Analysis of Irrigation, Crop Growth and Physiological Information in Substrate Cultivation Using an Intelligent Weighing System

Abstract

The online dynamic collection of irrigation and plant physiological information is crucial for the precise irrigation management of nutrient solutions and efficient crop cultivation in vegetable soilless substrate cultivation facilities. In this study, an intelligent weighing system was installed in a tomato substrate cultivation greenhouse. The monitored values from the intelligent weighing system’s pressure-type module were used to calculate irrigation start–stop times, frequency, volume, drainage volume, drainage rate, evapotranspiration, evapotranspiration rate, and stomatal conductance. In contrast, the monitored values of the suspension-type weighing module were used to calculate the amount of weight change in the plants, which supported the dynamic and quantitative characterization of substrate cultivation irrigation and crop growth based on an intelligent weighing system. The results showed that the monitoring curves of pressure and flow sensors based on the pressure-type module could accurately identify the irrigation start time and number of irrigations and calculate the irrigation volume, drainage volume, and drainage rate. The calculated irrigation amount was closely aligned with that determined by an integrated-water–fertilizer automatic control system (R² = 0.923; mean absolute error (MAE) = 0.105 mL; root-mean-square error (RMSE) = 0.132 mL). Furthermore, transpiration rate and leaf stomatal conductance were obtained through inversion, and the R², MAE, and RMSE of the extinction coefficient correction model were 0.820, 0.014 mol·m⁻²·s⁻¹, and 0.017 mol·m⁻²·s⁻¹, respectively. Compared to traditional estimation methods, the MAE and RMSE decreased by 12.5% and 15.0%, respectively. The measured values of fruit picking and leaf stripping linearly fitted with the calculated values of the suspended weighing module, and R², MAE, and RMSE were 0.958, 0.145 g, and 0.143 g, respectively. This indicated that data collection based on the suspension-type weighing module could allow for a dynamic analysis of plant weight changes and fruit yield. In summary, the intelligent weighing system could accurately analyze irrigation information and crop growth physiological indicators under the practical application conditions of facility vegetable substrate cultivation, providing technical support for the precise management of nutrient solutions.

1. Introduction

Commercial protected vegetable production is characterized as being “high-input and high-output” and occupies a crucial position in agri-food production. It is important for stabilizing the supply of agricultural products and promoting farmers’ income growth [1,2]. However, currently, facility vegetable production still predominantly soil-based in developing countries [3]. The yield and quality of vegetable crops cannot be guaranteed, and the sustainability of facility vegetable production is affected by factors such as soil-borne diseases and secondary salinization [4,5,6]. With the continuous increase in threats affecting food safety, the traditional vegetable production pattern faces severe challenges [7]. In recent years, the soilless cultivation technology, free from the constraints of soil, has developed and can overcome the obstacles of continuous cropping and soil quality deterioration. The soilless technology has been widely used in vegetable production in developed countries [8,9].

The substrate cultivation mode utilizes advanced cultivation systems and integrated water and fertilizer equipment to enable the precise regulation of the water and fertilizer supply in the root zone of crops and the efficient utilization of water and fertilizers [10], providing good water and fertilizer conditions for crops and ensuring high yield and quality. However, due to the limited space of crop root growth and poor buffering capacity of soilless substrates [11], the water content and fertilizer concentration in the root zone substrate fluctuates greatly. Therefore, it is necessary to develop precise water and fertilizer supply strategies to ensure the stable supply of water and nutrients for crop growth. Real-time acquisition of the crop root zone water content dynamics and water status is a prerequisite for developing reasonable water and fertilizer irrigation strategies. At present, crop root zone water dynamics is measured with soil moisture sensors or simulated using mathematical and physical models. Research has shown that the ECH2O moisture sensor (EC-5) could be used for irrigation decisions in the soilless production of lettuce [12]. Another study used GS3-Decagon sensors to measure the water and salt content of growth media irrigated with nutrient solutions and accumulating salt [13]. The substrate moisture content can be measured instantly using a soil moisture sensor at a single point [14]. It is impossible to understand the substrate water content in all parts of the growth channel comprehensively, and this causes some uncertainties for irrigation decisions. In addition, the measurement accuracy of the sensor is affected by the substrate type and content of salt and organic matter [15]. The substrate water content dynamics can be estimated by constructing a substrate water transport model. However, due to the high complexity of the model including a large number of parameters, this method is less applied to provide guidance for irrigation practices [16]. The surplus or deficit status of crop water and nutrients can be obtained by crop canopy image or spectral analysis, which can provide support for the management of nutrient solutions [17,18]. However, research on these methods is still in the early stages, with few applications in the actual production process. In recent years, the crop weighing system based on weighing sensors to estimate the crop water consumption process by the water balance principle has been widely studied by researchers. Due to the simplicity and operability of this method, it has been gradually accepted by greenhouse production managers. Choi Y.B [19] and his team developed an improved model specifically optimized for greenhouse environmental conditions. In the model, the transpiration rate was determined by the weighing system, and the main environmental parameters were integrated. Won Jun Jo et al. developed a transpiration model that could measure the actual crop transpiration rate and weight change data at 10 min intervals. The results indicated that the developed transpiration model could provide effective support [20]. Gremon Systems, Hungary, developed the Trutina weighing system with integrated smart sensors and a cloud-based data analysis platform to measure irrigation and drainage for irrigation strategies optimization. The Aquabalance system (Hoogendoorn, The Netherlands) could measure the weight of the substrate and the amount of liquid drained to determine when to start irrigation. Meanwhile, the system can also provide pH and EC (conductivity) measurements to further optimize the irrigation process and is suitable for hydroponics. The PRIVA-Groscale weighing system was used for accurate irrigation and fertilizer control. However, the system might have limitations and might not be suitable for actual greenhouse production conditions, which require flexibility in adjusting the irrigation regimes. Ruiz-Penalver L [21] developed a weighing system for potted crops to accurately measure parameters related to crop irrigation water use, including irrigation volume, evapotranspiration (ET), drainage volume, etc. Flow sensors were used for real-time monitoring of irrigation. Savitzky–Golay filter technology was employed to process the data and obtain reliable instantaneous ET values, providing accurate and dynamic data for precise irrigation control. However, this system is not applicable to substrate cultivation. Ofer H and his used a weighing system [22] to continuously monitor water in the soil–plant–atmosphere continuum under dynamic environmental conditions, analyzing the transpiration, biomass, stomatal conductance, and relative water content of the entire plant, providing an effective tool for crop phenotype diagnosis and screening. However, while some of these systems have achieved significant results in promoting crop growth, increasing productivity, and improving product quality. The applicability of the weighing system has beenlimited into potted plants., which was not suitbale for actual greenhouse pproduction conditions.

