Eco-friendly strategy for CO2 enrichment performance in commercial greenhouses based on the CO2 spatial distribution and photosynthesis

CO2 enrichment is an essential environmental control technology due to its significantly enhancing effect on crop production capacity. Despite being a key energy consumer in protected agriculture (i.e. greenhouse systems), CO2 enrichment remains at a low energy use efficiency level, highlighting the need for developing more energy-efficiency strategies for CO2 enrichment. Therefore, this study employed the computational fluid dynamics (CFD) simulation method to replicate the CO2 diffusion process resulting from CO2 enrichment in three commercial strawberry greenhouses with varying geometric characteristics. Based on the CFD-simulated CO2 concentration distributions, the leaf photosynthetic rate was calculated using a mathematical model group. The CO2 enrichment efficiency was then analysed by calculating the ratio of increased photosynthesis across the cultivation area to the amount of energy (in CO2 equivalent) used. The efficiency peaked when the average CO2 concentration was approximately 500 μmol mol−1, thereby providing guidance for determining the target concentration of CO2 enrichment in production. Although this study is limited as the CFD simulation only considered a typical short-period CO2 enrichment event, future research will provide a broader analysis by considering changes throughout the day.


Experimental greenhouses with CO 2 enrichment
To evaluate CO 2 enrichment performance under commercial conditions, three different greenhouses located in Aso City, Kumamoto Prefecture, Japan (32° 57′ 31 N, 131° 03′ 20.9′′ E) were selected according to their sizes and shapes (Fig. 1): Small Greenhouse (length, 40 m; width, 6 m; height, 3.2 m), Long Greenhouse (length, 84 m; width, 6 m; height, 2.9 m) and Long + Wide Greenhouse (length, 72 m; width, 14 m; height, 3.6 m).Table 1 contains the specific geometric information for the three greenhouses.During the experiment, all greenhouses were managed under conditions of normal commercial production 27 and equipped with a CO 2 generator (ZO451, FULTA ELECTRIC MACHINERY CO., LTD., Japan), which generated CO 2 gas by burning kerosene and released it directly into the space inside.In order to accurately control the CO 2 enrichment process, a CO 2 concentration monitoring device was employed in the experimental greenhouses.CO 2 sensors were arranged near the strawberry at the greenhouse centre to monitor the CO 2 concentration inside the greenhouse.In this way, the CO 2 concentration changes could be observed in real-time while controlling it manually or automatically by the on-off action of the CO 2 generator.In commercial production, the target of CO 2 concentration was 1000 μmol mol −1 , and the enrichment automatically stopped when the CO 2 concentration reached the target value.According to the farmer's experience and our in-site observation of the CO 2 concentration, ~ 10 min was needed for the concentration to reach the target value, so 10 min was chosen as the enrichment duration.During CO 2 enrichment, greenhouses were kept closed, and no other environmental control measures were applied.Strawberry plants (Fragaria × ananassa Duch.) were planted in plastic cultivation beds at a height of ~ 1.1 m.The plants were grown in a substrate with an approximate volume of 2 L per plant and irrigated using a nutrient solution.www.nature.com/scientificreports/

