Fluid Dynamics Simulation and Temperature Gradient Validation along a Greenhouse Temperature Gradient

Background: Climate change is increasing the vulnerability of horticultural crop cultivation and production. It is urgent to study such extreme weather phenomena (heatwave, drought, etc.), and in particular, to evaluate crop productivity according to temperature change. For this purpose, the crop physiological response to temperature change in simulated weather conditions was studied. However, there is a limitation in articial light wavelength, which requires experiments to be carried out in protected facilities or open elds. In this study, we simulated temperature differences with computational uid dynamics (CFD) in tunnel-type greenhouses. They can create temperature gradients and improve the accuracy of CFD with vertically and horizontally measured temperature proles. The growth and physiological response of Kimchi cabbage were examined and validated using a temperature gradient within a semi-closed plastic tunnel. Results: Correlation coecients of measured heights were: 1.120, 0.597, and 0.459. Root mean square error was below 0.1025, which means the CFD simulation values were highly accurate. The error analysis showed that it was possible to accurately predict temperature gradients change within a semi-closed tunnel-type greenhouse using CFD techniques. CFD results showed an average error of 0.597°C compared to eld monitoring results. The maximum temperature difference of GTG was 5.7°C, suggesting a well-controlled set point (6°C difference between outside conditions and inside conditions of GTG). In a cloudy day, the gradient temperature of GTG was well maintained by the set differential temperature (dT), which suggests that the set dT was not precisely and accurately performed in GTG of a sunny day. There was a signicant difference in the growth, net photosynthetic rate, transpiration rate, and Ci concentration along with temperature differences in GTG. Conclusions: CFD can simulate temperature gradient distribution in a tunnel-type greenhouse and predict the temperature difference for equipment with different specications. These facilities can be used in climate change-related studies, such as assessment of crop production area optimization, crop physiological response to temperature, vulnerability assessment of crop production under increasing temperatures, or extreme weather. the eld-monitoring results. This is due to the warm air in the eld greenhouse that was stagnated during the whole experimental period. At the 0.4 m height, the CFD results showed relatively higher air temperature compared to the eld monitoring data. This is probably caused by the airow pattern that was concentrated near the oor surface and the pipe inlets blowing hot air. Nevertheless, the difference between eld and CFD results were negligible. The CFD model was able to simulate the eld conditions with very high accordance. the temperature gradient in tunnel-type greenhouses and predict the temperature difference for equipment with different specications. It is expected that these facilities will be used to conduct climate change-related studies for assessments of crop production area optimization, crop physiological responses to temperature, and vulnerability assessment of crop production under increasing temperatures or extreme weather. The CFD simulation model can be used effectively to understand airow patterns and thermal distribution in the greenhouse temperature gradient. CFD simulation showed similar results to the eld experiment since the temperature difference was only ve degrees. CFD model was validated by computation of about 90 air temperature data that were compared with the eld measurements. This resulted in good agreement between the analysis of airow patterns and thermal distribution. After changing the boundary conditions near the entrance, it is possible to simulate the internal environment according to a preliminary intention. Then, the advanced ventilation system, including heating and ventilation structures, can be applied to an experimental greenhouse for maintaining internal environmental conditions. Using the greenhouse temperature gradient with the CFD modeling technique allows us to better comprehend the growth rate of plants such as cabbage.

Experimental studies using GCG and GTG, some aiming to study photosynthetic acclimation, a phenomenon that leads to decreased photosynthetic capacity under long-term exposures to elevated CO 2 , have been conducted [10]. However, there is a lack of information on the speci cation and performance of facilities and devices for temperature gradient implication in an open eld or protected tunnels. Scientists and engineers in agricultural, biological, and ecological sciences have developed a variety of controlled environmental systems for experimental research of plant responses to their environmental conditions [11]. A greenhouse is designed to create a microclimate that allows adequate crop growth and to isolate it from adverse external conditions. To develop a climate control scheme for a greenhouse, it is necessary to obtain a mathematical model, which describes accurately the dynamic behavior of climate variables [12]. The appropriate distribution of climatic parameters, such as temperature, humidity, and CO 2 concentration in greenhouses are the major factors that in uence crop growth uniformity. Thus, investigating the internal climatic pattern is of great importance for improving the design of greenhouse ventilation and cooling, or heating systems. Climatic gradients at plant level within the greenhouse are the main concern of growers. The prediction of such gradients was found to be di cult because the complex nature of such gradients.
