Utilization of irrigation, drainage, and electrical conductivity data for efficient use of nitrate in a soilless culture system


 Nitrate management in agricultural systems has mainly been established based on nitrate supply and the yield response curve. In the case of intensive fertilization systems such as soilless culture, the nitrate amount usually remains above the curve's optimal point. A surplus nutrient supply under these conditions could result in the excessive emission of chemical fertilizers. However, very few studies have developed a decision-making process for the efficient use of nitrate under the soilless culture system online. This study was conducted to develop an indicator related to the absorption of nitrate that can be applied in online systems utilizing the monitored irrigation and drainage amount data, electrical conductivity (EC), and the nitrate analysis data of irrigation and drainage. In the simulation, a stochastic change was generated for the nutrient absorption rate. The cultivation experiment verified the theoretical prediction, and a higher correlation of tomato yield with the nitrate absorption indicator was confirmed than with the nitrate supply amount. Also, the normalization of indicator and tomato yield showed dynamic time-series responses. The simulation and cultivation experiments showed that the indicator related to nitrate absorption estimated by online EC, irrigation, and drainage monitoring provides useful theoretical and experimental frameworks regarding efficient resource management decisions.


32
From the planetary boundaries perspective, the global nitrogen cycle has already transgressed the 33 boundary that humanity can operate safely (Steffen et al., 2015). Along with the growing interest in 34 sustainability, the soilless culture is recently receiving significant attention as one of the promising 35 approaches to manage fertilizers and water by the closed-loop system ( Balancing the yield increase while decreasing nitrogen consumption has long been challenging for 45 sustainable agriculture (Tilman et al., 2002;Zhang et al., 2015). Conventionally, nitrogen management 46 in plant production has mainly been established based on nitrate supply or concentration and the yield 47 response curve. An increase in the supply of nitrogen from a range of deficiencies leads to an increase 48 in yield (Engels et al., 2012). To date, quantitatively summarized nitrogen use and yield response had 49 been used as the primary decision-making process to use appropriate fertilizers (Pan et al., 2020). In an 50 open field agricultural production system, the cultivation fields represent a broad range of plant 51 nutritional conditions, from deficiency to toxicity. However, the yield response curve using the amount 52 of nitrogen supply is only sensitive within the range from deficiency to optimal conditions (Engels et 53 al., 2012). Therefore, under controlled nutrition conditions where nitrogen is managed mostly at 54 moderate or excessive levels, such as in soilless cultures, there are technical difficulties in the online 55 evaluation of efficient nitrogen use (Massa et al., 2011). Therefore, it may be challenging to solve the Thus, it could be expected that the extended application of these approaches to the utilization of nitrogen 82 absorption information is also feasible. 83 In the present study, the indicator related to the absorption of nitrate was investigated by the 84 simulation and experimental analyses. The error-provoking conditions, such as intermittent irrigation 85 control, subsequent fluctuations in nutrient concentration, and nonhomogeneous nutrient distribution in 86 the substrate, were simulated. Under these simulated conditions, the nitrate absorption indicator was 87 determined based on irrigation, drainage, EC, and nitrate concentration. The nitrate absorption indicator 88 was applied to the actual soilless culture system to analyze crop yield correlation with the nitrate 89 absorption indicator. In addition, to broaden the range of nitrate absorption of plants, some cultivation 90 experiment lines were subjected to LED inter-lighting conditions. 91

Simulation analysis of nutrient uptake estimation 93
The model used in the present study simulated the automated nutrient and water management of a 94 soilless culture system in which nutrient absorption, solar radiation, solar radiation-based irrigation 95 control, transpiration, and water content change in the substrate were included (Fig. 1a). The model 96 with cloud cover according to the solar altitude estimation equation was used for the simulation of solar 97 radiation change (Holtslag and Van Ulden, 1983): 98 where, + is the reduced solar radiation by cloud cover; 0 + is the incoming solar radiation at ground 100 level under clear skies, determined by the changes in solar altitude over time and location on the ground; 101 1 and 2 are the empirical coefficients; and is the total cloud cover. is a value between 0 and 1; 102 the closer to 0, the clearer the day and the closer to 1, the cloudier the day. In the simulation analysis, 103 dynamic weather changes were applied using the random-walk process method. In the soilless culture 104 system, the irrigation was controlled based on the integrated solar radiation of + and it followed the general greenhouse irrigation automation technique (Shin and Son, 2016). Nutrient and water transfers 106 were made by referring to the nutrient transport model under substrate conditions (Silberbush et al., 107 2005), and the interconnection between the models was based on Ahn and Son's soilless culture system 108 model (Ahn and Son, 2019). (3) 120 where, is an arbitrary coefficient for applying the multiplication factor to the nutrient absorption 121 rate. In the present study, was used to simulate the stochastic changes in the nutrient absorption 122 rate.
