A real-time on-site precision nutrient monitoring system for hydroponic cultivation utilizing LIBS

Abstract

Indoor hydroponic farming is an advanced cultivation technique with diverse sustainability benefits, such as facilitating local produce, minimizing transportation costs and emissions, and enabling year-round crop cultivation. To optimize crop growth for enhanced yield, improved crop quality, and reduced environmental footprint, precise monitoring and replenishment of essential nutrients within hydroponic systems is crucial. Current methods employed in most commercial farms for online nutrient supply monitoring is limited to pH and conductivity measurements. These techniques can only offer an indication of the overall change in the complex nutrient mixture and lack the capability to precisely identify the specific nutrient or quantify the nutrient content. Most of the existing techniques for measuring individual nutrient levels are expensive and invasive, necessitating sample preparation, frequent recalibration, and skilled personnel for operation. In this context, we propose and demonstrate a real-time, on-site monitoring system for the precise analysis of hydroponic nutrient supply based on laser-induced breakdown spectroscopy (LIBS). We also discuss the system design considerations, parametric optimizations, limit of detection (LOD), and limit of quantitation (LOQ) of key nutrient components such as potassium (K), sodium (Na), calcium (Ca), and magnesium (Mg), using the proposed approach. The detection range of the developed LIBS-based monitoring system can encompass the typical concentration range observed in hydroponic nutrient solutions used at agricultural farms. This technique offers rapid online monitoring of individual nutrient components, providing precise, real-time analysis and the potential to enable comprehensive automation capabilities for current and future hydroponic farms.

Graphical abstract

Background

Indoor vertical farming is an alternative to traditional agriculture methods, offering high productivity within a small footprint. This is a promising technology for sustainable food production, especially in constrained spaces or land-exhausted areas [1]. The indoor farming techniques ensure year-round crop supply through precise control of inputs, also often termed as ‘speed breeding’ techniques [2]. These facilitate rapid crop growth, thereby enhancing the crop yield within a stipulated time. Majority of the indoor farms employ hydroponic technology as an alternative to conventional soil-based cultivation [3, 4]. In this approach, the roots are either supported with grow substrates such as perlite, clay, and rock or suspended in water, which is supplemented with essential nutrient solutions. Within the hydroponic system, the essential inputs such as ambient temperature, light, water, nutrients, and oxygen are precisely controlled and systematically supplied to the plants. It also offers the modularity to be placed anywhere (indoors or outdoors) and in any spatial configuration (e.g., vertical columns, walls, horizontal or flat-bed system, etc.) depending on the available space and the type of crops being grown. Hydroponic cultivation allows more efficient use of space, water, and fertilizers, as well as better control of pest factors thereby increasing productivity, crop quality, and economic income [5].

Among the various factors influencing the hydroponic cultivation system, the composition of the nutrient solution is regarded as one of the most crucial in terms of crop yield and quality [6, 7]. The nutrient solution in a hydroponic system is an aqueous solution containing ions from soluble salts of essential elements. In general, there are 17 important nutrients for plants: (i) the macronutrients comprising nitrogen (N), phosphorus (P), potassium (K), calcium (Ca), sulphur (S), magnesium (Mg), carbon (C), oxygen (O), and hydrogen (H), and (ii) the micronutrients comprising iron (Fe), boron (B), chlorine (Cl), manganese (Mn), zinc (Zn), copper (Cu), molybdenum (Mo), and nickel (Ni). The amount of nutrients required by each crop depends on the individual crop species and the nutrient type. Nutrient levels beyond the sufficiency range of a crop can cause a decline in overall crop growth and a reduction in crop health either due to a deficiency or due to a toxicity. Nutrient deficiency occurs when an essential nutrient is not available in enough amounts to satisfy the crop growth requirements whereas toxicity occurs when a nutrient exceeds crop requirements, causing a decrease in growth or quality. Although nutrients in a hydroponic system are formulated in ratios that reflect the needs of the crop, the nutrient level will gradually deplete or change from the optimal value by transpiration during the crop growth. This rate of change will vary according to the crop type.

