The global food sector is responsible for about 26% of total anthropogenic greenhouse gas (GHG) emissions (Poore and Nemecek, 2018), of which roughly 12% can be attributed to the use of manure and synthetic fertilizers on agricultural soils (Smith et al., 2014). The excessive use of fertilizers also leads to eutrophication of aquatic systems, losses in biodiversity and comprised drinking water (Robertson and Vitousek, 2009; Steffen et al., 2015). Additionally, the current industrial agriculture is very resource intensive in terms of land and water consumption (Campbell et al., 2017; Springmann et al., 2018). Therefore, more sustainable ways for providing the growing world population with food are searched for (Jurgilevich et al., 2016; FAO, 2019; Gerten et al., 2020), including the reduction of GHG emissions from plant cultivation and fertilizer losses to the environment. As one recommendation is to shift human nutrition to a more plant-based diet (Poore and Nemecek, 2018; Springmann et al., 2020), minimizing environmental impacts from vegetable cultivation might become crucial in future. Greenhouse cultivation, especially in hydroponic systems, has the potential to grow vegetables in a very resource-efficient way (Gruda, 2009; Savvas et al., 2013), by maximizing yield per area and by minimizing water consumption and nutrient losses. However, there is little knowledge about fertilizer-derived GHG emissions from greenhouse vegetable cultivation (Gruda et al., 2019). Direct GHG emissions from fertilized plant cultivation mainly consist of the release of nitrous oxide (N₂O) and methane (CH₄) from soils or other growing substrates, while carbon dioxide (CO₂) emissions from the root zone are considered to be in balance with photosynthetic CO₂ fixation by the aboveground biomass (Smith et al., 2014). While CH₄ emissions mainly occur in flooded soils under anaerobic conditions, N₂O emissions also occur under well-aerated conditions. N-fertilizers (e.g., ammonium and nitrate) that are not immediately taken up by plants are available to microbial N-transformation processes, such as nitrification and denitrification, which are associated with the release of N₂O (Firestone and Davidson, 1989; Baggs, 2011). N₂O has a global warming potential about 300 times higher than CO₂ on a 100-year scale and, in addition, depletes the vital stratospheric ozone layer (Myhre et al., 2013). Due to the high amounts and dosage rates of N fertilizers as well as favorable climate conditions, N₂O emissions from greenhouse systems might be generally high (Gruda et al., 2019). Indeed, the few existing studies on hydroponically grown cucumbers (Daum and Schenk, 1996b; e.g., Daum and Schenk, 1996a, 1998) and tomatoes (e.g., Hashida et al., 2014; Yoshihara et al., 2014, 2016) revealed substantial N₂O emission rates, with average values of up to 70 g N₂O–N ha⁻¹ d⁻¹. These studies from only two teams, however, were selective and conducted under laboratory experimental greenhouse conditions and may not properly reflect N₂O emissions from commercial tomato and cucumber production. A more recent study on cucumbers cultivated in a large phytotron with cabins of 30 m², which was specifically constructed to measure gas fluxes in the root zones of multiple plants, found relatively low N₂O emission rates of ~17 g N₂O–N ha⁻¹ d⁻¹ during plant cultivation (Nett et al., 2019). Nevertheless, the same study also showed that under certain conditions, like the enhanced degradation of roots after fruit removal and shoot-cutting, N₂O emission rates can rise to high levels in the range of 180–390 g N₂O–N ha⁻¹ d⁻¹. In general, the N₂O release from growing media is known to depend on complex interactions of different variables (Butterbach-Bahl et al., 2013) and has been extensively studied for soils (Stehfest and Bouwman, 2006). The redox potential and oxygen status are of major importance for the microbial processes related to N₂O production (Davidson et al., 2000; Baggs, 2011). Sub-oxic hot-spots together with the supply of organic carbon (C), e.g., as root exudates or from decaying roots, typically increase denitrification rates (Morley and Baggs, 2010; Giles et al., 2017), yielding N₂O emissions from the reduction of nitrate or nitrite. In contrast, nitrifying microorganisms, which convert ammonium to nitrate with N₂O as a side product, typically favor aerobic conditions and are mostly independent of organic C supply (Firestone and Davidson, 1989). Further variables known to affect microbial N cycling are temperature and pH in the growing medium (Farquharson and Baldock, 2007). Because the influence of the different variables on N₂O emission rates can strongly vary over time and with plant growth stage (Daum and Schenk, 1996a), it is important to monitor the emissions regularly during the growing season. In this study we report for the first time seasonal data on N₂O and CO₂ emissions from the root zones of tomato and cucumber plants cultivated in a commercial production greenhouse using rock wool hydroponic systems. Our objectives were to (i) estimate whole-season N₂O emissions and N₂O emission factors from the amount of applied fertilizers, (ii) relate the N₂O emission rates to various influencing variables, including climate conditions inside the greenhouse and plant growth stage, (iii) assess how organic growing substrates alter N₂O and CO₂ emission rates from tomato cultivation compared to rock wool substrate, and (iv) determine how the re-use of rock wool substrate affects N₂O and CO₂ emission rates from cucumber cultivation compared to the use of factory-fresh rock wool. We hypothesized that (a) N₂O emissions from hydroponic vegetable cultivation vary widely over the growing season and are higher on a per unit area basis than for comparable soil-based crops, (b) organic growing substrates increase N₂O and CO₂ emission rates from the root zone of tomato plants, and (c) re-used rock wool with root residues from previous cultivation increases N₂O and CO₂ emissions from cucumber cultivation.