Measurement and data collection
During the experiment, various environmental parameters (temperature, humidity, solar radiation, CO 2 concentration, wind speed) within the greenhouses were measured to provide the data required for setting up the simulation environment and verifying model accuracy.These measurements were made in December 2019 and December 2020.Upon multiple repetitions, four replicates were made for the Small and Long Greenhouses and two replicates were made for the Long + Wide greenhouse, enabling model validation (for air temperature, air relative humidity and CO 2 concentration, see Section "CFD simulation accuracy validation").The experimental greenhouse was kept unventilated during the measurements.CO 2 concentrations were measured using CO 2 sensors (GMP 252, Vaisala, Finland; an accuracy of ± 40 μmol mol −1 and a measurement range of 0-2000 μmol mol −1 ) at six points inside each greenhouse.The location and number of sensors were determined by referring to previous studies on CFD analysis 21,22,29 .The air temperature was measured at the same position for CO 2 measurement using T-type thermocouples.The thermocouples were covered with aluminium shield which enables sun protection and natural ventilation.Figures 2 and 3 show the locations of each CO 2 and air temperature measurement point, respectively.The height of measurement points (1, 3, 4, 5 and 6) was 1.5 m, and Point 2 was 2.5 m.This arrangement enabled a more comprehensive grasp of the CO 2 and air temperature distribution at various locations inside the greenhouses.Solar radiation was measured using three pyranometers (PCM01-SD, PREDE, Japan), where two were located at the greenhouse centre and one was located near the greenhouse roof.
The relative humidity of greenhouse air was measured using two humidity sensors (HMP60, Vaisala, Finland; an accuracy of ± 3% and a measurement range of 0%-100%), and these sensors were located around the greenhouse centre, where one was at a canopy height and the other at a 1.5 m height.The generator outlet wind speed was measured using a thermal anemometer (VA21, IET, Japan; an accuracy of ± 3% and a measurement range of 0-40 m s −1 ).The above data were recorded using a data logger (MIJ-01, Environmental Measurement Japan) at a 10-s interval.Temperatures at various locations (soil, PO plastic film and cultivation bed) in the greenhouse were measured using a mobile type of temperature sensor (Tr-52i, T&D, Japan; an accuracy of ± 0.3 °C and a measurement range of − 60-155 °C).Before starting the experiment, all sensors were placed under the same standard environmental conditions and corrected for differences in each instrument.www.nature.com/scientificreports/

Basic equations
Commercial software (FLUENT, ANSYS) was used to simulate the microclimate and CO 2 diffusion processes inside the greenhouse.Due to its sealed nature, the airflow within the greenhouse was primarily driven by natural convection, with the temperature difference between different locations serving as the main influencing factor.
To verify the diffusion process of CO 2 gas inside the greenhouse, unsteady simulations were used in this study.The ideal gas law was chosen to simulate air density change caused by the temperature difference, after which the actual measured temperatures and radiations of various boundaries were used to create a realistic temperature and natural convection environment.The governing equation for physical quantity φ (mass, energy and monument) can be written as follows 30 : where x j and u j represent the coordinate (m) and velocity component (m s −1 ) in the jth direction, Γ φ pertains to the diffusion coefficient of φ, and S φ is the source term of φ.
The airflow inside the greenhouse was simulated using the standard k-ε model because it showed good accuracy in many dispersion simulations and greenhouse studies 15,16,[31][32][33][34][35] .The discrete ordinates (DO) model was used to calculate the solar radiation inside the greenhouse, and the canopy was considered a semi-transparent medium with an absorption coefficient of 0.46 and a refractive index of 2.77 36 .

Crop models
The structure of a canopy can affect airflow, inducing a blocking effect.To simulate this effect, the canopy was used as a porous medium-the most common method for simulating crop canopy in CFD studies 15,37 .Using this method, the canopy-induced resistance could be converted into a sink term in the momentum equation, written as follows: where ρ (kg m −3 ) is air density, LAD (m 2 m −3 ) pertains to leaf area density defined as the leaf area index divided by the canopy height, v (m s −1 ) refers to the wind speed and D represents the drag coefficient.Notably, the constant 0.32 was adopted for D in this study because it has been applied in many simulation studies on greenhouse environments and showed good accuracy for different crops 17,22,30,37,38 .
The strawberry leaf transpiration can cause mass-energy exchange between the crop canopy and greenhouse environment, thereby altering environmental parameters such as humidity and temperature around the canopy.To simulate these processes, the leaf transpiration rate was calculated using plant physioecological process-based models 39 , which were also used to calculate canopy-generated water vapour, which was added to the simulation as a source term: where Tr (kg m −2 s −1 ) is the leaf transpiration rate, g aw (mol m −2 s −1 ) is leaf-boundary-layer conductance to H 2 O, g sw (mol m −2 s −1 ) is stomatal conductance to H 2 O, VPD (kPa) is leaf-to-air vapour pressure deficit and P a (101.3kPa) is atmospheric pressure.
The heat absorbed by transpiration can be written in the following form and added to the simulation as a sink term of energy: where H (J kg −1 ) is the heat of vapourisation.

