FormalPara Contents

1 Introduction

In order to provide more stable production conditions, protected cultivation methods ranging from low-cost greenhouse solutions (low-tech GH, including low and walk-in tunnels, shelters, and net-houses) to highly technological systems such as hydroponic greenhouses (high-tech GHs) and vertical farming (VF) systems are used. High-tech GHs and VF systems aim to isolate crops from external climate for better control. While high-tech GHs have been widely adopted, interest in VF has grown in recent years due to its potential benefits (Kozai et al. 2020; van Delden et al. 2021; Erekath et al. 2024). However, its global viability and impact are debated. Some critics argue that VF still contributes little to food production (Stanghellini and Katzin 2024), although estimates are built on outdated data (Terazono 2020, based on figures from van Rijswick 2018). More recent figures (Chun 2020; Fang 2020; Tong and Yang 2020; Wood et al. 2020; ResearchAndMarkets 2021; Namkung 2022; Zhuang et al. 2022) suggest VF is expanding, with thousands of farms across the USA, Asia, and Singapore. Despite challenges, especially from the 2022 energy crisis, VF is expected to attract $30 billion in investments by 2030 (Orsini and Zauli 2023). Given setup costs ranging from 1500 to 3500 $ m−2 (Monzini and Orsini 2024), this could translate into 850–2000 ha of cultivated area. Though small compared to the 500,000 ha of greenhouse farming globally (Marcelis et al. 2019), VF’s rapid growth and technological appeal merit attention, especially when evaluating its sustainability. It should be further noted that, although a number of VF companies have struggled to overcome the sudden increase in energy costs occurring since 2022, such failures and bankruptcies could also have been forecasted, as pointed out by Bustamante (2024) and Guthman and Fairbairn (2024), who concluded that 2023 would have coincided with the so-called economic phase of trough of disillusionment that any industrial sector will face when innovation emerges. An interesting analysis of how Infarm (one of the world’s largest VF companies at the time) collapsed is included in Martington (2023), who stressed that ineffective decisions (mainly related to the persistence of a start-up mindset or unclear choices in human resources management) rather than the inefficiency of the technology per se were behind such failure. When another large VF company (Bowery Farming) collapsed during the fall of 2024 (Bradbury 2024), a pest outbreak across all its facilities was claimed to be the cause of the collapse. Such pest outbreaks seem avoidable, given the current available knowledge and phytosanitary control technologies. Although it is commonly accepted that claims about vertical farming—like any new technology introduced—must be substantiated by scientific evidence, it is even more important to provide and share with the scientific community a comprehensive picture of the knowledge currently available on the technology.

Based on this background, the purpose of this article is, therefore, to provide and integrate available information on the topic of VF (Fig. 1) in order to help researchers further pursue the identification and validation of the pros and cons. More explicitly, this review article aims to analyse (with a specific focus on lettuce as model leafy vegetable crop) available figures on yield and productivity, energy and water use efficiency, and overall environmental impact (expressed as global warming potential) in VF vs traditional cultivation systems (open field, low-tech and high-tech GHs) in order to define where and how different technological solutions available to growers offer the best trade-offs for vegetable crop production.

Fig. 1
Fig. 1
Full size image

A picture of a vertical farm where plants are grown under light-emitting diode (LED) lights in stacked shelves, where all the growth conditions can be exactly controlled (photo credit Leo Marcelis)

2 Yield and productivity of vertical farming (VF) systems

Although VF systems allow for the cultivation of essentially all types of crops (Righini et al. 2024), most current research and commercial facilities concentrate on leafy vegetables, and more specifically on lettuce (Lactuca sativa L.), due to their limited size, ease of cultivation, short cycle, high harvest index (namely, the marketable over the total biomass), and overall economic viability. From a physiological perspective, while common light use efficiency (LUE) at present stands at approximately 0.55 g dry weight (DW) mol−1 (Jin et al. 2022), the achievable LUE is approximately 0.8 to 1 g DW mol−1 for most crops (Stanghellini and Katzin 2024), including lettuce (Carotti et al. 2021), and also with a photoperiod of 24 h d−1 (as detailed in Pennisi et al. 2020a). However, increased light intensities may also augment dry matter concentration, as evidenced in Carotti et al. (2021), e.g. from 2.6 to 4.2% when Light intensity was increased from 200 to 750 μmol m−2 s−1, possibly as a consequence of concurrent limiting factors (e.g. since plant density, plant nutrition, or air humidity were not modified). Accordingly, it may be estimated that when a photosynthetic photon flux density (PPFD) of, for instance, 400 μmol m−2 s−1 is applied over a continuous photoperiod (24 h d−1), fresh Yield of above 500 kg FW m−2 y−1 (building on estimates by Carotti et al. 2021 and Jin et al. 2022, and reported in Table 1) is theoretically achievable, although far above what is attainable in current VF systems, where the adoption of lower PPFD values (in the range of 200–250 μmol m−2 s−1) is common. While under the current economic constraints, such theoretical yields are not viable, it should also be considered that such elevated performances could exacerbate the emergence of disorders, e.g. tip burn, that would ultimately affect the produce value.

Table 1 Theoretical yield of lettuce (Lactuca sativa L.), grown in a VF system at either 250, 400, and 500 μmol m−2 s−1, under constant photoperiod (24 h d−1), based on either average or maximum light use efficiency (LUE) values for shoot dry weight (DW) production, as identified by Jin et al. (2022). Dry matter was assumed to vary in response to increased light intensity (respectively featuring 2.6% at 250, 3.8% at 400, and 4.2% at 750 μmol m−2 s−1), according to Carotti et al. (2021).
Full size table

In commercial VF systems, leafy vegetables, and more specifically lettuce, are the most common products whose yield (although quite variable between sources in literature and from commercial operators) featuring respectively (on a cultivated surface basis) averages of 60 and 105 kg FW m−2 y−1 and medians of 48 and 70 kg FW m−2 y−1, as reported in Fig. 2.