Gas fluxes were calculated using the R software [version 3.6.3; R core team (2020)] and the R package “gasfluxes” [version 0.4-4; Fuss (2020)], with the latter automatically selecting for the best fit model from either linear, robust linear, and non-linear (HMR model) regressions. For flux calculation, the measured concentrations (in ppm) were transformed to μmol m⁻³ according to the ideal gas law under the assumption of SATP conditions (T = 25°C and p = 101.3 kPa). Because each chamber measurement always included two plants, the area (A) to which the fluxes referred was calculated as: A = 2 × Dp⁻¹, with Dp being the plant density (in m⁻²). Further input variables for flux calculation were the chamber air volume (in m³) and the time after closing the chamber (in h). The use of non-linear regression was restricted, as suggested by the gasfluxes package authors, by using the margin of uncertainty from the external standards (±10% for N₂O and ±2% for CO₂ and CH₄) as surrogate for the measurement precision of the GC system. The resulting gas fluxes (in μmol m⁻² h⁻¹) were further converted to g N₂O–N ha⁻¹ d⁻¹, g CH₄ ha⁻¹ d⁻¹ and kg CO₂ ha⁻¹ d⁻¹ based on molar masses. Cumulative emissions (kg N₂O–N ha⁻¹ and Mg CO₂ ha⁻¹) were calculated on the basis of daily N₂O and CO₂ emission rates by linear interpolation between sampling days and summing up daily emission rates over the study periods (trapezoidal method). N₂O emission factors (in %) were calculated through dividing cumulative N₂O emissions by the total amount of N (in kg N ha⁻¹) supplied in the nutrient solution during the study period. Yield-scaled N₂O emissions (in mg N₂O–N kgfruit−1) were calculated through dividing cumulative N₂O emissions by marketable yield (in kg ha⁻¹). CO₂ equivalents (in Mg CO₂ ha⁻¹) were calculated from cumulative N₂O emissions (in kg N₂O ha⁻¹) through multiplying by the 100-yr global warming potential of 298 for N₂O (Myhre et al., 2013).

