields.This method has the advantage of being easily adopted, but it also requires large amounts of land and must deal with natural fluctuations such as weather.Due to environmental factors such as temperature and humidity, this method of farming is not possible to use for all crops and/or in all locations.An alternative to open cultivation is to use greenhouses, which are a common way of producing high amounts of food in a limited amount of space.It also enables city-near production of food and has the advantage of allowing cultivation of food products that would not have grown well in the environment outside of the greenhouse.
Many fruits and vegetables are not adapted to grow in regions with colder climates, and countries in such regions therefore need to acquire them through either imports or by using greenhouses to artificially create good climate conditions.Sweden is a country in Northern Europe that is currently heavily reliant on imports to cover a large share of the fruit and vegetable demand (Swedish Board of Agriculture, 2016).While the domestic production is high for certain crops, such as carrots, potatoes and onions, which are well-suited for Swedish conditions, it is very low for many other crops such as tomatoes (Swedish Board of Agriculture, 2021).The domestic production of tomatoes currently covers only 17 % of the Swedish consumption (Swedish Board of Agriculture, 2021), which shows that a considerable amount is imported from other countries.Tomatoes are among the vegetables with the highest import volumes, and the import of tomatoes alone accounts for almost half of the economic value of Sweden's total vegetable imports (Swedish Board of Agriculture, 2016).Most of the imports (62 %) originate from the Netherlands (Swedish Board of Agriculture, 2016), where tomatoes are grown in greenhouses.
The share of domestically produced tomatoes decreases even further during the cold/winter season (Lööv et al., 2011), when the Swedish production of tomatoes is at its lowest.Historically there has been no production of tomatoes at all during the winter (Lööv et al., 2011), but this has started to change as several of the largest producers in Sweden have started growing tomatoes throughout the year in parts of their greenhouses (Elleholms tomater, 2022;Stotzer, 2020).
An increase in the domestic production of tomatoes could have several possible advantages.There are economic aspects, where increased domestic production may contribute to a more favorable trade balance by reducing the financial outflows associated with imports, but there are also possible advantages related to the shorter supply chains of domestic production.One such advantage is a decreased amount of food losses related to the longer transport distances (Chauhan et al., 2021) normally associated with imports, and another is related to the possible reduction in environmental impact from the decreased need of transportation.
This possible reduction in environmental impact could however be offset by higher environmental impact in other parts of the domestic production system.Therefore, it is important to thoroughly analyze the entire system to determine the total environmental impact.There are several different tools available for assessment of environmental impact, but the perhaps most widely used method is life cycle assessment (LCA).This method can be used to analyze multiple scenarios, which is important because tomato production can be conducted in various ways.Although greenhouses are generally a requirement in colder climates, there are still other aspects such as choice of lighting technology and heat supply that could significantly affect the environmental impact of the production system.
There have been several previous LCA studies conducted on both Dutch greenhouse tomatoes (Antón et al., 2012;Torrellas et al., 2012;Vermeulen and van der Lans, 2011) and on Swedish tomato production (Bosona and Gebresenbet, 2018;Röös and Karlsson, 2013).There are also some studies on tomato production in other countries that have a similar climate as Sweden, such as Norway (Naseer et al., 2022) and Canada (Dias et al., 2017).Normally such studies do not share the same system boundaries and assumptions, and they often consist of either a single case study or a limited number of scenarios, which can lead to difficulties when trying to make comparisons between different product systems.Due to this, a detailed analysis comparing multiple production scenarios for Sweden with import scenarios from the Netherlands would be useful when working towards limiting the future environmental impact related to the consumption of tomatoes in Sweden, especially as Swedish production of tomatoes during the winter season is still an emerging sector.