Therefore, an intelligent weighing system was deployed in tomato substrate cultivation in facilities for the online monitoring and analysis of irrigation and crop growth physiological features such as tomato leaf stomatal conductance and plant weight changes in greenhouse. The results of this study aim to provide theoretical and technical support for the precise nutrient solution irrigation management and efficient production of vegetables in facilities.

2. Materials and Methods

This experimental study was conducted from September 2023 to January 2024 in the solar greenhouse of the Intelligent Equipment Technology Research Centre, Beijing Academy of Agriculture and Forestry Sciences (BAAFS). The experimental equipment included a weighing system and an integrated-water–fertilizer automatic control system [23], both of which have independent intellectual property rights. The weighing system was also used as a substrate cultivation system for tomato cultivation in this experimental study and consisted of independent units. The integrated-water–fertilizer automatic control system was mainly used to manage the nutrient solution irrigation. Both the weighing system and the integrated-water–fertilizer automatic control system were connected with wireless communication.

2.1. Weighing System

The weighing system was developed by the Intelligent Equipment Technology Research Centre of Beijing Academy of Agricultural and Forestry Sciences. The weighing system consisted of 4 parts (Figure 1): a pressure-type weighing module, a suspension-type weighing module, a gateway, and a data storage module. The pressure-type weighing module included a stand, a pressure sensor, a cultivation tank, a flow sensor, and a data collector. The stand was placed on a level surface. The pressure sensor was installed on the bracket, with one set each. The cultivation tank was placed on top of the pressure sensors, with a drain channel at the bottom of the cultivation tank and a drainage pipe at the end of the drain channel. The flow sensor was configured at the bottom of the cultivation tank and connected to thedrainage pipe. The data collector was installed on the stand. The pressure sensor was used to collect the mass (W) of the cultivation tank and the built-in object, the cultivation tank was used to place the substrate for tomato growing, the flow sensor was used to monitor the volume of liquid discharged from the cultivation tank (V), and the data collector was used to collect the data from the pressure sensor and upload them to the gateway via Lora transmission. The suspension-type weighing module integrated a hook, a tension transducer, and a data collector and was suspended above the pressure-type weighing module. The lower hook connected to the plant-hanging rope and the upper hook connected to a steel wire rope securing the greenhouse skeleton were used to monitor the above-ground mass of the plant (PW). The gateway, used to transmit the monitoring data from the pressure-type weighing module and the suspension-type weighing module to the data storage module, was mounted on the greenhouse wall where the weighing system was located. The data storage module provided data, storage, visualization, and download functions, which was deployed in this study on the integrated-water–fertilizer automatic control system.

2.2. Parameter Analysis Using the Weighing System

2.2.1. Data Acquisition and Processing for Noise Reduction

The pressure-type weighing module used a pressure transducer (FSSB-50, Transcell Technology, Inc., Buffalo Grove, IL, USA) operating up to 50 kg, with a resolution of 2.0 mv/v ± 0.1% and an accuracy of ±0.02%. A hanging load cell (BAB-20MT, Transcell Technology, Inc., Buffalo Grove, IL, USA) for up to 20 kg, with a resolution of 2.0 mv/v ± 0.1% and an accuracy of ±0.02%, was used. The flow sensor (OSA-18S1, Hebei Province Ji Ou Speed Company, Shijiazhuang, China) was a 20 mL tipping bucket with a maximum allowable flow rate of up to 30 mL·s⁻¹ and a measurement accuracy of ±3%. The data acquisition frequency was 1 time·min⁻¹.

To mitigate sensor data fluctuations induced by supply voltage variations, which might lead to significant analytical errors, we employed variational mode decomposition (VMD) for data preprocessing. VMD [24], an advanced signal processing technique, decomposes the signal into intrinsic mode functions (IMFs) with distinct center frequencies, effectively isolating and removing noise components. This approach enhanced data quality and prevented errors in subsequent weight slope calculations caused by minor data variations.