All statistical analyses were performed in the R software (version 3.6.3). Linear mixed-effects models (LMMs) were done using the R package “lme4” [version 1.1-21; Bates et al. (2015)]. LMMs on N₂O and CO₂ emission rates from tomato cultivation in 2019 included sampling date and location inside the greenhouse (north or south block) as fixed factors, and growing bag identity as random intercept. LMMs on N₂O and CO₂ emission rates from cucumber cultivation in the late season of 2019 included sampling date, location inside the greenhouse (north or south block) and substrate (new or re-used rock wool) as fixed factors, and growing bag identity as random intercept. LMMs on N₂O and CO₂ emission rates from tomato cultivation on different substrates in spring 2019 included sampling date and substrate (rock wool, coir or perlite with wood fiber) as fixed factors, and growing bag identity as random intercept. Prior to LMM analyses, data were log(x + 1)- or sqrt-transformed to fulfill the requirements of LMMs (i.e., normality and homogeneity of variances). Regression analyses between mean gas (N₂O and CO₂) emission rates and climate parameters were done for tomato and cucumber cultivation in 2019 using the “lm” function from the R base package. Prior to analysis, N₂O fluxes were log-transformed to fulfill the assumptions of normality and homogeneity of variances, and all flux and climate data were studentized for better comparability between different units and scales. The sampling dates when no nutrient solutions were supplied to plants were excluded from regression analyses to avoid potential bias because gas emissions were strongly reduced, likely due to dry conditions strongly limiting microbial activity. Permutational ANOVAs were done to determine exact P-values for block and substrate effects on cumulative emissions from cucumber cultivation in the late season of 2019 (substrate and block effects) and tomato cultivation in spring 2019 (only substrate effects) using the R package “lmPerm” [version 2.1.0; Wheeler and Torchiano (2016)]. If a significant substrate or block effect was found (P_exact < 0.05), a Tukey HSD post-hoc test was done on the results from ordinary ANOVA on (1/x)-or sqrt-transformed data using a level of significance of α = 0.05.

Increasing the sustainability of food production is indispensable regarding current global changes in climate and population. One way of approaching this goal is to increase irrigation and fertilizer efficiency (Gerten et al., 2020). Greenhouse cultivation is known to be very resource-efficient, however, for its expansion potential trade-offs, such as GHG emissions from fertilizer application, need to be assessed (Gruda et al., 2019). In this study, we focused on determining the N₂O emissions from hydroponic tomato and cucumber cultivation under real production conditions. Although the CO₂ emissions from the root zone were about 10 times higher than the N₂O emissions converted to CO₂ equivalents (Table 3), the measured CO₂ emissions do not affect the greenhouse gas budget of plant cultivation because of the preceding photosynthetic CO₂ fixation (Smith et al., 2014). In contrast, CH₄ emissions would also affect the GHG budget of plant cultivation but were not traceable in this study. Possibly, CH₄ production was suppressed by rather aerobic conditions in the growing bags and by the high abundance of nitrate fertilizer (Le Mer and Roger, 2001). Despite a few gaps, the data collected during 2019 and 2020 was sufficient for estimating seasonal N₂O emissions from rock wool substrates, since the combined dataset from both years covers all different growth stages of the two vegetable plants. Remarkably, the N₂O emissions reported here are about 10 times smaller compared to findings from previous studies on rock wool-based hydroponic systems. Daum and Schenk (1996a) found that on average 1.2% of the applied N-fertilizer was emitted as N₂O during cucumber cultivation, and Hashida et al. (2014) reported that 4–8% of applied the N-fertilizer was emitted as N₂O during tomato cultivation. In contrast, in our study we found N₂O emission factors of 0.1–0.2% and 0.31% (Table 3) for the N supplied during the cultivation of cucumbers and tomatoes, respectively. These values are clearly below the general N₂O emission factor of 1% utilized by the Intergovernmental Panel on Climate Change for estimating N₂O emissions from crop cultivation on soils (IPCC, 2019).

