1. Introduction

Plants are affected by many types of biotic (insects, fungi, bacteria, viruses) and abiotic (salinity, drought, light, temperature) stress which can have a negative impact on agriculturally and economically significant florae [1–4]. Global climate changes make the abiotic stress effects more adverse. It is important to understand complex physiological and molecular mechanisms of plant stress response in order to engineer stress tolerance and improve crop yield. The complex responses to stress in plants involve multiple steps including the signal perception by stress sensors, generation of signaling molecules such as reactive oxygen species and abscisic acid, and modification of cellular ion contents of elements such as Ca, K, Na and Fe. The abiotic stress effects such as drought and salinity may be reduced by the timely application of relevant measures at the appropriate field locations. Therefore, early and rapid detection of various plant stresses has been the focus of extensive investigations

Currently available technologies for detecting plant stress include plant tissue and soil water content monitors via thermography [5,6], visible/near-infrared reflectance [7,8], and UV-visible fluorescence imaging [9–12]. These methods provide indirect information related to plant stress but do not directly measure the nutrient atomic and molecular contents. Other chemical analytical methods such as gas chromatography, mass spectrometry and nuclear magnetic resonance can provide such information on a laboratory scale but require time-consuming elaborate sample preparation techniques. Rapid detection with little sample preparation is necessary to scale the sensing technology to the field size. LIBS provides the corresponding advantages and is therefore a promising candidate to address these challenges.

LIBS measurements are performed by focusing a laser pulse onto the sample surface causing ablation or vaporization of the sample material forming a plasma plume. The hot plasma breaks down the ablated material into elemental components, and excites electrons into higher energy levels. Once the hot plasma expands and cools, the electrons return to the ground state emitting photons of characteristic frequencies from atomic constituents [13]. Based on these spectroscopic signatures, LIBS has been used to differentiate between tissues [14], identify bacterial strains [15,16], and determine soil pollution [17,18]. LIBS has also been used for the analysis of the chemical composition of plants [19–23]. Semi-quantitative and quantitative measurements of Ca, K, Na, Fe, Mn, Zn, Pb, N and other elements were achieved in various plant species such as grasses, maize, wheat, clover, cotton, soy, spinach, sunflower, lettuce, potato, coffee, pepper, mango and many others.

Within the broad applications of LIBS, there is much flexibility in the performance of the technique. Pulses used for ablation of small amounts of sample surface can be obtained from nano-, pico-, and femtosecond lasers. Femtosecond pulses are of particular interest because high powers can be obtained with smaller energies; therefore, resulting in less damage to the sample [13]. Additionally, shorter pulse lengths allow for less interaction of the pulse with the formed plasma, causing a reduction in broad continuum background noise [13]. Femtosecond laser pulses have been used in plant science applications [19,21,24], medical purposes, such as in dentistry cavity preparation [25] and LASIK surgery [26,27], and micromachining [28]. In particular, femtosecond laser pulses were found to yield more accurate results for applications where plant samples were employed due to the lower continuum background [24]. Here, we demonstrate the possibility of using LIBS for rapid detection of drought stress in wheat and gardenia. We performed daily monitoring of relative contents of Ca, K, Na and Fe using LIBS and identified spectroscopic markers of mild and severe stress. Our results may be readily extended to remote applications on the field scale using unmanned ground vehicle (UGV) and unmanned aerial vehicle (UAV) technologies.

2. Results

2.1 Drought stress in decorative and crop plants

We performed LIBS measurements on decorative indoor and cultivated outdoor plants such as gardenia and wheat, respectively. Gardenia plants are fragrant flowering evergreen shrubs or trees typically used as decorative plants. Wheat is a major cereal crop ranked third in the U.S. in overall farm revenues, and drought stress has been shown to cause substantial reduction in wheat crop yield [29]. We induced drought stress conditions by withdrawing water in these two plant species and measured the effects on watered (non-stressed) and non-watered (stressed) samples. The details of sample preparation and stress measurements are described in the Methods section. Figure 1(a) shows visual signatures of drought stress in gardenia which are revealed as wilting and color change in the non-watered (stressed) plant on day 31. The corresponding relative water content in % of day 0 is shown in Fig. 1(b). The water content stayed constant in the watered (non-stressed) plant and gradually decreased in the stressed plant. Similar results were obtained for wheat [Figs. 1(c) and 1(d)]. The constancy of the relative water content of the watered (non-stressed) sample plants pertains to the stability of the system. However, the visual signatures of stress were observed in both watered and non-watered wheat plants. The stress effects on the watered wheat were attributed to the indoor lighting conditions imposed during the LIBS measurements for the entire treatment period. Visual inspection could not clearly distinguish between these mild environmental lighting (mild stress) effects from the drought (severe stress). We were able to distinguish these types of stress using LIBS as described below.

