Plant-Driven Precision Irrigation in Aeroponics: Real-Time Turgor Sensing for Sustainable Lettuce Cultivation

Abstract

The narrow margin for irrigation error in aeroponics necessitates advanced control strategies beyond fixed timer-based approaches. This study evaluates a plant-driven irrigation method based on real-time leaf turgor feedback in aeroponic romaine lettuce (Lactuca sativa L. var. longifolia) cultivation. A leaf thickness–turgor sensor was interfaced with an Arduino Mega 2560 to activate misting events dynamically. Two identical aeroponic systems were operated in a fully controlled environment: a conventional timer-based control (TC) system applying mist every 10 min and an Arduino-controlled (AC) system triggered by turgor changes. Over two independent 37-day cultivation cycles, the AC strategy reduced total water use by an average of 15.9% and pump activations by 17.2% while improving water use efficiency by 17.8% and nutrient use efficiency for N, P, and K by an average of 17.8%, with no statistically significant differences in shoot biomass, height, or yield. Although root dry weight was significantly higher under TC, the AC treatment led to a 45.0% reduction in leaf nitrate accumulation and non-significant increases in phenolic content. These findings demonstrate the potential of turgor-responsive irrigation for enhancing sustainability, resource use efficiency, and the quality of produce in aeroponic systems, thereby supporting its broader integration into controlled-environment agriculture (CEA).

1. Introduction

Aeroponics is a soilless cultivation technique in which plant roots are suspended in air and intermittently misted with a nutrient-rich solution. This method offers higher oxygenation of the root zone compared to hydroponics, facilitating enhanced nutrient uptake, accelerated plant growth, and reduced water use [1,2,3]. Moreover, the absence of inert media reduces system weight and simplifies waste management, making aeroponics suitable for urban farming, vertical agriculture, and Controlled-Environment Agriculture (CEA) applications [4,5]. Additionally, water-use efficiency in aeroponics is consistently higher than in both hydroponics and soil cultivation, since nutrient solutions are delivered directly to the root surface with minimal losses [6].

Despite these advantages, aeroponics presents specific operational, economic, and technical challenges. Foremost among them is the narrow margin between optimal hydration and water stress due to the direct exposure of roots to air. Because there is no moisture-retaining substrate, irrigation frequency must be precisely regulated; even brief interruptions in misting caused by pump failure, nozzle clogging, or power outage can induce severe stress or irreversible root desiccation within minutes [7,8]. This fragility contrasts with hydroponic nutrient films and soil systems, where water retention provides some buffering capacity. In addition, aeroponics is associated with high energy demand for continuous pressurization and frequent misting, while the requirement for specialized hardware—high-pressure pumps, precision nozzles, and filtration units—increases both capital and recurring costs [9]. Operator skill requirements are likewise more demanding, as minor errors in calibration, programming, or scheduling can translate into significant crop losses. These constraints limit adoption, particularly in regions with unstable electricity supply or limited technical support [8,10].

Advanced aeroponic systems integrating Internet of Things (IoT) technologies are being developed to dynamically adjust misting frequency based on real-time monitoring, thereby overcoming the limitations of conventional fixed-timer systems [11,12]. These systems can interface with a wide range of sensors to monitor parameters such as temperature, humidity, light intensity, and nutrient reservoir levels, enabling low-latency, automated irrigation responses [10,13,14]. Remote access, alerts, and data logging further improve resilience by reducing response times to failures. Recent studies confirm that IoT-enabled aeroponics can improve water-use efficiency, reduce energy consumption, and stabilize production under variable environmental conditions [9,10,15].

However, most current IoT implementations still rely on indirect environmental proxies, such as vapor pressure deficit (VPD), which do not always linearly reflect plant water status and may vary with genotype, canopy microclimate, or developmental stage [16]. Beyond environmental monitoring, the next step in irrigation management involves plant-driven control, in which irrigation is triggered directly by real-time physiological signals from the crop itself. Among the available indicators, leaf turgor pressure has emerged as one of the most sensitive and reversible markers of plant water status, reflecting the balance between transpiration and water uptake [17]. The SG-1000 turgor sensor, which detects micrometer-scale variations in leaf thickness and translates them into voltage signals, has been widely studied in orchard and field crops.

When integrated with automated control systems, it enables closed-loop irrigation scheduling that synchronizes misting with actual plant demand [18,19].

This study explored the integration of an SG-1000 leaf turgor sensor into an Arduino-based control system to manage irrigation in an aeroponic lettuce cultivation setup. Romaine lettuce (Lactuca sativa L. var. longifolia) was selected for its shallow root system, fast growth, and sensitivity to irrigation dynamics [20]. Two aeroponic systems were operated in parallel: a conventional timer-based system (TC) and a sensor-controlled system (AC). Environmental conditions, including temperature, humidity, barometric pressure, and photosynthetically active radiation, were continuously monitored using digital sensors (BME280, MLX90614, and ambient light sensors). An ultrasonic sensor tracked real-time nutrient reservoir levels, while VPD and leaf VPD were calculated to estimate transpiration demand. In the treatment group, misting was triggered based on SG-1000 output using a defined turgor threshold, while in the control group, misting occurred at regular 10 min intervals regardless of plant status.

All sensor data and irrigation events were logged at 30 s intervals. At harvest, growth-related, physiological, and biochemical parameters—including nitrate content, total phenolic content (TPC), and antioxidant capacity (FRAP)—were assessed to compare treatment effects. Previous studies suggest that sensor-based control can significantly improve water-use efficiency and reduce nitrate accumulation in lettuce [21,22]. Sensor-driven irrigation also contributes to better growth performance and biochemical quality through improved precision in water application [23,24]. The present work aimed to validate whether turgor-driven irrigation can enhance resource-use efficiency while maintaining or improving crop quality under aeroponic conditions. By focusing on a low-cost, open-source implementation, this study contributes to the development of scalable precision irrigation technologies for CEA systems aligning with the increasing emphasis on low-cost operational IoT-based systems for sustainable agriculture [9,23] particularly in controlled environments.

2. Materials and Methods

2.1. Growth Environment and Aeroponic System

The study consisted of two sequential cultivation cycles, each lasting 37 days: the first from 25 February to 2 April 2025 and the second from 9 April to 15 May 2025. Both were conducted in a fully enclosed, insulated, and climate-controlled growth chamber room at the Hellenic Agricultural Organization “ELGO-DEMETER” based in Thessaloniki, Greece. Environmental parameters were maintained using a 24,000 BTU split-type air conditioning system (TCL Technology, Huizhou China), while air renewal was managed through motorized inlet shutters and an exhaust fan providing approximately two full air exchanges per hour in the room. To regulate relative humidity, a dedicated humidifier (RAM, HydroGarden Ltd., Coventry, UK) and dehumidifier (Morris, Thessaloniki, Greece) unit were used. The above experimental setup was designed in such a way to eliminate external variability, simulating an ideal controlled-environment agriculture (CEA) scenario [25].

Two identical X-Stream 120 aeroponic units (Nutriculture UK Ltd., Lancashire, UK) were used, each designed to hold up to 120 net pots. Each unit measures approximately 120 cm × 67 cm × 50 cm and features a 70 L reservoir equipped with a 1000 L/h misting pump (18 W) that delivers nutrient-rich mist through 18 mist nozzles directly over the root zone. The nutrient solution is drained back in the tank after each irrigation event to ensure maximum water efficiency. Each aeroponic unit was illuminated using one LUMii Black 720 W LED 6-Bar Fixture (Lumii, Highlight Horticulture Ltd., Nottinghamshire, UK), mounted above the plant canopy. The lighting system was calibrated to deliver a photosynthetic photon flux density (PPFD) of approximately 300 μmol·m⁻²·s⁻¹, measured at canopy level using a factory pre-calibrated PAR meter (SpotOn Quantum PAR Light Meter, Innoquest Inc., Woodstock, IL, USA). The lights operated under a photoperiod of 16 h light/8 h dark throughout the experiments. The estimated daily light integral (DLI) was 17 mol·m⁻²·day⁻¹.

The spectral output of the grow lights consisted of the following:

• Red light (600–700 nm): 40–45%;
• Green light (500–580 nm): 25–30%;
• Yellow light (580–600 nm): 10–15%;
• Blue light (400–500 nm): 20–25%.

The selected spectral composition was tailored to support optimal lettuce growth by targeting key photosynthetic and morphogenetic responses. Red light (600–700 nm), comprising 40–45% of the spectrum, is the primary driver of photosynthesis and promotes leaf expansion and biomass accumulation [26]. Blue light (400–500 nm, 20–25%) regulates stomatal conductance and enhances photomorphogenesis, contributing to compact plant architecture and improved chlorophyll synthesis [26]. Green (500–580 nm) and yellow (580–600 nm) wavelengths, together accounting for 35–45%, penetrate deeper into the canopy and aid in whole-plant light distribution, indirectly supporting photosynthesis in lower leaves [27,28,29]. This balanced spectrum aligns with the physiological requirements of lettuce, enhancing both productivity and uniformity under controlled environment conditions [30,31].

During the entire cultivation period, vapor pressure deficit (VPD) was actively maintained within physiologically optimal limits, averaging 0.7 kPa, as determined from real-time temperature and humidity readings. This stable VPD profile supported consistent transpiration rates and minimized abiotic stress across both irrigation treatments [32].

A schematic representation of the aeroponic setup is shown in Figure 1. Representative images of the experimental setups and sensor integration are presented in Figure 2.

2.2. Instrumentation and Data Logging System

A fully customized automation system was developed using an Arduino Mega 2560 (Arduino AG, Monza, Italy) microcontroller board to coordinate the collection of environmental and physiological data, execute irrigation control logic, and store datasets for subsequent analysis. The system integrated multiple analog and digital sensors through I2C, UART, and SPI communication protocols and was designed to operate autonomously throughout the cultivation cycles [11,13,14].