Settings for CFD simulation
An ICEM-created unstructured mesh was used in this study, and 0.8, 1.2 and 1.5 million elements existed for Small, Long and Long + Wide greenhouses, respectively.The minimum orthogonal quality values for Small, Long and Long + Wide greenhouses were 0.3, 0.23 and 0.28; the maximum aspect ratios were 12.6, 14.9 and 13.4; skewness values were 0.69, 0.76 and 0.71, respectively.
In the simulation, temperature data obtained from actual measurements were used for all the parts inside the greenhouse.The air outlet of the CO 2 generator was set as a mass flow inlet, and the airflow rate was obtained as follows: where W (kg s −1 ) is the mass flow rate, ρ a (kg m −3 ) is air density, A o (m 2 ) is the generator outlet area and v a (m s −1 ) is the airflow velocity at the generator outlet.
Table 2 contains the specific settings and values for each boundary inside the greenhouse.Table 3 presents the physical parameters of the various materials used in the simulation.The average solar radiation amounts received by the canopy during the measurement (10 min) were 365, 178 and 204 W m −2 in Small, Long and Long + Wide greenhouses, respectively. (1)

Scientific Reports
| (2023) 13:17277 | https://doi.org/10.1038/s41598-023-44200-9 www.nature.com/scientificreports/ The SIMPLEC method and second-order upwind discretisation were used in simulations.The convergence criteria of residuals were automatically set by the software.The time step was set to 1 s to balance the calculation speed and accuracy.

Model description
The photosynthetic rate of a single leaf (P) was simulated using plant-environment-combined mathematical models 13 .The model group comprised biochemical photosynthesis 40 , stomatal conductance 41 , and single-leaf transport models for CO 2 and heat.Table 4 contains the specific values of the main parameters used in the model.For the photosynthetic rate calculation, P was detained by the hyperbolic minimum of the Rubisco-limited rate (P c ) and RuBP-limited rate (P j ).
where θ A (0.99) represents the curvature in the transition from one limitation to the other, V cmax (μmol m −2 s −1 ) represents the maximum rates of carboxylation, C i (μmol mol −1 ) represents the leaf intercellular CO 2 concentration, Γ* (μmol mol −1 ) represents the CO 2 compensation point with no respiration, J (μmol m −2 s −1 ) represents (7) θ A P 2 − P P c + P j + P c P j = 0 (8)  where I (μmol m −2 s −1 ) represents the photosynthetic photon flux density, J max (μmol m −2 s −1 ) represents the light-saturated rate of electron transport, ϕ (mol mol −1 ) represents the initial slope of the curve corresponding to the apparent quantum yield of the electron transport at low light conditions and θ J (0.85) represents the curve convexity.
According to the gas diffusion theory of Fick's law, the relationship between leaf photosynthetic rate and CO 2 diffusion for a single leaf could be written as follows: where g ac represents the conductance of leaf-boundary-layer for CO 2 transport, g sc represents the stomatal conductance for CO 2 , and C a (μmol mol −1 ) and C s (μmol mol −1 ) represent the CO 2 concentration at the ambient air and leaf surface.
The leaf-boundary-layer conductance ( g ah ) was calculated by the following equation: where u represents wind speed and d leaf (0.064 m) represents the strawberry leave characteristic dimension.g ah can be expressed as g ac and g aw using g ac = g aw /1.37 and g aw = 1.08 g ah .In this study, the effect of CO 2 enrichment on greenhouse photosynthetic capacity could be demonstrated by using the CFD-simulated CO 2 concentration values to calculate the leaf photosynthetic rates.The model accuracy has been validated in a previous study (Kimura et al., 2020).

Spatial analysis of leaf photosynthetic rate
To quantitatively analyse the spatial distribution of the photosynthetic rate at the plane of 1.5 m height, which was close to the strawberry canopy inside each greenhouse, evenly distributed sampling points in each greenhouse were selected, separated by 2 m in the width direction and 1 m in the length direction (Fig. 4).
Due to the different areas of each greenhouse, 123, 255 and 511 sampling points were observed in Small Greenhouse, Long Greenhouse and Long + Wide Greenhouse, respectively.