Fig. 2.
Fig. 2.
Full size image

Yield productivity of lettuce (Lactuca sativa L.) in vertical farms (on a cultivated surface basis), according to scientific literature (white box, mean 59.7±5.8 kg FW m−2 y−1, median 48.3 kg FW m−2 y−1, n=54) and data declared by companies (grey box, mean 105.2 ± 15.2 kg FW m−2 y−1, median 69.5 kg FW m−2 y−1, n = 32). Round symbols refer to individual values (not all symbols are visible, due to overlapping), the horizontal line in the boxplot indicates the median, whereas the mean value is represented by the X symbol. Values leading to the boxplot of scientific reports were extrapolated from Incrocci et al. (2006); Li et al. (2016); Touliatos et al. (2016); Wang et al. (2016); Kuno et al. (2017); Hytönen et al. (2018); Zhang et al. (2018); Chen et al. (2019a and b); Meng et al. (2019); Meng and Runkle (2019); Pennisi et al. (2019a, b); Rouphael et al. (2019); Yan et al. (2019); Zou et al. (2019); Ahmed et al. (2020); Cammarisano et al. (2020); Esmaili et al. (2020); Gómez and Jiménez (2020); Kelly et al. (2020); Meng et al. (2020); Pennisi et al. (2020a, b); Spalholz et al. (2020); Zou et al. (2020); Bhuiyan and Van Iersel (2021); Cammarisano et al. (2021); Carotti et al. (2021); Chen et al. (2021); Hosseini et al. (2021); Jayalath and van Iersel (2021); Jin et al. (2021); Kusuma et al. (2021); Nguyen et al. (2021); Saengtharatip et al. (2021); Voutsinos et al. (2021); Yi et al. (2021); Zou et al. (2021); Blom et al. (2022); Nguyen et al. (2022); Smirnov et al. (2022); Zhuang et al (2022); Carotti et al. (2023); Cossu et al. (2023); Ertle and Kubota (2023a and 2023b); Gavhane et al. (2023); Jin et al. (2023); Kaufmann (2023); Matysiak et al. (2023); Meng and Runkle (2023); Carotti et al. (2024 and unpublished data). Commercial farm data are extrapolated from Armanda et al. (2019); Kozai et al. (2020); Martin (2023); Martin et al. (2023a); Joensuu et al. (2024); Martin et al. (2024). When multiple values were found in the same study, all were included in the boxplot. While all data from scientific reports are related to 1 m2 of cultivated surface, commercial farm sources were sometimes unclear on whether they referred to yield per cultivated area or occupied land by the farm. Since no data were available on the occurring interval between cultivation cycles, it was assumed that new plants were moved to the system on the same day that the previous cycle was harvested.

It could be considered that VF systems are actually exploring the volume of the cultivation environment more efficiently than greenhouses. In other words, while crops are grown on a single layer in a low-tech or high-tech GH, they can be cultivated on multiple layers or green walls in VF systems. While land surface use efficiency (L-SUE, expressed as kg FW m−2 y−1, and related to the floor area occupied by the VF facility) is often underestimated in experimental studies (where plants are cultivated in small-scale systems inside walk-in cultivation chambers within research centres), recent research has looked into the productivity of commercial farms. For instance, Martin et al. (2023a) analysed a Swedish large-scale VF company that grows lettuce on vertical walls and has a land-use index of 3.5 m2 m−2 (approximately 5900 m2 of cultivated land over a floor area of 1670 m2, which also includes paths, walls, equipment, and machineries), reaching L-SUE values of approximately 311 kg FW m−2 y−1. Similarly, another Swedish company that develops small-scale cultivation cabinets (that hosts 4.36 m2 of cultivated surface over a floor area of 0.87 m2) features a land-use index of 5 m2 m−2 and L-SUE of 275 kg FW m−2 y−1 (Martin 2023; Martin et al. 2024). Consistently, a recent study covering over 41 operational VF systems confirmed a land-use index of approximately 3.05±0.19 m2 m−2 and L-SUE of 350.4±22.3 kg FW m−2 y−1 (Zhuang et al. 2022). It should be noted, however, that based on numbers from Fig. 2 (that place VF productivity of cultivated area somewhere between 48 and 105 kg FW m−2 y−1), L-SUE values of approximately 300 kg FW m−2 y−1 would require land use-index values ranging from 3 to 6 m2 m−2. Nevertheless, from a productivity perspective, L-SUE values of VF systems seem to be much higher than those reported in some Dutch high-tech GHs with supplemental light (65 kg FW m−2 y−1, Raaphorst and van Tuyll 2023), while approximately 2-fold of those normally obtained in non-illuminated high-tech GHs (34 kg FW m−2 y−1, Raaphorst and van Tuyll 2023). Furthermore, they stand at approximately 50 to 100-fold of those normally experienced in low-tech GH and open field, respectively (Barbosa et al. 2015). While it can certainly be argued that this yield increase comes at high energy costs, the overall viability should be assessed against factors such as the local land cost (e.g. in highly populated city-states), the food supply alternatives (e.g. when long-distance or international import is the only option), or even the local impact that resource use (e.g. water, energy, nutrients) associated with greenhouse cultivation technologies may have (e.g. in harsh environments, or when resources are scarce and competition for alternative uses is high). To this extent, this study will further explore the resource use efficiency of VF.

3 Energy use efficiency

Energy use stands as the elephant in the room when addressing the financial viability and environmental sustainability of VF. On average, energy costs are mainly related to electricity for lighting (42–80%), followed by climate control (16–43%), while only approximately 0.4–15% is consumed by the other production facilities (e.g. cultivation systems, post-harvest facilities for packaging, or auxiliary practices, e.g. cleaning, as addressed in Yokoyama 2019; Monzini and Orsini 2024). With reference to total energy consumption and based on the cultivated surface, VF systems were recently estimated to consume between 400 and 1260 kWh m−2 y−1 (Talbot and Monfet 2024), based on both model and experimental data from eight previous studies. Energy consumption data from five commercial VF systems across Italy (Agricola Moderna, LocalGreen) and Sweden (Ljusgarda, Gronska, and IKEA) confirm values of energy use in the range of 800 to 1125 kWh m−2 y−1 (Martin et al. 2023a; Martin 2023 and personal communication, 2023; Maggioni, M., personal communication, 2022). However, rather than energy use values per unit surface, a common indicator of the performance of VF systems is the energy use efficiency (EUE), defined as the fresh biomass produced per energy input (e.g. kg FW kWh−1). EUE values may be referred to as either the produced fresh harvestable yield or to the produced dry biomass (with values being easily derived considering a standard dry matter content of 3.5% in lettuce, Stanghellini and Katzin 2024). While from a physiological perspective we acknowledge dry biomass to be more interesting (as for the LUE values, commonly expressed as g DW mol−1), for EUE values, we prefer to build our indicator on fresh harvestable yield values (kg FW kWh−1), as this may lead to easier economic evaluations (being the price of the produce based on its fresh biomass, as well as the cost of energy expressed by kWh). Accordingly, reference values from previous literature (that include both energy from lighting as well as from other VF operations) have been converted into harvestable fresh Yield per unit of energy consumed to extrapolate comparative values to be used for our analysis. A recent estimate of the theoretical EUE in VF systems suggests a value of 0.30 kg FW kWh−1 (Stanghellini and Katzin 2024), which, however, seems to be very difficult to achieve. When looking at Literature and available commercial VF systems data, the realised EUE are often found to be lower, with common values of approximately 0.08–0.13 kg FW kWh−1 (Fig. 3). These values also reasonably agree with Miserocchi and Franco (2024) who reported that values in Literature range from approximately 0.03 to 0.20 kg FW kWh−1.