2.2.1. Hardware and Sensor Integration

Ambient environmental parameters were monitored using a BME280 sensor (Gravity Series, DFRobot, Shanghai, China), which measured temperature (°C), relative humidity (%), and atmospheric pressure (hPa), connected via the I2C bus. Leaf surface temperature (°C) was monitored using a Gravity MLX90614 infrared temperature sensor (DFRobot, Shanghai, China), also interfaced through I2C and mounted at canopy height for non-contact thermal sensing. Light intensity was monitored using a Gravity Analog Ambient Light Sensor (DFRobot, Shanghai, China), positioned at canopy level to estimate light exposure and photoperiod status [6,7].

Plant physiological status was monitored via two SG-1000 turgor sensors (AgriHouse Inc., Berthoud, CO, USA), each installed on a fully expanded mature leaf of a representative plant in the Arduino-Controlled (AC) and Timer-Controlled (TC) systems, respectively. These sensors capture micrometer-scale fluctuations in leaf thickness, used as a proxy for turgor pressure, with a resolution of approximately 0.5 μm. Following the validated methodology [18,19], SG-1000 sensors were consistently mounted on fully expanded leaves at mid-canopy height. This placement yields a representative measure of plant water status by balancing between the rapid development of upper canopy leaves and the senescence of basal leaves. Previous studies have demonstrated that canopy position strongly influences the stability of leaf thickness–water status relationships, with mid-canopy leaves offering the most consistent signals [33]. To ensure reliable readings, only leaves free from shading or mechanical damage were selected, and sensors were securely attached to minimize artefacts [34]. Although only the SG-1000 sensor in the AC system was used to control irrigation events, the second sensor served as a passive monitoring device for physiological benchmarking under time-based irrigation [18,35,36,37,38].

To monitor nutrient reservoir levels, a waterproof ultrasonic distance sensor (A02YYUW, DFRobot, China) was mounted above the surface of each nutrient tank. The sensor emitted ultrasonic pulses and measured time-of-flight to the fluid surface, allowing real-time, non-contact estimation of solution depth with centimeter-level precision. These values were used to guide nutrient replenishment cycles, as described in Section 2.4.

Two 5 V relay modules (OEM, Shenzhen, China) were used to control irrigation pumps. Relay 1, triggered by physiological feedback from the SG-1000 sensor, controlled the misting pump of the AC system via digital pin 7. Relay 2, operated on a fixed-interval logic, activated the TC system pump via digital pin 8. A detailed summary of all sensors, control modules, and their respective Arduino connection protocols is provided in

Table 1. List of sensors, modules, and interface protocols used in the Arduino-based aeroponic irrigation system, including their function, communication type, and assigned microcontroller pins.

Sensor/Module	Function	Communication Protocol	Arduino Connection Pins
BME280 (DFRobot, China)	Ambient temperature, relative humidity, pressure	I2C	SDA: Pin 20, SCL: Pin 21
MLX90614 (DFRobot, China)	Leaf surface temperature (IR)	I2C	SDA: Pin 20, SCL: Pin 21
SG-1000 (AgriHouse Inc., USA)	Leaf thickness (turgor pressure proxy)	Analog	AC system: A2; TC system: A3
A02YYUW Ultrasonic Sensor (DFRobot, China)	Nutrient solution level in reservoir	UART	TX: Pin 16,18, RX: Pin 17,19
Gravity Analog Ambient Light Sensor (DFRobot)	Incident light intensity (PAR proxy)	Analog	A0
DS3231 RTC (Adafruit, USA)	Real-time clock for timestamping	I2C	SDA: Pin 20, SCL: Pin 21
Waveshare Micro-SD Storage Board (China)	Sensor and relay data logging	SPI	MOSI: 51, MISO: 50, SCK: 52, CS: 53
Relay Module 1 (OEM, China)	Misting pump control (AC system)	Digital Output	Pin 7
Relay Module 2 (OEM, China)	Misting pump control (TC system)	Digital Output	Pin 8

. All sensors and relays were connected to the central control unit using either serial communication protocols (such as I2C, UART, or SPI) or analog input channels, depending on the specific requirements of each device. Sensor readings and relay states were transmitted to the control unit through these interfaces, ensuring reliable data acquisition and device control. Data acquisition and control logic were synchronized with a DS3231 Real-Time Clock (RTC; Adafruit, New York, NY, USA) to ensure accurate temporal resolution. Data logging was executed through a Waveshare Micro-SD Storage Board (Waveshare Electronics, Shenzhen China), connected via SPI. The complete hardware configuration of the Arduino-based monitoring system is illustrated in Figure 3.

2.2.2. Software and Logging Architecture

The Arduino-based automation system was developed and programmed using the Arduino Integrated Development Environment (IDE, version 2.3.6), which served as the primary platform for code compilation, uploading, and real-time data monitoring. The code was written in Arduino C/C++, a simplified language based on standard C++ with added libraries and functions tailored for microcontroller development. Open-source libraries including Adafruit_BME280, DFRobot_MLX90614, RTClib, SPI, and SD were employed to facilitate sensor communication and peripheral control. The complete Arduino sketch was version-controlled and maintained in plain-text .ino format, ensuring transparency, reproducibility, and the potential for modification in future trials.

Throughout the experimental period, the Arduino Mega 2560 board was continuously connected via USB to a dedicated laptop. The laptop remained powered on with the Arduino Serial Monitor active to facilitate real-time data acquisition. The serial interface was also used for debugging, initialization confirmation, and validation of data logging events (e.g., SD card write success or error handling).

All program logic, including relay timing and conditional statements, was fully embedded on the microcontroller, enabling autonomous system operation independent of the computer connection. However, the real-time display of parameters via the serial console was crucial for verifying system integrity and for early detection of anomalies such as communication faults, sensor errors, or abnormal environmental readings.

At every loop iteration, the system executed a fixed sensor-reading sequence to maintain synchronized data streams. The BME280 provided air temperature, relative humidity, and barometric pressure; the MLX90614 infrared thermometer recorded both leaf and ambient temperatures; two SG-1000 turgor sensors produced analog voltage signals representing leaf water status; and dual A02YYUW ultrasonic sensors measured reservoir depth via UART communication. In addition, an analog ambient light sensor (A0 input) recorded incident light levels. All raw readings were immediately converted into engineering units (°C, %, hPa, V, mm) using scaling functions, while turgor voltages were simultaneously evaluated against the irrigation threshold (1.720 V).

Data, including timestamped records of ambient and leaf temperatures, relative humidity, pressure, light intensity, leaf turgor values, calculated vapor pressure deficit (VPD), calculated leaf vapor pressure deficit and the ON/OFF status of both irrigation relays were logged every 30 s to a file (log.txt) stored on the SD card, using a Waveshare micro-SD data logger with a DS3231 real-time clock (RTC) for timestamp accuracy.

Each log entry followed a structured format to facilitate later parsing and statistical analysis. Records were printed as plain-text lines in the following order:

• Timestamp;
• AirTemp (°C), RH (%), Pressure (hPa);
• LeafTemp (°C), LightLevel (ADC units);
• LeafVolt1 (V), LeafVolt2 (V);
• Tank1_Distance (mm), Tank2_Distance (mm);
• VPD (kPa), LeafVPD (kPa);
• Relay7 State (ON/OFF), Relay8 State (ON/OFF).

Data logging was triggered using a millis()-based timing condition to avoid blocking execution of the main control logic, ensuring high-frequency sensor updates and relay responsiveness. A fixed 2 s delay was also included at the end of each loop iteration to stabilize sensor communication without interfering with the 30 s logging interval. Sensor values and relay states were printed to the serial monitor for debugging and written to the SD card for offline analysis. The SD card interface was initialized at startup and checked before each logging cycle. If the card was unavailable or failed to initialize, the system printed a warning to the serial monitor without interrupting control logic. Non-blocking state machines were implemented for both relay channels, which allowed irrigation events to proceed independently of the sensor polling and SD logging routines. This ensured precise adherence to irrigation thresholds and timing intervals, even when communication lag or SD card latency occurred. The code structure employed non-blocking state machines for both relay channels, which preserved timing accuracy even under asynchronous sensor updates or I/O delays. The sensor-driven relay was triggered when SG-1000 voltage exceeded the programmed threshold of 1.720 V, with a 60 s activation window, a 240 s lockout period, and a 2 h failsafe safeguard. The timer-controlled relay operated strictly on a fixed 60 s ON/11 min OFF cycle. Relay states were logged in parallel with sensor data, enabling synchronization of irrigation history with physiological responses. This structure was chosen to ensure reliable long-term operation in an unmanned setting while maintaining compatibility with future control extensions such as wireless communication or cloud integration.

2.3. Irrigation Strategies

During the first three days after transplanting, both aeroponic systems were subjected to an identical root establishment phase to support seedling acclimatization. Mist irrigation was applied for 1 min every 10 min, delivered by their respective pumps, irrespective of plant water status. This ensured uniform hydration and root zone development before initiation of the experimental treatments.

Two independent irrigation strategies were implemented across identical aeroponic systems to evaluate the effects of dynamic, plant-driven control versus fixed-schedule irrigation on plant performance, physiological responses, and water use. Both systems operated under identical photoperiod, temperature, and nutrient conditions and followed a uniform early establishment protocol.

2.3.1. Sensor-Based Irrigation (Treatment Group)

Following the rooting phase, irrigation in the AC system was governed by real-time measurements from the SG-1000 leaf turgor sensor. A multi-stage calibration procedure was performed before the start of the experiment to determine a robust and physiologically meaningful activation threshold. Fully hydrated plants at early, mid, and late growth stages were monitored under stable environmental conditions to define baseline turgor voltages. Across all stages, stable voltages ranged between 1.700–1.715 V. During induced mild water stress (by suspending irrigation while holding environmental parameters constant), voltage readings rose due to leaf dehydration [35,36,37].

A threshold of 1.720 V was identified as the consistent inflection point corresponding to early, reversible turgor loss, characterized by subtle leaf wilting and reduced firmness. This threshold was integrated into the control logic such that when SG-1000 voltage readings exceeded 1.720 V, the Arduino activated Relay 1, initiating mist irrigation for 60 s. To prevent repeated triggering, a 4 min lockout period was implemented after each misting event.