Efficiency of CO 2 enrichment
In order to quantitatively analyse the impact of CO 2 enrichment on greenhouse overall photosynthetic capacity and energy utilisation efficiency, the efficiency of CO 2 enrichment (ECE, mmol mol −1 ) was calculated.ECE (10) Figure 5 shows the comparison of measured and CFD-simulated values of air temperature, relative humidity and CO 2 concentration at corresponding measurement points for three greenhouses.The mean absolute errors of air temperature, relative humidity and CO 2 concentrations were 1.3 °C, 6.9% and 79.2 μmol mol −1 , indicating that the CFD simulations achieved good accuracy.Considering the accuracy obtained in previous studies 21,22,29 , the error values were within the normal range for CO 2 simulation using CFD.These results demonstrate that the CFD model accurately simulated the diffusion process of CO 2 in different greenhouses.In addition, the simulation results maintained good accuracy with different greenhouses under varying environmental conditions, indicating that the CFD model had good adaptability.

CO 2 spatial distribution in the three greenhouses
Figure 6 demonstrates the CO 2 concentration distribution inside the Small Greenhouse after 10 min enrichment.A significant concentration difference in the height direction of the greenhouse could be seen, which was caused by the upward convection of CO 2 gas.Because CO 2 was produced by combustion, it had a much higher temperature (65 °C) than the greenhouse air (27 °C), thereby causing CO 2 gas to drift to the greenhouse's upper part and gather there.Therefore, this could limit the effects of CO 2 enrichment since the crops are located in  the lower regions.However, the difference in CO 2 concentration along the length direction was not so large due to its smaller size; the concentration at each location was maintained at ~ 1000 μmol mol −1 .This had positive significance for maintaining a uniform CO 2 concentration environment in the cultivation area.Figure 7 demonstrates the CO 2 concentration distribution inside the Long Greenhouse after 10 min enrichment.Given that this greenhouse used the same method to produce CO 2 , the same problem also occurred in the height direction near the CO 2 generator.However, unlike the Small Greenhouse, a significant CO 2 concentration gradient was observed along the length direction mainly due to the extremely long length (> 80 m) of the greenhouse.In the region near the generator, the CO 2 concentration could reach ~ 1500 μmol mol −1 , while in the region far from the generator, the CO 2 concentration did not significantly increase.The uniformity of CO 2 concentration distribution should directly influence the effect of CO 2 enrichment 7 .Such uneven distribution could reduce the overall effect of increasing yield, posing difficulties in applying a reasonable control strategy of CO 2 enrichment.In many modern greenhouses, environmental control devices are often regulated by real-time monitoring of environmental parameters, usually measured at one point (e.g.around greenhouse centre) 13 .For CO 2 enrichment, setting relevant ranges to control the usage of CO 2 generators is necessary, but uneven distribution makes it difficult to select the sampling points and target concentrations in greenhouses with longer lengths.
Figure 8 demonstrates the CO 2 concentration distribution inside the Long + Wide Greenhouse after 10 min enrichment.The application of CO 2 enrichment in this greenhouse produced quite different results from those in the previous two greenhouses.Notably, the CO 2 concentration in most areas of the greenhouse did not increase significantly, mainly due to the huge volume of the greenhouse, being 4 and 2.5 times the volume of the Small and Long Greenhouses, respectively.Thus, the current application of CO 2 for greenhouses with a large volume significantly failed to achieve good performance.
Because the strawberry canopy CO 2 concentration is considered the most important factor when evaluating the effect of enrichment, the spatial distribution of CO 2 concentration at the 1.5 m height plane was also analysed (Fig. 9). Figure 10 shows a statistical analysis of the CO 2 concentration within the plane, revealing that the CO 2 enrichment performance in the three greenhouses differed significantly.In the Small Greenhouse, the difference in CO 2 concentration in the cultivation area was small, and the overall concentration was maintained at 900-1100 μmol mol −1 .Given the absence of extremely high or low concentrations, the current measures of CO 2 enrichment in this greenhouse achieved a relatively ideal state.In the Long Greenhouse, although the median and average values of CO 2 concentration were both maintained at ~ 900 μmol mol −1 , a relatively ideal state, obvious extreme values in the highest and lowest CO 2 concentrations were observed.With the highest concentration reaching close to 1500 μmol mol −1 , the lowest value stayed at 400 μmol mol −1 level.Notably, most of the recommended target CO 2 concentration from the literature is between 600 and 1000 μmol mol −1 8,10,42,43 , and excessive CO 2 has a limited effect on promoting photosynthesis, indicating that the state for this greenhouse would reduce the efficiency of CO 2 enrichment.Therefore, the primary improvement requirement for this greenhouse type is to achieve a more uniform distribution of CO 2 concentration with the interior.In the Long + Wide Greenhouse, the problem was mainly that the effect of CO 2 enrichment was not noticeable.Although some areas had a CO 2 concentration of 1000 μmol mol −1 near the CO 2 generator, the overall CO 2 concentration level of the cultivation area remained around 400 μmol mol −1 .Therefore, the primary improvement should be increasing CO 2 concentration in the greenhouse.