2.2.2. Analysis of Irrigation Start -Stop Times and Number of Irrigations

In this experiment, the pressure-type weighing module of the weighing system was used as a substrate cultivation system to grow tomatoes, and irrigation and plant growth were the factors that caused an increase in the pressure sensor monitored values. However, the effect of plant growth on the increase in the monitored values of the pressure sensor was ignored when plant growth was small in a short period, and the hanging strategy reduced the effect of plant growth on the pressure-type weighing module. After irrigation was stopped, the pressure sensor values decreased with plant transpiration and excess nutrient discharge. Therefore, the extreme value detection method was used to identify the irrigation start and stop time and determine the number of irrigations.

The slope (Equation (1)) was calculated using any number of values at the consecutive time intervals at j − 1 and j moments recorded by the pressure sensor of the pressure-type weighing module:

$$K={\displaystyle \frac{W_j-W_{j-1}}{\Delta t}}$$

(1)

where K is the slope, W_j and W_j−1 are the values monitored by the pressure sensor of the weighing module at the j and j − 1 moments, respectively, in g; Δt is the time interval of data collection in this experiment. The j moment pressure sensor value between the occurrence of more than two consecutive times with K < 0 and K > 0, respectively, was minimal and was identified as the irrigation start time. On the contrary, There would be a very high value at the end of an irrigation event.

2.2.3. Analysis of Irrigation Volume, Drainage Volume and Drainage Rate

The irrigation volume for the m-th irrigation (I_m) was estimated by:

$$I_m=W_{sm}-W_{em}\pm D_{m1}\pm {ET}_{m1}$$

(2)

where W_sm and W_em are the sensor outputs of the pressure-type weighing module at the start and stop of the m-th irrigation, respectively, in g; ET_m1 is the evapotranspiration during the irrigation period, which could be neglected due to the short time of irrigation under soilless conditions, i.e., ET ≈ 0; and D_m1 is the drainage in mL, calculated by:

$$D_{m1}=D_{sm}-D_{em}$$

(3)

where D_sm and D_em are the monitored values of the flow sensor at the start and end of the irrigation cycle, respectively, in mL.

The daily irrigation volume (I_d) is the sum of the irrigation volume generated by all irrigation events on a day and is calculated by:

$$I_d=\sum I_m$$

(4)

Drainage volume (D): the drainage volume generated by the m-th irrigation event consists of two components, i.e., the discharge volume during the irrigation period and the discharge volume during the non-irrigation period, and is calculated by:

$$D=D_{m+1}-D_m$$

(5)

where D_m and D_m+1 are the monitored values of the flow sensor at the m-th and m+1-st irrigation start time, respectively, in mL.

The daily drainage (D_d) is the sum of the flows in mL produced by all the irrigation events of the day, calculated by:

$$D_d=\sum D$$

(6)

The drainage rate (D_Rate-m) for the m-th irrigation event is the ratio of the drainage volume to the irrigation volume for the corresponding irrigation event, in%, and is calculated by:

$$D_{Rate-m}={\displaystyle \frac{D}{I_m}}\times 100$$

(7)

The daily drainage rate (D_Rate-d) is the ratio of daily drainage volume to daily irrigation volume in %, calculated by:

$$D_{Rate-d}={\displaystyle \frac{D_d}{I_d}}\times 100$$

(8)

2.2.4. Analysis of Physiological Indicators of the Plant (Transpiration, Transpiration Rate, and Stomatal Conductance)

According to the principle of water balance, the daily transpiration of plants on day i, ET_d (mL), was calculated from the difference between the daily irrigation volume, the daily drainage volume, and the monitored value of the pressure sensor of the pressure-type weighing module at the time of 0:00 on the following day and calculated by:

$${ET}_d=I_d-D_d-(W_{i+1}-W_i)$$

(9)

where W_j and W_j−1 are the values of the pressure sensor at any j and j − 1 moments in the non-irrigation time period, respectively, in g; D_j and D_j−1 are the corresponding values recorded by the flow sensor, in mL.

In addition, the transpiration rate of the crop (ET_Rate), mL·h⁻¹, during the hours of non-irrigation was calculated by:

$${ET}_{Rate}={\displaystyle \frac{W_j-W_{j-1}+(D_j-D_{j-1})}{t_j-t_{j-1}}}$$

(10)

where W_j and W_j−1 are the values monitored by the pressure sensor of the pressure-type weighing module at any time j and j − 1 during the non-irrigation period, in g; D_j and D_j−1 are the corresponding values drainage recorded by the flow sensor, in mL.

Based on the determination of the transpiration rate, the stomatal conductance (g_sc) of the plant leaves, in mol·m⁻²·s⁻¹, was further determined by considering meteorological data and calculated by:

$$g_{sc}={\displaystyle \frac{{ET}_{Rate}\times P_{atm}}{18\times LA\times VPD}}$$

(11)

where g_sc is the stomatal conductance, in mol·m⁻²·s⁻¹; P_atm is the atmospheric pressure, corresponding to 101.3 kPa; VPD is the saturated vapor pressure difference, in kPa; LA is the total plant leaf area, m²; the coefficient of 18 is the molar mass of water, in g·mol⁻¹.