In addition, a failsafe routine was included: if no sensor-triggered activation occurred within a 1 h window, the system autonomously triggered one 60 s irrigation event to ensure minimum hydration [38,39].

Control logic, activation events, and corresponding environmental conditions were logged every 30 s. This allowed temporal alignment between physiological turgor signals, environmental drivers (e.g., VPD), and irrigation responses, supporting detailed post hoc evaluation of irrigation efficiency and plant responsiveness.

Calibration Experiment for Threshold Definition

A series of calibration experiments was conducted to characterize SG-1000 voltage dynamics and to identify a physiologically meaningful threshold for irrigation control. All tests were performed in the same controlled-environment growth room used for the main cultivation experiments, ensuring that sensor output reflected plant water status rather than environmental variability. Air temperature, relative humidity, and photoperiod were identical to those described in Section 2.2. Plants initially underwent the standard rooting regime (1 min misting every 10 min) before the specific calibration protocols were applied.

The experimental design was adapted from Seelig et al. [19], who used staged stress and detachment tests to characterize leaf thickness responses. Calibration was performed at three crop stages (day 4–10, 18–24, and 30–37) using SG-1000 sensors placed on representative mid-canopy leaves. Sensor voltage was logged every 30 s with the Arduino-based system. Four complementary sub-experiments were conducted:

(a)

Non-stress diurnal stability: Plants were irrigated with a high-frequency misting regime (1 min every 5 min, 24 h/day) to maintain constant hydration. The objective was to capture natural day–night oscillations in sensor output under conditions that excluded water limitation.

(b)

Maximum stress (leaf detachment): A mid-canopy leaf equipped with the SG-1000 sensor was excised at the petiole under stable room conditions. Voltage was logged continuously for ≥120 min to record both the immediate transient rise and the subsequent plateau.

(c)

Extended water-deficit stress (progressive dry-down): Irrigation was withheld from intact plants until visible wilting occurred. Voltage was logged continuously throughout the dry-down phase and after re-irrigation to evaluate recovery dynamics.

(d)

Mild, reversible stress (short-term mist suspension): Misting was temporarily suspended for 1–2 h under otherwise constant conditions. Sensor voltage was logged throughout the suspension period and compared against visual assessments of turgor loss and recovery.

Each sub-experiment was replicated at all three crop stages to test for developmental effects. By combining these approaches, the full physiological operating range of the SG-1000 sensor was determined under aeroponic conditions, providing the basis for threshold definition.

2.3.2. Timer-Based Irrigation (Control Group)

The timer-based irrigation system followed a fixed-interval misting protocol, representing conventional aeroponic irrigation practices. Relay 2 was programmed to activate the irrigation pump every 10 min for 60 s, independent of plant water status or environmental feedback [21,40].

To enable physiological monitoring under the timer-based strategy, a second SG-1000 sensor was installed on a representative plant in the TC system. While not used to trigger irrigation events, its continuous logging allowed assessment of plant water status under fixed-schedule irrigation.

Analysis of sensor readings showed that many misting events occurred while turgor voltage remained within the optimal hydration range (1.710–1.718 V), especially during nighttime or periods of low VPD. This indicated that irrigation was frequently initiated before physiological demand, potentially resulting in water and nutrient overuse [41].

The synchronized logging of sensor voltages, environmental conditions, and relay activations provided a robust dataset for evaluating the alignment (or mismatch) between irrigation inputs and real-time plant needs under both strategies.

2.4. Plant Material and Nutrient Solution

Lettuce seedlings (Lactuca sativa L. var. longifolia) were germinated in peat moss and transplanted into 50 mm net pots at the third true-leaf stage. Expanded hydroponic clay pebbles were used to stabilize the seedlings into the net pots and fill in the gaps. A total of 28 plants were transplanted into each aeroponic system, ensuring consistent plant density across both setups Each system received 40 L of a half-strength Hoagland solution prepared using reverse osmosis (RO) water, with a background EC of 0.02 mS·cm⁻¹ and mixing stock nutrients. Final EC was ~1.12 mS·cm⁻¹, and pH was adjusted to 6.0. The nutrient concentrations used in the half-strength Hoagland solution are listed in

Table 2. Macro- and micronutrient concentrations of full-strength and half-strength Hoagland solutions used in the experiment, showing formulation details relevant to nutrient uptake and solution management.

Elements	Full Strength (ppm)	½ Strength (ppm)
Nitrogen	210.0	105.0
Phosphorus (P)	31.0	15.5
Potassium (K)	235.0	117.5
Calcium (Ca)	200.0	100.0
Magnesium (Mg)	48.0	24.0
Sulfur (S)	64.0	32.0
Iron (Fe)	1.5	0.75
Manganese (Mn)	0.5	0.25
Zinc (Zn)	0.05	0.025
Copper (Cu)	0.02	0.01
Boron (B)	0.5	0.25
Molybdenum (Mo)	0.01	0.005

.

The EC and pH of the nutrient solution were monitored daily using a Bluelab Combo Meter (Bluelab Corporation Limited, Tauranga, New Zealand), calibrated according to the manufacturer’s instructions. To prevent nutrient depletion and maintain chemical stability throughout the cultivation cycle, both systems were refilled with 10 L of the same nutrient solution three times during each 37-day cycle, based on volume readings from ultrasonic sensors positioned above the reservoirs. This intermittent replenishment strategy not only avoided full depletion of the nutrient tanks but also helped stabilize the EC and pH levels, which tended to fluctuate due to water uptake and nutrient consumption by the plants [20,42].

2.5. Biochemical and Physiological Measurements

At the end of each cultivation cycle, plants from both aeroponic systems were harvested and divided for a comprehensive set of biochemical, physiological, and morphological analyses to assess treatment effects on crop quality and plant responses.

Biochemical analyses included nitrate content, total phenolic content (TPC), and antioxidant capacity. Nitrate content was quantified from 2.5 g of fresh leaf tissue extracted in distilled water and analyzed spectrophotometrically using the salicylic acid method, with absorbance measured at 410 nm [43]. TPC was measured via the Folin–Ciocalteu method following extraction in 80% methanol, with absorbance recorded at 760 nm and results expressed as mg gallic acid equivalents per g fresh weight (mg GAE/g f.w.) [44]. Antioxidant capacity was evaluated using the FRAP assay at 593 nm and expressed as μg ascorbic acid equivalents per g fresh weight (μg AAE/g f.w.).

Physiological and compositional parameters were also measured. Total soluble solids (°Bx) were determined using a digital refractometer. Root and shoot biomass were recorded in both fresh and dry states (g), and plant height at harvest was recorded in cm. Colorimetric measurements of lettuce leaves were taken using a calibrated colorimeter, recording L* (lightness), C* (chroma), and h° (hue angle) values to assess visual quality and pigment development. Approximately half of each sample’s fresh biomass was stored at −30 °C for subsequent biochemical analysis, while the remaining material was oven-dried at 60 °C for mineral content evaluation and dry weight determination.

2.6. Statistical Analysis

All collected data, including laboratory assay results, growth measurements, and sensor-derived variables, were processed and analyzed to compare the effects of the two irrigation strategies on water usage, environmental dynamics, plant water status (turgor responses), and overall crop quality. All quantitative data collected from physiological, biochemical, and growth-related parameters were subjected to statistical analysis to evaluate the effects of irrigation strategy (Arduino-Controlled vs. Timer-Controlled) and experimental repetition. A General Linear Model (GLM) framework was used, treating irrigation strategy (Treats) and cultivation run (Exp) as fixed factors, and replication (Reps) as a random factor nested within experiments. Analysis of Variance (ANOVA) was conducted to test for main effects and interactions, with significance determined at the α = 0.05 level. When ANOVA results were significant, mean separation was performed using Tukey’s Honestly Significant Difference (HSD) post hoc test to identify statistically distinct groups. Assumptions of normality and homogeneity of variances were checked using residual plots and Levene’s test, respectively. All analyses were carried out using Minitab Statistical Software (version 21.3, Minitab LLC, State College, PA, USA). For each parameter, the model’s R-squared and predictive R-squared values were reported to assess goodness-of-fit and predictive accuracy. Outliers and influential observations were evaluated via standardized residuals and Cook’s distance but retained if they did not materially alter the interpretation. Statistical grouping was indicated with letter codes in all relevant tables to aid in interpretation.

2.7. Water, Nutrient Use Efficiency and Energy Calculations

Water and nutrient use efficiency metrics were computed to evaluate the resource performance of each irrigation strategy. Total water consumption was quantified for each aeroponic system across the 37-day cultivation period by summing the cumulative volume of nutrient solution replenished, based on ultrasonic sensor data and manual refills recorded during the experiment.

Water Use Efficiency (WUE) was calculated as the ratio of fresh shoot biomass (in grams) to the total volume of water used (in liters) per cultivation cycle, expressed in g·L⁻¹. The fresh biomass of all harvested plants per system was aggregated and divided by the corresponding total water use to determine WUE values for each treatment and experimental run.

Nutrient Use Efficiency (NUE) was estimated for nitrogen (N), phosphorus (P), and potassium (K) by combining measured yield data with assumed macronutrient concentrations in the applied half-strength Hoagland solution. Specifically, concentrations of 105 mg·L⁻¹ for N, 15.5 mg·L⁻¹ for P, and 117 mg·L⁻¹ for K were used as reference values. Total nutrient input per system was calculated by multiplying the cumulative water use by the nutrient concentration for each macronutrient. NUE was then expressed as grams of fresh biomass produced per gram of nutrient supplied (g·g⁻¹).

These metrics allowed for direct comparisons of irrigation efficiency and nutrient uptake efficiency between the Arduino-Controlled (AC) and Timer-Controlled (TC) treatments, independent of absolute yield values.

Pump energy consumption was estimated from the logged number of irrigation events and the rated power of the submersible pump (18 W). Each 60 s activation consumed ~0.0003 kWh (18 W × 60 s), which was multiplied by the activation frequency.