Analysis of leaf photosynthetic rate in three greenhouses
The biochemical process of leaf photosynthesis should be considered when conducting CO 2 enrichment within greenhouses 7,13 .Figures 11 and 12 illustrate the spatial distribution and statistics of the leaf photosynthetic rate at the 1.5 m height plane for the three greenhouses, respectively, intuitively revealing the effect of increasing CO 2 concentration on crop photosynthesis.Given the constancy assumption of environmental parameters other than the CO 2 concentration, the distribution of leaf photosynthetic rate was attributed to the CO 2 concentration distribution.The photosynthetic rate was unevenly distributed in the Long and Long + Wide Greenhouses compared to the Small Greenhouse, possibly explaining why an insignificant increasing effect of CO 2 enrichment was observed in those greenhouses.Considering the overall photosynthetic rate variations in the greenhouses, the average values for the three greenhouses increased by 28.7%, 23.4% and 7.4%, respectively, for the Small, Long and Long + Wide Greenhouses, compared to their status under the condition of CO 2 concentration at 400 μmol mol −1 .As shown in Fig. 12, the difference in the photosynthetic rate at respective locations in the Small Greenhouse was extremely small, as most areas had photosynthetic rates staying above the 16 μmol m −2 s −1 level.
In the Long Greenhouse, the maximum value of the photosynthetic rate changed slightly compared to the Small Greenhouse, which can be attributed to the nonlinear relationships between the photosynthetic rate and CO 2 concentration, although an obvious variation existed in the uniformity of the photosynthesis rate.In addition, only half of the area had a photosynthetic rate exceeding 16 μmol m −2 s −1 , and the lowest photosynthetic rate  reached a level close to 13 μmol m −2 s −1 .In the Large + Wide Greenhouse, the uneven distribution of photosynthetic rate still existed, and the effect of increasing this rate significantly dropped to 12.5 μmol m −2 s −1 .More than half of the area had photosynthetic rates below 13 μmol m −2 s −1 .