CFD (Computational Fluid Dynamics) is based on aerodynamic approaches in terms of heat, gas, dust concentrations, and ow patterns. It can accurately analyze and gather both quantitative and qualitative data inside and outside the agricultural facility. Although the eld experiments in greenhouse conditions are highly important to understand the actual pattern, the speci c environmental factors are limited when the sensing equipment has been installed with a limited number of points.
The installed equipment can only measure within a limited area and the monitoring data can be distorted compared to the actual situation. The internal aerodynamic problems cannot be solved only by eld experiments. This is due to the invisible air ow that should be properly interpreted to understand the change of internal environmental factors. Therefore, it is possible to identify environmental problems, obtain improvement measures, and suggest structural enhancement plans through a process that analyzes the internal ow and thermal environment distribution inside the experimental greenhouse.
This could be done using a veri ed CFD simulation model and eld monitoring.
CFD has been actively used to solve internal environmental problems of agricultural facilities such as greenhouses and livestock houses. CFD model can be applied in various approaches in the agricultural sector [13]. It analyzes the internal greenhouse environmental factors according to the characteristics of the different entrances [14]. CFD can evaluate greenhouse wind pressure, too [15]. Internal environmental conditions in three different types of naturally ventilated greenhouses were analyzed using a two-dimensional transient CFD simulation [16].
There is a lack of information on GTG of CFD case study and validation of growth and physiological crop response to the temperature gradient. Thus, the crop physiological response to temperature change using arti cial weather was studied.
However, there is a limitation of GTG arti cial light wavelength, which requires experiments to be carried out in protected facilities or open elds. The objective of this research is to interpret the internal thermal pattern of the greenhouse temperature gradient and to establish its usability in greenhouse conditions. An indoor temperature and humidity monitoring were performed for three different levels of the greenhouse temperature gradient. A three-dimensional CFD model with the same structural con guration was designed for precise analysis of ventilation, heating, external environments, and internal structure. The CFD model was validated by comparing it with eld-monitored temperature distributions. It was used to analyze the internal environmental conditions of the greenhouse temperature gradient.

Methods
Speci cation of the greenhouse temperature gradient (GTG) There are three GTGs in the National Horticultural and Herbal Science Institute located in Wanju (35°16′N, 127°02′E, 32 m a. s. l.). The experiments were carried out in one of the GTGs ( Fig. 1-a and c). GTG's speci cations and characters are shown in Table 1. The GTG is a semi-open tunnel type with a plastic cover material (Ethylene Tetra Fluoro Ethylene (ETFE), 0.1 mm thin) that has light transiency greater than 89%. The tunnel is a galvanized round iron pipe designed in a Korean standard operation procedure, which means that the shade inside the GTG is below 10%. The GTG's temperature gradient was designed using the environmental control system equipped with a data logger (CR1000x, Campbell Sci. Co., USA), programable relays (SDM-CD16, Campbell Sci. Co., USA), six ventilation fans, and a heater ( Fig. 1-b). The controlling algorithm is shown in Fig. 2. It was used with the six ventilation fans (EK-4070, Hanil Co., Korea) and the hot-blast-heating style heater (Sunart-6000, NS Tech. Co., Korea). The programmable relays were controlled by a mechanical language (supplementary data: Table 1). The ventilation fan control pattern was divided into photo and dark periods with a radiation intensity of 250 w/m 2 (correctable). The ventilation fans were programmed as follows: fans were open and controlled in the above sequence according to the temperature difference change (dT). When the radiation intensity was 250 w/m 2 or less, the ventilation fans were controlled according to the change in dT. During the night, ventilation fans 4 and 6 were not used due to the rapid temperature decrease when the ventilation fans were open. Thus, to avoid rapid temperature decrease, ventilation fans 4 and 6 were not used during the dark-period. The activity of hot air heater was also divided into photo and dark periods based on the radiation intensity of 250 w/m 2 . This was done to control its activity in the same way as the ventilation fans. When the solar radiation was 250 w/m 2 or greater, the hot air heater was not operational regardless of dT. When the solar radiation was 250 w/m 2 or less, the hot air heater was working according to the change in dT.   Signi cance y *** *** *** *** *** *** z The letters represent the signi cance of Duncan's multiple range test at p = 0.05. Mean ± standard deviation. y ***P ≤ 0.001.