3. Results

3.1. Performance of the Arduino-Based Monitoring and Irrigation Control System

3.1.1. System Boot Sequence and Sensor Initialization

Upon power-up, the system executed a sequential initialization routine within the setup() function. Successful initialization of all hardware components was confirmed via serial messages, including the BME280 atmospheric sensor, MLX90614 infrared temperature sensor, DS3231 real-time clock (RTC), and the SD card interface via SPI. Each initialization call was encapsulated within conditional statements (e.g., if (!bme.begin()), if (!rtc.begin())), ensuring that any hardware failure produced a clear error message before entering the loop() function.

All sensors and peripherals were mapped to predefined pins and communication protocols. The BME280 sensor operated via I²C using SDA (pin 20) and SCL (pin 21), while the MLX90614 infrared sensor communicated over I²C at address 0x5A. The DS3231 RTC was also connected through I²C. The SG-1000 turgor sensor was connected to analog pins A2 (Arduino-Controlled system, AC) and A3 (Timer-Controlled system, TC), and the ambient light sensor was connected to analog pin A0. The A02YYUW ultrasonic sensor communicated via UART on digital pins 17 and 19 using a custom readA02YYUW () function. SD card data storage was managed via SPI, with the chip select defined as pin 53 and the Waveshare SD module interfaced through MOSI (pin 51), MISO (pin 50), and SCK (pin 52).

Relay control pins were defined using preprocessor macros (#define RELAY1 7 and #define RELAY2 8) and were initialized with appropriate mode settings using pinMode() during setup. This initialization procedure ensured that all components were fully functional before data acquisition and control processes began.

3.1.2. Sensor Readings and Data Acquisition Logic

The main data acquisition and control logic was implemented within the loop() function, which operated on a 30 s polling interval using a non-blocking millis()-based timing routine. Environmental data were collected through the following function calls: bme.readTemperature(), bme.readHumidity(), and bme.readPressure() for ambient conditions; mlx.getObjectTempCelsius() for leaf surface temperature; analogRead(A0) for light level; analogRead(A2) and analogRead(A3) for turgor sensor voltages; and readA02YYUW() for water reservoir distance measurements.

Turgor sensor voltages were converted from raw analog values using the formula voltage = analogRead (pin) ×* (5.0/1023.0) to enable comparison with predefined thresholds used for irrigation triggering. The system also computed vapor pressure deficit (VPD) using a dedicated calculateVPD() function, which took temperature and relative humidity as inputs. This function was executed twice per cycle—once for ambient VPD and once for physiological VPD using leaf temperature—enabling feedback-based irrigation scheduling.

3.1.3. Relay Activation Behavior

Relay 1 (Sensor-Triggered)

Relay 1 was controlled by a voltage threshold condition: if (leafVoltageV1 ≥ 1.720), where leafVoltageV1 represented the output from the SG-1000 sensor. Once this threshold was exceeded, the relay remained active for a 60 s irrigation pulse (relay7OnDuration = 60,000). A lockout period of 240 s (relay7OffDuration = 240,000) was enforced to prevent over-irrigation, using a relay7Cooling flag combined with millis() time tracking. To ensure hydration under prolonged low-turgor conditions, a failsafe condition triggered irrigation if millis()—relay7LastActivationTime exceeded one hour (relay7FailsafeDuration = 3,600,000).

This dynamic control logic ensured that irrigation events were closely coupled to real-time plant water status, while also preventing continuous activation or under-irrigation due to sensor noise or signal plateauing.

Relay 2 (Fixed Timer)

In contrast, Relay 2 operated on a fixed duty cycle, independent of any sensor input. It was activated for 60 s (relay8OnDuration = 60,000), followed by a 10 min off period (relay8OffDuration = 600,000). A state flag (relay8On) and millis() comparisons governed the switching logic. This relay served as the baseline control, mimicking conventional timer-based irrigation commonly used in aeroponic systems.

Throughout the experiment, Relay 1 responded dynamically to plant physiological signals, while Relay 2 maintained a static irrigation cycle, providing a comparative framework for assessing water and nutrient use efficiency.

3.1.4. Data Logging and SD Storage

Every 30 s, the system generated a new log entry in the log.txt file stored on the SD card. Using dataFile.print() commands, each log entry recorded the following variables: (1) timestamp from the RTC (rtc.now()); (2) all raw sensor readings; (3) calculated VPD values for both ambient and leaf conditions; (4) relay activation states (relay7On, relay8On); and (5) reservoir distance measured via the ultrasonic sensor.

Data were saved in plain-text CSV format to facilitate downstream analysis and visualization. This structure enabled full reconstruction of environmental conditions, plant responses, and irrigation behaviors over time. Across the two 37-day cultivation cycles, the total number of recorded rows exceeded 100,000, confirming the robustness and long-term stability of the logging system, memory buffer, and SD write protocol.

3.2. Calibration Results for Irrigation Threshold Definition

The calibration experiments designed following the staged approach described by Seelig et al. [19], established the operating range of the SG-1000 sensor under aeroponic conditions and identified the most physiologically meaningful irrigation threshold.

Under the non-stress high-frequency misting regime (1 min every 5 min, 24 h/day), voltages remained stable between 1.700 and 1.715 V at all growth stages. This range defined the baseline output of fully hydrated plants and confirmed that the sensor was able to capture natural diurnal oscillations without drift.

When leaves were detached at the petiole, voltages rose sharply and stabilized above 1.740–1.745 V within 120 min. This response was consistent across early, mid-, and late stages and represented the catastrophic upper bound of the sensor signal.

During progressive dry-down experiments, sensor voltages gradually increased as misting was withheld. Once values exceeded 1.735 V, visual wilting became pronounced, and recovery after re-irrigation was incomplete, indicating the onset of irreversible stress.

By contrast, short-term mist suspension (1–2 h) induced only moderate, reversible stress. In these cases, a reproducible inflection at 1.720 V coincided with the first visible signs of reversible wilting, characterized by slight leaf drooping and loss of firmness.

Taken together, the four sub-experiments revealed the sensor’s physiological dynamic range, from 1.700–1.715 V (hydrated baseline) to ≥1.735 V (irreversible stress). Within this window, 1.720 V consistently emerged as the earliest reproducible threshold, reflecting the transition from optimal hydration to the first reversible stress response. Unlike an intermediate value (e.g., 1.725 V), which could be mathematically interpolated but did not align with a clear plant signal, the 1.720 V threshold coincided robustly with a physiological event. Importantly, baseline values and threshold responses did not differ across growth stages, confirming that mid-canopy leaves provide stable and development-independent turgor signals, consistent with the observations of Seelig et al. [19]. A concise overview of the four calibration stages and their corresponding physiological interpretations is provided in

Table 3. Summary of calibration experiments for SG-1000 turgor sensor threshold definition under aeroponic conditions.

Experiment	Irrigation/Stress Treatment	Voltage Response (V)	Physiological Observation	Interpretation
Non-stress diurnal stability	High-frequency misting (1 min every 5 min, 24 h)	Stable 1.700–1.715	Natural day–night oscillations; no drift	Defines hydrated baseline range
Maximum stress (leaf detachment)	Leaf excision at petiole; stable environment	Rise to ≥1.740–1.745 within 120 min	Irreversible dehydration; no recovery	Defines catastrophic upper bound
Extended water-deficit (progressive dry-down)	Irrigation withheld until visible wilting	>1.735 → incomplete recovery after re-irrigation	Pronounced wilting; irreversible stress	Identifies onset of irreversible stress
Mild reversible stress (short-term suspension)	Misting paused 1–2 h under stable environment	Reproducible inflection at 1.720	First visible signs of reversible wilting (slight drooping, reduced firmness)	Defines operational threshold for irrigation control
Cross-test consistency	All sub-experiments repeated at early, mid, late stages	Baseline and threshold values consistent	Mid-canopy leaves provided stable signals	Confirms development-independent threshold

, which summarizes the experimental evidence supporting the selection of 1.720 V as the operational threshold.

3.3. Plant Growth and Biomass Accumulation

The overall growth and biomass partitioning of the lettuce plants were assessed through measurements of plant height, fresh plant weight, and root biomass at both fresh and dry weights.

3.3.1. Fresh Plant Weight (FPW)

The final Fresh Plant Weight (FPW), a primary indicator of overall yield, was evaluated to determine the impact of the irrigation strategies. The analysis of variance indicated that there was no statistically significant difference in FPW between the AC and TC treatments (F(1, 15) = 1.73, p = 0.208). This suggests that both irrigation methods were equally effective at producing total plant fresh biomass. The main effect of the experimental run (Exp) was also found to be non-significant (F(1, 15) = 1.28, p = 0.276), as was the effect of replication (Reps) within the experiments (F(5, 15) = 0.51, p = 0.766), indicating a high degree of uniformity in growth conditions and experimental setup across both time and space.

Furthermore, the interaction term Exp*Treats was not significant (F(1, 15) = 0.20, p = 0.662), demonstrating that the performance of the AC and TC treatments was consistent across both independent experimental runs.

Tukey’s Honestly Significant Difference (HSD) post hoc test for pairwise comparisons confirmed the ANOVA findings for the main treatment effect. The mean FPW for the TC treatment was 70.31 g, while the AC treatment produced a mean FPW of 68.29 g. Despite this minor numerical difference, both treatments were assigned to the same statistical group (‘A’), confirming the lack of a significant difference. When examining the interaction, all four combinations (Exp1-AC, Exp1-TC, Exp2-AC, Exp2-TC) were also grouped together, with means of 69.50 g, 70.83 g, 67.08 g, and 69.78 g, respectively.

The statistical model for FPW accounted for a relatively small portion of the total variance (R-sq = 27.67%), and the predictive R-squared value was 0.00%, suggesting that the factors included in the model were not strong predictors of the final fresh weight, which showed little variation in response to the treatments. One observation (Obs 19) was flagged as having a large residual, but this did not unduly influence the overall conclusions.