The greenhouse production in the Netherlands and Sweden exhibits distinct characteristics.A significant difference lies in the size of their greenhouse sectors: the Netherlands has 9688 ha of greenhouses (Paris et al., 2022), compared to Sweden's 291 ha (Swedish board of agriculture, 2021).Dutch greenhouses are generally designed for high production intensity, utilizing large, climate-controlled structures and substantial resource inputs to achieve high yields (Paris et al., 2022).Conversely, despite its higher latitude, Sweden's greenhouses consume less energy per unit area, at 2.1 GWh/ha (Swedish board of agriculture, 2021)) compared to the Dutch consumption of 3.1 GWh/ha (Swedish board of agriculture, 2021).This lower energy use in Sweden can be explained by a lower average production intensity, in addition to only 86% of the total greenhouse area being heated (Swedish Board of Agriculture, 2021).Furthermore, the Swedish greenhouse industry has largely transitioned from fossil fuels, with only 13% of the energy mix derived from these sources in 2020, while renewable fuels such as wood pellets and wood chips have become predominant (Swedish board of agriculture, 2021).Conversely, Dutch greenhouses primarily use natural gas, with renewable energy accounting for just 9.4% of energy usage.Nevertheless, this proportion is rapidly increasing at an approximate rate of 35% per year, particularly with the adoption of geothermal heat (Paris et al., 2022).Owing to their intensive production, Dutch greenhouses also have a substantial electricity demand, representing approximately 26% of their total energy needs, of which 58% is produced on-site via cogeneration (Paris et al., 2022).
The aim of this study is to assess the potential environmental impact of Swedish tomato consumption by comparing several year-round domestic production scenarios with scenarios representing import from the Netherlands.This is done to provide comparable LCA results for both production locations, using several scenarios and similar greenhouse designs for both locations, as well as to highlight the key factors that affect the environmental performance of the production in both countries.To the authors' knowledge there is no previous literature making a comprehensive analysis of this kind.

Material and methods
To compare Swedish production with import from the Netherlands, a greenhouse simulation software is used to simulate the inputs (e.g.heat, electricity and carbon dioxide supplementation) and the output for a theoretical greenhouse placed in each country.The chosen software was SIOM version 2.1.3.5 (System Integration and Optimization Model), which was originally developed by TNO (The Netherlands Organization for Applied Scientific Research) to be a tool for decision-support for greenhouse builders (Janssen et al., 2014).The results from the simulation software were combined with data from the EcoInvent 3.8 database (Wernet et al., 2016) and used to perform a LCA in the software SimaPro 9.3, analyzing the potential environmental impact for the greenhouse at both locations.This process is illustrated in Fig. 1.

Simulation software and study objects
Greenhouse simulation software can strongly increase speed and quality of the integral design of greenhouse systems, although this design is a complex process (TNO, 2022).The chosen software, SIOM, consists of three main components: the calculation models, the databases, and the graphical user interface, all of which are based on a modular and extensible software architecture.Current software includes models for generating a typical meteorological year for the location of interest and for simulating greenhouse climate, considering actions for different systems, crop growth and use of resources (e.g.energy and carbon dioxide supplementation).The software has been validated using data from Dutch greenhouses and is currently used by several companies which are members of the Hortivation foundation, mainly during the preliminary design of a greenhouse project (TNO, 2022).
In this study, SIOM was used to simulate a typical Venlo type greenhouse (DutchGreenhouses, 2022) with tomato crop for two different areas: Trelleborg in Sweden and Bleiswijk in The Netherlands.Both are in regions where a large share of the total greenhouse area of each country is located, which is the reason that these specific locations were chosen.The two greenhouses were equipped with the same technical systems (heating, screening, CO 2 and artificial lighting) which were equally controlled, with the only difference being the capacity of the heating system (120 W/m 2 for Trelleborg and 80 W/m 2 for Bleiswijk).The calculations were made for two lighting systems, a lighting system with LED lamps of 60 W/m 2 electrical power and photosynthetic photon efficacy (PPE) of 2.5 μmol/J, and high-pressure sodium (HPS) lamps of 80 W/m 2 electrical power and PPE of 1.8 μmol/J, giving a total D. Danevad et al. of four simulation scenarios.Artificial light was applied when the outside radiation was below 350 W/m 2 between 07:00-19:00, assuming 12 h of illumination.Both greenhouses were hydroponic systems with water recirculation.The simulated crop was tomatoes with a cultivation period of 50 weeks, from mid-September until end of August the coming year.