Tomato grew with a high planting density in the greenhouse and had more than five fruit spikes. The radiation each plant leaf received from bottom to top varied due to leaf shading. Therefore, an extinction coefficient k (k = 0.89) was introduced to take into account the effects of leaf shading in stomatal conductance calculation [25]:

$$g_{sc}={\displaystyle \frac{{ET}_{Rate}\times P_{atm}}{18\times k\times LA\times VPD}}$$

(12)

2.2.5. Analysis of Changes in Plant Weight

The suspension-type weighing module monitored the plant’s above-ground mass (PW), which increased following plant growth and decreased after removing growing points and fruit harvesting. Therefore, the daily rate of change in plant weight and fruit harvest quality could be obtained from the change in plant weight values (ΔPW) recorded by the suspension-type weighing module. In this case, the magnitude of the change in plant weight (ΔPW) and the daily rate of change in plant weight (ΔPW_Rate-d) were calculated by:

$$\Delta PW={PW}_j-{PW}_i$$

(13)

$${\Delta PW}_{Rate-d}={\displaystyle \frac{{PW}_j-{PW}_i}{j-i}}$$

(14)

where PW_i and PW_j are the values recorded by the suspension-type weighing module at 6:00 a.m. on day i and day j, respectively, in g.

Harvesting or stripping of leaves was carried out during non-irrigation hours. These operations were carried out in sequence, the time was recorded, and the mass of harvested fruit FW_m or of stripped leaves LW_n was obtained from the difference between the values recorded by the suspension-type weighing module at the moments before and after the operation of harvesting or stripping of leaves, according to the following Equations:

$${FW}_m={PW}_{m+1}-{PW}_m$$

(15)

$${LW}_n={PW}_{n+1}-{PW}_n$$

(16)

where PW_m and PW_m+1 are the values recorded by the suspension-type weighing module at the moment before and after the m-th fruit picking, in g; PW_n and PW_n+1 are the values recorded by the suspension-type weighing module at the moment before and after the n-th leaf stripping, in g.

2.3. Substrate Cultivation and Management of the Tomatoes

The tomato variety used in this experimental study was Jingcai 8 (Beijing Modern Farmer Seedling Technology Co., Ltd., Beijing, China). The cultivation substrate was cocopeat strips (Qingdao Ranmei Trading Co., Ltd., Qingdao, China), with a bulk density of 0.14 g·cm⁻³, a water-absorbing expansion volume of 100 cm × 15 cm × 10 cm, a field water holding capacity (θ_c) of 0.455 cm³·cm⁻³, a leachate EC ≤ 0.70 mS·cm⁻¹, and pH of 5.8–6.8. The nutrient solution was based on the Yamazaki tomato formula (1978); drip irrigation was conducted with a spacing of 25 cm and a drip head flow rate of 2 L·h⁻¹. The irrigation frequency was determined by the cumulative light radiation method, and the irrigation amount was determined based on the substrate moisture sensor method, that is, when the cumulative light radiation value was ≥0.35 MJ·m⁻² [26,27], the irrigation amount was determined by the current values of the substrate moisture sensor. The set target substrate moisture content was calculated by:

$$I=1000\times ({\theta }_1-{\theta }_2)\times V\times P/\eta$$

(17)

where I is the calculated irrigation volume, in mL; θ₁ is the set target value of the substrate moisture content, in cm³·cm⁻³; θ₂ is the current value recorded by the substrate moisture sensor, in cm³·cm⁻³; V is the volume of the cultivated substrate, in cm³; η is the coefficient of the effective use of the irrigation water, which was considered to be 0.9; and P is the substrate wetting ratio, which was considered to be 100%.

In this experimental study, the weighing system was also used as a cultivation system, and the specific layout in the greenhouse was the same as for conventional substrate cultivation, with 0.05 m spacing between adjacent weighing systems in the same row to avoid their mutual influence. Tomato seedlings with four leaves and one apical bud were selected for planting, and four plants were planted on each set of substrate strips of the weighing system at a density of 4.76 plants·m⁻². A substrate moisture sensor (EC-5, METER Group, Inc., Pullman, WA, USA) was installed in the center of the substrate strips of each set of weighing system tanks to monitor the moisture content of the substrate. A miniature weather station (Jinzhou Sunshine Weather Co., Ltd., Jinzhou, China) was installed in the middle of the greenhouse at a distance of 2 m from the ground to monitor the greenhouse air temperature T, the relative humidity RH (PTS-5), and the solar radiation Rn (TBQ-2). The substrate moisture sensor and the miniature weather station communicated with the integrated-water–fertilizer automatic control system, and the sampling frequency was 1 time·min⁻¹. In order to further explore the adaptability of crop growth and obtain physiological information based on the weighing system, two irrigation treatments were set up in this experimental study, at 0.85θ_c (T1) and 0.55θ_c (T2), corresponding to different substrate moisture contents of irrigation threshold, and a control treatment corresponding to 1.0θ_c (CK). The irrigation treatments were applied when the second ear of fruit entered the expansion stage, with three replications of each treatment in randomized rows. All treatments were managed consistently and monitored daily in the field, with four spikes of fruit remaining until the end of experiment (6 November) and three to four fruits per spike.

2.4. Measurement and Calculation of Indicators

The vapor pressure deficit (VPD) was calculated using the method recommended by FAO-56 [28] and was used to analyze leaf stomatal conductance (Equations (11) and (12)). The equation is as follows:

$$VPD=0.611\times (1-RH)\times exp\left[{\displaystyle \frac{17.502\times T}{240.97+T}}\right]$$

(18)

where RH is the relative humidity of the air, in %; T is the temperature of the air, in °C.