3.3.2. Plant Height (PH)

The final Plant Height (PH) was measured as an indicator of plant architecture and vertical growth. The ANOVA results showed that the main effect of irrigation treatment was not statistically significant (F(1, 15) = 1.24, p = 0.282). This finding indicates that the on-demand AC strategy supported vertical growth to the same extent as the continuous TC strategy. The main effects of Exp (F(1, 15) = 2.67, p = 0.123) and Reps (F(5, 15) = 0.34, p = 0.882) were also non-significant, further underscoring the consistent growing conditions throughout the study.

The interaction between experiment and treatment (Exp*Treats) was likewise found to be non-significant (F(1, 15) = 0.25, p = 0.621), confirming that the influence of the irrigation treatments on PH was stable and repeatable across both experimental runs.

Pairwise comparisons using Tukey’s HSD test corroborated the ANOVA results. The mean PH for the TC treatment was 19.2 cm, statistically identical to the mean PH of 18.9 cm for the AC treatment, with both being placed in the same statistical group (‘A’). The analysis of the interaction term also showed no separation among the four group means (Exp1-AC: 19.1 cm, Exp1-TC: 19.5 cm, Exp2-AC: 18.8 cm, Exp2-TC: 18.9 cm).

The overall model for PH explained only 28.11% of the observed variability (R-sq = 28.11%), with a predictive R-squared of 0.00%, indicating that plant height was not strongly influenced by the tested factors. This suggests that under both irrigation regimes, the plants reached a similar architectural state by the time of harvest.

3.3.3. Root Fresh Weight (RFW)

Analysis of the Root Fresh Weight (RFW) was performed to assess the impact of irrigation on the development of the root system’s fresh biomass. The main effect of irrigation treatment (Treats) was found to be statistically non-significant (F(1, 15) = 2.68, p = 0.122). The mean RFW for the TC treatment was 13.67 g, compared to 12.91 g for the AC treatment; however, Tukey’s HSD test grouped both treatments together (‘A’), confirming the absence of a significant difference.

In contrast to most other growth parameters, the main effect of the experimental run (Exp) was highly significant (F(1, 15) = 221.18, p < 0.001). Across both irrigation treatments plants in Experiment 2 had a significantly greater mean RFW (16.77 g) compared to those in Experiment 1 (9.81 g), indicating substantial baseline variability in root system size between the two runs. The replication factor (Reps) was non-significant (F(5, 15) = 0.98, p = 0.460).

Crucially, the interaction term Exp*Treats was not significant (F(1, 15) = 0.39, p = 0.542). This indicates that although the absolute RFW differed between experiments, the relative performance of the AC and TC treatments was consistent. In both experiments, the TC treatment produced numerically higher root fresh weights, but these differences did not achieve statistical significance within either run. The statistical model for RFW was very strong, explaining 93.86% of the variance (R-sq), driven largely by the significant effect of the experimental run.

3.3.4. Root Dry Weight (RDW)

In contrast to the fresh weight metrics, the analysis of Root Dry Weight (RDW), which reflects the actual biomass accumulated in the roots, revealed a significant effect of the irrigation treatment (F(1, 15) = 5.05, p = 0.04). This finding points to a fundamental physiological difference in root system development between the two strategies.

Tukey’s HSD post hoc test clearly separated the two treatment groups. The Time-Controlled (TC) treatment resulted in a significantly higher mean RDW of 1.10 g, placing it in statistical group ‘A’. The Arduino-Controlled (AC) treatment produced a mean RDW of 0.91 g, placing it in group ‘B’. This represents a 21.0% increase in root dry matter accumulation under the TC strategy compared to the AC strategy.

Similar to RFW, the main effect of the experimental run (Exp) was highly significant (F(1, 15) = 99.46, p < 0.001), with plants in Experiment 2 (mean RDW = 1.44 g) having substantially more root dry mass than those in Experiment 1 (mean RDW = 0.57 g). The effect of replication was non-significant (F(5, 15) = 0.71, p = 0.63). The interaction term Exp*Treats was also non-significant (F(1, 15) = 0.03, p = 0.87), confirming that the TC treatment consistently promoted greater RDW accumulation across both experimental runs. The model for RDW was robust, with an R-squared value of 87.81%. One observation (Obs 15) was identified as having a large residual, but its removal did not alter the significance of the main findings. Summary statistics for plant growth parameters under both irrigation treatments are presented in

Table 4. Growth performance metric of romaine lettuce under different irrigation strategies, including mean fresh plant weight (g), root fresh weight (g), plant height (cm), and root dry weight (g) across treatment and cultivation runs.

Experimental Run	Fresh Plant Weight (g)	Plant Height (cm)	Root Fresh Weight (g)	Root Dry Weight (g)
Exp1-TC	70.83 a	19.5 a	10.05 b	0.67 b
Exp2-TC	69.78 a	18.9 a	17.29 a	1.53 a
Exp1-AC	69.50 a	19.1 a	9.57 b	0.47 b
Exp2-AC	67.08 a	18.8 a	16.24 a	1.35 a
LSD	4.24	0.37	1.38	0.25
Mean (Exp1)	70.17 ns	19.3 ns	9.81 ***	0.57 ***
Mean (Exp2)	68.43 ns	18.8 ns	16.77 ***	1.43 ***
Mean (TC)	70.31 ns	19.2 ns	13.67 ns	1.10 *
Mean (AC)	68.29 ns	18.9 ns	12.91 ns	0.91 *

AC = Arduino-Controlled, TC = Time-Controlled. (A) Means not sharing the same letter differ significantly at p < 0.05. (B) ns: not significant (p > 0.05), * significant (p ≤ 0.05), *** highly significant (p ≤ 0.001). Two-way ANOVA (Exp x Treats) was performed.

.

3.4. Leaf Quality and Phytochemical Analysis

To assess the impact of irrigation on the nutritional and sensory quality of the lettuce, several phytochemical and colorimetric analyses were conducted.

3.4.1. Total Soluble Solids (TSS)

The Total Soluble Solids (TSS), measured in °Brix, is an important indicator of flavor, particularly sweetness. The ANOVA results showed that the irrigation treatment had no significant effect on TSS (F(1, 15) = 0.42, p = 0.526). Tukey’s comparison confirmed this, with the mean TSS for the AC treatment (3.12 °Brix) being statistically identical to the mean for the TC treatment (2.99 °Brix).

A highly significant difference was detected for the experimental run (Exp) (F(1, 15) = 41.57, p < 0.001), indicating that external factors varying between the two runs had a major impact on plant sugar metabolism. Plants in Experiment 2 had a significantly higher mean TSS (3.68 °Brix) than those in Experiment 1 (2.43 °Brix). The effect of replication was non-significant (p = 0.439).

The interaction term Exp*Treats was not significant (F(1, 15) = 0.00, p = 0.966), showing that the non-significant effect of the irrigation treatments was consistent across both experiments. The model for TSS explained 75.85% of the observed variance, primarily due to the strong effect of the experimental run.

3.4.2. Total Phenolic Compounds (TPH)

The concentration of Total Phenolic Compounds (TPH), which are health-promoting secondary metabolites, was measured to assess the nutritional quality of the lettuce. The main effect of the irrigation treatment was not statistically significant (F(1, 15) = 3.15, p = 0.096). While not meeting the conventional alpha level of 0.05, this p-value suggests a potential trend. The AC treatment yielded a numerically higher mean TPH concentration (0.328 mg GAE/g) compared to the TC treatment (0.307 mg GAE/g). However, Tukey’s HSD test did not separate these means, placing both in statistical group ‘A’.

The main effect of the experimental run (Exp) was highly significant (F(1, 15) = 34.55, p < 0.000), with plants from Experiment 2 (0.353 mg GAE/g) showing significantly higher TPH levels than those from Experiment 1 (0.283 mg GAE/g). The interaction term Exp*Treats was not significant (F(1, 15) = 1.49, p = 0.240), indicating a consistent treatment effect across the experiments. The statistical model explained 76.59% of the variance in TPH levels.

3.4.3. Total Antioxidant Activity (FRAP)

The total antioxidant capacity of the leaves was determined using the Ferric Reducing Antioxidant Power (FRAP) assay. The analysis of variance indicated that there was no significant difference in FRAP values attributable to the irrigation treatment (F(1, 15) = 0.75, p = 0.399). The mean FRAP value for the AC treatment was 77.29, while the TC treatment had a mean of 80.97; Tukey’s test confirmed these were not statistically different.

A significant effect of the experimental run (Exp) was found (F(1, 15) = 19.54, p < 0.001), with Experiment 2 showing significantly higher antioxidant activity (mean = 88.49) than Experiment 1 (mean = 69.77). The interaction between experiment and treatment was not significant (p = 0.739). The model for FRAP had a moderate R-squared value of 60.48% but a very low predictive power (R-sq (pred) = 0.00%), suggesting caution in its application. Two unusual observations (Obs 20 and 23) with large residuals were identified by the model diagnostics.

3.4.4. Nitrate Concentration

The analysis of leaf nitrate concentration yielded the most striking and statistically robust result of the entire study. The irrigation strategy was found to have a highly significant main effect on nitrate accumulation (F(1, 15) = 356.64, p < 0.001). This massive F-value underscores the profound impact of the watering regime on this critical food safety and quality parameter.

Tukey’s HSD post hoc test starkly illustrated this difference. The Time-Controlled (TC) treatment resulted in a mean nitrate concentration of 1019.3 ppm, which was assigned to statistical group ‘A’. The Arduino-Controlled (AC) treatment produced a significantly lower mean of 560.8 ppm, placing it in group ‘B’. This constitutes an 81.7% increase in nitrate accumulation under the TC strategy compared to the AC strategy.