corresponds to a random-walk process that increases or decreases with a certain The where, is the transpiration rate (L min -1 ), and are the empirical coefficients, is the 127 extinction coefficient in the plant canopy, is the leaf area index (LAI), and is the vapor 128 pressure deficit (VPD). The LAI is a fixed value for the simulation. The tomato leaf area used in the 129 LAI calculation was estimated by measuring the nondestructive leaf area of the cultivated tomato at the 130 same time as the measured environmental data used for simulation verification (Carmassi et al., 2007). 131 The VPD was simulated to be shifted by the random-walk process between 0.5 and 2.0 kPa to apply the 132 stochastic fluctuation for transpiration. For the simulation of EC based on the nutrient solution supply 133 method, the EC was calculated by an empirical equation for converting the equivalent concentration 134 into EC (Savvas and Adamidis, 1999). Under the simulated conditions, the day nutrient absorption index 135 for total nutrients (DNAIEC) and nitrate (DNAINO3) were calculated as the difference between the daily 136 nutrient inflow into the substrate and the outflow from the substrate: 137 where, and are day after DNAI calculation and present day, respectively, , , and 140 are the daily EC (ds/m), nitrate concentration (mM), and volume of the irrigated nutrient solution 141 (L), respectively and , , and are the daily EC, nitrate concentration, and volume of 142 the drained nutrient solution, respectively. We analyzed the effects of the nutrient absorption rate and 143 drainage ratio in the substrate using simulation analysis and the correlation of DNAIEC (a.u.) and 144 DNAINO3 (mmol) with total nutrient absorption and nitrate absorption, respectively. The main 145 parameters used in this simulation are summarized in Table 1.

Inter-lighting treatments for disturbance application on nutrient absorption 172
In the present study, inter-lighting was used as a factor that could affect the absorption of nutrients. 173 In tomato cultivation, inter-lighting can affect the production of photosynthetic assimilates, which can 174 be a factor in increasing yield (Tewolde et al., 2016). Thus, the treatment of inter-lighting can have a 175 significant effect on plant growth and nitrate absorption. The treatments consisted of three lines of inter-176 lighting (Inter1-3) and four lines of control (L1-4) for a total of seven measured hanging gutter lines. 177 The treatment for inter-lighting started on 87 DAT. Inter-lighting performance (LT080, Luco Corp., 178 Korea) has a photosynthetic photon flux density (PPFD) of 168 μmol m -2 s -1 , and the distance from the 179 tomato was 10 cm. The inter-lighting irradiation time was adjusted to two experimental conditions. For 180 the first experimental condition, irradiation was applied for 12 h from 22:00 to 10:00 the next day, 181 following the results of tomato inter-lighting by Tewolde et al. (2016). However, after the initial inter-182 lighting treatment, apparent stress symptoms such as chlorosis and scorch were observed. Therefore, 183 the second light irradiation condition was adjusted on 106 DAT, and the irradiation was conducted for

DNAIEC and DNAINO3 in the simulation analysis 194
By applying the random-walk process to the cloud cover, the simulation results showed a change 195 in the solar radiation and substrate moisture content (Fig. 2a). The changes in nitrate concentration and 196 EC in the substrate were simulated due to the nutrient supply, plant nutrient uptake, and drainage 197 generation based on the water content of the substrate (Fig. 2b and c). In an actual soilless culture system, 198 the probability of various outcomes under different environmental conditions could be happened. 199 Therefore, a stochastic change was applied, and the simulation was replicated with a change in the rate 200 of nutrient absorption from various pathways for nutrient absorption changes (Fig. 2d). Average nutrient 201 absorption factors of 0.88 and 0.47 were calculated during the simulation iterations, and the changes in 202 the nutrient absorption factor of various distributions between approximately 0.2 and 1.2 were simulated. 