 

Fig. 1 Visual signatures of drought stress and water content. Photographs of gardenia (a) and wheat (c) plants on the first and last days of the treatment periods for both watered and non-watered treatment groups. Corresponding plots showing relative water content (% of day 0) for gardenia (b) and wheat (d) plants for the entire treatment periods.

Download Full Size | PDF

2.2 Atomic spectral signatures of plant stress

The schematic of LIBS measurements of plant stress is shown in Figs. 2(a) and 2(b) for the non-stressed and stressed plants, respectively. The high intensity beams of femtosecond laser pulses were focused on leaf samples placed on the translation stage that was scanned during the spectral measurements to replace the ablated areas for consecutive data points. We observed significant changes in the LIBS spectra taken from the stressed versus non-stressed plants over a time span of approximately one month for gardenia and two weeks for wheat plants [Fig. 2]. The LIBS spectra consisted of a set of narrowband peaks on top of a broadband background. The observed peaks were compared to the previous work [29–33] and assigned according to the NIST atomic spectral line database [34] based on the macro and micronutrients acquired by plants from the soil and air [35] [Figs. 2(c) and 2(h)]. We identified Na, Ca, O and Fe in gardenia [Fig. 2(c)] and Na, K, Ca, O and Fe in wheat [Fig. 2(h)]. The K peak was absent in gardenia [Figs. 2(d) and 2(f)] and in the non-stressed [Figs. 2(i) and 2(k), black lines] and mildly stressed [Fig. 2(i), red line] wheat. However, it was clearly identified in the severely stressed wheat [Fig. 2(k), red line] and therefore can be used as a spectroscopic marker for the detection of drought stress. Similarly, the Fe peaks were absent in the non-stressed gardenia [Figs. 2(d), red and black lines, and 2(f), black line] and wheat [Figs. 2(i) and 2(k), black lines] plants, but were present in the stressed gardenia [Fig. 2(f), red line] and wheat [Figs. 2(i) and 2(k), red lines]. The observed wavelengths for each designated peak from Fig. 2 can be found in Table 1.

 

Fig. 2 Atomic spectral signatures of abiotic stress in gardenia and wheat. Schematic of the LIBS experiment on a non-stressed (a) and stressed (b) plant. A high intensity beam of femtosecond laser pulses is focused onto the surface of a plant leaf generating emission of light from atomic components of laser-induced hot plasma. LIBS spectra of gardenia (c) – (g) and wheat (h) – (l) plants: (c) and (h) show LIBS signals collected in the selected full spectral range on the first (black line) and last (red line) days of the treatment for the stressed plants; spectral regions of interest show relative changes in the LIBS signals of watered (d, e, i, j) and non-watered (f, g, k, l) plants on the first (black line) and last (red line) days of treatment. Peak assignments of major LIBS signals are shown in (c) and (h). Na, K and Fe peaks which are used for detection of plant stress are highlighted. The shown spectra are averages of twenty LIBS spectra from each sample.

Download Full Size | PDF

Table 1. Peak assignments of the observed LIBS signals based on the NIST atomic spectral database [34].

View Table

Na peaks were observed in all LIBS spectra. The Na peak intensities varied depending on the presence and degree of stress. Figure 2(e) shows no significant change in the intensity of the Na peak in the control case of the non-stressed gardenia. However, the corresponding Na peak intensity significantly increased in the stressed gardenia plant [Fig. 2(g)]. Similar changes were also observed for the mildly and severely stressed wheat as shown in Figs. 2(j) and 2(l), respectively. Larger peak intensity difference was observed for the Na peak of the severely stressed compared to the mildly stressed wheat. These results indicate that the relative change of the Na peak intensity can be used as a spectroscopic marker of plant stress.

