Plants Emit Informative Airborne Sounds Under Stress: Authors

bioRxiv preprint doi: https://doi.org/10.1101/507590. this version posted December 2, 2019.
The copyright holder for this preprint (which was

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
Plants emit informative airborne sounds under stress
Authors:
5 I. Khait1,§, O. Lewin-Epstein1,§, R. Sharon3,1,§, K., Saban1, R. Perelman1, A. Boonman2, Y. Yovel2,†,
L. Hadany1,†,*
Affiliations:
10 1 School of Plant Sciences and Food Security, Tel-Aviv University, Tel-Aviv, Israel
2 School of Zoology, Tel-Aviv University, Tel-Aviv, Israel
3 School of Mathematical Sciences, Tel-Aviv University, Tel-Aviv, Israel
† Equal contribution
§ Equal contribution
15
*Correspondence: Email: [email protected]; Tel: +972-36409831
20
Keywords:
Plant bioacoustics, phytoacoustics, sound, airborne, signaling, communication, drought stress,
machine learning
25
1
bioRxiv preprint doi: https://doi.org/10.1101/507590. this version posted December 2, 2019. The copyright holder for this preprint (which was
Abstract
Stressed plants show altered phenotypes, including changes in color, smell, and shape. Yet, the
30 possibility that plants emit airborne sounds when stressed – similarly to many animals – has not
been investigated. Here we show, to our knowledge for the first time, that stressed plants emit
airborne sounds that can be recorded remotely, both in acoustic chambers and in greenhouses.
We recorded ~65 dBSPL ultrasonic sounds 10 cm from tomato and tobacco plants, implying that
these sounds could be detected by some organisms from up to several meters away. We
35 developed machine learning models that were capable of distinguishing between plant sounds
and general noises, and identifying the condition of the plants – dry, cut, or intact – based
solely on the emitted sounds. Our results suggest that animals, humans, and possibly even
other plants, could use sounds emitted by a plant to gain information about the plant's
condition. More investigation on plant bioacoustics in general and on sound emission in plants
40 in particular may open new avenues for understanding plants and their interactions with the
environment, and it may also have a significant impact on agriculture.
45
2
Introduction
Plants exhibit significant changes in their phenotypes in response to stress. They differ visually,
with respect to both color and shape, from unstressed plants [1-4]. They also emit volatile
organic compounds (VOCs), e.g. when exposed to drought or herbivores [5, 6]. VOCs can also
50 affect neighboring plants, resulting in increased resistance in these plants [7, 8]. Altogether,
plants have been demonstrated to produce visual, chemical and tactile cues, which other
organisms can sometimes respond to [9-12]. Nevertheless, the ability of plants to emit airborne
sounds – that could potentially be heard by other organisms – has not been sufficiently
explored [11, 13, 14].
55
Plants exposed to drought stress have been shown to experience cavitation – a process where
air bubbles form, expand and explode in the xylem, causing vibrations [15, 16]. Yet, these
vibrations have always been recorded by connecting the recording device directly to the plant
[16, 17]. Such contact-based recording does not reveal the extent to which these vibrations
60 could be sensed at a distance from the plant, if at all [17-19]. Thus, the question of airborne
sound emission by plants remains unanswered [17, 20, 21].
Many animals, including herbivores and their predators, respond to sound [22-24]. Recently,
plants were also demonstrated to respond to sounds [13, 25-27], e.g., by changing gene
65 expression of specific genes [26, 27], or by increasing sugar concentration in their nectar [28]. If
plants are capable of emitting informative airborne sounds, these sounds can potentially trigger
a rapid effect on nearby organisms, including both animals and plants. Even if the emission of
the sounds is entirely involuntarily, and is merely a result of the plant’s physiological condition,
nearby organisms that are capable of hearing the sounds could use them for their own benefit.
70
We therefore set to examine whether plants emit informative airborne sounds, which may
serve as potential signals or cues to their environment. Here we show for the first time that
plants indeed emit airborne sounds, which can be detected from several meters away.
Moreover, we show that the emitted sounds carry information about the physiological state of
3
75 the plant. By training machine learning models we were able to achieve high accuracy in
distinguishing between drought-stressed, cut, and control plants, based only on the sounds
they emit. These results demonstrate the potential in studying phyto-acoustics and suggest that
this form of communication may play an important role in plant ecology and evolution, and
may have direct implications for plant monitoring in agriculture.