Fig. 3
Fig. 3
Full size image

Energy-use efficiency (EUE; kg harvestable fresh biomass per unit of energy use) of lettuce (Lactuca sativa L.) in vertical farms, according to scientific literature (white box, 0.13±0.01 kg FW kWh−1, median 0.12 kg FW kWh−1, n=34) and data declared by companies (grey box, mean 0.08±0.01 kg FW kWh−1, median 0.08 kg FW kWh−1, n=6). Round symbols refer to individual values (not all symbols are visible, due to overlapping), the horizontal line in the boxplot indicates the median, whereas the mean value is represented by the X symbol. Values leading to the boxplot of scientific reports were retrieved from experimental and modelling studies by Eaves and Eaves (2018); Graamans et al. (2018); Zhang and Kacira (2020); Blom et al. (2022); Kobayashi et al. (2022); Eaton et al. (2023); Gavhane et al. (2023); Stanghellini and Katzin (2024). Company data are extrapolated from Martin (2023); Martin et al. (2023a); Joensuu et al. (2024), as well as personal communications from Benjamin Franchetti (2022) and Marco Maggioni (2022). When multiple values were found in the same study, all were included in the boxplot. All values considered included energy consumed along the whole plant cycle.

Conversely, open field and low-tech GHs tend to require much less energy, featuring higher EUE in the range of 0.9–3.2 kg FW kWh−1 for open field (Canakci and Akinci 2006; Kuswardhani et al. 2013) and 2–10 kg FW kWh−1 for low-tech GH (Kuswardhani et al. 2013; Hedau et al. 2014). This makes these cultivation technologies more suitable for reducing electricity use for food production, wherever they are applicable. Indeed, as a consequence of both the quality standards required by the market as well as adverse environmental conditions in many production regions, high-tech GHs are also commonly adopted, although they result in increased energy requirements in comparison to low-tech GHs and open field. For instance, high-tech GHs with artificial Lighting producing lettuce in the Netherlands had an EUE of 0.197 kg FW kWh−1 (Raaphorst and van Tuyll 2023), closed greenhouses that produce tomatoes in Saudi Arabia had an EUE estimated in the range of 0.125 kg of tomatoes per kWh (Tsafaras et al. 2022), while a commercial greenhouse producing potted herbs in Sweden had an EUE (on a cradle to farm-gate perspective) of 0.03 kg FW kWh−1 (Martin et al. 2023b). Under such adverse climate conditions of these countries (hot in Saudi Arabia and cold and dark in Sweden), local VF systems (with EUE of 0.08–0.13 kg FW kWh−1 as mentioned above) may turn out to be similar or even more energy efficient than local high-tech GHs or imports. Furthermore, Graamans et al. (2018), using simulation models for greenhouse and for potential dry matter production of lettuce crops, estimated that energy requirements for producing lettuce in a vertical farm or in a semi-closed high-tech GH in the United Arab Emirates would require similar purchased energy (in MJ per kg of DW produced). They estimated that in these greenhouses, energy is mainly needed for dehumidification and cooling, while in the VF system, it is mainly for lighting. As a consequence of the adopted coefficients of performance (COP) for the different energy types, the amount of purchased energy (in kWh per kg of DW produced) was much lower in the semi-closed high-tech GH versus the vertical farm in which light-emitting diodes (LEDs) with relatively low efficacy (52%) and very high light intensity (500 μmol m−2 s−1) were used.

Further consideration should, however, be given to the stability and consistency of crop yield across the seasons. When plants are grown in VF systems, productivity is overall stable throughout the year (unless disruptive events emerge, which should be less likely to occur as technology and skill evolve). Conversely, constant production in greenhouse systems tends to be difficult to achieve and would also be energy-intensive. In Europe, low-tech GHs are spread across the Mediterranean, but their use is only common during cooler seasons, while no production takes place in summer months (Cajamar 2022). Costs for cooling would be prohibitive either in terms of water use (when evaporative cooling is adopted) or energy (when mechanical cooling is applied) (Tsafaras et al. 2022). On the other hand, in northern countries (e.g. the Netherlands or Sweden), summer cultivation is common, while winter production is limited—also in high-tech GHs—by the elevated costs for heating and lighting, with yields in December in a greenhouse with supplementary lighting being approximately half (i.e. 55%) of those achieved in May (Raaphorst and van Tuyll 2023). Again, while it may be arguable that a high-tech GH with full climate control and lighting may lead to productivity similar to VF, this will also lead to increased costs for energy and other resources. To this end, looking at yearly yield only (rather than their monthly trends) may become misleading when addressing stable crop production in a specific region.

Accordingly, when data of monthly averages of daily productivity (g m−2 d−1) are plotted (Fig. 4), it clearly emerges how yield increases and its variability across seasons is reduced when supplemental lighting is adopted in high-tech GHs (Fig. 4a), but this comes at additional energy use (Fig. 4b) (and therefore reduced EUE in darker months, Fig. 4c) and never reaches the stable production achieved by VF systems. Indeed, since the market demand for vegetables tends to be stable across the year (Martin et al. 2023b), any comparison of annual values may present some flaws when most GH productivity is concentrated in a few summer months. While it may be argued that society should move toward seasonal products to boost food systems’ sustainability, consumption trends confirm the demand to be quite stable across seasons. For instance, in urban centres of the UK, approximately 70% of local consumers were found to expect to be able to purchase salad items at any time of the year (Hospido et al. 2009). In Indiana (USA), ready-to-eat lettuce mixes were recently reported to be available year-round, although prices double in winter versus summer months (Philocles et al. 2020). Similarly, in Australia, lettuce is also available at any time of the year, although prices in April are double those reported in November, based on average data from 2016 to 2018 (Hort Innovation 2019). Accordingly, if the global market demand is constant, the supply of leafy vegetables to cover market needs during cooler or darker seasons would require a much larger area devoted to GH or much more energy being used to power additional lighting or ultimately an increase of imports from countries with more light access, which would result in increased transport impacts. While a transition to more seasonal and local dietary habits remains the priority, VF hereby shows potential for contributing to food system sustainability under the current consumption scenario.

Fig. 4
Fig. 4
Full size image

Monthly averages of daily fresh yield (a), energy use (b), and associated energy use efficiency (EUE), both in full chart (c) and in a magnified section to visualise differences between categories (d), across months in different systems, including high-tech greenhouses (GHs) without (dotted line) or with (dashed line) supplementary lighting (Graamans et al. 2018; Raaphorst and van Tuyll 2023) and VF (solid line, with values considered stable across the year and derived from references cited in Fig. 2).