203 DNAIEC, total nutrient absorption, DNAINO3, nitrate absorption, and the correlation analysis between 204 DNAIEC and DNAINO3 were measured (Fig. 3a-c). DNAIEC and DNAINO3 showed correlations in 205 different ranges depending on the nutrient absorption factors. Specifically, they showed a very high 206 positive correlation. However, the coefficient of determination was higher on the side that had higher 207 nutrient absorption. Based on the average drainage ratio during the simulation period, the DNAIEC, NO3-208 nutrient absorption coefficient of determination showed that the coefficient of determination (R 2 ) was 209 decreased as the drainage ratio was decreased (Fig. 3d). In addition, the decrease in the R 2 value with a 210 decrease in drainage ratio was greater in the low nutrient uptake magnification distribution. By monitoring during the DNAI calculation period, nitrate supply, average discharged 214 concentration of nitrate, discharge amount, irrigation amount, yield, and partial factor productivity were 215 summarized (Fig. 4). During the experimental period, the cumulative amount of nitrate supplied was 216 relatively slowly increased until 100 DAT. However, a change in the nitrate supply rate was observed 217 over time (Fig. 4a). The nitrate concentration in the drainage decreased during half of the measurement 218 period. However, repetitive trends of increase and decrease were observed over time (Fig. 4b). The 219 discharge amount of nitrate in the drainage was different among treatments (Fig. 4c). Different irrigation 220 amounts were also observed for each treatment, and there was a difference between 262 L (minimum) 221 and 296 L (maximum) (Fig. 4d). The apparent difference was not observed in the cumulative yields and 222 was not different for each treatment during the early stage of DAT; however, it was observed that the 223 deviation increased with increasing DAT (Fig. 4e). A difference in the partial factor productivity of 224 each gutter line was observed from 176 (minimum) to 219 kg yield/kg N use (maximum) (Fig. 4f). 225 The DNAINO3 value slowly increased, similar to the initial nitrate supply amount. In particular, the 226 DNAINO3 value rapidly increased from 95 DAT to 105 DAT (Fig. 5a). However, the deviations between 227 each treatment were large after 105 DAT. DNAIEC showed a similar tendency to that of DNAINO3 (Fig.  228   5b). The R 2 between DNAIEC and DNAINO3 was 0.98, which showed a very high positive correlation. 229 The cumulative nitrate supply amount and tomato yield showed a high level of correlation at 87 DAT 230 and 93 DAT during the initial period of the experiment compared to the other monitoring days (Fig.  231   6a). However, the tendency was different from each other, moving from a positive to a negative 232 correlation. At 87 DAT and 93 DAT, the median positive and median negative relationships were 233 analyzed, respectively. The average nitrate concentration of discharge and yield showed a high level of 234 correlation at 87 DAT and 93 DAT compared to that of the other monitoring days (Fig. 6b). At 87 DAT 235 and 93 DAT, the median negative and median positive relationships were analyzed, respectively. 236 DNAIEC and DNAINO3 had a high negative correlation at 105 DAT and 112 DAT, which was different 237 from the nitrate supply amount and nitrate drainage concentration (Fig. 6c). However, contrary to the 238 yield, the correlation between the non-destructively estimated leaf area and DNAINO3 showed the 239 median positive relationship (Fig. 7). 240

Normalized DNAIEC, DNAINO3, and tomato yield in the cultivation experiment 241
The normalized DNAIEC and DANINO3 showed similar trends for each treatment during the 242 experiment (Fig. 8). In particular, the normalized yield of Inter1-3 with inter-lighting decreased after 243 the experimental treatment started. In contrast, the tendency of the normalized DNAIEC and DNAINO3 244 of the Inter1-3 treatment increased after the treatment started. The initial cumulative yield was relatively 245 higher in the L1 gutter line than that of the other lines; however, it decreased to the median level up to 246 100 DAT after which followed a tendency of increasing normalization values of DNAIEC and DNAINO3 247 of L1. The highest cumulative yields were found for the L2 gutter line, except for DAT 87 and DAT 248 100. The normalized DNAIEC and DNAINO3 of L2 were lower at the beginning of the experiment and 249 the higher-level values were monitored at 100 DAT, where a decrease in the normalized yield of L2 250 was found. The lowest values were observed at 105 DAT and 112 DAT, where high coefficients of 251 determination of DNAIEC and DNAINO3 were analyzed. Overall, the normalized values for yield in the 252 L1, L3, and L4 gutter lines remained high. When they were increased, the DNAIEC and DANINO3 253 decreased. In addition, when the normalized value for yield was shown to decrease, the DNAIEC and 254 DNAINO3 increased. 255 256