80
Results
To investigate plants’ airborne sound emissions, we first constructed a reliable recording
85 system, where the plants were recorded within an acoustic box, and then tested the system in a
greenhouse.
Plants emit airborne sounds when stressed

To investigate plants’ ability to emit airborne sound emissions, we first constructed a reliable
90 recording system, in which each plant was recorded simultaneously with two microphones (see
Fig. 1a for illustration, and Methods for details), within an acoustically isolated anechoic box.
We recorded tomato (Solanum lycopersicum) and tobacco (Nicotiana tabacum) plants under
different treatments – drought stress, cutting (of the stem), and controls. We focused on the
ultrasonic sound range (20-150 kHz), where the background noise is weaker.
95
We found that plants emit sounds, and that both drought-stressed plants (see Methods) and
cut plants emit significantly more sounds than plants of any of the control groups (p<e-6,
Wilcoxon test for each of the 12 comparisons with Holm-Bonferroni correction). Three controls
were used for each plant species and treatment: recording from the same plant before
100 treatment (self-control), recording from an untreated same-species neighbor plant (neighbor-
control, see Methods), and recording an empty pot without a plant (Pot). The mean number of
sounds emitted by drought-stressed plants was 35.4±6.1 and 11.0±1.4 per hour for tomato and
tobacco, respectively, and Cut tomato and tobacco plants emitted 25.2±3.2 and 15.2±2.6
4
sounds per hour, respectively (Fig. 1b). In contrast, the mean number of sounds emitted by
105 plants from all the control groups was lower than 1 per hour. Our system did not record any
sound in the Pot control (Fig. 1b).
Figure 1. Plants emit remotely-detectable ultrasounds under stress. (a) Acoustic box setup. In each recording,
110 three plants are placed inside an acoustic box with two directional microphones oriented at each plant. Using two
microphones helps eliminating false detections resulting from electrical noise clicks of the recording system and
cross-plant interference. (b) Mean number of sounds emitted during 60 minutes of recording by tomato and
tobacco plants under two treatments, drought stress and cutting. Three control groups were used – empty pots,
and two groups of untreated plants: self-control – the same plant before treatment, and neighbor-control –
115 untreated plants that shared the acoustic box with treated plants. All treatment groups emitted significantly more
sounds (p<e-6, Wilcoxon test with Bonferroni correction) than all control groups: self-control (Meanself−control < 1
for all plant-treatment combinations) and neighbor control (Meanneighbor−control < 1 for all plant-treatment
combinations). The system did not record any sound from pots without plants during the experiments
(Meanpots = 0). 20 ≤ 𝑛 ≤ 30 plants for each group. (c) Examples of time signals of sounds emitted by: a drought-
5
120 stressed tomato, a drought-stressed tobacco, a cut tomato, and a cut tobacco. (d) The spectra of the sounds from
(c).
How does a stressed plant sound? Figs. 1c-d show examples of raw recorded time signals and
125 their spectra as recorded from drought-stressed and cut plants. The mean peak sound intensity
recorded from drought-stressed tomato plants was 61.6±0.1 dBSPL at 10 cm, with a mean peak
frequency of 49.6±0.4 kHz (frequency with maximal energy), and the mean intensity recorded
from drought-stressed tobacco sounds was 65.6±0.4 dBSPL at 10 cm, with a mean frequency of
54.8±1.1 kHz. The mean peak intensity of the sounds emitted by cut tomato plants was
130 65.6±0.2 dBSPL at 10 cm with a mean peak frequency of 57.3±0.7 kHz (frequency with maximal
energy), and the mean intensity of the sounds emitted by cut tobacco plants was 63.3±0.2
dBSPL at 10.0 cm distance with a mean frequency of 57.8±0.7 kHz. The distributions of sound
peak intensity and the maximum energy frequency of cut and drought-stressed plants are
shown at Fig. 1c. Spectrograms of raw recorded sounds from cut and drought-stressed plants
135 are shown at Supporting Information Fig. S1.