Ca peak intensities also varied as a response to stress in gardenia. No Ca peaks were detected in the LIBS spectra of the non-stressed wheat [Figs. 2(i) and 2(k), black lines]. Therefore, Ca peaks can also be used as spectral signatures of plant stress. However, similar Ca peaks were observed due to both mild [Fig. 2(i), red line] and severe [Fig. 2(k), red line] stress. Therefore, Ca cannot be used to distinguish between these different stresses. Figure 2 shows qualitative results of plant stress detection. Semi-quantitative information may be obtained from the analysis of peak ratios as described below.

2.3 Temporal response to drought stress

We investigated the temporal response of gardenia and wheat plants to drought stress by analyzing the daily evolution of LIBS signals. Although slight signal fluctuations were observed in all LIBS peaks for the non-stressed plants, the most significant changes were observed for the stressed samples. We established three different approaches for the plant stress detection using LIBS by monitoring the temporal evolution of nutrient LIBS signals, plasma temperatures and relative nutrient contents. Figure 3 shows the temporal evolution of the normalized average peak intensities for the selected Na, Ca and Fe elements observed in the LIBS spectra of gardenia and wheat. The effects of the drought stress in the LIBS spectra became more significant at the end of the treatment periods. In order to quantify the degree of these changes Fig. 3(a) shows the root-mean-square (rms) values of the difference between the average signal intensities of the stressed and not-stressed samples. This analysis shows that the 589 nm Na I peak had the greatest change over the course of the treatment. Temporal evolution plots of Na, Ca and Fe show that the major changes in nutrient LIBS signals happened during the last two and five days of the treatment periods for the gardenia and wheat, respectively. The two gardenia sets with one plant per treatment group showed similar behavior.

 

Fig. 3 Temporal evolution of nutrient LIBS signals and plasma temperature. (a) Difference in average peak intensities (rms) between watered and non-watered plants for the selected Na, Ca and Fe elements observed in the LIBS spectra of gardenia (yellow bars) and wheat (green bars) plants. (b) Difference in average plasma temperatures (rms) between watered and non-watered plants. Temporal evolution of nutrient LIBS signals and plasma temperatures for the watered (blue circles) and non-watered (orange circles) gardenia (i) – (iv) and wheat (I) – (IV) plants during the entire treatment periods. Dashed lines show quadratic curves-of-growth of specified peaks for the non-watered samples.

Download Full Size | PDF

As a second approach of plant stress detection, the plasma temperatures were calculated using Ca I signals for each day of the treatment period. Figure 3(b) shows rms values of the difference in plasma temperature between the stressed and non-stressed samples. The plasma temperatures remained constant until the last day of the treatment cycle for gardenia and randomly fluctuated during the last five days for wheat. This suggests that the plasma temperature may be used as another indicator of plant stress. However, care must be taken in more precise semi-quantitative use due to the lack of correlation between the plasma temperature and the stress-induced changes of nutrient LIBS signals. To obtain semi-quantitative information about the changes in relative nutrient contents we analyzed the LIBS peak ratios.

Temporal evolution of relative nutrient contents were obtained by taking the ratios of the nutrient LIBS signals and plotted over the entire treatment periods [Fig. 4]. Significant changes in the relative nutrient contents are clearly seen during the last five days of the treatment periods for both gardenia and wheat plants in Figs. 4(i) – 4(iv) and Figs. 4(I) – 4(IV), respectively. By comparison, the absolute values of the nutrient LIBS signals showed significant changes only for the last two days for gardenia. This demonstrates the advantage of using peak ratios for more precise detection of plant stress. Two Ca peaks were chosen as a control to insure the accurate measurement of the relative contents [Figs. 4(iv) and 4(IV)]. Rms values of the difference in peak ratios between the watered and non-watered plants for the selected Na, Ca and Fe peaks are shown for gardenia (yellow bars) and wheat (green bars) in Fig. 4(a). The peak ratios which involve Na show a large difference between the watered and non-watered samples. Conversely, the ratio between two Ca peaks shows a negligible difference. Additionally, the ratio between Ca and Fe peaks is small for wheat but larger for gardenia. These results imply that indeed the Na content is changing significantly more than that of other elements due to the drought stress experienced by the plants. The relative contents of Na/Fe and Na/Ca are larger and smaller, respectively, in gardenia than in wheat. This behavior indicates different possible mechanisms of stress response in these two different plant species.