Cfd Case Study And Speci cation Of Gtg
The CFD model is a tool for numerical analysis of non-linear partial differential equations for uid ow phenomena with Navier-Stokes equations. It has been used to analyze heat transfer, multiphase ow analysis, and chemical reactions along with the uid ow. The greenhouse temperature gradient was designed for analysis of indoor environments using spaceclaim (ver. 19.2., ANSYS Inc., USA).  were calculated. They were compared with the eld monitored data and CFD-computed data.
Using the validated CFD model, the internal air ow, temperature distribution, and ventilation e ciency were evaluated at heights of 0.4 m, 0.8 m, and 1.2 m from the oor bottom. TGD (Tracer Gas Decay) method was used to evaluate regional ventilation e ciency. The method tracks gas concentration reduction after opening the ventilation system. Thus, the regional ventilation e ciency of CFD was calculated using TGD to reduce CO 2 concentration at the initial condition to 2,000 ppm. The portable Li-COR gas exchange system (LI-6400XT, LI-COR, Inc., USA) maintained the ow rate and photosynthetic photon ux at 300 µmol/s and 500 µmol/m 2 /s, respectively. The leaf chamber block temperature was set at 25 °C and the relative humidity at 40-60%. All measurements of fully unfolded leaves with green tissues were replicated three times.
The water use e ciency was calculated as a ratio between the net photosynthetic rate and the transpiration rate. The CO 2 assimilation rate (A) is a function of the intercellular CO 2 concentration (Ci) curve (A-Ci curve) and the A response to the light intensity curve. They were measured in 52 and 59 DAT Kimchi leaves, respectively, using a portable photosynthesis system (Li 6400XT, LI-COR Co., Inc., USA). The fully expanded leaves were clamped onto a 6-cm 2 leaf chamber and the light was provided by a sensor head (6400-02B LED, LI-COR Co., USA). The relative humidity in the leaf chamber ranged from 70 to 80% and the block temperature was maintained at 25 °C. For the An-Ci curve, gas exchange responses to the various CO 2 concentrations ranged from 50 to 1,500 µmol/mol. They were measured at 500 µmol/m 2 /s photosynthetic photon ux (PPF) with a 10% ratio of blue light to maximize the stomatal aperture. The light response curves were measured at irradiance levels of 1800, 1500, 1200, 1000, 800, 500, 250, 100, 50, and 0 µmol/ m 2 /s PPF with 400 µmol/mol CO 2 concentration. Three replicates per treatment were performed. The calculation of CO 2 , light saturation, and compensation points were done with a regression analysis using a descriptive model (exponential rise to maximum, three parameters).

Pro les of air temperature in a semi-closed plastic tunnel
The outside GTG environment factors: radiation intensity, air temperature, relative humidity, wind speed and direction, and precipitation in the cloudy and sunny days are represented in Fig. 5. The total integral solar radiation was 0.5 and 9.4 MJ/m 2 /d for the cloudy and sunny days, respectively. The maximum radiation intensity was approximately 25 and 452 W/m 2 , respectively. The daily mean temperature of the cloudy day outside GTG was 7.1 °C, while for the sunny day it was 6.1 °C. The relative humidity in the cloudy day was 90.4%, the average wind speed was 2.5 m/s with southwest direction, and the precipitation was 11.2 mm/d. However, the relative humidity outside GTG in the sunny day was 69.1%, the wind speed was 1.1 m/s, and the wind direction was mainly southeast.