The leaf area (LA) is the sum of the leaf areas in a plant (m² per plant) and is calculated by:

$$LA={\displaystyle \frac{{\sum }_{i=1}^n{LA}_i}{10000}}$$

(19)

where LA_i is the area of a single leaf in cm² and is calculated by:

$${LA}_i=\left\{\begin{array}{c}0.3782ab, 0 \mathrm{cm}<a<20 \mathrm{cm}\\ 0.3184ab, a\ge 20 \mathrm{cm}\end{array}\right.$$

(20)

where a is the length of the leaf (from the base of petiole to the tip), in cm; b is the width of the leaf (the widest length perpendicular to the main vein of the leaf), in cm.

Stomatal conductance (g_sc) was measured using an LI-6400XT portable photosynthesis system. Under clear weather conditions, the functional leaves of the plants were measured every 2 h between 09:00 and 17:00 to obtain the stomatal conductance values of the leaves.

The masses of stripped leaves (LG_n) and picked fruit (FG_m) were measured using an electronic balance (precision: 0.01 g) each time the leaves were removed or the fruits were harvested.

2.5. Evaluation of Indicators

Coefficient of determination (R²) was a statistic used in regression analysis to evalute the agreement of a model fits the data.

$$R^2=1-{\displaystyle \frac{{\sum }_{i=1}^n{(y_i-{\widehat{y}}_i)}^2}{\sum _{i=1}^n{(y_i-{\overline{y}}_i)}^2}}$$

(21)

where y_i is the i-th measurement; $\widehat{y}$_i is the average of the i-th estimation; $\overline{y}$_i is the average of all estimations.

Root mean square error (RMSE) was a commonly used measure of the difference between the predictions of a model and the actual estimations.

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{(y_i-{\widehat{y}}_i)}^2}$$

(22)

Mean absolute error (MAE) was a common measure of the accuracy of a forecast model.

$$MAE={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n\left|y_i-{\widehat{y}}_i\right|$$

(23)

2.6. Data Analysis

All data were processed using Microsoft Excel 2022. Data denoising and irrigation pattern recognition were performed through Python programming, followed by graphical visualization using Origin 8.5 software.

3. Results

3.1. Meteorology

Figure 2 shows the environmental data of the greenhouse, including air temperature (T), relative humidity (RH), and solar radiation (Rn) from November 9 to 11. In the figure, it can be seen that the average daily air temperature in the three days was 19.30 °C, 15.80 °C, and 17.09 °C, respectively, the average daily relative humidity was 56.77%, 60.86%, and 56.36%, respectively, and the average daily radiation was 96.04 W·m⁻², 83.01 W·m⁻², and 97.17 W·m⁻², respectively. The air temperature and solar radiation intensity showed similar trends. They both started to increase rapidly from 08:00, reached a maximum at 11:40, and then started to decrease; the temperature decreased slowly, and the solar radiation tended to stabilize after 18:00. The relative humidity showed the opposite trend to that of the air temperature. The average daily mean temperatures on 9–11 November were 26.66 °C and 12.00 °C, 22.01 °C and 9.72 °C, and 25.32 °C and 9.43 °C, respectively and provided suitable temperature conditions for tomato growth.

3.2. Analysis of Irrigation Information

In this study, the values recorded by the pressure-type weighing module and the flow sensor for the CK (1.0θ_c) treatment were used to calculate irrigation-related information such as irrigation start-stop times of irrigations, number of irrigations, irrigation volume, and drainage rate.

3.2.1. Processing for Noise Reduction

Figure 3 shows the raw data and VMD noise reduction results for the values recorded by the pressure sensor on 10 November. Compared to the raw data recorded by the pressure-monitoring sensor, the data after the VMD noise reduction treatment were smoother, and the data of the non-irrigation period and the irrigation cycle showed consistent results. This suggested that the VMD noise reduction processing improved the smoothness of the values of the pressure sensor and provided data quality assurance for the analysis of irrigation information.

3.2.2. Irrigation Scheduling Parameters: Timing and Frequency

The monitoring curve of the pressure sensor of the pressure-type weighing module on 10 November (Figure 3) was obtained using the extreme value detection method, and the irrigation start and stop times were obtained and are shown in

Table 1. Irrigation start and stop times based on the pressure-type weighing module and the solenoid valve switching times, recorded by the integrated-water–fertilizer automatic control system on 10 November.

	Irrigation Recognition a		Solenoid Valve b
	Start Time	Stop Time	Open Time	Close Time
1	8:28:00	8:40:00	8:27:43	8:33:36
2	9:19:00	9:21:00	9:18:11	9:20:27
3	9:57:00	10:05:00	9:56:52	9:58:42
4	10:32:00	10:39:00	10:30:59	10:32:40
5	10:58:00	11:03:00	10:57:14	10:58:37
6	11:22:00	11:28:00	11:21:21	11:22:44
7	11:45:00	11:51:00	11:44:40	11:46:03
8	12:08:00	12:13:00	12:07:08	12:08:23
9	12:30:00	12:36:00	12:29:13	12:30:36
10	12:54:00	13:00:00	12:53:54	12:55:35
11	13:21:00	13:26:00	13:20:20	13:22:01
12	13:49:00	13:55:00	13:47:59	13:49:49
13	14:26:00	14:33:00	14:25:05	14:26:46
14	15:21:00	15:29:00	15:20:52	15:22:24
15	16:58:00	17:05:00	16:57:14	16:58:29

Note: a—Based on the irrigation start and stop times detected by the pressure-type module. b—Based on the solenoid valve switching time detected by the integrated-water–fertilizer automatic control system.