 

Fig. 4 Temporal evolution of relative nutrient contents. (a) Difference in peak ratios (rms) between watered and non-watered plants for the selected Na, Ca and Fe elements observed in the LIBS spectra of gardenia (yellow bars) and wheat (green bars) plants. Temporal evolution of relative nutrient contents for the watered (blue circles) and non-watered (orange circles) gardenia (i) – (iv) and wheat (I) – (IV) plants during the entire treatment periods.

Download Full Size | PDF

3. Discussion

Stress within a plant is any unfavorable condition or substance which negatively affects or inhibits the plant’s metabolism and may be short-term or long-term [36]. Plants react differently to stress events depending upon the stress response phase. Certain stress effects can be partially offset for by acclamation, adaptation and repair mechanisms for minor or acute stress. Chronic or strong stress can cause irreversible damage or may lead to death of the plant [36]. On a small scale, visual inspection of crops may be performed to determine the overall health of field crops. However, for large areas of vegetation, visual inspection is time consuming and sometimes impractical [10,12,37]. The LIBS approach has many advantages for rapid detection of plant stress that allow for this technique to be expanded to the field scale. LIBS can be performed on whole plants without any sample preparation. Additionally, LIBS requires few optical components. Because of these key features, LIBS has the potential to be performed remotely via drone/UAV, allowing for rapid monitoring of vast crop terrain.

Due to the relative simplicity of the method and the ability to monitor elemental contents in plants on the laboratory and field scales, LIBS has the potential to offer insights into the complex mechanisms of plant tolerance and stess response such as the relative K and Na level dynamics, Ca accumulation and transport, and many others. Optimal K⁺/Na⁺ ratio is crucial for plant metabolism and plays an important role in the osmotic adjustment and stress tolerance under drought and salinity [8,39]. For example, a significant increase of K⁺ and Na⁺ was reported under drought [39–41]. High tolerance to the combined drought and salinity stress was previously related to the lower Na⁺/K⁺ ratio [42]. Our results indicate an increase in the relative content Na/K ratio in wheat. Drought and salinity can also affect Ca2⁺ content [4,41,43]. Further, water deprivation has been shown to damage the mechanism controlling Fe uptake which leads to increased Fe concentration in the chloroplasts [44]. LIBS may provide a better understanding of the role of these ions in the stress response. Although the LIBS technique is relatively simple, the interpretation of the data can be complex due to the many factors that effect the sample matrix and plasma conditions. Therefore, the observed LIBS signals are due to the contributions of the changing elemental contents and to the changes in the sample matrix brought about by the induced drought stress. Both factors may be considered as signatures of the drought stress and may be used for the rapid detection of stress in plants.

Our experimental implementation of LIBS can be further improved by optimizing several parameters such as the lens-to-sample (LTS) distance which plays a minor role for long focal length lenses due to the gradual nature of the focus and increase in the focal volume [13]. Therefore, the LTS can be optimized for stand-off field applications. The detector sensitivity and spectral efficiency are additional control parameters which determine the effectiveness of LIBS in monitoring plant stress. Higher resolution detectors are useful for deconvoluting overlapping atomic species in the LIBS spectra and may provide more precise measurements. Moreover, high resolution detectors allow for monitoring of a larger number of atomic species. LIBS signal strength can be increased using detectors with greater sensitivity and better collection efficiency over large distances or in the presence of background. The benefits of the improved resolution and sensitivity must be balanced with the augmented size and cost of the detector in order to maintain feasibility of field applications. Here we used a lightweight inexpensive spectrometer (OceanOptics HR2000) which is suitable for the use on a UAV. For comparison, we also performed LIBS measurements using a high resolution spectrometer (IsoPlane, Princeton Instruments) and obtained similar results (not shown). In addition to compact spectrometers, compact femtosecond lasers are also available which are suitable for UAV use; however, they are limited in power output. More powerful lasers can be used to increase the signal strength of LIBS or to generate femtosecond plasma filaments which can be used for filament-induced breakdown spectroscopy (FIBS) [45]. FIBS may provide similar information as LIBS with the advantage of longer standoff detection range. Field-scale filament-based environmental analysis was previously performed on a vehicle platform (TERAMOBIL) [46]. It is envisioned that similar technology may be developed for agricultural applications.

In addition to the monitoring of drought stress, the LIBS method can be expanded to include elemental signals of other types of stress and to include other economically meaningful plant species. LIBS may be performed on a large scale in different environments and may be used for comparative field studies between different countries [47]. LIBS may be used to determine spectroscopic signatures of a plethora of stressors for different crops, thereby allowing for greater understanding of the mechanisms of, and improved response to, stress factors found in plant life.