The sounds of plants reveal their condition

To test whether we could identify different plant conditions and species based on their sound
140 emissions, we trained a regularized machine learning classifier. We divided the sounds to four
groups, with all the combinations of two plant types – tomato and tobacco, and two treatments
– drought or cutting. The treatments were applied to the plants before the beginning of the
recording. A binary classifier was trained to distinguish between two equal-size groups (“pair”)
in each comparison (Tomato-Dry vs Tomato-Cut; Tobacco-Dry vs Tobacco-Cut; Tomato-Dry vs
145 Tobacco-Dry; Tomato-Cut vs Tobacco-Cut). For cross validation, the model was tested only on
plants that were not a part of the training process (see Methods for more details). We used a
Support Vector Machine (SVM) as the classifier with several methods of feature extraction –
basic [29, 30], MFCC [31], and a scattering network [23] .
6
150 The SVM classifier with scattering network for feature extraction achieved the best results:
~70% accuracy for each of the four pairs (Fig. 2b red line), significantly better than random
(p<e-12 for each pair, Wilcoxon rank sum test with Holm-Bonferroni correction). The same
classifier was trained to discriminate between the electrical noise of the system (see Methods)
and the sounds emitted by the plants (Tomato vs Noise, Tobacco vs Noise) and achieved above
155 98% accuracy for both (Fig. 2b). The results were also robust to the dimension of the
descriptors and the scattering network specific parameters (Fig. S2). Using the SVM classifier
with MFCC as the input features, the results were still significantly better than random (p<e-4
for each pair, corrected as above), and even when we only used 4 basic features the results
were significantly better than random for 5 of the 6 pairs (p<e-6 for each of them, adjusted; Fig.
160 2b). However, Scattering network performed better than either MFCC (p<0.05) or Basic (p<e-6)
for all the pairs, using Wilcoxon sign rank test.
Figure 2. Plant condition and species can be detected from a distance by listening to its sound emissions. (a) The
165 recorded sounds intensity peak and the max energy frequency for the four groups – drought stressed tomato
plants, cut tomato plants, drought stressed tobacco plants and cut tobacco plants. (b) The accuracy of sound
classification achieved by different feature extraction methods, with an SVM classifier. The best results were
obtained using the scattering network method for feature extraction (red line, p < e-12 for each pair). Using MFCC
for feature extraction the results were also highly significant (black dashed line, p<e-4 for each pair) and even basic
170 methods for feature extraction allowed for better-than-random classification (gray line, p<e-6 for each pair apart
from one case: Tobacco dry vs. Tobacco cut, which was bot significant with the basic method). The comparisons
Tomato vs Elect. Noise and Tobacco vs Elect. Noise related to electrical noise of the system. Training set size of the
7
two groups in each pair was equal (400 < sounds for each pair, see Table S2), and significance levels for each pair
were calculated using Wilcoxon rank sum test with Holm–Bonferroni correction for multiple comparisons.
175
Plant stress can be identified from its sounds in a greenhouse

Finally, we tested the acoustic behavior of plants in a more natural setting, a greenhouse. A
major challenge we had to address was distinguishing between sounds generated by plants and
180 the general noises in a greenhouse that were absent in the acoustic box (e.g. wind, rain, air-
conditioning, construction). We first constructed a greenhouse noises library, by recording in
the empty greenhouse, without any plant inside. We then trained a convolution neural network
(CNN) model to distinguish between these greenhouse noises and the sounds of dry tomatoes
recorded in the acoustic box (see details in Methods section). The model obtained from this
185 training achieved 99.7% balanced accuracy in a cross-validation examination (see Figure 3 and
figure S3).
After we succeeded filtering greenhouse noises, we could use ”clean” plant sounds to classify
plant condition in the greenhouse. Results from the acoustic box recordings showed that
190 drought-stressed plants emitted significantly more sounds than control plants (Fig. 1). We thus
used the number of tomato sounds emitted within an hour of recording to distinguish between
drought-stressed plants and control plants, in the greenhouse. Each plant was recorded either
one day after watering (control) or five days after watering (drought-stressed). This
classification method, which only counted plant sounds, achieved ~84% accuracy, significantly
195 distinguishing between drought-stressed and control tomato plants (p<e-5, Fisher exact test;
see further details in figures S4 and S5).
We continued to study the acoustic manifestation of the dehydration process, by recording

tomato plants for ten consecutive days. Each tomato plant was watered, and then placed in the
200 greenhouse for ten days of recording without watering. We used the CNN model mentioned
above (figure 3a) to separate tomato sounds from greenhouse noises. We then counted the
tomato-classified sounds recorded on each day from each tomato. These results revealed a
8
consistent acoustic pattern: the plants emit very few sounds when irrigated, the number of
sounds per day increases in the following 4-6 days, and then decreases as the plant dries (see
205 figure 3e).