An advanced hydroponic system should be able to diagnose such nutrient deficiencies and toxicities by means of specialized measurement tools. Current hydroponic systems use conventional methods based on the electroconductivity and pH of the nutrient solution to get a feedback loop to replenish the nutrients [8, 9]. However, these measurements are unable to provide the exact composition and concentration of the individual nutrients and their depletion levels, but rather only give an overall indication of the change in the complex nutrient mixture [7]. This limits its capability to successfully replenish the right nutrients needed for ideal crop growth. For optimal replenishment, accurate monitoring of the amounts of each individual nutrient component in the hydroponic nutrition solution is essential. This optimization ensures the precise composition of the regenerated nutrient solution, thereby maximizing the quality and productivity of the crops. Specific nutrient measurements such as inductively coupled plasma-optical emission spectroscopy (ICP-OES), inductively coupled plasma-mass spectroscopy (ICP-MS), atomic absorption, high performance liquid chromatography (HPLC), etc., are expensive offline measurements, and often require the use of “wet chemistry”, requiring dedicated skilled workforce, equipment, and tools that are generally not available in conventional indoor farms [10]. While some attempts have been made in the past to automate the existing “wet chemistry” laboratory to provide online analysis to the hydroponic system, such modalities almost always require the use of chemical reagents and involve tedious sample preparation [7, 10,11,12,13,14,15,16].

Recently, real-time monitoring of hydroponic nutrient solutions using ion-selective electrodes and the Internet of Things (IoT) have been explored [8,9,10,11, 17,18,19,20]. In this method, ion-selective features of specialized materials were used to generate electrical signals that can be detected and quantified and hence specifically monitor the concentration of individual nutrients in the solution [14, 15, 17, 21,22,23,24,25]. However, not all elements in the nutrient solution can be detected as the selective electrodes for many of the elements have not been developed yet [8, 9]. Moreover, some electrodes for nutrients, such as Ca, show poor sensitivity, which further reduces over time [8, 23, 24, 26]. Most of these electrodes need frequent recalibration as the continuous monitoring of the nutrient solution leads to oxidation of the electrode surfaces leading to reduced detection sensitivity over time. Furthermore, these measurements are also greatly influenced by environmental conditions such as the temperature of the nutrient solution, and signal drift of electrodes for specific ions [8]. As an alternative, some indoor farms opt for visual inspection of symptoms displayed on crops to interpret nutrient deficiency and toxicity. However, interpreting nutrient stress symptoms remains challenging due to the complex combination of plant-nutrient reactions [27]. For instance, nitrogen (N) and sulphur (S) deficiency symptoms are similar depending upon placement, growth stage, and severity of deficiencies or multiple deficiencies and/or toxicities can occur at the same time [27].

In this context, we propose a LIBS-based measurement system, as a new tool for nutrient monitoring and smart nutrient management for hydroponic systems. LIBS is a rapidly evolving, in situ technique that can be used for the quantitative analysis of elemental compositions in various samples [28,29,30,31]. The analysis of the characteristic atomic, ionic and molecular emission spectra from the laser-induced plasma makes it possible to classify and measure elementary components in the sample. The key advantages of this technique include minimal or no sample preparation, versatility of the materials that can be analyzed, and real-time measurement capability. LIBS on liquid samples poses specific challenges such as excessive splashing, as well as rapid quenching of the plasma intensity leading to a shorter plasma lifetime [28]. This impacts the signal intensity and consequently the detection capability and sensitivity of liquid LIBS. However, optimization of the sampling configuration can mitigate these potential challenges [32]. A more detailed discussion on the state-of-the-art of liquid LIBS analysis, potential experimental strategies, and challenges can be found elsewhere [33, 34]. Here, we detail the system design considerations, experimental parameter optimization, detection limits for various nutrient components, and the potential of the proposed approach to accurately estimate the nutrient concentrations. The prospects of the proposed LIBS-based technique for online monitoring of specific nutrients, along with the potential to integrate automation and online monitoring features in forthcoming hydroponic farms, are also discussed.

Methods

Figure 1 shows the schematic of the LIBS-based nutrient solution monitoring system developed. A small amount (10 mL) of the nutrient solution from the hydroponic system is pumped to a quartz cell using a micro-pump and temporarily stored there by closing the outlet valve of the quartz cell for LIBS analysis. Once the measurements are completed, the nutrient solution in the cell is pumped back to the hydroponic system. The micro-pump is equipped with an inlet filter, which filters out suspended solids or algae in the nutrient solution. The cell is made of polished square quartz windows of 2 cm length. Nanosecond laser pulses from an Nd-YAG laser (Quantel, Q-smart 850) at the wavelength of 1064 nm and a repetition rate of 10 Hz are used to generate the plasma for the LIBS measurements. It should be noted that the high optical absorption rate, and minimal scattering and splashing at the excitation wavelength of 1064 nm enables the generation of plasma with distinct line emissions with a longer lifetime [35].