4. Methods

4.1 Sample preparation

Four gardenia plants were purchased from a local nursery and kept indoors at approximately 72° F (day and night) with an irradiance of approximately 9 W m⁻² from conventional fluorescent tubes. Out of the four gardenia plants, two were randomly chosen to be stressed (received no water) and the other two not to be stressed (received ~10 ounces of tap water daily). The plant mass was recorded daily using a digital scale (Cen-Tech) as shown in Fig. 1(b). Leaves from the end of the branch were selected to ensure approximately equal maturity and treatment levels. Twenty LIBS spectra were taken on both sides of the central vein in the leaf. Each spectrum was normalized to the LIBS signal at 500 nm for convenience. All twenty spectra were then averaged to obtain a single averaged spectrum for each plant/leaf on each day of the treatment period. Therefore, each individual spectrum shown in Fig. 2 is an average of twenty normalized spectra. Due to uneven distribution of the level of drought stress experienced by each leaf, the spectral intensities showed daily fluctuations. These intensities were determined from the maximum values of the normalized averaged spectra for each observed peak. The sample leaves were separated from the plant stem using scissors and were mounted onto the xyz-translation stage to insure sample flatness and uniformity of LTS distance. No additional sample preparations were made. This methodology may be extended for measuring LIBS spectra of leaves in situ by using an auto-focusing system. Fresh masses of gardenia plants on day 1 were 6595 ± 5 g and 5545 ± 5 g for the watered and non-watered plants, respectively.

The wheat was grown in 0.47 L black plastic pots (Dillen Products, Middlefield, OH). We utilized metro mix 900 for the potting media (Sun Gro Horticulture Canada Ltd, Vancouver, BC), starting on 13 Nov 2014, and fertilizing weekly with approximately 450 ppm nitrogen (Peters Professional 20N-8.74P-16.6K, Scotts Co., Marysville, Ohio). Plants were grown in a plastic greenhouse without light exclusion, with temperature set points of 85° F day and 75° F night. Ten wheat pots (with ~10 plants per pot) were randomly selected to receive water and the other 10 pots were selected not to receive water. Every pot containing wheat plants in the watered (mild stress) group received water daily to keep the soil moisture level constant. There was a mild lighting stress in the watered wheat group. Plants from the non-watered (severe drought stress) group received no water. Their masses were recorded daily using the same digital scale for all pots containing wheat plants separately and were then averaged for the group [Fig. 1(d)]. One sample leaf was selected from each pot (middle leaf of an individual wheat sprout) that represented the state of the whole pot and then all spectra from the 10 representative wheat leaves were averaged and plotted as a single averaged spectrum. Average wheat masses on day 1 were 296 ± 26 g and 274 ± 29 g for the watered and non-watered treatment groups, respectively.

4.2 Laser-induced breakdown spectroscopy (LIBS)

LIBS experiments were performed using a femtosecond laser system consisting of a mode-locked Ti:Sapphire laser oscillator (Tsunami, Spectra Physics) operating at ~800 nm center wavelength, with a pulse duration of ~35 fs FWHM and an amplifier (TSA, Spectra Physics) operating at a repetition rate of 1 kHz. The samples were placed on a xyz-translation stage to ensure that a fresh surface was ablated. Translation of the sample in the beam path took place at a rate of approximately 1 cm/s. The incident beam was attenuated to 300 mW (0.3 mJ pulse energy) by a neutral density filter and was then focused onto the sample surface by a 50 mm focal length lens. A color glass filter was used to suppress the pump beam reflection. The LIBS spectra were collected at a 45° angle and measured using an Ocean Optics HR2000 spectrometer as shown in Fig. 5. No time delay between excitation and spectral measurements was employed. All LIBS measurements were performed at ambient conditions. LTS distance was optimized daily utilizing a reference sample to insure system stability.

 

Fig. 5 Schematic of the laser-induced breakdown spectroscopy (LIBS) performed on a leaf sample (inset).

Download Full Size | PDF

4.3 Plasma temperature measurement

Assuming that the plasma is in local thermal equilibrium (LTE) and optically thin [13,48], the Boltzmann distribution method was used to determine the plasma temperature. The intensity of a given LIBS spectral line is: