World agriculture consumes approximately 70% of the fresh water withdrawn per year to irrigate only about 17% of the world’s cropland [1
]. This amount of irrigated land is slowly expanding due to the increased human food requirements and the effects of global warming [1
]. The application of right agricultural practices and supporting policy solutions is then crucial; in particular, water efficiency in crop irrigation can be largely improved by introducing more accurate systems to indicate the actual water need of the crop [3
is evolutionarily well-suited to dry climates, but prolonged water scarcity or fluctuating water soil availability severely affects berry quality [4
], reduces yield and compromises economic viability of the crop. Increasing water scarcity could lead to a more frequent use of irrigation for an affordable crop production [4
] but, generous watering can reduce the quality of the fruit, through a decrease in colour and sugar content, and can imbalance the acidity and interfere with the flavonoid development [5
]. Due to the large dependence of berry quality parameters on soil water availability, irrigation should be accurately regulated through the development of new methods of accurate irrigation scheduling based on plant “stress sensing” to achieve a more environmentally sustainable viticulture with a reasonable fruit quality [6
Coffee is a traditional and widely consumed beverage in many countries and has a high impact in the economical and social development of producer countries. The process of growing coffee plants is based on a constant and balanced supply of nutrients to each part of the tree. A poor distribution of nutrients may cause different diseases, such as chlorosis and deformation of leaves, among others, which directly affects the production. Colombia, one of the most important coffee growers in the world, has increased its production in around 10% during last year up to total production, in 2016, estimated in 4.2 millions of bags of processed coffee [9
]. However, the negative effects of “el Niño” phenomenon on the production are becoming more evident. In view of this, there exists a need for implementing new strategies and technologies in order to mitigate the effect of long periods of drought by means of a proper control of the water status of the plantations. Furthermore, the measurement of plant hydration is a very important factor which may have implications on fertilization, presence of weeds and the identification of seasons, i.e., when the environmental conditions can produce either excess or defect of water content. This governs the dynamics of flowering and fruit growth, as well as the presence of plagues and diseases [10
]. Irrigation control is pivotal as it can be used to increase planting density, to increase crop yield, and to affect fertilization management [11
Traditionally, the establishment of the amount of water needed to irrigate a crop has been solved by using climatic mean values (potential evapotranspiration or recently, crop evapotranspiration), and by monitoring either the soil water content or plant water status. The first method does not consider inter-annual variations, commonly found in semi-arid regions like the Mediterranean. In the second one, the size and development of the plant root system constitutes a limitation in the calculation of the irrigation requirement of the plant due to the spatial variations of the soil water availability. Therefore, the direct monitor of the plant water status is the only way to accurately adjust the water dose required by the crop. Methods to obtain plant water status are mainly based on the measurement of water potential or relative water content [15
]. Water potential (Ψ) describes the energy status of water in plants, it is expressed as potential energy per unit volume and its units are those of pressure, MPa or bars. The most widely used method to measure Ψ is the so called Scholander pressure-chamber technique. On the other hand, the relative water content (RWC) is the amount of water per unit weight of water at full hydration. The calculation of RWC is based on the following ratio: (fresh weight − dry weight)/(saturated weight − dry weight) [18
]. These methods are considered destructive techniques precluding repetitive measurements in a given tissue and, therefore, they are not suitable for studying dynamic water changes within the same plant tissue or organ. For this reason, and during the past decades, there has been a challenge to find non-destructive or non-invasive techniques [19
Resonant ultrasonic spectroscopy (RUS) [25
] is a well know technique to obtain the elastic constants of solid materials from the analysis of the resonant frequencies of different modes of vibration of samples having a well defined geometry and free boundary conditions [26
]. In a similar way, and for the case of plates, air-coupled ultrasound has been used to excite and sense thickness resonances with a similar purpose: to obtain elastic constants [29
]. In this sense, this technique can be considered as non-contact RUS (NC-RUS), though there are significant differences with conventional RUS (e.g., no free boundary conditions are considered in this case). NC-RUS has also been applied to excite and sense thickness resonances in plant leaves and to determine some of their properties (thickness, density, elastic modulus, mechanical damping) [35
]. Moreover, it has been demonstrated that there is a close relation between the parameters extracted from the ultrasonic resonance of the leaves and their relative water content and water potential. In particular, as leaves become dehydrated the variation in the frequency of the first thickness resonance (
) with the relative water content follows a decreasing sigmoid whose point of inflection is located at the turgor loss point [36
]. More recently NC-RUS has also been proven as a technique for the dynamic determination of leaf water status [40
]. So far, we have applied the NC-RUS technique to more than 50 plant species, where the only requirements is that the leaf surface must be larger than the acoustic beam width and that leaf surface must be relatively flat over the section defined by the acoustic beam width.