The main effects of Exp (F(1, 15) = 63.36, p < 0.001) and Reps (F(5, 15) = 3.97, p = 0.017) were also significant. However, the interaction term Exp*Treats was non-significant (F(1, 15) = 0.00, p = 0.959), demonstrating the powerful consistency of this treatment effect. A closer look at the interaction means from Tukey’s test is revealing:

• Exp 1-AC: 463.5 ppm (Group D);
• Exp 2-AC: 658.1 ppm (Group C);
• Exp 1-TC: 923.3 ppm (Group B);
• Exp 2-TC: 1115.3 ppm (Group A).

This detailed breakdown shows that while baseline nitrate levels varied between experiments, the AC treatment consistently produced lettuce with significantly lower nitrate levels than the TC treatment within each respective run. The TC strategy consistently pushed nitrate accumulation to much higher levels. The statistical model for nitrates was exceptionally strong, explaining 96.70% of the total variance (R-sq = 96.70%). Results for phytochemical composition and nitrate accumulation in lettuce leaves are shown in

Table 5. Physicochemical and nitrate composition of romaine lettuce leaves grown under the Arduino-Controlled and Timer-Controlled irrigation, including total soluble solids (TSS), total phenolic content (TPH), antioxidant activity (FRAP), and nitrate concentration (ppm).

Experimental Run	Total Soluble Solids (°Brix)	Total Phenolic Compounds (mg GAE/g)	Antioxidant Activity (FRAP)	Nitrate Concentration (ppm)
Exp1-TC	2.37 b	0.279 b	70.89 bc	923.3 b
Exp2-TC	3.61 a	0.335 a	85.93 ab	1115.3 a
Exp1-AC	2.50 b	0.286 b	68.65 c	463.5 d
Exp2-AC	3.73 a	0.370 a	85.93 ab	658.1 c
LSD	0.57	0.038	11.70	94.56
Mean (Exp1)	2.43 ***	0.283 ***	69.77 ***	693.4 ***
Mean (Exp2)	3.68 ***	0.353 ***	88.49 ***	886.7 ***
Mean (TC)	2.99 ns	0.307 ns	80.97 ns	1019.3 ***
Mean (AC)	3.12 ns	0.328 ns	77.29 ns	560.8 ***

AC = Arduino-Controlled, TC = Time-Controlled. (A) Means not sharing the same letter differ significantly at p < 0.05. (B) ns: not significant (p > 0.05), *** highly significant (p ≤ 0.001). Two-way ANOVA (Exp x Treats) was performed.

.

3.5. Leaf Colorimetric Analysis

The visual quality of the lettuce was objectively quantified using the CIELAB color space, measuring lightness (L*), chroma (C*), and hue angle (h°).

3.5.1. Leaf Lightness (L*)

The L* value, which indicates the lightness of the green color from black (0) to white (100), was not significantly affected by the irrigation treatment (F(1, 15) = 0.17, p = 0.689). The mean L* value for the AC treatment (41.5) and the TC treatment (41.0) were statistically identical, as confirmed by Tukey’s HSD test.

A significant difference was observed for the experimental run (Exp) (F(1, 15) = 10.59, p = 0.005), with leaves from Experiment 1 being significantly lighter (mean L* = 43.2) than those from Experiment 2 (mean L* = 39.2). The lack of a significant Exp*Treats interaction (p = 0.840) indicates that the irrigation strategies did not influence leaf lightness in either experiment. The model for L* was relatively weak, explaining 47.62% of the variance with a predictive R-squared of 0.00%.

3.5.2. Leaf Chroma (C*)

The C* value, representing the intensity or saturation of the green color, was similarly unaffected by the experimental factors. The main effect of the irrigation treatment was not significant (F(1, 15) = 0.01, p = 0.931), with both AC (mean C* = 30.8) and TC (mean C* = 31.0) treatments producing leaves of statistically identical color saturation. The main effects of Exp (p = 0.233), Reps (p = 0.340), and the interaction Exp*Treats (p = 0.266) were all non-significant. This demonstrates that neither irrigation strategy had any discernible impact on the vibrancy of the leaf color. The statistical model for C* was very weak, explaining only 37.70% of the variance and having no predictive power.

3.5.3. Leaf Hue Angle (h°)

The hue angle (h°), which specifies the shade of the color (where values around 120–130° represent a typical green), showed no significant main effect of the irrigation treatment (F(1, 15) = 1.37, p = 0.260). The mean h° for the AC treatment was 124.6° and for the TC treatment was 124.1°, both falling into the same statistical group.

Interestingly, the interaction term Exp*Treats approached the threshold of statistical significance (F(1, 15) = 4.36, p = 0.054). This suggests a potential trend where the effect of the treatment on the shade of green may have differed between the two experimental runs. However, a closer examination with the more conservative Tukey’s HSD test on the four interaction groups found no significant differences among their means (Exp1-AC: 124.5°, Exp1-TC: 125.0°, Exp2-AC: 124.7°, Exp2-TC: 123.1°). Therefore, it is concluded that the irrigation strategy did not consistently or significantly alter the fundamental hue of the lettuce leaves. Comparative colorimetric measurements across treatments are compiled in

Table 6. Colorimetric parameters of lettuce leaves, including lightness (L*), chroma (C*), and hue angle (h°), as affected by irrigation strategy and cultivation run, assessed using CIELAB color space.

Experimental Run	Leaf Lightness (L*)	Leaf Chroma (C*)	Leaf Hue Angle (h°)
Exp1-TC	43.1 a	31.1 a	125.0 a
Exp2-TC	38.8 a	30.9 a	123.1 a
Exp1-AC	43.3 a	33.1 a	124.5 a
Exp2-AC	39.6 a	28.5 a	124.7 a
LSD	3.42	5.83	1.50
Mean (Exp1)	43.2 **	32.1 ns	124.8 ns
Mean (Exp2)	39.2 **	29.7 ns	123.8 ns
Mean (TC)	41.0 ns	30.8 ns	124.1 ns
Mean (AC)	41.5 ns	31.0 ns	124.6 ns

AC = Arduino-Controlled, TC = Time-Controlled. (A) Means not sharing the same letter differ significantly at p < 0.05. (B) ns: not significant (p > 0.05), ** moderately significant (p ≤ 0.01). Two-way ANOVA (Exp x Treats) was performed.

.

3.6. Chlorophyll Index and Quantum Yield Response

Chlorophyll-related parameters were assessed to evaluate potential differences in leaf pigmentation and photosynthetic performance under the two irrigation strategies. The Chlorophyll Concentration Index (CCI) served as a proxy for pigment accumulation, while Chlorophyll Quantum Yield (QY) reflected the efficiency of photosystem II.

3.6.1. Chlorophyll Concentration Index (CCI)

The analysis of CCI revealed a significant influence of the experimental run (Exp) and the interaction between experiment and irrigation treatment (Exp*Treats), while the main effect of treatment (Treats) alone was not statistically significant (p = 0.506). Experiment 2 produced significantly higher CCI values compared to Experiment 1 (p < 0.001), suggesting environmental or temporal factors played a major role in chlorophyll accumulation. Although TC and AC treatments yielded similar average CCI scores, their interaction with each experiment revealed distinct performance: the TC treatment in Experiment 2 had the highest mean CCI (21.58), significantly greater than all other combinations. This indicates that under certain environmental conditions, constant irrigation may enhance chlorophyll retention or synthesis more effectively than sensor-driven misting. However, the absence of a main treatment effect suggests that both systems generally support similar leaf pigmentation, with variations largely driven by environmental run-to-run differences.

3.6.2. Chlorophyll Quantum Yield (QY)

The General Linear Model analysis indicated no statistically significant effects of treatment, experiment, or their interaction on Chlorophyll Quantum Yield (QY). The irrigation strategy (Treats) had a negligible influence (p = 0.964), and neither experimental run (Exp) nor the interaction term (Exp*Treats) approached significance. The model’s low R² value (42.56%) and a predictive R² of 0.00% suggest high variability and minimal explanatory power. These findings indicate that both the Arduino-Controlled (AC) and Time-Controlled (TC) systems maintained similar levels of photosynthetic efficiency in lettuce leaves. The consistent QY across treatments underscores the physiological stability of the plants under both irrigation regimes, suggesting that water delivery timing had little impact on the photosystem II efficiency during the measurement period. These results support the conclusion that while irrigation mode influences nitrate levels and potentially secondary metabolites, it does not disrupt fundamental photosynthetic performance under controlled environmental conditions. Leaf chlorophyll content and quantum yield values are presented in

Table 7. Leaf chlorophyll concentration and photosynthetic efficiency metrics, comparing Chlorophyll Concentration Index (CCI) and Quantum Yield (QY) under sensor-based and timer-based irrigation treatments.

Experimental Run	Chlorophyll Concentration Index (CCI)	Quantum Yield (QY)
Exp1-TC	20.49 b	4058.83 a
Exp2-TC	21.58 a	4047.83 a
Exp1-AC	20.81 b	4050.42 a
Exp2-AC	21.06 ab	4058.83 a
LSD	0.54	105.24
Mean (Exp1)	20.65 ***	4054.50 ns
Mean (Exp2)	21.32 ***	4053.33 ns
Mean (TC)	21.03 ns	4053.21 ns
Mean (AC)	20.93 ns	4054.63 ns

AC = Arduino-Controlled, TC = Time-Controlled. (A) Means not sharing the same letter differ significantly at p < 0.05. (B) ns: not significant (p > 0.05), *** highly significant (p ≤ 0.001). Two-way ANOVA (Exp x Treats) was performed.

.

3.7. Water and Nutrient Use Efficiency

Water consumption throughout the 37-day cultivation period ranged from 35.48 to 43.14 L depending on the irrigation strategy and experimental run. In both experiments, the Arduino-Controlled (AC) irrigation strategy resulted in lower total water usage compared to the Time-Controlled (TC). Specifically, in Experiment 1, the AC system consumed 35.48 L versus 42.11 L for the TC system, while in Experiment 2, the values were 36.23 L and 43.14 L, respectively. This corresponds to a reduction in water use of 15.7% in the first run and 16.0% in the second run.