The daily mean temperature in the cloudy day was 7.5 °C-8.8 °C in zone A of the GTG ( Table 3). The temperature difference between zone A and B was approximately 3.0 °C. The maximum temperature difference of GTG was 5.7 °C, which means a well-controlled set point (6 °C difference between GTG outside and most-inside GTG conditions). The particular reason that the average daily temperature measured at the farthest distance from the outside was lower (about 0.7 °C) compared to the previous position is the installation of a fan that leads to temperature difference and heat loss. Also, the temperature difference between day and night occurred according to the house distance similar to the daily average temperature result.
On the sunny day, the daily mean temperature close to the outside of the GTG was 6.3 °C-8.0 °C, 8.3 °C-9.2 °C, and 9.8 °C-10.9 °C in A, B, and C zones, respectively. In the C zone, which was 15.6 m inside GTG, or away from the outside environment, the daily mean temperature was up to 4.3 °C higher than that of A zone. Besides, the temperature difference between zone A and zone B was about 2.9 °C and the temperature difference between zone B and zone C was 1.7 °C. The temperature gradient in the sunny day was not well pronounced inside the GTG. The temperature difference between the photo and dark periods was similar to the daily mean temperature trend. However, the temperature gradient manifestation in the GTG was reduced in the sunny day. Even after a month of continuously collecting and analyzing data, it was obvious that the GTG temperature gradient is manifested according to the distance from the outside environment (data not shown).

Validation Of Measured Data And Cfd Simulation Values
The CFD model was validated by comparing eld monitoring data with CFD-computed results. The error of the CFD model was on average 1.120 °C with a standard deviation of 0.640 (Fig. 7). The air temperature of CFD simulation at 0.4 m height was relatively higher than the eld temperature. Comparing the eld experiment to the CFD model in terms of temperature and ventilation ow, it seems that the air entering through the pipe inlet of the CFD model was well supplied. CFD modeling was performed to provide hot air heated at the one side through pipe inlet based on the design capacity. The generated heat was relatively lower than the desired capacity, which was caused by the pressure generated in the eld experiment. In the  Cfd Simulation In Regional Thermal Condition Figure 12 shows the thermal distribution of the experimental greenhouse temperature gradient measured at a selected period for CFD model validation. Figure 13 shows the thermal distribution computed by the CFD model at the same initial conditions, which represents almost the same air temperature distributional pattern. Temperature gradients in eld experiments showed 4.60 °C, 4.52 °C, and 5.62 °C at 0.4, 0.8, 1.2 m heights, respectively. The measurements were based on the temperature difference between six points near the inlet opening and six points near exhaust fans. The CFD-computed results also showed almost similar results. They con rmed that the air temperature can be maintained between 13 °C-15 °C through the half greenhouse. The air ow pattern computed by CFD showed relatively parallel results to the greenhouse. This was achieved by creating a negative pressure formed by the exhausting fans, which were installed on the other sidewall (Fig. 14). Since the incoming wind in the eld experiment has entered in irregularly through the open wall, the CFD-computed air ow pattern and the thermal gradient were more uniform.
GTG application for the growth and photosynthesis characteristics of Kimchi cabbage The growth of Kimchi cabbage was evaluated according to three different GTG temperature zones (A, B, and C). The daily mean temperatures in zone A, B, and C during the Kimchi cabbage cultivation period (60 days after transplanting) were 20.3 °C, 22.7 °C, and 25.1 °C, respectively (data not shown). The maximum leaf length, leaf number, leaf area, fresh weight, stomatal conductivity, and leaf temperature were signi cantly different. However, there was no signi cant difference in leaf width and SPAD value ( Table 6). Zone A growth conditions are closest to the outside. It had the lowest growth rate (leaf growth, leaf area, and fresh weight) among treatments and the stomatal conductivity decreased by 98.9% compared to other higher-temperature zones in GTG. In particular, the stomatal conductivity of Kimchi cabbage leaves increased signi cantly along the temperature gradient. There was a signi cant difference in the net photosynthetic rate, transpiration rate, and Ci concentration along the temperature gradient, too ( Table 7). The net photosynthetic rate increased in zones B and C where the temperature was high, while WUE of the Kimchi cabbage grown at low temperature (zone A) was 4.3 times higher than in the zone C. Signi cance y ns ** ** * * ns ** *** There was a difference in the maximum saturation of Ci concentration, the maximum photosynthetic rate according to Ci concentration, and the exponential decay model . The regression analysis showed that the Kimchi cabbage in zone B was the highest, followed by the Kimchi cabbage grown in zone A. The area with the lowest maximum photosynthetic rate was zone C (Fig. 8-a). Among the regression equations of the A-Ci curve of zone A, Kimchi cabbage coe cients a, b, and c were − 4.7, 30.9, and 0.005, respectively. The coe cient of determination (r 2 ) was 0.9959.