The simulation results provided information about tomato yield, heat requirements, electricity usage, and the amount of carbon dioxide supplementation for the four simulations.This information was then used as an input for the life cycle assessment.The yields were designed to be approximately the same for the different simulations, making comparisons easier.

Goal and scope definition
The objective of this LCA is to: 1. Calculate the potential environmental impact of Swedish tomatoes and compare with Dutch tomatoes imported to Sweden.2. Identify the major hotspots of the production chains.3. Investigate how different options for production of heat, electricity and CO 2 affect the results.
The functional unit (FU) of this study is 1 kg of fresh grade tomatoes, in bulk, at wholesale depot in Sweden.The system includes construction and operation of the greenhouse, including inputs of energy and other resources, as well as transportation to the wholesale depot (see Fig. 2).Details of this system are altered to create several scenarios for both production locations, where different combinations of energy and CO 2 sources are analyzed to calculate the potential environmental impact of each scenario.Each production scenario is also calculated using both LED and HPS lighting systems.The LCA is performed using a retrospective attributional approach (JRC-IEA, 2010) and uses standard LCA procedure (ISO, 2006).This approach enables the evaluation and comparison of tomatoes from both locations as if they were produced in real, current systems and allows for the assessment of the total environmental impact of the studied systems, including the identification of hotspots that highlight potential areas of improvement for future assessments.
The different combinations of energy supply are based on technologies commonly used in each production location, as well as a few combinations that the authors expect would decrease the total environmental impact of the system.An overview of all 16 production scenarios included in this study is shown in Table 1.The details of each scenario are described in upcoming sections.

Table 1
Names and descriptions of all production scenarios for both locations (NL=The Netherlands, SE=Sweden, NG = natural gas, wc = wood chips).(Magaud, 2019) was used as a basis for the life cycle inventory model.The dataset includes information for all inputs required for tomato production in a Dutch setting, as well as the outputs.The dataset was then modified to fit all production scenarios for both locations used in this study.
For the LCI, all production scenarios used the greenhouse construction (e.g.loadbearing structures and covering materials) included in the EcoInvent dataset.The yield of the EcoInvent dataset is lower than the yields of this study, which comes from the calculations in SIOM, so the dataset was adapted to better represent the included production scenarios.This was done by adjusting the resource requirements based on the yield of each scenario, according to calculation information found in the documentation of the EcoInvent dataset.
Besides the greenhouse construction, the dataset also includes all the required information about inputs and outputs for greenhouse operation.Some of these in-and outputs were expected to have very small variations for the included production scenarios and were therefore not analyzed in detail in this study.To create a more complete system they were still included, but in an unmodified state.This includes the use of stone wool as cultivation substrate, production and use of pesticides, planting, production of tomato seedlings, emissions occurring inside the greenhouse, and land use.Other parts of the dataset were modified to better represent the studied product system.This includes processes such as heat and electricity supply and a few of the transports.Several new processes were also introduced to the dataset, including the introduction of liquid CO 2 and several new transports.The modifications and the new additions to the dataset are described in detail in subsequent subsections.

Energy use, fertilizers and carbon dioxide supplementation.
Using a LED lighting system decreases the electricity requirements, due to more efficient lighting when compared to a HPS system, while also increasing the required heat input as the LED lighting system is more efficient and therefore generates less heat.The needed inputs of electricity and heat for the production scenarios, as well as information on carbon dioxide supplementation and tomato yield, are provided by the SIOM simulations and can be found in the Results section.