Fig. 1

Schematic representation of the LIBS-based hydroponic nutrient solution monitoring system

A plano-convex lens (L1) with a focal length of 50 mm is employed to focus the laser beam inside the quartz cell containing the liquid nutrient sample. The resulting plasma emission from the sample is collected in the backward geometry (180° collection scheme) using the same lens (L1). The collected emission is transmitted through the dichroic mirror and focused by another plano-convex lens (L2) to be directly coupled to the entrance slit of a high-resolution Czerny–Turner spectrometer (Kymera 328i, Andor). The spectrometer is equipped with a grating consisting of 1200 grooves per mm and an intensified charge-coupled device (ICCD) camera (iStar, Andor). The spectrometer disperses the collected light using the grating, which is then captured by the ICCD camera. The spectral range monitored during the measurements is 350–800 nm. To efficiently capture the LIBS signal, the acquisition is appropriately gated, and the delay between the laser pulse and detection is controlled by the timed triggering of the spectrometer using the laser source’s synchronization output signal. The data acquisition software plays a crucial role in integrating the laser triggering and spectral collection processes, enabling real-time nutrient quantification. The recorded LIBS spectra were processed using Andor SOLIS software. Emission peak identification was achieved by comparing emission peaks with the NIST database. Signal-to-background (S/B) ratios were calculated for each measurement to assess the quality of the acquired spectra. The details relevant to the calculation of the S/B ratio were earlier reported by our group [33]. The S/B ratio was calculated using a custom software code written in Python and each emission spectrum presented is an average of 200 measurements unless stated otherwise. Liquid samples for the calibration were prepared according to standard protocols, including dilution and homogenization steps to ensure uniformity. Analytical grade sodium chloride (NaCl), potassium chloride (KCl), magnesium sulfate (MgSO₄), and calcium chloride (CaCl₂) were procured from Sigma Aldrich. Stock solutions of each analyte were prepared by dissolving the respective solute in deionized (DI) water and were subsequently stored in standard volumetric flasks. To attain solutions of varied concentrations required for the analysis, serial dilution was carried out. Since the measurements are done on samples pumped out from the grow tray into a sample cell in an isolated measurement system, external lighting or plant growth lighting does not affect the measurements.

Results and discussion

Parametric investigations conducted for optimization of the system showed that the analytical performance of the LIBS-based nutrient monitoring system relies on various system parameters. The primary controllable parameters influencing the LIBS analytical performance are the laser pulse energy and the delay between the laser pulse and the spectrometer gating. Optimizing these experimental parameters has the potential to minimize background noise and improve overall system performance.

As LIBS utilizes a high-energy laser pulse to create the laser-induced plasma, the plasma properties and the analytical performance of the system are highly dependent on the exciting pulse energy. Optimizing the laser pulse energy is thus crucial for achieving enhanced performance. To determine the optimal laser pulse energy, LIBS spectra generated from a sample solution containing 400 ppm Na were recorded at a fixed delay time of 500 ns and gate width of 1000 ns, employing different laser pulse energies ranging from 5 to 25 mJ. Figure 2a displays the recorded Na analyte signal corresponding to different laser pulse energies, and Fig. 2b illustrates the Na analyte S/B ratio. The data in Fig. 2 represent an average of over 200 measurements. From the figures, it is evident that the signal intensity increased with the rise in the exciting laser pulse energy and reached an optimal value at ~ 15 mJ. At higher laser pulse energies, the plasma absorbs more energy, elevating the plasma temperature, and causing an increase in continuum background emission. Consequently, this leads to a reduction in the S/B ratio. Additionally, beyond the optimum laser energy, the intensity of the discrete line emission began to decrease, which is attributed to the shielding effect and self-absorption within the plasma [35,36,37].

Fig. 2

a The intensity variation in the atomic line signal of Na (589.14 nm) and b the corresponding S/B ratio at different laser pulse energies

It is well-known that plasma emission is dominated by the continuum background emission (so-called Bremsstrahlung radiations) and discrete line emissions [35, 37]. These emissions originate and decay at different times and rates [35]. To capture an optimum LIBS signal, it is required to properly select an optimum detection window. In general, when a nanosecond laser pulse induced plasma is formed, the continuum background dominates in the first several microseconds, which decays much faster than the discrete line emissions that emerge at a later time. Thus, controlling the detection gate delay allows for the acquisition of LIBS signals with minimal background and thus an enhanced S/B ratio. To investigate the influence of the delay between laser pulse and the detector gating on the LIBS signal obtained from the present configuration, the gate delay was varied from 0 to 2200 ns, and the LIBS signals were recorded. The laser energy was kept at the optimum value of 15 mJ in all the measurements. Figure 3 shows the variation in the atomic line signal of Na and the corresponding S/B ratio at different gate delays. As seen from the figure, there is a significant dependence of the S/B ratio on the gate delay, with the optimal value identified to be around 300–500 ns. It should be noted that a similar optimal gate delay value is observed for K. In contrast, Mg and Ca required shorter gate delays (10–20 ns) for optimal signal capture.

Fig. 3

a The intensity variation of the atomic line signal of Na (589.14 nm) and b the corresponding S/B ratio with the delay between the laser pulse and detector gating. The laser pulse energy was kept at the optimum value of 15 mJ

Following the optimization of the LIBS signals, the proposed nutrient monitoring system was calibrated for the elements K, Na, Ca, and Mg as representative constituents of the nutrient solution. The calibration was done using known concentrations of the salt solutions in DI water. For Ca and Mg, this ranged from 200 to 1000 ppm, for Na from 1 to 100 ppm, and for K from 1 to 200 ppm. The calibration curves at the optimized experimental conditions are generated by recording the LIBS signals of these elements across various concentrations. Figure 4 shows the calibration curves for K (766.5 nm), Na (589.14 nm), Ca (422.64 nm), and Mg (518.37 nm). Each data point represents an average of 200 measurements and the error bars indicate the standard deviations. The solid lines represent linear fits to the data points. These results indicate that the LIBS signal intensity has a linear dependence on the analyte concentration in the concentration ranges investigated. The high values of the determination coefficient (R²) (values close to 1.0) for the linear fits demonstrate the reliability of the LIBS-based system in measuring the concentrations of these nutrients in a hydroponicnutrient solution. The limit of detection (LOD) and the limit of quantitation (LOQ) of the system are the most important parameters for any detection system. The LOD defines the lowest concentration of the analyte that can be detected with the developed LIBS system but is not necessarily quantified as an exact value; whereas, LOQ defines the lowest concentration of analyte that can be quantitatively determined with acceptable precision and accuracy. The LOD and LOQ can be estimated using the equations [28, 38]:

$${\text{LOD }} = { 3}\sigma /{\text{S,}}$$

(1)

$${\text{LOD }} = { 10}\sigma /{\text{S,}}$$

(2)

where σ is the standard deviation of the background and S is the sensitivity, which is given by the ratio of the LIBS signal intensity to the sample concentration. The parameter S can be determined from the slope of the calibration curve [28, 32]. Using these equations, the LOD (LOQ) values for K (766.5 nm), Na (589.14 nm), Ca (422.64 nm), and Mg (518.37 nm) were found to be 660 ppb (2.2 ppm), 439 ppb (1.46 ppm), 102 ppm (339 ppm), and 78 ppm (263 ppm), respectively. The LOD and LOQ values observed in this investigation are appropriate for the proposed application and are comparable to the values reported in similar works in the literature. For instance, Zhang et al. reported LOD values for Na in aqueous solutions in the range of 1 ppm [39]. Other investigations such as those by Cremers et al. reported LOD values for Ca and Mg in the range of 1–100 ppm [40]. It may be noted that the typical concentration range of nutrients in indoor farms falls within the detection range of the proposed system.

Fig. 4

Calibration curves of K (766.5 nm), Na (589.14 nm), Ca (422.64 nm), and Mg (518.37 nm)

The feasibility and accuracy of the proposed system for measuring the concentration of various elements in a solution is assessed using a representative complex nutrient solution containing known concentrations of K, Na, Ca, and Mg. The specific concentrations chosen were as follows: Na and K at 100 ppm, and Ca and Mg at 600 ppm. This mixed salt solution allowed us to ensure comprehensive detection even in a complex nutrient mixture. Matrix-matching was not employed in this measurement as our goal was only to evaluate the LIBS performance in a simple DI water matrix. Figure 5a shows the LIBS spectra (averaged over 200 acquisitions) of the complex nutrient solution, where the atomic emission lines from K, Na, Ca, and Mg are identified. From the acquired LIBS spectra, the S/B ratio values of the elemental emission lines are calculated, and the concentration of each element is estimated using the corresponding calibration curves.