The purpose of this paper is to review the main requirements of a NC-RUS sensing system to measure thickness resonances in plant leaves, to propose two different transducer/sensor solutions for two particular cases: Vitis vinifera and Coffea arabica leaves and to test the possibilities of the proposed solution to determine leaf parameters, RWC and Ψ both in lab and field conditions. Moreover, this same solution can be used for other species whose leaves present thickness resonances in these frequency bands
2. Description of a NC-RUS System for Plant Leaves and Main Design Parameters and Specifications
shows a schematic representation of the main elements of a NC-RUS system to measure thickness resonances in plant leaves. These elements can be grouped in four categories (sensors, electronics, software and structural elements):
Sensors. A couple of wideband and high sensitivity air-coupled ultrasonic transducers (transmitter: Tx and receiver: Rx).
Electronics. A pulser/receiver to excite Tx and to filter, amplify and digitize the electrical signal in Rx. If an analog pulser/receiver is used, then a digital oscilloscope or a similar device is required to digitize the received signal.
PC and software. Including: (i) software to control the electronics and display the results, includes a graphical user interface (GUI) and (ii) the software to solve the inverse problem and extract leaf parameters from the measured resonance.
Structural elements. Including: (i) a system to hold sensors in the right position, (ii) a sample holder that allows the right location of the leaf in-between the ultrasonic sensors and (iii) any system to isolate the measurements from the influence of environmental conditions.
The main design parameters of an NC-RUS system to measure thickness resonances in plant leaves, the elements affected and the specifications to be met are summarized in Table 1
2.1. Size of the Measurement Area and Geometry of the Ultrasonic Field (Beam)
Size of the leaf area where measurements are performed coincides with the ultrasonic beam section (see Figure 1
) and the size of the beam section is slightly smaller than the size of the transducers aperture (depending on the transducer-leaf distance). As the beam section must be completely included within the leaf, this imposes an upper limit for transducers size. In addition, as obtained leaf properties are averaged values over the measurement area, it is then convenient to take the largest section possible while avoiding any major inhomogeneity like large veins or largely curved parts. As an example, Figure 2
shows acceptable beam size and location point of the measurement area for a few examples.
It must be also be considered that the analysis of the spectra of the thickness resonances is performed assuming plane wave and normal incidence. Therefore, to achieve a wavefront of the incident acoustic beam on the leaf surface as plane as possible, the transducer surface must present a piston like vibration.
2.2. Centre Frequency and Frequency Bandwidth
The main requirement for frequency band of the NC-RUS system for the study of the leaves of a given species is that this band must include the whole spectra of the first thickness resonance for all leaves of this species. In general, the leaf resonance spectra (magnitude and phase) are well defined by taking a frequency band or window defined by: Magnitude spectrum peak value—6 dB. In addition, as the value of not only varies from leaf to leaf, but is also variable for a given leaf (depends on the degree of development, the water content, etc.), then, the frequency band of the NC-RUS system must be large enough to include all these variations.
With the purpose of illustrating the typical requirements, Figure 3
presents some spectra of the first thickness resonance of some leaves of different species that are rather representative of the different situations found. Measurements and theoretical calculations are obtained following the procedure explained in [33
is normally located within the frequency range 0.1–1.0 MHz, where the lower values normally correspond to soft leaves of herbaceous species like Arabidopsis thaliana
or Lactuca sativa
. The 6dB relative bandwidth of the resonances observed in Figure 3
is about 70% for the cases where the resonance peak is strongly attenuated (like in Ficus carina
and Nicotiana tabacum
), and between 25% and 35% for those cases where the resonance peak is less attenuated (Coffea arabica
, Vitis vinifera
and Citrus reticulata
As an example of the typical variability of
from leaf to leaf (with all leaves under similar conditions) measurements in 30 leaves of Viburnum tinus
and Arabidopsis thaliana
were performed. The obtained relative variation in
was 6% (490 ± 30 kHz) and 8% (157 ± 13 kHz), respectively. This range of variation can be considered representative of the behaviour of most of the species. To illustrate the magnitude of the variation in
with the degree of leaf development, leaves of three different Vitis vinifera
cultivars planted in 10l pots at CSIC-Madrid were measured in the period May–September. Results are shown in Figure 4
a. This variation is due to the change in both the leaves thickness and in the cell wall elastic modulus. Finally, as an example of how RWC affects
, Figure 4
b shows some result obtained for Viburnum tinus
leaves: when RCW decreases from 1.0 to 0.7, relative variation in
is 33%. Similar results were found for Coffea arabica
(relative variation in
of 26%) and Vitis vinifera
(relative variation in
of 20%). All these variations must be taken into account in the design of the transducers for NC-RUS for a given species.