4 Environmental performances

From a life cycle assessment (LCA) perspective, a common unit adopted for environmental impact assessment is greenhouse gas (GHG) emissions (in terms of carbon dioxide equivalents per functional unit). The functional unit for foods typically refers to the weight of the edible portion. Therefore, the environmental impacts for lettuce are often shown as kg CO2-eq kg−1 FW. The increase of technology (e.g. from open-field to GH to VF) is generally associated with increases in overall emissions per kilogram of produce (Blom et al. 2022). However, few studies have compared the environmental impacts of GH and VF systems with land-based production methods. When examining the results from roughly 200 LCA analyses and scenarios in 29 studies of lettuce production on a farm-to-gate perspective (as summarised in Fig. 5), it emerges that while GHG emissions tend to be substantially lower when plants are grown in open-field or low-tech GH systems, VF and high-tech GH systems feature comparable values (means of 2.87 and 2.38 kg CO2-eq kg−1 FW and medians of 0.83 and 1.98 kg CO2-eq kg−1 FW, respectively). It should be noted, however, that variability across systems and within each system is rather large (especially in VF systems), with impacts highly dependent on both technologies used as well as the local environment where farming takes place (local climate for GH, but also energy sources for VF, with lower impacts observed in regions where highest share of renewable energies is integrated). Accordingly, future research should look at the main causes of the observed impact variations and address them when choosing the optimal technology for a specific context.

Fig. 5
Fig. 5
Full size image

GHG emissions on a cradle-to-gate perspective of lettuce (Lactuca sativa L.) grown in open field (white box, mean 0.176±0.02 kg CO2-eq kg−1 FW, median 0.127 kg CO2-eq kg−1 FW, n=44, from Hospido et al. 2009; Maraseni et al. 2010; Davis et al. 2011; Venkat 2012; Hall et al. 2014; Romero-Gamez et al. 2014; Li et al. 2020; Beaumont De Oliveira et al. 2021; Blom et al. 2022; Casey et al. 2022; Martin et al. 2023a; Maynard et al. 2023; Sandison et al. 2023), low-tech greenhouses (low-tech GH, white box with diagonal Lines, mean 0378±0.07 kg CO2-eq kg−1 FW, median 0.219 kg CO2-eq kg−1 FW, n=17, from Hospido et al. 2009; Romero-Gamez et al. 2014; Bartzas et al. 2015; Fusi et al. 2016; Blom et al. 2022; Martin et al. 2023a), high-tech greenhouses (high-tech GH, grey box, mean 2.38±0.26 kg CO2-eq kg−1 FW, median 1.98 kg CO2-eq kg−1 FW, n=36, from Hospido et al. 2009; Stoessel et al. 2012; Yrjänäinen et al. 2013; Räsänen et al. 2014; Hallikainen 2019; Silvenius et al. 2019; Li et al. 2020; Blom et al. 2022; Sandison et al. 2023; Song et al. 2022; Martin et al. 2023a; Maynard et al. 2023; Joensuu et al. 2024), and vertical farm (VF, grey box with diagonal Lines, mean 2.87±0.38 kg CO2-eq kg−1 FW, median 0.82 kg CO2-eq kg−1 FW, n=121, from Shiina et al. 2011; Hallikainen 2019; Avgoustaki and Xydis 2020; Li et al. 2020; Wildeman 2020; Beaumont De Oliveira et al. 2021; Blom et al. 2022; Casey et al. 2022; Sandison et al. 2023; Song et al. 2022; Martin 2023; Martin et al. 2023a; Maynard et al. 2023; Joensuu et al. 2024; Martin et al. 2024). Round symbols refer to individual values (not all symbols are visible, due to overlapping), the horizontal line in the boxplot indicates the median, whereas the mean value is represented by the X symbol. When multiple values or scenarios were found in the same study, all were included in the boxplot. When figures on dry matter were missing, standard values were adopted (e.g. lettuce dry matter content of 3.5%, similar as used in Stanghellini and Katzin 2024). Highest values observed in the VF category are associated with scenarios with high impact energy sources (e.g. coal) or old LED technologies (e.g. references prior 2019).

Across LCA results of open-field cultivation of lettuce, and contrary to common consumers’ belief, lower GHG emissions were observed when conventional protocols were adopted instead of organic ones (respectively 0.19 vs 0.27 kg CO2-eq kg−1 FW, Venkat 2012), mostly due to increased yields and lower energy use in the farm. Moreover, fertilisation strategies (0.003 to 0.013 kg N kg−1 FW in the USA and Spain, respectively) were also shown to dramatically affect the environmental impact of open-field lettuce production (resulting in 0.08 vs 0.18 kg CO2-eq kg−1 FW) (Casey et al. 2022). When greenhouse technologies were adopted, differences could be ascribed to the LCA inventory or reference used (e.g. Martin et al. 2023a, b, with impacts ranging 0.11 to 0.65 kg CO2-eq kg−1 FW in low-tech GH in Sweden and Italy, respectively). Adoption of renewable energy instead of current sources was estimated to reduce GHG emissions from 1.21 to 0.79 kg CO2-eq kg−1 FW in low-tech GH (−34%) and from 1.45 to 0.75 kg CO2-eq kg−1 FW (−48%) in high-tech GH (Blom et al. 2022). Similarly, transitioning from coal (ranging 1.32 to 3.95 kg CO2-eq kg−1 FW) or gas (1.18 to 3.08 kg CO2-eq kg−1 FW) to solar (0.17 to 1.96 kg CO2-eq kg−1 FW) or wind (0.01 to 1.87 kg CO2-eq kg−1 FW) was suggested to substantially decrease GHG emissions in high-tech GH (Hallikainen 2019). Recent publications, however, have also reported the potential of VF to feature low GHG emissions (e.g. Li et al. 2020; Martin et al. 2023a; Joensuu et al. 2024), whenever the impacts of the farm are reduced (e.g. by integrating renewable energies or waste valorisation—as also feasible in GH horticulture) or shorter food supply chains are established (enabling to reduce food miles, e.g. at northern latitudes or within dense cities). Studies have found that VF systems can produce crops with GHG emissions of 0.06 kg CO2-eq kg−1 FW using wind energy in Sweden (Hallikainen 2019), 0.36 kg CO2-eq kg−1 FW through photovoltaic energy and nutrient recovery in Singapore (Li et al. 2020), 0.48 kg CO2-eq kg−1 FW using grid-supplied wind energy across different locations (Casey et al. 2022), 0.33 kg CO2-eq kg−1 FW in a VF system fully powered by renewable energies in the UK (Sandison et al. 2023), 0.66 kg CO2-eq kg−1 FW in a wind-powered modular VF system that shares impacts with the multifunctional building where it is hosted (Martin 2023), and 0.74 and 0.78 kg CO2-eq kg−1 FW, respectively, in a hydro-wind-powered large scale (Martin et al. 2023a) or modular (Martin et al. 2024) VF system in Sweden. Conversely, when scenarios for VF systems employ energy sources with high carbon emissions (e.g. coal) or surpassed technologies (e.g. LEDs with low efficacy), large GHG emissions were observed (e.g. 5.3 to 22.0 kg CO2-eq kg−1 FW in Hallikainen 2019 or 17.8 kg CO2-eq kg−1 FW in Casey et al. 2022). To this extent, it may be worth quantifying if and how products from alternative production methods to vertical farms compare, as further discussed in the next section.