The regression coe cients a, b, and c of the A-Ci curve of zone B were − 5.4, 35.4, and 0.004, respectively (r 2 = 0.9755). In the zone C, the coe cients a, b, and c were − 5.9, 33.9, and 0.004, respectively (r 2 = 0.9985). The CO 2 compensation point/saturation point of Kimchi cabbage leaves in zones A, B, and C were 33/369, 41/405, and 48/461 µmol/mol, respectively. As the light intensity increased, the photosynthetic rate increased in all treatments (Fig. 8b). The light curve of zone C, which had the optimal temperature for Kimchi cabbage, was most pronounced. The maximum photosynthetic rate decreased when the temperature decreased below the optimal growth temperature of Kimchi cabbage. The Kimchi cabbage light compensation/saturation point in zones A, B, and C were − 21/647, 8/468, and − 18/1008 µmol/ m 2 /s, respectively.

Discussion
Intergovernmental Panel on Climate Change (IPCC) published climate change (CC) models that simulate air temperature suggesting that it will rise with 1 °C to 4 °C by the end of the 21st century due to increased greenhouse gas emissions in the atmosphere. When designing and implementing a GTG, the representative concentration pathway (RCP) 8.5 CC scenario of the Korean Meteorological Administration should be taken into account. This report con rmed that after 100 years, the temperature will increase by approximately 6 °C from the current average level without any mitigation policy applied. As a result of the temperature deviation pro ling of GTG, the temperature difference set in the control program was predicted very accurately (Figs. 4 and 5, Table 3). There are many approaches regarding the temperature gradient and CO 2 concentration change in semi-open type greenhouses. The research of climate change impact on crop production using GTG showed that the temperature gradients in a partially shaded large greenhouses were created with evaporative cooling pads [17]. Those facilities had well-controlled set points. There were similar works on gradients of temperature, humidity, and CO 2 for fanventilated greenhouses [18]. These studies focused on the night-time thermal gradients in the greenhouse using computational thermal dynamics [19]. Other similar studies have been conducted on sunlit plant growth chambers with controlled-environment [11]. There are also studies regarding the construction of large-scale semi-eld facilities to study genotypic differences in deep-root growth and resource acquisition [20]. Those studies claimed that accuracy and precision are important for the device or facility's performance. Therefore, this has become a major task for the assessment of crop impacts and vulnerability due to climate change.
In studying the response of agriculture research elds to climate change, it is necessary to rely on growth models for the study of crop physiological and biochemical reactions. Also, the prediction of crop production should be done in terms of statistical and mathematical models, which to be calibrated and veri ed. In such cases, a chamber using arti cial light is mainly used, or agricultural statistics for national surveys are occasionally used. However, calibration and veri cation methods can lead to errors in the calculation of actual crop biomass. GTG was used in various crops as utilization equipment for correcting and verifying such errors. Using of GTG experimental facilities under high temperature and drought conditions led to signi cantly and consistently increased pH due to decreases in malic acid of grapevine [21]. When plants were grown under elevated CO 2 , the largest increase was observed in leaf (white) and root (red) growth. An increase in the mean temperature with 4 °C had no consequences on the vegetative growth of grapevine. Yield and yield-related traits were unaffected by the climate change scenario [3]. Previous research revealed that an increase in air temperature, CO 2 concentration, and precipitation signi cantly disturbed plant growth and reduced the yield of horticultural crops [1,5,[7][8][9][22][23][24][25]. Especially, using the extreme weather simulators [26] have documented the impact of moderate and extreme climate change scenarios on growth, morphological features, photosynthesis, and fruit production of hot pepper. Similar work has dealt with the temperature gradient and air-soil as a factor in the optimization of net photosynthesis in whole plants [27]. Finally, stem juice production of the C4 sugarcane (Saccharum o cinarum) is enhanced by growth at doubleambient CO2 and high temperature [28].