There are three combinations of energy sources that are included for both production locations, while some are specific to only one of the locations.The combinations included for both locations are combustion of wood chips, either by co-generation or to obtain only heat, as well as using residual heat from nearby industries.Residual heat is included as an option as there are large amounts of residual heat generated in both Sweden and the Netherlands (Miró et al., 2015).However, this study did not assess the availability of residual heat at the specific locations under consideration.In the LCA, the combustion scenarios are represented by EcoInvent datasets (Bauer, 2019;Treyer, 2019aTreyer, , 2019b)), while residual heat is assumed to be used without upgrading and with negligible environmental impact.In the scenarios where electricity is not generated it is assumed that the electricity is obtained from the electricity grid.The characteristics of the electricity production in the two countries are very different.Swedish electricity comes mainly from a combination of hydropower and nuclear power, with an increasing share of wind power, while the majority of Dutch electricity production (68.3 %) comes from combustion of fossil fuels (Eurostat, 2022).In this study, EcoInvent datasets describing the respective country's national electricity mix (Treyer, 2019c(Treyer, , 2019d) ) are used to represent the grid electricity.Transformation losses are included in all electricity related datasets.For the Dutch production location, natural gas combustion is included in two different ways: only heat production (Heck, 2019) with electricity from the grid, as well as co-generation of heat and power (Ecoinvent, 2019a).Natural gas is not simulated for the Swedish production location as the share of tomato-producing greenhouses using fossil fuels in Sweden is very low (Swedish Board of Agriculture, 2018).
The use of carbon dioxide supplementation is a way of increasing the yield of a greenhouse.The CO 2 can be provided in various ways, depending on type of energy supply.In this study, for the production scenarios using natural gas it is assumed that the carbon dioxide from the exhaust gases is used to cover the need for carbon dioxide supplementation.For the other scenarios, it is assumed that liquid carbon dioxide is purchased and used.The liquid CO 2 is assumed to be of biogenic origin, based on the properties of CO 2 produced by Linde Gas, which is one of the larger suppliers of industrial gases in Sweden.Their CO 2 is mainly retrieved as a byproduct from fermentation of wheat (Linde Gas, personal communication, April 22, 2021).It is assumed that the CO 2 in the Netherlands is also biogenic, to enable a fair comparison.The liquid carbon dioxide is represented by an EcoInvent dataset (Hischier, 2019) for liquid carbon dioxide obtained as a waste gas from other production processes.It is assumed that the carbon dioxide is produced in the country where it is intended to be used, and therefore the dataset was modified to use either Dutch or Swedish electricity.
The amount of fertilizer applied is highly dependent on the desired yield.This study uses a closed water system, which results in very low nutrient losses.It is assumed that the same amount of fertilizer is required per kg of product in each scenario, which makes the amount of fertilizer identical for all scenarios.Sonneveld and Voogt (2009) describe that a greenhouse producing 60 kg/m 2 of tomatoes yearly requires 1185 kg of nitrogen, 284 kg of phosphorus and 2044 kg of potassium per hectare.For this study, these values were recalculated and used as an approximation for the amount needed per kilogram of product, which was then adapted to the functional unit.These values were then applied for all production scenarios.The production of the nutrients is represented by EcoInvent datasets for inorganic nitrogen, phosphorus, and potassium fertilizers.Emissions occurring during the use of the fertilizers is only accounted for through the default emission data available in the main EcoInvent tomato production dataset that was modified for this study.The two main reasons that this is not explored in further detail is that the use of fertilizers is not a part of the main focus of the paper, and because the amount of fertilizers used is almost identical in each location and scenario.
2.2.2.3.Transports.Several transports are described in the EcoInvent dataset for Dutch greenhouse tomatoes.These were left untouched, except for fertilizer transports as the higher yield resulted in more fertilizers being transported.The amount of transport required was increased based on the increase of mass, assuming the same transport distance for the production locations in both countries.