5 Opportunities for reducing food miles and increasing supply chain resilience

In earlier studies (mostly relying on models instead of actual on-farm analyses), it was advocated that GH could perform better than VF when comparing the GHG emissions (Graamans et al. 2018; Blom et al. 2022). While some of the assumptions of these studies are today outdated, in part due to significant improvements in VF technology or refining of the models, impacts associated with post-harvest handling of the produce were also often underestimated or excluded. More specifically, the possibility to counterbalance increased energy needs and associated emissions by relocating horticultural productions in areas where more favourable environments exist is not substantiated by the energy costs associated with transport and cold chain preservation (accounting for approximately 0.71±0.41 g CO2 for every km travelled by 1 kg of fresh lettuce, according to average values from Blom et al. 2022; Casey et al. 2022; and Martin et al. 2023a). Similar findings were also reported in Maynard et al. (2023), where transportation may have an important contribution to the overall carbon footprint of fresh products (approximately 1.08 g CO2 for every km travelled by 1 kg of fresh lettuce), despite dissimilar vehicular sizing and packaging sizing, compared to the previously cited studies (primarily located in Europe) for shipping of produce in the USA. All these values are however substantially higher (approximately 10-fold) than the estimate included in Stanghellini and Katzin (2024, approximately 0.074 g CO2 kg−1 km−1), which originates from previous estimates from Mulholland et al. (2023), accounting for 0.052 to 0.062 g CO2 kg−1 km−1, associated with lorry transport, plus an additional 15% computed by Yang et al. (2021), as a consequence of cooling (rounded at +20% in Stanghellini and Katzin 2024). It should be noted, however, that such impact values are not tailored to food transport, but come instead from standard data on lorry transportation of any material (Mulholland et al. 2023) and truck cooling (Yang et al. 2021), and therefore a possible reason for the impact discrepancy may come from specific features of the lettuce crop (relatively low specific weight and density, e.g. for the case of ready-to-eat salads) that may ultimately increase transport impacts (due to a reduced loading capacity per single lorry), as compared with other crops. Accordingly, when plotting data from references specifically on lettuce (Blom et al. 2022; Casey et al. 2022 and Martin et al. 2023a, retrieved from LCA ecoinvent databases), the GHG emissions (expressed as kg CO2-eq kg−1 FW) may be quantified to be approximately 0.03882 + (0.00025*km of transport), as expressed by the linear relationship evidenced in Fig. 6. Therefore, we hereby propose a function for calculating how much distance may turn VFs more environmentally sound than low-tech GHs. For instance, building on the reference values provided in Martin et al. (2023a) (e.g. VF in Sweden—where energy contributes to a small carbon footprint—and a low-tech GH in Italy on a cradle-to-gate perspective featuring GHG emissions at production stage respectively of 0.96 and 0.65 kg CO2 kg−1 FW), it may be evinced that a threshold of approximately 1100 km as the distance radius that would turn local production in VF as a more sustainable option vs import from distant low-tech GH. However, in locations where the electricity grid for running a VF is less based on renewable energy, this threshold will be larger. While it may be argued that high-tech GH may also allow production in adverse environments (therefore reducing post-harvest impacts), their yield will still present seasonal variations (as displayed in Fig. 4), while their resource use performances and environmental impact will be highly dependent on the integrated technologies (e.g. when artificial lighting is used in darker months, or semi-closed GHs are adopted to reduce water loss) required in the local conditions. Given the limited literature on resource use and environmental impact in lettuce production within semi-closed and artificially illuminated high-tech GH, it turns out to be difficult here to extend the comparative analysis of which distance would make them more sustainable than alternative imports from low-tech GH options.

Fig. 6
Fig. 6
Full size image

Linear relationship between transportation of fresh food produce in refrigerated trucks in response to distance travelled (in km) and estimated greenhouse gas (GHG) emissions (as kg CO2-eq kg−1 FW). Values retrieved from Blom et al. (2022, black triangles), Casey et al. (2022, grey circles), and Martin et al. (2023a, b, white squares). Regression equation: y=0.00025x + 0.03882; R2 = 0.82.

As vertical farms may lead to shorter supply chains if located close to consumers, this may also lead to knock-on effects such as reduced food loss and waste (namely the losses in the produced biomass that do not reach consumers, e.g. during production and post-harvest stages until retail). Although comparative assessments of food losses and waste in VF vs high-tech GH are not yet available, some general considerations may be drawn from existing Literature. More specifically, while studies that address losses and waste in lettuce specifically grown in high-tech GHs are still lacking, some insights from studies that assess the impacts of conventional systems can be drawn to highlight commonly observed figures. For instance, losses and waste in open-field grown lettuce were estimated to range 24–34% in Sweden, resulting in GHG emissions up to 0.94 kg CO2 kg−1 of fresh lettuce marketed (Eriksson et al. 2012; Strid and Eriksson 2014). Similarly, Castro-Granados et al. (2019) evidenced losses and waste accounting for approximately 40% of lettuce grown on organic or conventional farms in Costa Rica. Lettuce loss at the production stage was estimated at approximately 33% in California (Baker et al. 2019) and 38% in South Africa (Le Roux et al. 2018). Siñol and Martin (2025) also found that large shares of food losses are produced in conventional supply chains for mixed salad bags (primarily composed of lettuce). Due to the recent emergence of the sector, data on food losses specifically occurring in VF are not yet available. A recent survey was however implemented to address losses and waste across leafy green producers from Italy, Spain, France, Portugal, Germany, Greece, Turkey, and Egypt, including both VF (three cases, each with annual production of 9–16 t y−1) and a number of open-field urban farming projects (including 78 commercial initiatives of medium scale, each producing 1–5 t y−1 and 20 small-scale community-driven projects, each producing 0.5–1 t y−1) (Tonini 2024). In the study, the system boundaries included the product from its ready-to-harvest stage until it was sold to consumers (excluding the consumption phase), including the three stages of primary production, post-harvest and distribution, and retail. While all systems shared a short food-supply chain (due to limited distance between production and consumption centres), VF systems distributed their produce through local groceries, whereas open-field initiatives featured direct selling through in-farm shops, farmers markets, and citizen subscription programmes. Overall losses ranged 4–9% in VF and 14–26% in open field agriculture projects. While for VF losses at the production phase were mainly associated with high quality standards required by retailers, in open field farming, damages from abiotic and biotic stressors at cultivation resulted in losses by reduced productivity as well as waste associated with shorter shelf-life. Besides, VF products are commonly marketed as packaged ready-to-eat mixes, where the need for keeping the mix balanced across different crops may ultimately increase losses at the packaging stage (Siñol and Martin 2025). Conversely, customers from community-driven open-field urban farming projects were less influenced by the visual quality of the produce, which ultimately limited waste at retail stage. Overall, reduced losses and waste in VF results from reduced time between production and consumption, the establishment of supply agreements with retailer or consumers (also thanks to the elaboration of demand-driven supply models, e.g. Sugimura et al. 2022), the implementation of models for optimised production and improved plant performances, as well as to the avoidance of losses in response to external abiotic or biotic stresses (Monzini and Orsini 2024; Tonini 2024).