Fig. 5

a LIBS spectra of the mixed solution containing K, Na, Ca, and Mg. Atomic emission lines from these elements are identified. b Comparison of the concentrations of the elements measured using the proposed monitoring system with the actual concentration values. The error bars indicate standard deviation

Figure 5b shows a comparison of the concentrations of K, Na, Ca, and Mg determined by the proposed system and the actual concentration values. The estimated concentrations of K, Na, Ca, and Mg are 89 ± 3 ppm, 92 ± 4 ppm, 564 ± 11 ppm, and 520 ± 13 ppm, respectively. It can be seen that the estimated concentrations are very close to the actual values. From these results, the relative error can be calculated using the equation:

$$\text{Relative error }\left(\text{\%}\right)= \left(\frac{(\left|\text{AC}-\text{EC}\right|)}{\text{AC}}\right)\times 100,$$

(3)

where AC is the actual concentration and EC is the estimated concentration using the proposed LIBS system. The relative error for these elements were found to be between 6 and 13% (K ~ 11 %, Na ~ 8 %, Ca ~ 6 %, Mg ~ 13 %), which is comparable to or better than the values measured in previously reported nutrient monitoring systems [8, 9, 23]. The low relative error values indicate the proposed system’s high sensitivity and low variability in measurements. These results show that the proposed nutrient monitoring system can be used in the development of unmanned large-scale hydroponic farms.

Figure 6 illustrates the conceptual framework of the proposed LIBS-based automated real-time on-site nutrient monitoring system for implementation in smart unmanned hydroponic farms. The algorithm for the overall monitoring process and control systems is illustrated. The automated nutrient replenishment system will be primarily based on the difference between the (crop specific) reference library values and current measured values. In the algorithms being developed, it is important to implement proper thresholding conditions and revise these thoroughly to make sure that the error margin is eliminated from the replenishment regulation criterion. Also, it is to be noted that for several plant species, the optimal elemental composition encompasses a reasonably broad range, and the system may only need to be activated if the measured values are beyond this range (and not for a minor deviation from a mean value). Farm users should be able to set optimal nutrient values for each crop type at specific growth stages through the user interface. For reliable measurements, the sampling will need good filter systems to make sure that the debris does not affect the measurements and good isolation and environmental control would be required at the sampling zone in the measurement system. After each measurement, the cell can be cleaned by flushing it with a small volume of DI water to prevent errors caused by trace elements or calcium deposition on the cell walls in subsequent measurements. The proposed system could be integrated to both indoor and outdoor hydroponic farms. Being a non-invasive inline measurement technique, the proposed liquid LIBS-based approach can help in nutrient replenishing without waste generation. In addition to the liquid cell demonstrated, the use of different sampling configurations such as a microfluidic channel could improve the detection sensitivity. In hydroponic systems utilizing drip irrigation, automated nutrient injection systems can be programmed to deliver specific nutrient solutions tailored to the requirements of various plants or growth stages. By utilizing such monitoring systems to adjust nutrient levels, hydroponic growers can achieve higher yields, better quality crops, and embrace more sustainable production practices. Additionally, the integration of data logging systems to continuously monitor nutrient levels and environmental conditions within the hydroponic system can enable optimum nutrient dosing. It is envisaged that this automated real-time monitoring system would be a better alternative to existing technologies that require complex sample preparation for precisely analyzing the individual nutrients.

Fig. 6

Block diagram and algorithm of the proposed LIBS-based automated real-time on-site nutrient monitoring system

Conclusions

In summary, we have proposed and demonstrated a real-time, on-site nutrient monitoring system for hydroponic cultivation utilizing liquid LIBS. The developed system stands apart from conventional monitoring systems that focus solely on pH and electroconductivity values or rely on labor-intensive wet chemical methods. It provides a distinct advantage by enabling rapid, on-site assessment of precise levels of individual nutrients within the hydroponic solution. This capability facilitates accurate estimation of the required application loads of variable-rate fertilizers in hydroponic cultivation, thereby enhancing the efficient utilization of crop fertilizers. The system allows for reliable measurement of all essential nutrients and its sensitivity covers the typical concentration range found in commonly used hydroponic nutrient solutions. With relative errors for different elements ranging from 6 to 13% (K ~ 11 %, Na ~ 8 %, Ca ~ 6 %, Mg ~ 13 %), the system demonstrates improved accuracy compared to previously reported nutrient monitoring methods. The proposed system can be applied in various hydroponic setups, from small-scale indoor farms to large commercial operations, offering particular benefits in urban environments where precise nutrient management is crucial. The current system can be further improved in terms of detection sensitivity and refining sampling methods to ensure consistency and accuracy for micronutrients. Future research should focus on these improvements and explore integration with automated hydroponic technologies to create a fully automated nutrient management system. By enhancing detection sensitivity, refining sampling approaches, and enabling real-time nutrient adjustments, the system is envisaged to have the potential to revolutionize efficient nutrient management in hydroponic farming. This will lead to higher yields, more efficient resource use, and a significant reduction in labor costs, making it a valuable tool for modern agriculture.