For the Swedish production location, liquid CO 2 is assumed being transported from the city of Norrköping, where a large Swedish biorefinery is located, to the greenhouse location in the city of Trelleborg.The total transport distance is 487 km.Due to a lack of specific data, the transport distance for Dutch liquid CO 2 is assumed to be the same to maintain comparability, even though such a distance suggests the origin point would likely extend beyond the Netherlands borders.The same truck type (Valsasina, 2019) is assumed to be used for both production locations.
The tomatoes are then transported from the grower in Trelleborg/ Bleiswijk to Helsingborg, which is the first location of the distribution network for several Swedish retail chains (Röös and Karlsson, 2013).It is assumed that this transport is performed by a refrigerator truck (Ecoinvent, 2019b), to ensure that the tomatoes are transported at a cool and constant temperature.Based on previous research, there are losses during the transport and storage stage amounting to 1% of the Swedish tomatoes and 2% of the Dutch tomatoes (Röös and Karlsson, 2013).These losses were included in this study.

Life cycle impact assessment
The impact assessment used the CML-IA baseline V3.07 method and included the five impact categories global warming potential (GWP), acidification potential (AP), eutrophication potential (EP), human toxicity potential (HTP) and abiotic depletion of fossil fuels (ADPf).The selection of the first three categories is due to the significant contribution of food production to each (Poore and Nemecek, 2018), making their analysis particularly relevant.The latter two indicators were chosen to provide a more comprehensive view of the environmental impacts, encompassing both toxicity and resource use.The CML method has historically been common in life cycle assessments of tomato production, (Torres Pineda et al., 2021), and the chosen categories provide a wide range of information about the environmental impact of the system under study.

Results
The results from the SIOM simulations are shown in Table 2 and includes tomato yield, electricity use, heat use and carbon dioxide supplementation for the four simulation scenarios.The results show that there are significant differences in energy use between the scenarios using LED lighting and the scenarios using HPS lighting.This was expected, as LED lighting uses less electricity but also emits less heat to the environment.Due to this there is a need to provide more heat from another source, which explains the lower electricity supply and higher heat supply for the LED scenarios.Although the variations in total energy usage between the two countries are small, the Swedish simulation scenarios do have a slightly higher energy usage per produced kg of tomato.
The results of the life cycle assessment are split into Figs.3-6, containing four production scenarios each.
Fig. 3 shows all the scenarios where natural gas (NG) is used, and therefore only includes Dutch production scenarios.The results show that the production scenarios using LED lighting always have a lower impact than the corresponding HPS scenario.The main reason for this is that the increased amounts of electricity in the HPS scenarios have a higher impact than the avoided heat supply, when compared to the LED scenarios.There is also a notable difference between the cogeneration scenarios and the scenarios using electricity from the grid, where the cogeneration scenarios generally perform better due to lower total impact from the heat and electricity production.
The results for production scenarios using a combination of electricity from the grid and heat from combustion of wood chips are shown in Fig. 4. The results again show that all the LED production scenarios have a lower impact than the corresponding HPS scenarios, although there are some scenarios where the difference is very small.The Swedish production scenarios perform better for these impact categories mainly due to the Swedish electricity mix having a lower impact per unit of electricity than the Dutch electricity mix.
Fig. 5 shows the results for all scenarios where electricity and heat are acquired through a cogeneration process using wood chips.
Production scenarios using LED lighting again has a consistently lower impact than the corresponding HPS scenarios, although the difference is smaller than in some of the earlier scenarios.A major difference compared to earlier figures is that the impact from electricity is less dominant in the categories ADPf, GWP and HTP, which makes transports and production of liquid carbon dioxide stand out more.For the Dutch scenarios, the impact of transporting tomatoes from the Netherlands to Sweden contributes 28-31% and 21-23% of the total environmental impact for the ADPf and GWP categories, respectivelya reflection of the total transportation distance, which is notably longer than for Swedish domestic production.Liquid carbon dioxide has a significant impact for the ADPf, GWP and HTP categories in all scenarios, where it contributes with 20-23%, 28-30%, and 40-46% of the total impact, respectively.The impact from the electricity production is still dominant in the AP and EP categories.