6 Quantifying the contribution of solar energy on VF sustainability

In order to reduce VF electricity dependency on the grid and cut down energy costs, solar-generated electricity from photovoltaic (PV) panels can be integrated into the facility. Previous studies addressed the dimensioning of a PV system for covering the energy needs of a vertical farming facility. Cossu et al. (2023) stated that a PV surface ranging from 5 to 14 times the land surface occupied by the VF area (cultivation chamber plus other facilities for production) is needed to cover its energy needs, resulting in 8 to 23 times the cultivated area of the system. From the study by van Delden et al. (2021), it emerged a lower value (around 3 m2 per m2 of cultivated land, which translates into approximately 28 m2 per m2 of land occupied in a scenario of a vertical farm with nine stacked layers), although the study only considered the energy needed to cover the lighting system requirements. Such values can, however, be greatly dependent on several factors, e.g. the geographical location of the facility, which highly influences the solar radiation potentially reaching the surface of the PV modules as well as their mounting features, the environmental temperature, which affects PV efficacy, and the adopted VF technology, which may affect the energy needs. When plotting the energy consumption reported in Martin et al. (2023a) for a commercial-scale VF facility with the PV power potential provided by Global Solar Data 2.0, Fig. 7 shows how the PV panel surface needed for covering the energy requirement of the facility can range from 2.0 to 6.0 m2 m−2 of cultivated area according to different regions of the world. In warmer climates (where water scarcity may require using VF or high-tech GH instead of open field production), more energy is generated (although both PV panels and heating, ventilation, and air conditioning (HVAC) systems may turn less efficient due to excessive heat, Decardi-Nelson and You 2024), whereas at northern latitudes, a larger land surface is required to host PV panels. However, it should be noted that VF systems normally contain within their premises also post-harvest processing facilities, offices, and other operational rooms. Accordingly, despite the vertical cultivation, the overall covered land surface tends to be greater than the cultivated area (e.g. 7000 vs. 5940 m2 in Martin et al. 2023a). While this should not be considered a drawback limited to VF systems as it is also common for GH to have such facilities, it may avail additional surface for PV panel installation.

Fig. 7
Fig. 7
Full size image

Categorization of world subdivisions (e.g. provinces, states, countries) according to the photovoltaic (PV) panel surface needed (land occupied) for producing the energy used (considering a cradle-to-grave perspective) per unit of growing surface in a VF (cultivated area). The PV power potential for each subdivision was retrieved from Global Solar Data 2.0 (https://globalsolaratlas.info/map?c=11.523088,8.4375,3), and vertical farm energy requirements were estimated building on Martin et al. (2023a) (generated with Map Chart, https://www.mapchart.net/). Classes of surface needed are 2.0–2.4 m2 m−2 (in green), 2.5–3.5 m2 m−2 (in yellow), 3.6–6.0 m2 m−2 (in red), while grey regions had no available PV power potential.

7 Water use efficiency

Among the main advantages of VF systems, the possibility to substantially save water is considered one of the main drivers to choose the technology in water-scarce environments. The maximum achievable water use efficiency (WUE, expressed as g FW L−1 H2O, based on harvestable yield and total water supplied through irrigation) for lettuce has been estimated by Stanghellini and Katzin (2024) to be 850 g FW L−1 H2O, building on data on crop harvest index (De Pinheiro Henriques and Marcelis 2000). Under experimental settings, however, to date, the maximum WUE value reported is 160 g FW L−1 H2O when water from the dehumidification unit was recovered and recirculated to the irrigation system (Carotti et al. 2023). Looking at the reasons for the distance from potential values, in the study, all VF elements that consume water were included in the calculation (e.g. irrigation and HVAC, but also technical operations, e.g. for washing trays) (Carotti et al. 2023). Furthermore, the research was performed within an experimental setting where frequent door opening could have resulted in increased atmospheric water loss through ventilation, as acknowledged within the paper. Alternative figures from commercial operations (that rarely adopt dehumidified water recovery, due to design issues that may affect product safety due to volatile emissions, e.g. from piping system, and elevate costs for achieving air tightness) exist, placing WUE values at roughly 110 g FW L−1 H2O in Sweden (Martin et al. 2023a). When plotting available literature data on WUE values in different farming systems (Fig. 8), it emerges that VF (with average WUE of 140 g FW L−1 H2O) may potentially enable substantial water saving, despite the large WUE variability across case studies that may have hindered more evident differences versus, e.g. high-tech GH.

Fig. 8
Fig. 8
Full size image

Water use efficiency (WUE) of lettuce (Lactuca sativa L.) in open field (white box, mean 15.3±4.63±4.6 g FW L−1 H2O, median 9.6 g FW L−1 H2O, n=16, from Hospido et al. 2009; Maraseni et al. 2012; Barbosa et al. 2015; Bartzas et al. 2015; Goldstein et al. 2016; van Ginkel et al. 2017; Michelon et al. 2020; Maynard et al. 2023), low-tech greenhouses (low-tech GH, white box with diagonal Lines, mean 27.7±5.15.1 g FW L−1 H2O, median 15.9 g FW L−1 H2O, n=24, from Hospido et al. 2009; Bartzas et al. 2015; Kirnak et al. 2016; Muharomah et al. 2020; Çerez and Şahin 2023; Maynard et al. 2023; Li et al. 2024), high-tech greenhouses (high-tech GH, grey box, 68.5±2.1 g FW L−1 H2O, median 66.5 g FW L−1 H2O, n=20, from Hospido et al. 2009; Barbosa et al. 2015; Goldstein et al. 2016; Chen et al. 2019c; Maynard et al. 2023; Wang et al. 2023), and vertical farm (VF, grey box with diagonal Lines, mean 140.7±71.0 g FW L−1 H2O, median 60.7 g FW L−1 H2O, n=88, from Ohyama et al. 2020; Casey et al. 2022; Carotti et al. 2023; Martin et al. 2023a, b; Maynard et al. 2023; Joensuu et al. 2024; Martin et al. 2024). Round symbols refer to individual values (not all symbols are visible due to overlapping), the horizontal line in the boxplot indicates the median, whereas mean value is represented by the X symbol. When multiple values or scenarios were found in the same study, all were included in the boxplot. An outlier for VF (625 g FW L−1 H2O, reported in Casey et al. 2022 that considered a commercial VF in a shipping container and elaborated a LCA inventory building on standard water consumption data declared by the manufacturer) was included in the analysis although not visible in the chart for better visualisation of differences across cultivation systems.