257
In a conventional soilless culture system, irrigation is stopped at night and commences again after 258 sunrise. Thus, the water content in the substrate is continuously decreased by the VPD at night (Choi 259 and Shin, 2020; Stradiot, 2001). During the day, the rate of transpiration increases due to solar radiation. 260 Daytime irrigation compensates for the transpiration during night and daytime and generates drainage 261 with irrigation exceeding the transpiration rate (Shin and Son, 2016). Therefore, the water content of 262 the medium decreases during the night, then increases in the daytime due to daytime irrigation, with 263 these repeated saturation patterns measured after reaching field capacity (Stradiot, 2001). In the present 264 study, the daily changing pattern in the substrate water content showed that the general moisture 265 management pattern of the soilless culture system was well reflected (Fig. 2a). EC or nutrient 266 concentration in the medium showed dynamic fluctuations due to the nutrient supply, drainage, 267 transpiration, and nutrient uptake (Shin and Son, 2016;Stradiot, 2001;Van Noordwijk, 1990). Fig. 2b  268 and c showed that the daily changing patterns in the variation range of EC and nutrient in the medium 269 were good representations of dynamic fluctuations due to nutrient solution supply, drainage, 270 transpiration, and nutrient absorption. In Fig. 3a and b, DNAINO3 and nitrate absorption had a high 271 positive correlation, as did DNAIEC and total nutrient uptake. Therefore, DNAIEC and DNAINO3 reflect 272 the tendency of nutrient uptake to change with a high probability despite the error-provoking factors in 273 the measurements of the cultivation system. In addition, the coefficient of determination was high when 274 the absorption factor was high compared to the coefficient of determination when the absorption factor 275 was low. At very low drainage, the coefficient of determination was decreased (Fig. 3d). The change in 276 nitrate or total nutrient uptake via DNAINO3 or DNAIEC is difficult to detect during the early stages of 277 growth or when the amount of nutrient absorption is small. Nitrate accounts for a high proportion of the 278 nutrient composition, and the difference in the ratio for several standard compositions is not large (De 279 Rijck and Schrevens, 1998b). Therefore, fluctuations in nitrate can greatly contribute to fluctuations in 280 EC (Massa et al., 2011). In Fig. 3c, the high positive correlation between DNAIEC and DNAINO3 showed 281 that EC is associated with nitrate uptake. The amount of nitrate uptake is very important for the efficient 282 During the cultivation experiment in the present study, a maximum deviation of 36 L was observed 292 for the cumulative irrigation between the gutter line of each treatment, which affected the difference in 293 the cumulative supply of nitrate ( Fig. 4a and d). It was confirmed that the difference in the partial factor 294 productivity was up to 43 kg yield/kg N via moderate deviation between the gutter line of each treatment 295 (Fig. 4f). However, the difference in the amount of nitrate supplied in each gutter line did not show a 296 high correlation with the final yield and there was no consistent trend (Fig. 5a). The average nitrate 297 concentration in drainage shows the nitrate concentration in the root zone, which is also important index 298 for nitrogen-crop yield management (Xiong et al., 2017); however, it did not show a high correlation in 299 the present study (Fig. 6b). The relationship between the supply of nitrogen fertilizer in a previous study 300 and the increase in production is seen in the low supply range starting from deficiency (Engels et al., 301 2012). However, if the fertilization amount is increased to a certain level, a diminishing return is 302 observed under the relationship between fertilizer supply and yield response (Tilman et al., 2002). In 303 the area of a diminishing return, there is no significant response to the supply of fertilizer. In the soilless 304 culture system, nitrogen is generally managed under a moderate or excessive range. Therefore, efficient 305 nitrogen management based on fertilizer supply is difficult in soilless cultures. In these areas, actual 306 nutrient absorption might be a more direct indicator than the fertilizer supply rate. 307 In this experiment, DNAIEC and DNAINO3 showed a high negative correlation with a cumulative 308 yield at