. On 10 November, the experimental plot underwent 15 scheduled irrigation events. The initial irrigation commenced at 08:28:00 and concluded at 08:40:00, while the final irrigation cycle was initiated at 16:58:00 and terminated at 17:05:00. In this study, the switching time of the solenoid valve recorded by the integrated-water–fertilizer automatic control system was used as the real value to further analyze the accuracy of the irrigation start and stop times and the number of irrigations. As shown in Table 1, the irrigation start time was delayed by 6 s to 68 s compared with the solenoid valve opening time, and the irrigation stop time was delayed by 33 s to 396 s compared with the solenoid valve closing time. This indicated that the calculated value of the irrigation start time was more accurate than that of the irrigation stop time. The acquisition frequency of the weighing module was the main reason for the start time error, and the larger irrigation stop time error was probably related to the transport of irrigation water in the substrate. Furthermore, the results in Table 1 show that the number of irrigation events obtained by the two methods was the same, indicating that the number of irrigation events could be efficiently determined based on the monitored value of the pressure sensor of the pressure-type weighing module.

Previous studies have shown that the substrate water content can be monitored using moisture sensors, and that substrate water content curves can provide direct feedback on the irrigation process. Figure 4 shows that the substrate water content and pressure sensor monitored values had identical trends, which also showed that information such as irrigation start and stop time and the number of irrigation events could be obtained online through the pressure sensor of the pressure-type weighing module.

3.2.3. Irrigation Water Fluxes: Irrigation Volumes and Drainage Characteristics

The values recorded by the pressure and flow sensor corresponding to the irrigation start and stop times were used to obtain the irrigation and drainage volumes, as shown in Figure 5. Figure 5b shows the pressure and flow sensor data between the first two irrigation start times (Figure 5a). The start and stop times of this irrigation correspond to the values recorded by the pressure sensor and flow sensor of 17,917 g and 18,274 g and 428 mL and 428 mL, respectively, from which the current irrigation volume of 357 mL was obtained (Figure 5b). The value recorded by the flow sensor at the second irrigation start time was 428 mL, corresponding to a discharge of 0 mL during the first irrigation cycle (Figure 5c). Figure 5c shows the irrigation and drainage volumes corresponding to each irrigation in Figure 5a. It can be seen that the first irrigation volume was the largest and then showed a trend of decreasing and then gradually increasing, which was closely related to the results obtained with the method used in this study to determine the irrigation frequency and volume. In addition, only the fourth to seventh and last irrigations were monitored for drainage volume, and the return drainage volume from the fourth to seventh irrigation showed a trend of first increasing and then decreasing. This might be because irrigation in this study was determined based only on the current substrate water content and field water-holding capacity, without considering the role of meteorological parameters in the daily change in crop water demand.

The quality of the above irrigation volume analysis was verified using the irrigation volume determined based on the substrate moisture sensor method from 9 to 11 November, and the results are shown in Figure 6. The calculated value of the irrigation volume based on the pressure sensor of the pressure-type weighing module agreed well with the value of the irrigation volume determined using the substrate moisture sensor, with R² up to 0.923, and the mean absolute error (MAE) and root-mean-square error (RMSE) were 0.105 mL and 0.132 mL, respectively, indicating that the irrigation volume could be obtained with good accuracy using the pressure-type weighing system. Therefore, the pressure-type weighing module could be used to obtain the daily irrigation volume and the daily drainage volume, and the daily drainage rate could be further calculated to provide a reference for nutrient solution management. In this experiment, the daily irrigation volumes from November 9 to 11 were 2885 mL, 2005 mL, and 2309 mL, respectively, the daily drainage volumes were 260 mL, 64 mL, and 0 mL, respectively, and the daily drainage rates were 9.01%, 3.19%, and 0%, respectively. The drainage rate consistently remained below 15%, suggesting inadequate irrigation that failed to leach the accumulated salts in substrate. Therefore, optimizing the irrigation management is essential to maintain optimal tomato growth conditions.

3.3. Plant Physiological Characteristics: Transpiration Rate and Stomatal Conductance of the Leaves

Based on the irrigation and drainage volumes in Figure 3 and Figure 5, the plant evapotranspiration during the irrigation cycle on 10 November can be calculated to be 433.25 mL and the daily evapotranspiration on that day to be 561.25 mL, respectively.

Figure 7 present the variation in plant transpiration rate under different irrigation treatments from 9 to 11 November. The trend of the plant transpiration rate under different irrigation treatments was consistent, i.e., on each single day, the curve showed a single peak with a ‘stable–increasing–decreasing–stable’ pattern, with the peaks on the 9th and 11th days close to and higher than that on the 10th day. This was similar to the trend of the changes in radiation (Figure 2). In Figure 7, it can be seen that the transpiration rate for the CK treatment was higher than those for the T1 and T2 treatments, especially during the irrigation period (daytime). The maximum transpiration rates for the CK, T1, and T2 treatments were 112.0 mL·h⁻¹, 82.8 mL·h⁻¹, and 76.0 mL·h⁻¹, respectively, on 9 November. This is related to the higher transpiration rate of plants with a higher substrate water content. From the above, it can be seen that the pressure-type weighing module has a good ability to determine the transpiration rate of plants under different irrigation treatments.