9
10
Figure 3. Acoustic detection of plant condition in the greenhouse. (a) Illustration of the procedure used to train a
classifier that distinguishes between tomato sounds and greenhouse noises. A greenhouse noises library was first
210 generated, by recording inside an empty greenhouse for several days. Using this library and the library of tomato
sounds recorded in an acoustic box, we trained a convolution neural network (CNN) classifier to distinguish
between tomato sounds and greenhouse noises. (b) Illustration of the recordings in the greenhouse. Tomato
plants were recorded in the greenhouse. The recorded sounds were filtered using the trained CNN classifier,
leaving only the tomato-classified sounds. (c) Confusion matrix showing the success of the trained CNN classifier in
215 distinguishing between tomato sounds and greenhouse noises. Balanced accuracy score of ~99.7%. (d) Confusion
matrix showing the success in distinguishing between dry tomato plants and irrigated tomato plants, based on one
hour of recording inside a greenhouse. The condition of the plant (dry / irrigated) was here decided based on the
number of recorded tomato-classified sounds: if above three the plant was classified as “dry”, and otherwise as
“irrigated”. Balanced accuracy score of ~84% (p<e-5; Fisher exact test). (e) The number of tomato-classified
220 sounds per day during dehydration. Tomato plants (𝑁 = 21) were recorded in the greenhouse for ten consecutive
days without watering (starting one day after watering). The recorded sounds were then filtered using the trained
CNN classifier, leaving only the tomato-classified sounds. We find significant difference in the amount of sounds
between the following consecutive days: 1-2 and 2-3 (Wilcoxon signed-ranks tests, corrected for 9 comparisons
between pairs of consecutive days using Holm-Bonferroni method; The ‘*’ markings in the figure represent
225 adjusted p-values < 0.05).
Discussion
Our results demonstrate for the first time that plants emit remotely-detectable and informative
230 airborne sounds under stress (Fig. 1). The plant emissions that we report, in the ultrasonic
range of ~20-100 kHz, could be detected from a distance of 3-5m (see Methods), by many
mammals and insects (when taking their hearing sensitivity into account, e.g., mice [32] and
moths [24]). Moreover, we succeeded in differentiating between sounds emitted under two
different stress conditions – dry and cut – with accuracy of ~70% using supervised machine
235 learning methods (Fig. 2). Finally, we were able to filter plant sounds effectively from
greenhouse noises (accuracy of ~99.7%, Fig. 3b, c) and distinguish drought-stressed and control
plants in a greenhouse, based only on the sounds they emit (accuracy of ~84%, Fig. 3d, e).
These findings can alter the way we think about the Plant Kingdom, which has been considered
to be almost silent until now [20].
11
240
Our work can be extended in several ways. First, our results can be generalized to other species
of plants from different families. In a preliminary study, we successfully recorded sounds from
additional plants from different taxa, e.g., Mammillaria spinosissima cacti and Henbit
deadnettle (Fig. S6). We thus expect that many plants have the ability to emit sounds, but the
245 exact characteristics of these sounds, and the similarity between groups, are yet to be
identified. Third, future studies could explore the sounds emitted under different plant states,
including other stress conditions such as disease, cold, herbivores attack, or extreme UV
radiation, and other life stages, such as flowering and fruit bearing. Once a large library of plant
sounds is constructed, it could be analyzed by modern tools to obtain additional insights.
250
A possible mechanism that could be generating the sounds we record is cavitation – the process
whereby air bubbles form and explode in the xylem [15, 16]. Cavitation explosions have been
shown to produce vibrations similar to the ones we recorded [15, 16], but it has never been
tested whether these sounds are transmitted through air at intensities that can be sensed by
255 other organisms. Regardless of the specific mechanism generating them, the sounds we record
carry information, and can be heard by many organisms. If these sounds serve for
communication a plant could benefit from, natural selection could have favored traits that
would increase their transmission.