ECE analysis in three greenhouses
Figure 13 demonstrates the change in average CO 2 concentration in the 1.5 m height plane and CO 2 use efficiency under different amounts of pure CO 2 gas (mol) enriched in the greenhouse.Although most researchers have provided an estimated range of suitable CO 2 concentrations 600-1000 μmol mol −1 8,10,42,43 , clarifying the changes in greenhouse photosynthetic capacity and energy use efficiency under different CO 2 usage conditions is needed for improved performance and efficiency.The average CO 2 concentration over the greenhouse plane exhibited an obvious linear relationship with CO 2 usage (Fig. 13).The resulting slope gradually became smaller as the volume increased because the greenhouses had volume variations.Another important factor affecting the concentration change is that the inhomogeneity (standard deviation) of CO 2 concentration in the cultivation planes of the Small and Long + Wide Greenhouses was almost constant despite variations in the amount of CO 2 usage.In contrast, the inhomogeneity in the Long Greenhouse increased significantly as the CO 2 usage increased.Their geometric features could cause this kind of difference, and it mainly depended on the area and aspect ratio (length: width).Due to its smaller area, the Small Greenhouse exhibited a more uniform distribution of CO 2 concentration, resulting in a smaller constant inhomogeneity.The Long + Wide Greenhouse had a similar aspect ratio  but a larger area than the Small Greenhouse.As a result, it had a larger inhomogeneity but also at a constant.The main reason for this is that in the cultivation plane, CO 2 diffusion can be considered a two-dimensional transport along with the greenhouse length and width directions.Therefore, greenhouses with similar aspect ratios have similar transport conditions, leading to constant inhomogeneities in both the Small and Long + Wide Greenhouses.However, the Long + Wide Greenhouse had a much larger area, which would cause a more serious concentration difference in the cultivation plane and result in a higher standard deviation in CO 2 concentration.In contrast, the Long Greenhouse had an extremely higher aspect ratio than the other two greenhouses, which would limit the diffusion in the width direction.Therefore, CO 2 diffusion can be seen as a one-dimensional transport along with greenhouse length.In addition, due to the dominance of CO 2 gas generation over CO 2 diffusion in terms of performance, CO 2 gas tended to accumulate near the generator, particularly at the front.Consequently, as the amount of CO 2 usage increased, the standard deviation of CO 2 concentration expanded, reflecting the accumulation effect.
Notably, the ECE generally decreased as the CO 2 usage increased.The two main reasons for the change in efficiency are as follows.First, a nonlinear relationship existed between leaf photosynthetic rate and CO 2 concentration.When the CO 2 concentration was initially low, increasing it could markedly enhance the photosynthetic rate, but as the concentration continued to rise, the effect gradually decreased 44 , resulting in reduced ECE.In response to this problem, maintaining the concentration in the cultivation area at a lower level seems to be more efficient under the current enrichment method.Second, the spatial distributions of CO 2 concentration and photosynthetic rates within greenhouses are a potential reason.When the amount of usage was small, CO 2 gas was mainly concentrated in the upper part of the greenhouses, so an increase in the concentration around the cultivation area (at 1.5 m height) was insignificant.As a result, the ECE became relatively low.As the amount of CO 2 usage increased, more CO 2 gas diffused from the upper part of the greenhouses to the cultivation area, and then efficiency was improved.Nevertheless, when further CO 2 enrichment was continued, a significant portion of the CO 2 gas was allocated to increasing the concentration in noncultivation spaces.This allocation is likely to have adverse effects on overall ECE.

Eco-friendly strategy for CO 2 enrichment in commercial greenhouses
By analysing the CO 2 , leaf photosynthetic rate distribution and ECE variations in different commercial greenhouses, the following problems were identified for the use of CO 2 enrichment in commercial production: The first is the uneven distribution of CO 2 in the cultivation area.Using circulation fans to provide air movement seems to be an ideal method for solving this problem 45 .However, in this study, the cultivators did not use circulation fans in the greenhouses, possibly due to anticipated pest control difficulties.In addition, when using circulation fans, the airflow created by the circulation fans must be maintained at proper strength to avoid high wind speed damage to crops 46 .
Second, the existing enrichment method could not achieve a desirable increase in CO 2 concentration in large-scale greenhouses.Therefore, burning more fuel to produce more CO 2 from the generator seems the easiest way to solve this problem.However, simply increasing the CO 2 supply from one generator may not be an ideal improvement measure because it would induce a serious uneven distribution of CO 2 , same with the Long Greenhouse.Therefore, increasing the number of CO 2 generators and reasonably arranging their locations may be a better choice.In this case, more generators can ensure sufficient CO 2 supply, and reducing the spacing between generators could effectively prevent evident CO 2 unevenness inside the greenhouse.
Third, the limited distribution of CO 2 at the height direction is another problem.Because the high temperature of CO 2 gas caused this problem, lowering its temperature is the most direct solution.Zhang et al. 14 proved the effectiveness of this measure in improving the distribution of CO 2 in the height direction using CFD simulation.
Finally, through the quantitative analysis of ECE under different amounts of CO 2 usage, setting the target CO 2 concentration at a lower level of approximately 500 μmol mol −1 may be a better choice from an energyefficiency perspective.
The above discussion of improvement measures was based on continually using the existing CO 2 enrichment method.Various improvement measures may improve the application effect to a certain extent, but fundamental improvement is not guaranteed.Therefore, to make a qualitative change in the performance of CO 2 enrichment in commercial production, the enrichment method needs to be fundamentally changed.A recently proposed new enrichment method (i.e.crop-local enrichment) may be an ideal solution 14,24,26,27 , as it enables CO 2 gas transport to the cultivation area while directly releasing it into the crop canopy, thereby creating the highest CO 2 concentration environment just around the strawberry canopy while ensuring good uniformity in the cultivation area.