Figure 8 shows the results of the comparison between the values of leaf stomatal conductance calculated using the pressure-type weighing module and the values measured by the photosynthesizer. Compared to the measured values, the calculated values were low when using the conventional method and high after introducing the correction with the extinction coefficient. The stomatal conductance calculated by these two methods was linearly fitted to the values measured by the photosynthesizer, and the coefficient of determination R² was 0.820 for both methods. However, the slope of the linear fit corrected with the extinction coefficient was closer to 1 (1.03) than that determined with the conventional method (0.92), and MAE and RMSE were reduced from 0.016 mol·m⁻²·s⁻¹ to 0.014 mol·m⁻²·s⁻¹ and 0.020 mol·m⁻²·s⁻¹ to 0.017 mol·m⁻²·s⁻¹, respectively. Compared with those of the traditional method, the MAE and RMSE of stomatal conductance estimation were reduced by 12.5% and 15.0%, respectively. It can be concluded that introduction of the extinction coefficient corrected the stomatal conductance algorithm, allowing for the in situ analysis of stomatal conductance of tomato leaves based using the weighing system.

3.4. Change in Plant Weight

Plant weight dynamics exhibit phase-dependent responses to cultivation practices: growth processes (fruit growth, fruit set, and fruit development) contribute to plant weight, whereas growing tip removal and fruit harvesting induce plant weight reduction. Figure 9 illustrates the irrigation treatment-specific plant weight trajectories recorded by the suspension-type weighing module throughout the tomato growth cycle. As the tomato growth process progressed, the plant weight in all three irrigation treatments showed a gradual increase, and then a slow decrease at mid to late harvest, and the daily change rate in plant weight varied from large to small in the following order: flowering and fruiting stage, pre-mid fruit expansion (before reaching the growth point), mid-late fruit expansion, harvesting stage. This is consistent with the pattern of plant weight change during tomato growth and development. After the irrigation treatment was initiated on 19 October, the daily rate of change in plant weight was greatest for the CK treatment, followed by the T1 treatment. In the mid-early and middle-late stages of fruit expansion, the daily rate of change in plant weight for the T1 and T2 treatments was lower than that for CK by 12.24% and 31.66% and 18.67% and 30.74%, respectively. Among these rates, that for the T2 treatment was significantly lower than that for the CK and T1 treatments at the early stage of fruit expansion. In addition, the daily management records showed that the plants reached the growth point on 6 November, and the first fruit was harvested on 30 November. As shown in Figure 9, plant weight decreased for all three irrigation treatments on 6 and 29 November, which coincided with the above-mentioned management activities. Compared to CK, both T1 and T2 treatments showed smaller decreases, which may be related to the weight of the growing tip and fruit picking. The above results indicated that the suspension-type weighing module provided direct feedback on plant weight changes and was applicable to different irrigation treatments (substrate water content).

Figure 10 shows a linear regression analysis comparing the mass of plant components (fruits, leaves and removal of growing tips) measured by the electronic scales with the plant weight values (FWn and LWn) derived from the suspended weighing module. The measurements were recorded before and after each experimental operation, yielding a coefficient of determination (R²) of 0.958. Error analysis revealed close agreement between the two measurement methods, with MAE and RMSE values of 0.145 g and 0.143 g, respectively. These metrics demonstrated the suspension-type weighing module’s capability for accurate dynamic monitoring of plant weight changes. When integrated with production records, this system enables the real-time tracking of critical agricultural parameters, including cumulative fruit yield and growth pattern.

4. Discussion

4.1. Effect on Irrigation Data Processing Accuracy of Noise Reduction

Despite weighing systems being important devices for obtaining key information, the accuracy of the load cell’s data directly affect the reliability of data acquisition for subsequent analyses [29]. However, disturbances caused by supply voltage fluctuations are unavoidable in practical applications, resulting in frequent small-amplitude fluctuations in load cell data. If these noises are not effectively processed, they would be amplified in the subsequent data analysis process [30]. This will lead to significant errors in the judgment of irrigation start and stop times, the calculation of the irrigation amount, and the analysis of plant transpiration and palnt weight changes, seriously affecting the accurate assessment of crop growth status. Huang et al. improved the accuracy of the weighing evapotranspiration measurement by noise reduction through outlier detection [31]. Jiménez-Carvajal C used Savitzky–Golay filtering noise reduction technology to process data and obtain reliable instantaneous ET values, deepened the research of ET change law in the irrigation period, and provided accurate dynamic data support for precise irrigation regulation [32]. In our study the data were pre-processed using the variational modal decomposition (VMD) technique, which is based on advanced signal processing theory and is able to accurately identify and remove the noise components by decomposing complex signals into a number of intrinsic modal functions (IMFs) with well-defined center frequencies. By examining the noise-reduced values obtained from the pressure-type module on 10 November, it can be seen that the data curves became smoother after the VMD process, and the data showed a good consistency in both non-irrigation periods and irrigation cycles. It not only provides a solid data quality guarantee for the accurate analysis of key data such as irrigation information and plant transpiration but also ensures the scientific and reliable nature of agricultural production decisions based on these data.