260 We have shown that plant sounds can be effectively classified by machine learning algorithms.
We thus suggest that other organisms may have evolved to classify these sounds as well, and
respond to them. For instance, many moths – some of them using tomato and tobacco as hosts
for their larvae [33, 34] – can hear and react to ultrasound in the frequencies and intensities
that we recorded [22-24]. These moths may potentially benefit from avoiding laying their eggs
265 on a plant that had emitted stress sounds. We hypothesize that even some predators may use
the information about the plant’s state to their benefit. For example, if plants emit sounds in
response to a caterpillar attack, predators such as bats [35] could use these sounds to detect
attacked plants and prey on the herbivores [36], thus assisting the plant. The same sounds may
12
also be perceived by nearby plants. Plants were already shown to react to sounds [13, 25-28]
270 and specifically to increase their drought tolerance in response to sounds [37, 38]. We
speculate that plants could potentially hear their drought stressed or injured neighbors and
react accordingly.
Finally, plant sound emissions could offer a novel way for monitoring crops water state – a
275 question of crucial importance in agriculture [39]. More precise irrigation can save up to 50% of
the water expenditure and increase the yield, with dramatic economic implications [39, 40]. In
times when more and more areas are exposed to drought due to climate change [41], while
human population and consumption keep increasing [42], efficient water use becomes even
more critical, for both food security and ecology. Our results, demonstrating the ability to
280 distinguish between drought-stressed and control plants based on plant sounds, open a new
direction in the field of precision agriculture.
We demonstrated for the first time that stressed plants emit remotely detectable sounds,
similarly to many animals, using ultrasound clicks not audible to human ears. We also found
285 that the sounds contain information, and can reveal plant state. The results suggest a new
modality of signaling for plants and imply that other organisms could have evolved to hear,
classify and respond to these sounds. We suggest that more investigation in the plant
bioacoustics field, and particularly in the ability of plants to emit and react to sounds under
different conditions and environments, may reveal a new pathway of signaling, parallel to
290 VOCs, between plants and their environment.
Materials and Methods

295 Plants materials and growth conditions
Tomato – Solanum lycopersicum ‘Hawaii 7981’ [43] – and tobacco – Nicotiana tabacum
‘Samsun NN’ – were used in all the experiments. All the plants were grown in a growth room at
13
25 °C and kept in long-day conditions (16 h day, 8 h night). The plants were tested in the
experiments 5-7 weeks after germination.
300
Recording protocol
In the acoustic box:
The recordings were performed in a 50 × 100 × 150 𝑐𝑚3 acoustically isolated box tiled with
acoustic foam on all sides to minimize echoes. Two cable holes, 2 cm radius each, were located
305 in two corners of the box and covered with PVC and acoustic foam. Inside the acoustic box
were only the recorded plants, 6 microphones connected to an UltraSoundGate 1216H A/D
converter (Avisoft). The PC and all the electricity connections were in the room outside the
acoustic box. Two USB cables connected the PC to the 1216H device inside the box, through the
holes. There was no light inside the acoustic box.
310 The recordings were performed using a condenser CM16 ultrasound microphone (Avisoft),
digitized using an UltraSoundGate 1216H A/D converter (Avisoft), and stored onto a PC. The
sampling rate was 500 KHz per channel. We filtered the recordings above 20 KHz. A recording
started only when triggered with a sound which exceeded 2% of the maximum dynamic range
of the microphone. Two microphones were directed at each plant stem, from a distance of 10
315 cm. Only sounds that were recorded by both microphones were considered as a “plant sound”
in the analysis afterwards. The frequency responses of the microphones can be found in Avisoft
website.
In the greenhouse:
320 The recordings were performed in a greenhouse in Tel-Aviv University. Inside the greenhouse
were only the recorded plants. The recordings were performed using the same hardware and
setting, as mentioned in the acoustic box recording section, apart from the acoustic box itself.
325 Data pre-processing

Data processing was performed off-line using matlab code we developed (MATLAB 8.3, The
MathWork Inc.), with the following steps: 1. Identifying the microphone that had recorded the
14
highest intensity peak at the moment the recording started. 2. Selecting the sounds that were
detected by two microphones oriented at the same plant at the same time, and saving them for
330 further analysis. The recordings produce 6 channel wav files. A processed recording includes a
short section of 1ms before and 1ms after the peak of the recorded sound that triggered the
system to record. “Greenhouse noise” sounds were obtained when the Greenhouse included
only acoustic equipment without plants or pots, by the two microphones (thus not including
“Electrical noise”).