Compared to zone B, the fresh weight was lower in zone C, which had the highest temperature in the GTG. This was due to the approximately higher temperature (with about 3 °C) compared to the Kimchi cabbage optimal cultivation temperature (average 20 °C), leading to partly physiological damage to the head formation ( Table 6). The combination of elevated temperature and CO 2 concentration stimulated leaf dry weight reduction of Kimchi cabbage. It decreased about 3-fold compared to that only exposed to elevated CO 2 . A drastic decrease was also observed in stomatal conductance, internal CO 2 , and photosynthesis rate [22]. A temperature increase from 28 °C to 32 °C resulted in less leaf calcium, magnesium, and manganese concentrations (dry weight basis) in pakchoi without signi cant changes in shoot dry weight, suggesting potential negative effects of the elevated temperature on pakchoi leaf nutrient status [23]. The moderate and extreme climate change (CC) treatments led to physiological disorders in terms of tip burn leaves, while the control plants with normal climate condition applied by weather of primary production area had the greatest yield among the studied plants. The yield of moderated treatment 65% reduction observed in moderate CC treatment plants. The CC scenarios retarded the growth of Kimchi cabbage and negatively affected leaf morphology, photosynthesis e ciency, and yield [29]. In general, Kimchi cabbages are harvested around 50 to 100 days after sowing, depending on the cultivar, cropping type, cultivation season, and the prevailing weather conditions. Kimchi cabbage is represented as a cool-season crop, thus that will not occur head formation when the daily mean air temperature is over 25 °C. The net photosynthetic rate increased in hightemperature zones, while WUE of Kimchi cabbage grown at low temperature (zone A) was 4.3 times higher than in the zone where the temperature was highest ( Table 7). The e ciency of photosynthetic system improved under warm temperature.
However, WUE in cool season crops is normally strengthen by optimal temperatures [27,29,30]. Regarding the A-Ci and light curves (Fig. 8), the maximum net photosynthetic rate of Kimchi cabbage in zone B was highest. It was followed by the Kimchi cabbage grown in zone A. The area with the lowest maximum photosynthetic rate was zone C. Previous research revealed that the maximum carboxylation rate, maximum rate of electron transport, and triose phosphate utilization rate of the higher temperature treatment plants were signi cantly lower than those of the control [7,8,22,26,29,31,32]. It was con rmed that the GTG can be applied to climate change response and adaptation studies. Therefore, it is expected that this study can also be used to evaluate the vulnerability and impact of climate change in open-ended vegetables, especially by utilizing GTG.

Conclusions
The GTG was well-designed to create a temperature gradient. The cultivation experiment of Kimchi cabbage, a lowtemperature crop, was useful to study the growth and physiological response to the temperature gradient. CFD can simulate the temperature gradient in tunnel-type greenhouses and predict the temperature difference for equipment with different speci cations. It is expected that these facilities will be used to conduct climate change-related studies for assessments of crop production area optimization, crop physiological responses to temperature, and vulnerability assessment of crop production under increasing temperatures or extreme weather. The CFD simulation model can be used effectively to understand air ow patterns and thermal distribution in the greenhouse temperature gradient. CFD simulation showed similar results to the eld experiment since the temperature difference was only ve degrees. CFD model was validated by computation of about 90 air temperature data that were compared with the eld measurements. This resulted in good agreement between the analysis of air ow patterns and thermal distribution. After changing the boundary conditions near the entrance, it is possible to simulate the internal environment according to a preliminary intention. Then, the advanced ventilation system, including heating and ventilation structures, can be applied to an experimental greenhouse for maintaining internal environmental conditions. Using the greenhouse temperature gradient with the CFD modeling technique allows us to better comprehend the growth rate of plants such as cabbage.

Availability of data and materials
The raw data from all experiments, as well as the material used in this manuscript, can be obtained from the corresponding author upon reasonable request.      Outside environmental conditions during the measurement of air temperature for validation of CFD simulation data. a: solar radiation, b: air temperature, c: relative humidity, d: wind speed, e: wind direction, and f: precipitation.