When accounting for yield, Water Use Efficiency (WUE) defined as grams of fresh biomass per liter of water used was significantly higher under the AC treatment. In Experiment 1, WUE increased from 47.07 g·L⁻¹ (TC) to 54.85 g·L⁻¹ (AC), representing an improvement of 17.4%. In Experiment 2, WUE improved from 45.95 g·L⁻¹ (TC) to 51.83 g·L⁻¹ (AC), a gain of 18.2%.

To assess Nutrient Use Efficiency (NUE), average nutrient concentrations for half-strength Hoagland solution were assumed: 105 mg·L⁻¹ for nitrogen (N), 15.5 mg·L⁻¹ for phosphorus (P), and 117 mg·L⁻¹ for potassium (K). Nutrient uptake efficiency was consistently higher under the AC strategy in both experiments. In Experiment 1, NUE-N increased by 17.6%, NUE-P by 18.0%, and NUE-K by 17.6% compared to the TC control. In Experiment 2, the corresponding increases were 19.3%, 16.9%, and 17.1%, respectively.

These results confirm that turgor-feedback-based irrigation not only conserves water but also improves the utilization efficiency of key macronutrients in aeroponic romaine lettuce production under controlled environmental conditions. Detailed water and nutrient use efficiency values under each treatment and run are shown in

Table 8. Water and macronutrient use efficiency in romaine lettuce production, including Water Use Efficiency (WUE), and Nutrient Use Efficiency (NUE) for Nitrogen (N), Phosphorus (P), and Potassium (K) under each treatment and experimental run.

Experimental Run	Total Water Use (L)	Yield (g)	WUE (g·L−1)	NUE-N (g·g−1 N)	NUE-P (g·g−1 P)	NUE-K (g·g−1 K)
AC Experiment 1	35.48	1946	54.85	525.1	3572.7	473.9
TC Experiment 1	42.11	1982	47.07	446.4	3025.0	402.9
AC Experiment 2	36.23	1878	51.83	514.1	3491.4	463.1
TC Experiment 2	43.14	1954	45.95	430.9	2987.4	395.3
Mean (AC)	35.86	1912	53.34	519.6	3532.0	468.5
Mean (TC)	42.63	1968	46.51	438.7	3006.2	399.1

. Percentage improvements in WUE and macronutrient uptake efficiency under sensor-based irrigation are summarized in

Table 9. Percentage improvements in WUE and NUE for nitrogen, phosphorus, and potassium under Arduino-controlled irrigation compared to Timer-Controlled strategy, reported separately for each cultivation run and averaged overall.

Experimental Run	WUE (%)	NUE-N (%)	NUE-P (%)	NUE-K (%)
First Run	+17.4	+17.6	+18.0	+17.6
Second Run	+18.2	+19.3	+16.9	+17.1
Average	+17.8	+18.5	+17.5	+17.4

.

3.8. Irrigation Frequency Across Cultivation Runs

The number of pump activations was used as a proxy for irrigation frequency to compare the two water management strategies. In both cultivation runs, the Arduino-Controlled (AC) system was programmed to automatically log each pump activation onto an SD card via the embedded Arduino code, enabling precise monitoring of irrigation frequency throughout the experiments.

In Run 1, the AC system recorded 2903 pump activations through the total cultivation cycle, compared to 3506 activations in the Time-Controlled (TC) system. This corresponds to a reduction of 603 activations, or approximately 17.2% fewer events under the AC strategy. In Run 2, the AC system registered 2974 activations, while the TC system logged 3592 activations, again yielding a reduction of 618 activations, or 17.2%. These consistent reductions in pump operation reflect the efficiency of the sensor-driven irrigation approach in triggering misting events only in response to real-time plant water status, thereby avoiding unnecessary irrigation cycles. Beyond water-saving implications, this reduced activation frequency also contributes to lower mechanical stress on the pump system and extends hardware longevity. The close agreement between runs further reinforces the robustness, consistency, and repeatability of the Arduino-based turgor-feedback control strategy across seasonal or operational variation. The number of irrigation events logged for each system during both cultivation cycles is presented in

Table 10. Total number of irrigation events recorded per system during two 37-day cultivation cycles, highlighting pump activation frequency and corresponding reductions under sensor-based irrigation.

Irrigation System	Arduino-Controlled	Time-Controlled	Difference (-)	Reduction (%)
Measured Activations (Run 1)	2903	3506	603	17.2
Measured Activations (Run 2)	2974	3592	618	17.2
Average	2938.5	3549.0	610.5	17.2

.

3.9. Comparative Environmental Performance of Hydroponic and Aeroponic Irrigation Strategies

To provide a more comprehensive understanding of the environmental footprint associated with each irrigation strategy, a simplified environmental impact assessment was conducted comparing the two setups under discussion with a standard Nutrient Film Technique (NFT) hydroponic setup. This assessment was based on experimentally measured water, electricity, and nutrient inputs for the aeroponic systems and literature-derived values for the hydroponic system, combined with established emission factors. All systems were assumed to utilize a half-strength Hoagland nutrient solution. Based on experimentally recorded water consumption, the Arduino-Controlled (AC) aeroponic system was the most resource-efficient, requiring an average of 18.8 L·kg⁻¹ of nutrient solution per kilogram of fresh lettuce yield. The Time-Controlled (TC) aeroponic system used 21.7 L·kg⁻¹, demonstrating lower efficiency. For comparison, standard hydroponic Nutrient Film Technique (NFT) systems typically require around 20 L·kg⁻¹, with literature reporting values ranging from 20 ± 3.8 L·kg⁻¹ when realistic operational factors are considered. These include elevated ambient temperatures, increased surface evaporation from the thin nutrient film, and periodic system flushing [45,46,47]. In nitrogen terms, literature on hydroponic NFT-grown lettuce indicates biomass nitrogen concentrations averaging 2–5% of dry mass [48] (i.e., ~28–37 g N·kg⁻¹ DM), and a typical head yielding ~55 g DM per kg fresh weight. This translates into approximately 1.6–2.0 g N·kg⁻¹ fresh mass under ideal conditions. Experimentally derived nitrogen inputs are calculated at ~1.97 g·kg⁻¹ for AC and ~2.28 g·kg⁻¹ for TC values that align well with theoretical NFT use. The AC system also exhibited significantly lower nitrate accumulation in edible tissues. Mean nitrate content in AC-grown lettuce was 560.8 mg·kg⁻¹, compared to 1019.3 mg·kg⁻¹ in the TC system. This 45% reduction enhances food safety and aligns comfortably within regulatory nitrate limits set by the European Commission (e.g., 2500–4500 mg·kg⁻¹ for fresh lettuce). Literature supports this nitrate accumulation values for hydroponically grown lettuce, which vary significantly with nitrogen management and light conditions [49,50]. The lower nitrate levels in the AC system also reduce the risk of environmental contamination through post-harvest leaching [42]. In addition, pump energy use was derived from the logged number of irrigation events (Table 10) and the rated load of the submersible pump (18 W). With each 60 s misting pulse consuming ~0.0003 kWh, the Time-Controlled (TC) system averaged 1.07 kWh per crop cycle, whereas the Arduino-Controlled (AC) system averaged 0.88 kWh per crop cycle (

Table 11. Comparative environmental footprint per kilogram of fresh lettuce yield, summarizing water use, pump energy consumption, and nitrate accumulation across hydroponic (NFT), time-based aeroponic (TC), and Arduino-controlled aeroponic (AC) systems.

Category	Unit	Hydroponic (NFT) (Estimated)	Time-Controlled Aeroponic (TC) (Experimental Avg.)	Arduino-Controlled Aeroponic (AC) (Experimental Avg.)
Water Use	L·kg−1	23.8	21.7	18.8
Pump Energy Use	kWh/cycle	15.98	1.07	0.88
Nitrate Accumulation	mg·kg−1	~1019.3 (Typical)	1019.3	560.8

). This corresponds to a 17.8% reduction in energy use per cycle, in line with the observed reduction in irrigation activations. For comparison, a conventional Nutrient Film Technique (NFT) system operating with a continuously running 18 W recirculation pump over a 37-day crop cycle would require 15.98 kWh per cycle, highlighting the markedly lower energy demand of both aeroponic strategies. These findings confirm that integrating plant-responsive irrigation control not only reduces the number of pump activations but also translates into measurable energy savings, thereby lowering the operational footprint of the AC approach. Overall, these findings demonstrate that integrating plant-responsive, sensor-based irrigation not only conserves water and energy but also contributes to a lower agrochemical footprint and improved crop quality. This resource-based environmental footprint approach provides valuable insights into the sustainability potential of smart aeroponic systems within controlled-environment agriculture (CEA). A comparative analysis of environmental metrics across irrigation strategies and hydroponic reference systems is provided in Table 11.

4. Discussion

The current study introduces a novel, plant-driven irrigation strategy for aeroponic cultivation, leveraging real-time turgor measurements from SG-1000 leaf thickness sensors integrated with an open-source Arduino-based control (AC) platform. Unlike conventional timer-based systems, which apply irrigation at fixed intervals regardless of plant status, our approach dynamically adjusts irrigation events in direct response to the physiological needs of the crop. This represents a significant advancement in the field of smart sensor-based systems for crop monitoring and management [11,34,35,37].

Aeroponics is already recognized for its resource efficiency, but the absence of buffering capacity makes precise irrigation control essential. Traditional timer-based methods are inherently limited, as they cannot adapt to rapid changes in plant water demand caused by environmental fluctuations or developmental stage. By contrast, the system developed in this work continuously monitors leaf turgor, a direct indicator of plant water status, and triggers irrigation only when it is regarded necessary. This closed-loop feedback mechanism ensures that water and nutrients are supplied in synchrony with actual plant requirements, minimizing waste and environmental impact.

An important aspect of this study was the calibration of the SG-1000 sensor to define a physiologically meaningful irrigation threshold. Through a series of staged experiments including non-stress diurnal stability, leaf detachment, extended dry-down, and mild reversible stress, the full operating range of the sensor was characterized under aeroponic conditions. These tests consistently identified 1.720 V as the inflection point marking the transition from optimal hydration (1.700–1.715 V) to the first signs of reversible wilting. This reproducibility across developmental stages supports the robustness of the threshold used in the AC control logic and confirms that mid-canopy leaves provide stable, development-independent signals [19].