DAT 105 and DAT 112 (Fig. 6c). The balance of tomato vegetative and reproductive growth 309 has an important effect on fruit yield. Tomatoes can be biased to vegetative growth when nitrogen is 310 absorbed excessively, which can lead to a decrease in yield (Sainju et al., 2003). In this cultivation 311 experiment, the analyzed DNAINO3 and yield showed a negative correlation. In contrast to the DNAINO3 312 and yield relationship, leaf area and DNAINO3 showed a positive correlation (Fig. 7), and this could be 313 attributed to the balanced growth of tomatoes. For the nutrient absorption phenomenon, nutrients are 314 also stored in vacuoles in addition to the structure of the plant to maintain ionic homeostasis (Amtmann 315 and Leigh, 2009). However, from a plant stoichiometric point of view, nutrient accumulation 316 dominantly contributes to the growth of the plant structure and has a high relationship with the relative 317 growth rate (Ågren, 2008). In the present study, DNAIEC and DNAINO3 showed a high correlation with 318 cumulative yield compared to nitrate supply or nitrate concentration, which is a traditional indicator of 319 agronomic resource management. This was presumed to be based on the result reflecting the change in 320 the absorption amount based on the plant growth state of each treatment. The point at which a high 321 determination coefficient was observed was the period during which the drainage ratio was increased 322 (Fig. 6c), and these results were consistent with the simulation results (Fig. 3d). The difference in tomato 323 production for each treatment is shown as a result of the difference in the micro-environment of each 324 cultivation space and plant growth status according to the micro-environment of each cultivation space. 325 In the present study, inter-lighting was applied to some gutter lines to change the growth and absorption 326 of nutrients by creating additional disturbances in the micro-environment compared to the control 327 treatment. Data normalization was performed to analyze the relative changes in yield, DNAIEC, and 328 DNAINO3 between each treatment during the experimental period. Overall, DNAIEC and DNIANO3 329 increased when the normalized value of the yield decreased. In particular, in Inter1-3, the value of the 330 normalization of the yields after treatment decreased simultaneously and a simultaneous increase in 331 DNAIEC and DNAINO3 was observed (Fig. 8). Nitrate and the light environment of plants are in a 332 physiologically close relationship (Lillo and Appenroth, 2001). The relationship between light and 333 nitrate uptake is already well-known and there is an experimental case reporting nitrate uptake increases 334 in LED supplemental lighting treatment (Wojciechowska et al., 2016). Even though in the present study, 335 the LED inter-lighting treatment application showed light stress symptoms, it was found that the 336 increase of the normalized DNAINO3 in the inter-lighting treatment was prominent (Fig. 8). Similar 337 trends were observed in DNAIEC, which showed changes in the total nutrient absorption in the 338 simulation. 339 From a conventional agronomical point of view, the relationship between nitrogen fertilizers and 340 plant production can be defined primarily as an increase in plant yield with increasing fertilizer supply 341 (Pan et al., 2020;Tilman et al., 2002). In the soilless culture system, nitrate generally remains within a 342 moderate or excessive range. Thus, nitrate in these ranges can often be managed in excess. However, 343 this study's theoretical and experimental results provide a technical framework to utilize nitrate 344 absorption indicators in the soilless culture system online. 345

346
The present study showed that in systems where intensive nitrate management is administered, 347 such as soilless culture systems, nitrogen management based on nitrate supply or concentration might 348 have some restrictions in proper resource management. DNAIEC and DNAINO3 also showed a high 349 positive correlation, and thus are expected to improve the technological ease of applying the DNAI NO3 350 in the system online. Furthermore, the normalized DNAINO3 responded to changes in the normalized 351 yield for each gutter line treatment during the cultivation period based on the relatively high correlation. 352 The time-series response of the normalized DNAINO3 shows potential usability as an onsite decision-353 support technique for efficient yield-promoting nitrate management. Although this study may be limited