The lower N₂O emissions compared to previous studies could be due to various factors depending on the hydroponic setup. Especially the irrigation rate (Abalos et al., 2014; Yoshihara et al., 2014) and the draining of the substrate could have decreased the emissions by providing more aerobic conditions, thereby decreasing N₂O production from denitrification. The irrigation technique has been found to be a major influencing factor on N₂O emissions from field-based tomato cultivation (Kennedy et al., 2013; Ye et al., 2019). Indeed, we also found strongly increased N₂O emissions from accidentally water-logged growing bags (Table 5; Supplementary Figure 8A), with up to 40 times higher average N₂O emission rates, underpinning the critical role of precise irrigation and oxygen supply to the root zone in minimizing N₂O production. Another factor limiting N₂O emissions might have been the slightly acidic nutrient solution (pH ~ 5.6) that was supplied to the plants. The activity of nitrifying and denitrifying bacteria is typically highest under neutral and slightly alkaline conditions and decreases with lower pH values (Farquharson and Baldock, 2007). Furthermore, it cannot be excluded that the measurement chambers used here are relatively prone to gas leaking, because the installation and tightening of the chambers in the production greenhouses is very difficult and smaller leaks might have been missed. Nevertheless, mostly linear increases of N₂O and CO₂ concentrations in the chambers over the measurement period of 1 h and extremely high N₂O emission rates found in water-logged substrates (Supplementary Figure 8) indicate that the chamber measurements generally worked well. On the other side, it is possible that we rather over-estimated the N₂O emission rates, because the shown gas fluxes were measured during daytime and extrapolated to 24 h. Additional measurements at different daytimes exhibit that the gas emission rates decrease during nighttime (Supplementary Figure 9), when no nutrient solution was supplied and when the temperature inside the greenhouse is lower. Remarkably, comparable N₂O emission rates and N₂O emission factors were reported by Kennedy et al. (2013) for field tomato cultivation, amounting to 0.5 and 0.8% of the applied N-fertilizer from drip-fertigated and conventionally fertilized crops, respectively. Llorach-Massana et al. (2017) found that lettuce crops on perlite bags emitted 0.7-0.9% of the applied N-fertilizer as N₂O. Similarly, low N₂O emission rates were found by (Nett et al., 2019) for cucumber cultivation on substrate-filled pots.

The study of Nett et al. (2019) also showed that N₂O emission rates can strongly increase if sufficient organic C is available in the substrate, as demonstrated by a peak of N₂O emissions following the degradation of roots after cutting shoots. The N₂O emissions from hydroponic systems are probably mainly due to denitrification, as nitrate is typically used as primary N-fertilizer in such systems (de Kreij et al., 2003). Because microbial denitrification is a predominantly heterotrophic process depending on the supply of organic C (Baggs, 2011), the degradation of plant residues can increase the N₂O production by denitrifying microorganisms (Chen et al., 2013), likely also by limiting oxygen availability due to increased C mineralization (Morley and Baggs, 2010). Previous studies (Hashida et al., 2014; Kazuhiro Shoji, 2014) found a strong increase of N₂O emissions from the long-term use of rock wool substrate. However, in this study the N₂O emissions were only slightly increased in re-used rock wool compared to fresh rock wool growing bags used for cucumber cultivation (Figure 3A; Table 1). Nevertheless, root biomass remained from the previous cultivation in the re-used rock wool growing bags (Supplementary Table 2). Possibly, the effect of re-used substrate was overlaid by the strong block effect, which was due to the delayed planting of the north greenhouse block. Because of technical restrictions, the smaller cucumber plants in the north block received the same amount and composition of nutrient solution as the larger plants in the south block. Consequently, the lower water and nutrient demand of cucumber plants in the north block might have resulted in higher moisture and nitrate contents in the growing bags, yielding increased N₂O emission rates compared to the south block (Figure 3A). In contrast, we could find a clear effect of the presence of organic material in the growing substrate of tomato plants (Figure 4A; Table 4), showing that the use of coir and perlite with wood fiber growing bags increased N₂O emissions almost by 50% compared to rock wool growing bags after 5 months of cultivation in June (Table 5). This effect might even be higher at the end of the growing season, as emissions from rock wool growing bags increased in July and remained at a relatively high level until the end of October (Figure 1A).