335
Acoustic-box Drought stress experiment

Each plant was recorded twice: first before drought treatment (“self-control”), and again after
it. In the first recording, all the plants were healthy and their soil was moist. Then, for 4-6 days,
340 half of the plants were watered while the other half were not, until the soil moisture in the pots
of un-watered plants decreased below 5%. Then, the plants were recorded again at the same
order. In each recording session three plants were recorded simultaneously for one hour and
each triplet of plants included at least one watered and one un-watered plant to allow
“neighbors-control” – watered plants that were recorded while sharing the acoustic box with
345 un-watered plants. Soil moisture content was recorded using a hand‐held digital soil moisture
meter ‐ Lutron PMS‐714.
Acoustic-box Cut stress experiment

350 The experiment followed the experimental design of the drought stress experiment described
above, but drought stress was replaced with cutting of the plant. Here the pot soil was kept
moist for all the plants throughout the experiment. The plants included in the treatment group
were cut with scissors close to the ground right before the recording started. The severed part
of the plant, disconnected from the roots, was recorded. We used the same controls of the
355 drought stress experiment.
15
Greenhouse experiment
In each recording session three plants were recorded simultaneously for one hour. All the
360 recorded plants were grown in a growth room, and were brought to the greenhouse only for
the recording session. Each plant was recorded either one day after irrigation (control plants) or
4-6 days after irrigation (drought-stressed plants).
365 Classifying sounds recorded in the acoustic box

Our classification method was composed of two main stages. First, we extracted various
acoustic features from the raw recorded signals (see Data Pre-processing section). Second, we
trained a model to classify plant sounds into classes based on the feature representation
obtained in the first stage. We used three methods of feature extraction: (a) Deep scattering
370 Network, as described in Andén and Mallat [44], red dotted line in Fig. 2b. This method extends
MFCC while minimizing information loss. We used the implementation by ScatNet [45], with
Morlet wavelets. The results were robust to the dimension of descriptors and the scattering
network specific parameters: number of layers used; time support of low pass filter; and Q-
Factor (Fig. S2). The values of the specific parameters used in this work are shown at Table S1.
375 (b) MFCC feature extraction (dashed black line in Fig. 3b). We used the Ellis Dan
implementation [31]. (c) Basic features. The basic features we used were energy, energy
entropy, spectral entropy, and maximum frequency (gray line in Fig. 2b) [29, 30]. We used SVM
with Radial kernel with the LIBSVM implementation as classifier. After feature extraction, we
used Z-score for normalization of the features and PCA to reduce the dimensionality of the
380 problem. We used only the training set to choose the number of features.
During the training process, we leave all the emitted sounds of one plant out for cross
validation. Then we constructed the training set such that the two compared groups would be
of the same size. We repeated the process so that each plant constructed the testing group
exactly one time. The accuracy of the classification was defined as the percentage of correct
16
385 labeling over the total size of the testing set [46, 47]. The numbers of plants in each group are
shown at the Table S3.
Classifying sounds recorded in the greenhouse

390 For classification of greenhouse noise and drought-stressed tomato sounds, we used a
convolution neural network implemented using Python 3.6 with keras package and tensorflow
backend. The network receives the vectors of the processed recordings (2ms long time signals,
see Data pre-processing section) as input. The network is composed of three 1-D convolution
layers, each followed by a maxpooling layer, and one fully-connected layer afterwards. At the
395 end, we add a layer of size 1 with sigmoid activation. The model is trained with binary
crossentropy loss function and 'adam' optimizer. This model achieved a balanced accuracy
score of ~0.997 over leave-one-plant-out cross validation (namely in each iteration we trained
a model on the sounds from all plants except one plant, and then we validated the model on
this latter plant).
400
17
Statistical analysis
For statistical analysis of the number of sound emissions for the treatment and the control
405 groups (Fig. 1b) we used the Wilcoxon rank-sum test. To compare our classifier to random
result (Fig. 2b), we used the binomial probability distribution function (PDF) and calculate the
probability to get the classifier accuracy or higher randomly for each group. To compare the
results obtained when using scattering network for feature extraction to the results obtained
when using MFCC or basic feature extraction methods (Fig. 2b), we used Wilcoxon sign rank
410 test with Holm-Bonferroni correction. To test the success in distinguishing between drought-
stressed and control plants (Fig. 3) we used Fisher's exact test.