On the other hand, while it is true that GH may also allow for substantial water saving as reported by Stanghellini and Katzin (2024), this is generally associated with elevated increases in energy use, similar to those experienced in VF. When Tsafaras et al. (2022) obtained tomato WUE at 500 g FW L−1 H2O in a semi-closed greenhouse in Saudi Arabia, this came at the energy cost of approximately 0.125 kg FW kWh−1, much below average EUE values for traditional tomato cultivation in low-tech (15 kg FW kWh−1) and high-tech (0.19-0.24 kg FW kWh−1) GH (D’Amico et al. 2023; Maureira et al. 2022). A recent estimate by Righini et al. (2023) proposed that VF could allow tomato production with theoretical WUE values of 730 g FW L−1 H2O, at the cost of very low EUE of approximately 0.02 kg FW kWh−1 (proxy from energy use of 146.3 MJ kg−1). While comparative assessment between semi-closed GH and VF is still difficult (since limited data on VF-grown tomatoes exist and literature is scarce on semi-closed GH production of lettuce), some considerations can be drawn. Tomato productivity (and its consequent high WUE and EUE performances) in GH has increased by 130% in four decades (Heuvelink et al. 2025). This increase is due to the application of new technology (e.g. soilless systems, higher greenhouse transmissivity, and CO2 enrichment), increased cultivation knowledge, and improved cultivars (more productive and resistant to diseases) (Marcelis et al. 2019). In current high-tech GHs, tomato plants may have a stem length of over 10 m at the end of the production season. Current VF technologies are usually designed for small-sized crops (e.g. leafy vegetables and herbs), and for these environments, research on tomato cultivation has mostly used small-sized cultivars (Langenfeld and Bugbee 2023; Zhuang et al. 2024), suggesting that research effort should be placed toward breeding as well as improved management strategies (e.g. dynamic spacing, Karpe et al. 2024) that may enable to largely increase crop performances—therefore allowing for a better comparison vs current GH standards.

Another important water-related feature in VF concerns the implications that the technology may have over the post-harvest water footprint. It is specifically the case of the so-called “ready-to-eat” produce (e.g. greens that are washed and packaged, and therefore ready for consumption), which today represents the major share of VF production. When this business strategy is pursued, VF systems typically have within their premises a packaging unit that enables immediate processing after harvest. While ongoing regulatory framework normally requires “ready-to-eat” food to be washed prior to packaging, national regulations (e.g. in Italy within the National Competitivity Law of 2023, Comegna 2023) are adapting to the innovation by allowing their marketing without washing, whenever the production and post-harvest processing is fully automated and no harmful products (e.g. pesticides) are distributed over the crop. Regulations to discipline VF are still limited globally, but national efforts are emerging (Japanese Minister of Agriculture, Forestry and Fisheries, 2019); Banker 2020; US Senate Committee on Nutrition, Agriculture and Forestry 2023). As automation becomes further integrated in VF systems, thus avoiding human interaction with the produce during cultivation, and as pesticide-free production is achieved, washing is often avoided in current VF products. Given that washing generally requires an average water use of approximately 3.9 L H2O kg−1 FW (Ölmez 2013; Lehto et al. 2014), if regulatory changes allow defining VF products as ready for consumption without further washing, it will be possible to achieve significant water saving. While these regulations may also be applied to high-tech GH-grown products, full GH automation is less likely to occur on a global scale anytime soon; therefore, washing will most probably remain a standard procedure in most high-tech GHs producing for the ready-to-eat sector.

8 Where and how vertical farming may reshape horticulture?

From a sustainability standpoint, whenever the climate is favourable and non-limiting resources (water and nutrients) are available, low-tech solutions (e.g. open field cultivation or unheated greenhouses) present unbeatable efficiency in the use of energy and therefore allow to minimise environmental impact, although at the expenses of increased use of natural resources (mainly land, water, and nutrients). Conversely, both VF and high-tech GH allow for elevated performances in harsh environments or to implement water-efficient fresh food production when the only other alternative is food import. To this end, it is important to highlight that recent reports (Martin et al. 2023a; Joensuu et al. 2024) have evidenced that VF may turn more environmentally sound than the overall impacts of GH cultivation plus import (as displayed in Fig. 6). In the case of Sweden, for instance, a break-even point of roughly 1100 km for imported conventional production may therefore be calculated. Besides, other reasons may drive choice toward VF whenever local resources become available (e.g. within the so-called concept of circular cities) from urban flows (Chowdhury and Asiabanpour 2024a). Urban agglomerates release a number of residual wastes (CO2, organic matter, nutrient-rich water, among others) that, while functional to food production, are often neglected when logistics makes it unfeasible to transfer and use them into traditional rural agriculture systems. Besides, VF could also benefit residential and office spaces with their own waste flows (e.g. residual heat, Blom et al. 2024). VF design that integrates urban flows into food production may indeed further stimulate overall food system sustainability, as recently described, e.g. by Martin et al. (2022), Righini et al. (2022), Song et al. (2022), D’Ostuni et al. (2022 and 2023), and Cossu et al. (2023). When choosing between GH and VF, specific considerations should also be elaborated for climatic regions where specific environmental stressors (e.g. solar radiation, water, temperatures) limit the adoption of low-tech solutions for food production. Since overall economic, social, and environmental sustainability results from a combination of availability and accessibility of all production factors, the selection of which among the available technologies will work better in a specific location requires a thorough assessment.

A further element that has so far not received sufficient consideration is the scenario that will supposedly emerge as the use of fossil fuels (e.g. burning natural gas for heating requirements) may dramatically decrease in the next years to meet zero carbon targets of greenhouse cultivation (e.g. in the Netherlands, van Tuyll et al. 2022). Today, a common strategy in greenhouses is to recover CO2 generated by burners and use it to boost plant photosynthesis. In the future, to sustain crop production, growers will need alternatives to supply CO2 to the plants. Growers will have to either purchase CO2 or recover it from the air or from the digestion of biogenic resources, which will increase the costs for supplementation of CO2. It was recently estimated that CO2 costs may elevate (e.g. in the Netherlands, approximately 0.15 Eur kg−1, as the average price between its delivery by piping systems or truck, Raaphorst and van Tuyll 2023), as emissions from non-biogenic CO2 suppliers decrease as of 2026. Therefore, it is deemed crucial to avoid CO2 losses toward the external environment through tight insulation (as common in both closed-GH and VF). Indeed, when windows are kept closed in a GH, climate control (and specifically cooling) may become energy-intensive, as solar radiation builds up temperatures. As of today, however, comparative studies for GH and VF that forecast CO2 costs under fossil fuel-free technologies are not yet available. Local regulations are therefore expected to significantly impact CO2 costs in the upcoming years, and this will likely affect the choice between VF and GH.