Conclusion
Spatial distribution of photosynthesis and CO 2 within greenhouses under the CO 2 enrichment was visualised using a CFD simulation-photosynthetic model.These distributions were characterised by the geometric features of greenhouses.Namely, the average value of CO 2 concentration and photosynthetic rate decreased due to greenhouse scale from Small to the Large + Wide tests design issues, and the uniformity of their distribution deteriorated as the aspect ratio increased.In particular, for the Long Greenhouse, due to an imbalance of capacities of CO 2 gas generation and transport, some areas had concentrations below and above the target concentration.However, most areas in the Long Greenhouse had a photosynthetic rate same as that of the Small Greenhouse, which can be attributed to the saturated relationship between photosynthetic rate and CO 2 concentration.Furthermore, ECE peaked when the average CO 2 concentration in the cultivation area was ~ 500 μmol mol −1 for Long and Long + Wide Greenhouses, and a subsequently continuous increase in the CO 2 concentration would reduce efficiency.Considering the CO 2 enrichment problems observed in commercial production, setting the target CO 2 concentration to around 500 μmol mol −1 seems reasonable.Other measures, such as daytime running of circulation fans and crop-localised CO 2 enrichment, could also be better choices, although they are yet to receive widespread applications in commercial greenhouses.

Figure 1 .
Figure 1.Photographs of the three greenhouses of Small Greenhouse (a), Long Greenhouse (b) and Long + Wide Greenhouse (c) used in this study.

Figure 2 .
Figure 2. Dimensions of three greenhouses: Small Greenhouse (a), Long Greenhouse (b) and Long + Wide Greenhouse (c).The height for CO 2 and air temperature measurement points 1, 3, 4, 5 and 6 was 1.5 m and Point 2 was 2.5 m.The height of the air relative humidity measurement Point 1 was around the canopy, and Point 2 was 1.5 m.

5 Figure 4 .
Figure 4. Schematic diagram of sampling points for photosynthetic rate in the greenhouse.The height of the sampling points was 1.5 m, and each point was separated by 2 m in the width direction and 1 m in the length direction.

Figure 5 .
Figure 5. Relationship between measured and simulated values for air temperature (a), air relative humidity (b) and air CO 2 concentration (c) in the three greenhouses (Small, Long and Long + Wide Greenhouses).The shaded area indicates the 95% confidence interval.

Figure 6 .
Figure 6.CO 2 concentration distribution inside the Small Greenhouse after min enrichment.

Figure 7 .
Figure 7. CO 2 concentration distribution inside the Long Greenhouse after 10 min enrichment.

Figure 8 .
Figure 8. CO 2 concentration distribution inside the Long + Wide Greenhouse after 10 min enrichment.

Figure 9 .
Figure 9. CO 2 concentration distribution inside the 1.5 m height plane after 10 min enrichment in three greenhouses (Small, Long and Long + Wide Greenhouses).

Figure 10 .
Figure 10.Box plot for CO 2 concentration at the 1.5 m height plane 10 min after enrichment for three greenhouses (Small, Long and Long + Wide Greenhouses).

Figure 11 .
Figure 11.Distribution of the leaf photosynthetic rate at the 1.5 m height plane after 10 min enrichment in three greenhouses (Small, Long and Long + Wide Greenhouses).

Figure 12 .
Figure 12.Box plot for leaf photosynthetic rate at the 1.5 m height plane 10 min after enrichment for three greenhouses (Small, Long and Long + Wide Greenhouse).

Figure 13 .
Figure 13.The relationships of the average CO 2 concentration at the 1.5 m height plane with standard deviation shown as the colour band (a) and the efficiency of CO 2 enrichment (ECE) (b) with the amount of CO 2 usage in three greenhouses (Small, Long and Long + Wide Greenhouses).

Table 1 .
Geometric information for the three greenhouses.

Table 2 .
Specific settings for each boundary in Small, Long and Long + Wide greenhouses.

Table 3 .
Specific parameters for the materials used in the simulation.

Table 4 .
13ecific values of physiological parameters used for leaf photosynthetic rate calculation in this study13.rate of electron transport dependent on light intensity, K c (μmol mol −1 ) and K o (μmol mol −1 ) represent the Michaelis constants for carboxylation and oxygenation (mmol mol −1 ), respectively, and O (mmol mol −1 ) represents the O 2 concentration. the