The scenarios using residual heat and electricity from the grid are shown in Fig. 6.As in all previous diagrams the LED production scenarios have a lower impact than the corresponding HPS scenarios.The Swedish production scenarios consistently have a lower impact than the Dutch scenarios, due to the lower impact of the Swedish electricity mix compared to the Dutch mix.The Swedish production scenarios presented in this figure have a lower impact than the previous scenarios, which is due to the assumption that the residual heat comes with no environmental burden.
When comparing all production scenarios for both locations, the scenarios using LED lighting are shown to always have a lower total impact than their corresponding HPS scenarios.The main reason for this is that the increased environmental impact for the heat supply was never enough to outweigh the reduced impact from the decreased electricity requirements.The Swedish scenario using LED and residual heat has a consistently low impact for all impact categories, where it has the lowest contribution for ADPf, GWP, AP and EP, and the third lowest for HTP.The Dutch scenario using LED and cogeneration of natural gas has the lowest contribution for HTP, followed by the corresponding HPS scenario.
Transports generally have a relatively low impact when compared to the total environmental impact.They only have a high share of the impact for a few of the import scenarios where the total impact is low thanks to lower impact from the use of energy.The combination of impact category and production scenario where transports have the highest percentage is ADPf and Dutch production using LED lighting and cogeneration of wood chips, where the share reaches 31.3 % of the total impact.The only other combinations where the transports account for more than 20 % of the total impact are the ADPf and GWP categories for the remaining Dutch scenarios using cogeneration of wood chips.The transports' share of the impact is generally lower for the Swedish production scenarios, even though these generally had a lower total impact than the Dutch scenarios.This is mainly due to the avoided importrelated transports.The impact from transports is the same for all Swedish scenarios, so the scenario with the highest share is automatically determined as the one with the lowest total impact for each impact category.The combination of impact category and Swedish production scenario that has the highest transport-related impact is ADPf and production using LED lighting and residual heat, where the share is 12.5% of the total impact.
The use of electricity accounts for a large share of the total impact for most scenarios.There are a total of 80 combinations of production scenarios and impact categories visualized in the bars of Figs.3-6, and electricity accounts for the majority of the impact in exactly 40 of them.In general, the impact is higher for the Dutch scenarios as most of them use either natural gas or Dutch electricity, while the impact is lower for scenarios using Swedish electricity or electricity originating from cogeneration of wood chips.In contrast, heat generally accounts for a small share of the total impact.The few exceptions are for the ADPf category and Dutch production using natural gas and grid electricity, as well as the AP and EP categories for Swedish production using wood    chips and grid electricity.The supply of liquid carbon dioxide consistently accounts for a high share of the impact for the HTP category, as well as high shares for ADPf and GWP for production scenarios where the total impact related to these impact categories is low.

Discussion
It is clear from the results that the choice of energy sources is a crucial factor for the environmental impact of a greenhouse of this kind, where the inputs of energy are relatively high.This is in accordance with previous research (Antón et al., 2012;Bosona and Gebresenbet, 2018;Naseer et al., 2022;Röös and Karlsson, 2013), but it becomes especially clear in this study where two very different national electricity mixes are included.There are several possible methodological choices that can be made when selecting the types of electricity to include in life cycle assessment studies.In this study the EcoInvent datasets for the respective country's national electricity mix were used.This choice has a positive impact for several of the Swedish production scenarios, as the impact from the Swedish electricity mix is lower than the Dutch.The national electricity grids operate on an international market, which is well-represented by the EcoInvent datasets as they include emissions from imported electricity.This should however be viewed as a snapshot of the current situation, and the results could change when/if the Dutch electricity mix transitions into having a lower environmental impact.The environmental performance of the Swedish electricity mix can also improve in the future, but as it is already comparatively good there are fewer options for improving it further.