Integrating this experimentally validated threshold ensured that irrigation events were triggered at the earliest point of physiological stress, avoiding both premature misting and delayed responses. To further ensure robustness of the threshold definition, the control algorithm incorporated system-level safeguards, including a 4 min lockout to prevent overshooting and a failsafe hourly irrigation pulse to prevent under-watering. These design features guaranteed that plants were never exposed to prolonged stress, even if transient fluctuations occurred around the 1.720 V threshold. Such safeguards support the physiological reliability of the design in the absence of direct correlation with gas-exchange parameters.

Over the course of two 37-day cultivation cycles, the implementation of the sensor-based irrigation system led to a substantial reduction in both water consumption and pump activations when compared to the conventional timer-controlled method. More specifically, water use decreased by 15.7% and 16.0% in each cycle, while the number of pump activations was reduced by 17.2% [16]. These reductions highlight the system’s ability to precisely synchronize irrigation events with the actual needs of the plants, rather than relying on fixed schedules [19,39].

The enhanced alignment between irrigation and plant demand not only conserved water but also contributed to a notable increase in Water Use Efficiency (WUE), which improved by 17.4% and 18.2% across the two cycles. Similarly, Nutrient Use Efficiency (NUE) for nitrogen, phosphorus, and potassium increased by approximately 17.5%, reflecting the benefits of delivering nutrients in closer accordance with plant uptake patterns [51]. The reduction in irrigation frequency also resulted in lower energy consumption per kilogram of yield, a saving that becomes increasingly significant at larger operational scales [11].

These resource savings were achieved without compromising yield or shoot quality: shoot biomass and plant height remained statistically equivalent between treatments. Root dry weight, however, was significantly higher under the timer-controlled strategy, suggesting that continuous irrigation favored greater belowground dry matter allocation. One possible explanation is that the frequent and uniform misting under TC minimized transient episodes of water deficit, thereby maintaining sustained assimilate flow to the root system. By contrast, AC plants experienced slightly longer intervals between misting events, sufficient to maintain shoot turgor but not strong enough to induce compensatory root growth. This reduced need for exploratory root development likely shifted carbon allocation toward maintaining shoot function, consistent with the observed balance in growth performance. This finding contrasts with some previous precision irrigation studies that reported stronger root development under demand-driven regimes [35,52], and may reflect the inherently high misting frequency in aeroponics. Nevertheless, the absence of differences in root fresh weight indicates that functional water uptake capacity was preserved under both strategies. The 45% reduction in leaf nitrate content under the sensor-based regime (560.8 ppm vs. 1019.3 ppm; p < 0.001) is a notable result which supports the view that on-demand irrigation can prevent luxury nitrate accumulation and enhance food safety. In the TC system, irrigation was frequently triggered while leaves were already in an optimal turgor state, leading to oversupply of nutrient solutions and excessive nitrate assimilation. By contrast, AC supplied water and nutrients only when turgor indicated actual demand, thereby moderating uptake and reducing leaf nitrate concentrations. This effect also suggests that nutrient uptake was better synchronized with metabolic assimilation capacity, preventing accumulation of unassimilated nitrate in vacuoles. The stability of chlorophyll levels across treatments indicates that nitrogen was still available in sufficient amounts for photosynthetic pigments, highlighting that reduced nitrate was due to improved utilization efficiency rather than limitation. Other quality parameters, such as total phenolic content, antioxidant capacity, and sensory traits, were preserved, confirming that the proposed AC system maintained both nutritional and market value. The slight but non-significant increase in phenolics under AC may be explained by mild, reversible stress priming. Shorter hydration intervals can act as metabolic signals that stimulate phenylpropanoid pathways, thereby promoting phenolic accumulation without causing yield penalties [52,53].

By contrast, constant hydration in TC prevented such signaling, leading to slightly lower values. The lack of differences in TSS further supports that carbohydrate accumulation is mainly governed by light intensity and photoperiod, with irrigation mode exerting only minor influence under non-limiting water supply. Environmental variation across runs (e.g., light intensity and VPD) exerted a stronger influence on secondary metabolism and sugar accumulation.

Importantly, no evidence of impaired photosynthetic performance was observed, as chlorophyll quantum yield and shoot growth remained stable. The Exp × Treats effect noted for chlorophyll index, with TC outperforming AC in one run, likely reflects subtle environmental interactions rather than a consistent treatment advantage. The maintenance of QY despite reduced nitrate under AC demonstrates that photosystem II efficiency was not compromised, reinforcing that nitrogen supply remained adequate for chlorophyll biosynthesis and electron transport [54].

This stability underscores the robustness of lettuce physiology under both irrigation regimes and confirms that the AC strategy sustained hydration above the reversible stress threshold identified in calibration experiments.

A further perspective is provided by the environmental performance comparison. When benchmarked against hydroponic NFT systems, the Arduino-Controlled (AC) aeroponic strategy demonstrated superior resource efficiency, requiring only 34.3 L of water and 0.88 kWh of electricity per crop cycle, compared to 41.9 L and 1.07 kWh for the Timer-Controlled (TC) system, and ≈15.98 kWh for NFT. Moreover, nitrate accumulation was 46% lower in AC compared to TC, and remained below typical hydroponic values [55,56], underscoring its potential contribution to improved sustainability and food safety. A key innovation of the present study is the seamless integration of plant-based sensors for real-time physiological feedback with a robust, low-cost, and open-source control platform. Through combining direct, continuous monitoring of plant water status with an accessible AC-based architecture, the proposed system enables precise, adaptive irrigation management that is both scalable and readily adoptable by digital agriculture specialists [57]. A customized Arduino platform, programmed for non-blocking timing and continuous sensor polling, proved reliable and accessible, with integrated VPD calculations providing additional environmental context. Failsafe mechanisms ensured resilience, enabling advanced crop status monitoring that aligns with the goals of next-generation controlled-environment agriculture.

The deployment of leaf-contact turgor sensors must be considered in relation to system context and planting density. Although their use has been reported in open-field and greenhouse experiments, these applications often face challenges such as mechanical disturbance, heterogeneous canopies, and exposure to variable weather conditions, which limit stability and long-term reliability [58].

By contrast, Controlled-Environment Agriculture (CEA) and vertical farming systems provide conditions that are more favorable, as cultivation rooms are compact, standardized, and hygienic, thereby facilitating sensor placement, minimizing obstruction risk, and enabling systematic inspection. Within these environments, scalability for high-density plantings is more realistically achieved through a zone-level replication strategy, in which a limited number of strategically positioned sentinel sensors govern irrigation across canopy sections, supplemented by redundancy to mitigate single-point failure. The economic feasibility of this approach lies in avoiding sensor-per-plant installation, making costs proportionally low relative to the demonstrated gains in water, nutrient, and energy efficiency [59].

Operational safeguards, such as the failsafe hourly irrigation pulse employed here, further support robustness in dense canopies where occasional obstruction or sensor dropout may occur. Emerging solutions, including multiplexed wiring, wireless sensor arrays, and hybrid systems combining plant-based signals with environmental proxies, may reduce installation complexity and further lower costs. Non-contact alternatives such as imaging and thermal sensing also represent complementary tools for large-scale adoption. Taking together, these considerations indicate that the scalability of plant-driven irrigation using turgor sensors is most promising within commercial CEA systems, where the balance of environmental control, crop value, and resource savings justifies investment.

Building on the demonstrated advantages of integrating real-time plant feedback with open-source control systems in aeroponic cultivation, this work highlights a promising pathway towards more sustainable and precise crop management. The system’s ability to dynamically align irrigation with actual plant needs resulted in significant improvements in water and nutrient use efficiency, energy savings, and crop quality, all while maintaining yield, underscoring the transformative potential of customized smart sensor-based systems for advancing resource-efficient agriculture.

Nevertheless, several important considerations remain for future research and broader application. The generalizability of the system’s performance across different crop species and under varying environmental conditions warrants further investigation, as physiological responses and optimal irrigation thresholds may differ among plant types. Refinement of the control logic, potentially through the application of machine learning algorithms, could further enhance the adaptability and precision of irrigation scheduling. Expanding the sensor suite to include ion-selective probes, root-zone temperature monitoring, and additional environmental parameters would provide a more comprehensive understanding of plant–environment interactions and support even finer control of resource delivery. Moreover, regulated deficit irrigation strategies could be explored as a means to biofortify crops and stimulate the accumulation of health-promoting secondary metabolites. To fully assess the broader impact and sustainability of such smart systems, future work should incorporate detailed life cycle assessments and techno-economic analyses. Finally, in-depth characterization of root system architecture, including parameters such as diameter, suberization, and lignification, would offer valuable insights into the physiological mechanisms underlying plant adaptation to dynamic irrigation regimes.

Overall, it can be concluded that the current study demonstrates that the integration of plant-based sensors with accessible, open-source control platforms is capable of offering significant advances in terms of efficiency and sustainability for aeroponic systems. By directly linking irrigation to plant physiological status, the present approach sets a new benchmark for smart crop monitoring and management and establishes a new standard for innovation and efficiency in controlled-environment agriculture.

5. Conclusions

The current study demonstrates that turgor feedback-based irrigation, enabled by an Arduino and SG-1000 sensor, can substantially reduce water use by 15.7–16.0% (average 15.9%) and pump activations by 17.2% while lowering nutrient inputs in aeroponic lettuce cultivation without compromising yield or visual quality. The sensor-driven system improved resource use efficiency, reduced nitrate accumulation by 45%, and showed potential for enhancing phytochemical content. While root dry weight was lower under dynamic irrigation, shoot growth and photosynthetic function remained unaffected. These results support the adoption of plant-driven control strategies as scalable, sustainable solutions for precision agriculture in controlled environments. Future work should expand testing across crops and explore integration with adaptive algorithms to optimize performance further.