Finally, there may be cases where the optimal farm strategy combines both GH and VF technologies. Such a model—that already finds application, e.g. in flower production (Park et al. 2022)—consists of having initial production stages in a VF (where tight environmental control allows for improved plant establishment and crop uniformity), while successive phases take place inside GH systems. In a similar fashion, VF has been proposed as a viable technology for horticultural seedling production (Blom et al. 2022). The possibility to tightly modulate environmental factors independently and customise light properties features clear advantages for plant nurseries, enabling growers to tailor plant morphological features to meet farmer demand (Niu and Masabni 2021).

9 A need for more reliable information on VF

The year 2023 was marked by the concurrent failure of a number of major VF companies (some of which have then re-entered the market after financial adjustments), altogether with the understanding that a number of existing and emerging technologies (e.g. improved lighting systems and dynamic protocols, optimised cultivation systems that ease air circulation, use of renewable energies, among others) may substantially shape the sector’s sustainability, as also recently discussed in Kaiser et al. (2024). To this end, the speed at which innovation will be generated is highly linked to the successful collaboration between research bodies, policymakers, and commercial operators. Among successful initiatives that contributed to raising awareness and the uptake of agricultural innovation, a clear role has been recently exerted by hackathons and open competitions (Hemming et al. 2019; Orsini et al. 2019; Righini et al. 2022). International bodies are currently looking at the potential of VF’s technology (Monzini and Orsini 2024), and indicators for VF sustainability are being developed to validate its application in different global contexts, also within emerging economies (Paucek et al. 2023; Appolloni et al. 2024).

The journey ahead is dense with challenges and uncertainties that can be overcome only if strong collaboration emerges between players, public administrations, investors, and academia, with lessons learned from past experiences allowing for the integration of economic with both social and environmental sustainability dimensions. For VF to be sustainable, research must address both technological innovation (including automation and AI, horticultural lighting, climate control technologies, monitoring and phenotyping tools and sensors) (Chowdhury et al. 2023; Bhamare and Bansal 2024; Rathor et al. 2024), as well as foster a systemic vision (at whole farm level and within the larger context of urban symbiosis and circularity, as well as its integration within the supply chain) (Martin et al. 2022; Akintuyi 2024; Chowdhury and Asiabanpour 2024b). Plant breeding must address produce diversification, quality improvements, and optimisation of functional traits that consider peculiar conditions of VF environment (e.g. localised water and nutrient distribution nearby roots that allows for increased partitioning toward shoots, photosynthetic performances under elevate PPFD and CO2, limited vertical dimension between cultivation shelves, among others). As these innovations become established, comparative indicators for site-specific identification of food production systems that allow for reduced impacts (as those suggested in Fig. 6 for analysing transport distance thresholds) will become easier to develop and apply.

VF systems, with their controlled and technically advanced nature, offer a unique opportunity to enhance transparency regarding production efficiencies and environmental impacts. To fully realise this potential, it is essential for the industry to adopt standardised metrics, particularly for resource use efficiency. These metrics should clearly relate product output to the resources used, providing a more robust basis for comparing different systems. Additionally, it is crucial to distinguish between cultivated (or illuminated) area and the overall occupied land to accurately assess land use efficiency, which will vary depending on the number of stacked growing layers and the optimisation of unused spaces. To ensure the widespread adoption of these standardised metrics and improve data transparency, stronger collaboration between research and industry is essential. Researchers can play a pivotal role in developing these metrics and establishing data protocols, while the industry can provide real-world insights and feedback, helping to tailor metrics to practical, on-the-ground applications, facilitating the development of resource- and land-use-efficient systems, while also enabling reliable comparative data with traditional agricultural systems, such as greenhouses and open-field farming.

10 Conclusions

The purpose of this article is to provide and integrate available information on vertical farming in order to help researchers further pursue the identification and validation of the pros and cons. More explicitly, this review article aims to analyse (with a specific focus on lettuce as model leafy vegetable crop) available figures on yield and productivity, energy and water use efficiency, and overall environmental impact (expressed as global warming potential) in vertical farming vs traditional cultivation systems (open field, low-tech and high-tech greenhouses) in order to define where and how different technological solutions available to growers offer the best trade-offs for vegetable crop production. The manuscript aims, therefore, to provide quantitative information, tools, and metrics (e.g. greenhouse gas emissions, based on transportation distance, resource-use indicators, land use for photovoltaics integration) that may allow for evaluating which technology is better fitted to the local socio-economic and climate conditions.

Water and land use optimisation, potential for achieving high and stable yields, and possibilities for local food production in harsh environments or densely populated areas are among the main advantages of vertical farms over low-tech greenhouses, although such advantages may, to some extent, also be realised by high-tech greenhouses. As energy use is elevated (in both high-tech greenhouses and vertical farming systems), environmental impacts (e.g. greenhouse gas emissions) are high, although renewable energy sources (e.g. wind, photovoltaics) may substantially reduce burdens. On-site integration of photovoltaic technology may also contribute to vertical farming’s energy needs, although the ratio of land used for hosting photovoltaics vs the cultivated surface may vary significantly across the globe in response to varying incident solar radiation. While this will affect land use efficiency, it should be noted that photovoltaic panels may also be placed on unfertile soils or concrete environments, which tend to be available across urban and periurban settings. As production takes place in proximity to large consumption centres, transport impacts are reduced, as well as food losses and wastes, and circular resource flows may be established. In recent years, a large body of scientific literature on vertical farming sustainability has been released, accompanying the growth of commercial experiences and the emerging awareness of regulatory bodies and the general public. This is leading to unprecedented advances in the comprehension of optimised strategies by both academia and industry, with a rapid flow between research discoveries and innovations being taken up by commercial operators. To this end, vertical farming may certainly benefit from the knowledge gained during the evolution of the greenhouse sector, while also allowing for further advances whenever vertical farming technology overcomes challenges that are specific to greenhouses. While it remains unlikely that in several world regions vertical farms will substitute greenhouses, it may still provide a complementary option for fresh food producers in specific environments where greenhouse production is unfeasible or energy-intensive. Our review shows that key research priorities should include the production of data and insights on resource use and environmental impact across different technologies, on food production and post-harvest strategies, including consideration of quantification of food waste and losses.