2.3. Dynamic Range and SNR
Though transmission loss at resonance (see Figure 3
) is typically between 35 and 45 dB, smaller figures are obtained for herbaceous leaves (between 25 and 32 dB). In most cases, the spectrum of the first thickness resonance is well defined by taking a frequency band around the thickness resonance given by 6 dB loss respect to the peak value. This means that the minimum value of the modulus of the transmission coefficient to be measured is >−60 dB.
2.4. Separation between the Sensors and the Leaf
Separation between transducers and leaf (∆L) must be large enough so that the through transmitted signal does not overlap with the reverberations in the transducer/leaf air-cavity. Therefore the time for the ultrasonic signal to cross twice the distance between transducer and the leaf (∆t) must be larger than the duration of the through transmitted pulse (δt). δt depends on the centre frequency and bandwidth of both the transducers and the leaf thickness resonance. Typically, δt < 40 μs, then: ∆L > 14 mm. On the other hand, separation between transducers and leaf must be kept as short as possible to minimize the attenuation in the air, and any possible interference in the air path. To minimize the size of the beam section on the leaf surface, the leaf can be located at the natural focal length of the transducers which is located at a2/λ, where a is the radius of the transducer aperture and λ is the wavelength of the radiated beam.
2.5. Time of Measurement
The time to take one measurement must be small enough to allow for fast and in situ measurements. Given that the separation between Tx and Rx is normally smaller than 60 mm, the time to take one measurements is smaller than 180 μs. If several signals are to be acquired to take an average and improve SNR, then this time will be increased. Time to take this averaged measurement will then be mainly determined by the pulse repetition frequency (PRF) of the pulser/receiver and the number of samples to average. PRF values between 100 and 1000 Hz and averaging between 10 and 100 samples are normally used, this implies that the elapsed time will be between 1 and 0.1 s, respectively. However, the most time consuming stage will be the processing of the signal and the solution of the inverse problem to extract leaf parameters. Time of execution of the inverse problem code can be reduced by reducing the length of the digitized resonance spectra. For lengths below 100 points and inverse solution codes written using, relatively low speed, interpreted languages (like Matlab or Python) it is possible to obtain execution times below 10 s, which is quite acceptable for this application.
2.6. Portability and Robustness
Portability requirements for lab measurements are reduced; however, this is not the case for field applications. In these cases, the PC must be a laptop or a tablet, the electronics must be powered by batteries and the sensors must be embedded on a portable holder. The most demanding robustness requirements also correspond to field measurements as the influence of the environmental factors on the measurement must be reduced. In particular, an easy way to locate the leaf between transducers and some protection against possible strong winds must be provided. In addition, resistance of sensors to air moisture and temperature must be also considered.
Main specifications for an NC-RUS system, including sensors, electronics, software and structural elements, to measure plant leaves have been reviewed. In particular, requirements for air-coupled ultrasonic transducers for NC-RUS measurements in Coffea arabica
and Vitis vinifera
leaves have been determined. Following these design criteria, transducers were produced using 1–3 connectivity piezocomposites and matching to the air as described in [41
]. For Coffea arabica
leaves, centre frequency of the transducers is located at 350 kHz with a peak sensitivity of −26 dB and the operation bandwidth covers the frequency range 200–450 kHz, which corresponds to the frequency band 20 dB below the main peak. For Vitis vinifera
leaves, the centre frequency is located at 650 kHz with a peak sensitivity value of −29 dB. The useful bandwidth covers the frequency range 350–900 kHz. Using these transducers and commercially available and general purpose electronic equipment to drive the transmitter transducer (400 V amplitude semicycle of square wave) and to amplify the received signal (+40 dB) it has been possible to measure Coffea arabica
and Vitis vinifera
leaves under different conditions and to establish a relationship between
, RWC and Ψ which confirm the possibility to use this technique to obtain accurate information of the crop irrigation needs. For Vitis vinifera
decreases from 580 kHz to 460 kHz when Ψ varies from 0 bar (RWC = 1) to −25 bar (RWC = 0.78). In addition, in vivo measurements on trees subjected to water stress (20 days drought) revealed that variations in the predawn and midday
values were consistent with the variations observed in Ψ. Predawn
varies from 550 kHz (day 0) to 470 kHz (day 20), while the predawn Ψ varies from 0 bar (day 0) to −12 bar (day 20). In detached Coffea arabica
is about 350 kHz (RWC = 1) and decreases to 225 kHz at RWC = 0.7. The unique ability of the proposed NC-RUS system to register changes of the plant water status under conditions of free leaf transpiration constitutes a tool of paramount importance in order to maximize water use efficiency in crop plants. Applying this ultrasonic system in agriculture, water consumption could decrease by adjusting the irrigation doses to the plant water necessity. The adjustment of irrigation doses by NC-RUS on V. vinifera
and on C. arabica
could avoid both scarce watering, that could decrease production, and generous watering, that could reduce quality.