Although the results are mainly valid for greenhouses with high inputs of resources and high output of tomatoes, several parts of the analysis are of interest even for greenhouses with lower inputs and yields.Besides the electricity, factors that changed between scenarios in this study were mainly limited to yield, energy use (both amounts and sources), transport needs and use of liquid carbon dioxide.The need for transportation is dependent on the distance between production location and the intended consumption location the transportation is independent of the design of the greenhouse.The results of this study imply that the relative importance of transport-related emissions would likely increase in a less energy-intensive lower-yield greenhouse (e.g. a greenhouse that doesn't operate during the colder months), as the transport would account for a larger share of the total emissions and would therefore be more interesting to assess in detail for such scenarios.
The other aspects of the production have a minor importance when compared to the impact from the energy supply, unless energy sources with very low impacts are used.Also, the choice of using cogeneration or not can have an impact on the results, as seen in the Dutch scenarios where natural gas were either used only for heat generation or for cogeneration.The scenarios using cogeneration of natural gas had a 17-22% lower GWP than the scenarios using natural gas only for heat generation.
The choice of another transport mode could have some impact on the results, as transport modes such as trains and container ships can have a lower impact per transported unit.In this study it was assumed that all tomatoes were transported by a refrigerator truck, both for the imported tomatoes and the domestic production.Using rail transports for a share of the imported tomatoes would likely have decreased the total impact of those scenarios, although it would only have made a notable difference for the scenarios with low total impacts.
As previously noted, this study did not assess the availability of residual heat at the specific locations under consideration.However, due to the large availability of residual heat in both Sweden and the Netherlands, this was still assumed as a realistic option and was therefore included.Another important factor related to the use of residual heat is the characteristics of heat.Greenhouses need a consistent and steady heat supply, indicating that not all residual heat sources may be adequate.This study assumes that the available residual heat can fulfill the total heat requirement of the greenhouse, but actual production setups might require a backup boiler to provide additional heat during times of peak demand.
The use of liquid carbon dioxide had a high share of the total impact for a few of the scenarios and impact categories.The authors found only one source of a complete life cycle inventory data for liquid carbon dioxide, which was the one in EcoInvent.There were some partial LCA results available for the GWP of the Swedish production facility in Norrköping (Linde Gas, personal communication, April 22, 2021), but these were not included due to lack of information on the other environmental impact categories included in this study.The system boundaries of the data did not include as many processes as the EcoInvent dataset, but the considerably lower impacts of the partial LCA results of the Swedish facility suggests that a sensitivity analysis using other datasets would have been interesting.As the emissions from CO 2 supplementation had a relatively high share of the impact for several scenarios, it could have affected the results if the choice of another dataset had been possible, which makes it an interesting topic to pursue in future studies.
Several of the production scenarios have higher environmental impacts than what is usually found in similar studies.This is mainly due to the type of greenhouses simulated for both locations, which requires relatively high inputs of heat, electricity and carbon dioxide supplementation.This is also combined with production throughout the whole year, which while quite common in the Netherlands is still quite rare in a Swedish context.The higher impact caused by more energy intensive production and/or imports is the price that must be paid to have access to fresh fruit and vegetables all year round, and therefore it is important to assess which scenarios that have the lowest impact.The other options would be to avoid fresh food that is out of season, or to use greenhouses that are not operating during the colder months (seasonal or extended seasonal production), but that has been analyzed in earlier research and is outside the scope of this study.
There is a large variety in the possible production scenarios that could have been included in the study.The included combinations were chosen as they were either common in one or both countries, or as they were believed to have the potential of leading to a lower environmental impact.Including additional combinations, as well as introducing scenarios with variations in yield and amount of inputs to the system, could create an even more comprehensive image of the environmental impact of tomato consumption and could be an interesting topic for further studies of this kind.In addition to this, it would also be useful to compare these results with a combination of seasonal domestic production and imports.For that kind of scenario, the environmental impact for the domestic production would most likely be lower, but this could potentially be offset by higher impact of the imported crop.