In general, we could hardly find a correlation between the greenhouse climatic conditions and the N₂O emission rates (Table 2). On the contrary, the CO₂ emissions clearly reflected the changes in temperature over the growing season, as expected for the general microbial activity involved in the decomposition of labile organic C (Davidson and Janssens, 2006). Despite the known temperature sensitivity of N₂O emissions (Grant and Pattey, 2008) the lowest daily mean temperatures found in the greenhouse may have already been at the optimum for denitrification (Farquharson and Baldock, 2007) or a higher share of N₂O was further reduced to molecular nitrogen (N₂) with increasing temperature (Maag and Vinther, 1996). Similarly to temperature, no clear effect of the supplied amount of nutrient solution was visible in our study. The irrigation frequency was adjusted during the cultivation period according to temperature and solar radiation (Supplementary Figure 10), whereby strong fluctuations in moisture contents inside the growing bags should have been avoided. In hydroponic systems, the N₂O emission rate was found to strongly depend on plant growth stage (Daum and Schenk, 1996a; Hashida et al., 2014). In line with this, we found that the onset of harvest can temporarily increase N₂O emissions from tomato cultivation, potentially by altering plant C allocation with more C substrates translocated to roots. The increase of N₂O emission rates at the end of the growing season from both, tomato and cucumber cultivation, can be explained by the accumulation of senescent roots delivering C substrates needed for denitrification. In this way, the higher N₂O emission factors from tomato cultivation compared to cucumber cultivation (Table 3) might be explained by the higher root biomass of tomato plants (Supplementary Table 2), with a potential further increase in organic C due to the re-use of collected drain solution. However, the higher emission rates in September and October might also be related to a lower plant N uptake, considering the negative relation of photosynthetically active radiation and N₂O emission rates found by Yoshihara et al. (2016). In addition, observed plant diseases at the end of the growing season, like the Agrobacterium rhizogenes (“Crazy Roots”) infection of tomato plants or the mildew on cucumber plants, could also have affected microbial N₂O production by increasing C allocation to roots.

In conclusion, our study demonstrated that hydroponic systems offer a possibility to cultivate vegetable crops with low N₂O emissions if optimal conditions are provided. In particular, high moisture contents together with high C availability in the root zone should be avoided to minimize N₂O production from denitrification. In addition, other reduction possibilities for GHG emissions from greenhouse cultivation should be considered as well. The production of rock wool is associated with high CO₂ emissions, which could be avoided by using alternative (bio-degradable) substrates (Dannehl et al., 2015; Kennard et al., 2020). However, there is more research needed on the interaction of different substrates and GHG emissions as well as yield and quality. Considering the moderate increase in N₂O emissions when using organic-based substrates in our study, there might still be a high reduction potential compared to GHG emissions from rock wool production. Similarly, the GHG emissions from fertilizer production might be reduced by utilizing fertilizers recycled from waste streams. First investigations show that recycling fertilizers are suitable for hydroponic tomato production without increasing N₂O emissions (Halbert-Howard et al., 2020). Other measures that can contribute to reducing GHG emissions from greenhouse cultivation include avoiding heat losses by improved greenhouse insolation, using alternative heating and electricity sources, installing energy-efficient lamps, and using renewable sources for CO₂ enrichment in the canopy (Gruda et al., 2019). Taken all these measures together, hydroponic greenhouse cultivation could help to ensure sustainable vegetable production by reducing the distance between producers and consumers. Because of the high yield to area ratio and the possibility to control most environmental conditions, hydroponic or even aquaponic systems seem to be very promising for food production in urban areas or regions with otherwise adverse climatic conditions.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

DS and KH initiated the study. MG and KH managed greenhouse technics and plant cultivation during the study period. SK and MG conducted the GHG measurements. SK analyzed the gas samples, evaluated the gas flux data, and prepared the manuscript. All authors were involved in the planning, critically read the manuscript, and provided their feedback.

KH is the managing director of the company “Fontana Gartenbau GmbH” and MG is employed by this company. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.