4.2. Analysis of Plant Physiological and Ecological Information

Plant transpiration, an extremely important physiological activity during crop growth, is closely linked to key physiological processes such as water metabolism, nutrient transport, and photosynthesis [33]. Currently, the commonly used methods for continuously monitoring the crop water status and inferring water stress conditions monitor transpiration rate and stomatal conductance [34]. In this study, based on the principle of water balance, the accurate calculation of plant transpiration and transpiration rate was achieved by using the difference between the irrigation volume and the drainage volume monitored by the intelligent weighing system and the pressure sensor value of the pressure-type weighing module. The results of the study showed that the transpiration rate of plants under different irrigation treatments displayed a similar trend, i.e., a single-peak curve with a ‘steady–increase–decrease–steady’ pattern on each single day, similar to the results of a previous study [35], and this trend was very similar to the trend of light radiation. The transpiration rate with the CK treatment was significantly higher than that with the T1 and T2 treatments during the irrigation period (daytime) from November 9 to 11. There was a positive correlation between the substrate water content and plant transpiration, i.e., the higher the substrate water content, the greater plant transpiration [36]. The effect of the plant transpiration process on the growth and development of the crop itself is multifaceted. On the one hand. The tension created by transpiration promotes the uptake and transport of water and nutrients by the root system, providing the necessary nutrients for plant growth. The amount of transpiration also affects stomatal conductance [37]. When the water supply in the plant is sufficient, transpiration proceeds normally, and the stomata can maintain a certain degree of opening to meet the needs of physiological processes such as photosynthesis. The real-time measurement of stomatal conductance under plant level is an effective means of determining the actual crop water staus, which could be estimated by simulation models up to now. Some current models assume uniform light exposure of the canopy during stomatal conductance estimation, which is obviously inconsistent with the actual plant growth status [38]. The canopy extinction coefficient refers to the degree of absorption and scattering of light as it passes through the plant canopy. It is widely used in crop photosynthetic rate estimation models In our study, the extinction coefficient was introduced into the stomatal conductance estimation model to account for the effect of leaves shading on stomatal conductance, and the MAE and RMSE of stomatal conductance estimation were reduced by 12.5% and 15.0%, respectively, compared with those of the traditional estimation methods.

4.3. Analysis of Growth Information

Plant weight rate is an important agronomic parameter that reflects crop growth [39]. Common methods for determining plant growth rate include the image processing measurement method and the dry weight method. The image processing method can only measure the appearance of morphological indicators of the plant but cannot determine the overall growth status of the plant. The results of the dry weight method are stable and reliable and are not affected by changes in moisture content. However, it requires destructive sampling and does not allow for the continuous observation of the same plant [40]. Ofer Halperin used a weighing system to continuously monitor plant growth, but it had the limitation of a fixed irrigation rate, which is not suitable for greenhouse production environments that require the flexible adjustment of the irrigation rate. In this study, the suspension-type weighing module was used to monitor above-ground plant mass changes in real time, successfully obtaining key information such as the amount of plant weight change, the daily rate of plant weight change, and the quality of the harvested fruit. The suspension-type weighing module was used to quantify crop growth and fruit harvest and overcome the influence of transpiration- and irrigation-induced changes on substrate weight. However, there are still some limitations, such as the inability to monitor growth during pruning due to the small plant weight. A comparison of plant weight growth under the different irrigation treatments showed that the CK treatment presented the highest daily rate of change in plant weight, followed by the T1 treatment and the T2 treatment. In the early-expansion and mid-late fruit expansion periods, the daily rate of plant weight change for plants under T1 and T2 was 12.24%, 31.66%, 18.67%, and 30.74% lower than that of CK, respectively. In particular, the T2 treatment values were significantly lower than the CK and T1 treatment values in the early stage of fruit enlargement. This result clearly indicated that the water content of the substrate played a key role in plant growth [41]. An adequate water supply could provide a favorable environment for plant physiological activities, promoted plant growth and development, and increased plant weight. Insufficient water supply could limit plant growth, resulting in crop growth inhibition [42].

5. Conclusions

An intelligent weighing system wad deployed to carry out the analysis of nutrient solution irrigation management and crop growth physiology based on the system. The results demonstrated that the monitoring data from the pressure-type weighing module could be used to determine nutrient solution irrigation parameters for substrate cultivation. These parameters included star = stop times of irrigation, the number of irrigation cycles, the irrigation volume, the drainage volume, and the drainage rate. Additionally, the system could provide crop water physiology information such as plant transpiration volume and transpiration rate, as well as leaf stomatal conductance. The suspension-type weighing module could dynamically invert the amount of changes in plant weight dynamically and obtain fruit yield. The MAE and RMSE of the calculated and measured values of irrigation volume and leaf stomatal conductance, which were 0.105 mL and 0.132 mL and 0.014 mol·m⁻²·s⁻¹ and 0.017 mol·m⁻²·s⁻¹, respectively. The intelligent weighing system could be used to obtain information related to substrate cultivation and irrigation, as well as calculate physiological parameters closely related to crop growth accurately. The analysis method could also be embedded in the control module to characteristic the crop growth indexes online.