To contextualize the findings, it is beneficial to examine them within a broader environmental context.The Swedish production scenario that utilized LED lighting, residual heat, and grid electricity had the lowest GWP of the studied, at 0.37 kg CO2-eq.per kg of tomato.In contrast, the Dutch production scenario using HPS lighting, natural gas for heating, and grid electricity had the highest GWP, at 3.2 kg CO2-eq.per kg of tomato.The total consumption of tomatoes in Sweden is 96.1 million kg per year (Swedish board of agriculture, 2019), averaging about 9.6 kg of tomatoes per person.When these figures are combined with the results, the annual per capita GWP for Swedish tomato consumption based on the production scenarios studied ranges from 3.6 to 30.7 kg CO2-eq.At first glance, these numbers might appear negligible; however, they must be contextualized.On average, a Swedish individual is responsible for emissions of roughly 8 tons of CO2-eq.annually (SOU, 2022), with approximately 2.2 tons attributable to the average Swedish diet (Moberg et al., 2020).Moreover, the total annual emissions per person need to be reduced to less than 1 ton of CO2-eq.by the year 2050 (SOU, 2022).Understanding and mitigating the emissions from food production is a crucial part of achieving our sustainability goals.
It should be noted that the impact on the environment, as calculated by the methodology applied in this study, is not the only important criterion for the increase of domestic production in Sweden and the development of the greenhouse sector in general.Other parameters, such as the nutritional adequacy of the country, public health and the positive effects on the economy from the development of a sector are also important factors that would need to be considered in discussions about increased domestic production.
Although not directly applicable, the results of this study can also be seen as indicative for crops other than tomatoes that are grown in similar conditions.

Conclusions
This research fills a gap in the literature by providing detailed information on how several greenhouse design choices affects the environment.Domestic production of tomatoes in Sweden is compared with imports from the Netherlands, and factors that are important for limiting the environmental impact are highlighted for multiple production scenarios.As illustrated by the results, the choice of energy sources is a crucial factor when it comes to the environmental impact of the studied systems.
The results clearly showed that the production scenarios using LED lighting systems consistently had a lower impact than similar production scenarios using HPS lighting systems, when the same country and energy sources are used for both scenarios.It was also shown that Swedish domestic production has the potential to decrease the environmental impact of tomatoes consumed in Sweden, when compared to import from the Netherlands.The exception to this is for the human toxicity potential impact category, where the Dutch scenarios using cogeneration of natural gas had lower impacts than all other scenarios.For all other impact categories, when comparing the same kinds of energy sources for both locations, the Swedish production scenarios showed better results compared to the Dutch scenarios mainly due to either the lower impact of the Swedish electricity mix or due to the shorter transport distances.
That is not to say that consumption of Swedish tomatoes is always preferable over Dutch, as several of the better Dutch production scenarios had a lower impact than the worst Swedish production scenarios for some impact categories.Additionally, there were also a few combinations of Dutch production scenarios and impact categories that had a lower impact than the corresponding Swedish scenario if only production was considered, but when the impact from the transport to Sweden was added the Dutch scenarios reached a higher total environmental impact.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 4 .
Fig. 4. Life cycle assessment results of production scenarios using heat from combustion of wood chips and electricity from the grid.(NL = the Netherlands, SE = Sweden, wc = wood chips, grid = electricity grid).

Fig. 5 .
Fig. 5. Life cycle assessment results of production scenarios using energy from cogeneration of wood chips.The scale of AP has been changed in this diagram to improve readability.(NL = the Netherlands, SE = Sweden, wc = wood chips, cogen.= cogeneration).

Fig. 6 .
Fig. 6.Life cycle assessment results of production scenarios using residual heat and electricity from the grid.(NL = the Netherlands, SE = Sweden, Residual = Residual heat).

Table 2
Tomato yield, energy input and carbon dioxide supplementation of the four simulation scenarios.