415 Acknowledgements: We thank Daniel Chamovitz, Gal Chechik, Tuvik Beker, and Judith Berman
for comments on the paper; Guido Sessa, Doron Teper, Guy Sobol, Yura Pupov, Rotem
Shteinshleifer, Odelia Pisanty, Eilon Shani, and Meirav Leibman-Markus for helping with plants
materials; Yoel Shkolnisky, Marine Veits, Ilia Raysin, Uri Obolski, Yoav Ram, Eyal Zinger, Yael
Gurevich, Eylon Tamir, Yuval Sapir, Yaara Blogovski and Ruth Cohen-Khait for comments on the
420 way. The Titan Xp used for this research was donated by the NVIDIA Corporation.
Funding: The research has been supported in part by ISF 1568/13 (LH), and by the Manna
Center Program for Food Safety and Security fellowships (IK), Bikura 2308/16 (LH, YY), Bikura
425 2658/18 (LH, YY).
Author Contributions: LH and IK conceived the study. LH, YY and IK, designed the research. IK,
RP, KS and OLE performed the experiments. RS, OLE , IK, AB and LH analyzed the data. YY and
430 LH supervised the experiments. IK, OLE and RS contributed equally to the study. LH and YY
18
contributed equally to the study. All authors discussed the results and took part in writing the
manuscript.
435 Competing interests: Authors declare no competing interests.
Data and materials availability: The data will be deposited on Dryad upon acceptance.
440
References
1. Hsiao TC. Plant responses to water stress. Annual review of plant physiology.
1973;24(1):519-70.
445 2. Jackson RD. Remote sensing of biotic and abiotic plant stress. Annual review of
Phytopathology. 1986;24(1):265-87.
3. Mahlein A-K. Plant disease detection by imaging sensors–parallels and specific demands
for precision agriculture and plant phenotyping. Plant disease. 2016;100(2):241-51.
4. Potters G, Pasternak TP, Guisez Y, Palme KJ, Jansen MA. Stress-induced morphogenic
450 responses: growing out of trouble? Trends in plant science. 2007;12(3):98-105.
5. Holopainen JK, Gershenzon J. Multiple stress factors and the emission of plant VOCs.
Trends in plant science. 2010;15(3):176-84.
6. Paré PW, Tumlinson JH. Plant volatiles as a defense against insect herbivores. Plant
physiology. 1999;121(2):325-32.
455 7. Heil M, Karban R. Explaining evolution of plant communication by airborne signals.
Trends in Ecology & Evolution. 2010;25(3):137-44.
8. Dolch R, Tscharntke T. Defoliation of alders (Alnus glutinosa) affects herbivory by leaf
beetles on undamaged neighbours. Oecologia. 2000;125(4):504-11.
9. Karban R. Plant behaviour and communication. Ecology Letters. 2008;11(7):727-39.
19
460 10. Falik O, Mordoch Y, Quansah L, Fait A, Novoplansky A. Rumor Has It…: Relay
Communication of Stress Cues in Plants. PLoS ONE. 2011;6(11):e23625. doi:
10.1371/journal.pone.0023625.
11. Chamovitz D. What a plant knows: A field guide to the senses of your garden-and
beyond. New York: Scientific American/Farrar, Straus and Giroux; 2012.
465 12. Lev-Yadun S. Defensive (anti-herbivory) coloration in land plants: Springer; 2016.
13. Hassanien RH, HOU T-z, LI Y-f, LI B-m. Advances in effects of sound waves on plants.
Journal of Integrative Agriculture. 2014;13(2):335-48.
14. Gagliano M, Mancuso S, Robert D. Towards understanding plant bioacoustics. Trends
Plant Sci. 2012;17(6):323-5.
470 15. Tyree MT, Sperry JS. Vulnerability of xylem to cavitation and embolism. Annual Review
of Plant Biology. 1989;40(1):19-36.
16. Cochard H, Badel E, Herbette S, Delzon S, Choat B, Jansen S. Methods for measuring
plant vulnerability to cavitation: a critical review. Journal of Experimental Botany. 2013:ert193.
17. De Roo L, Vergeynst LL, De Baerdemaeker NJ, Steppe K. Acoustic emissions to measure
475 drought-induced cavitation in plants. Applied Sciences. 2016;6(3):71.
18. Bailey NW, Fowler-Finn KD, Rebar D, Rodríguez RL. Green symphonies or wind in the
willows? Testing acoustic communication in plants. Behavioral Ecology. 2013:ars228.
19. ten Cate C. Acoustic communication in plants: do the woods really sing? Behavioral
Ecology. 2013;24(4):799-800.
480 20. Gagliano M. Green symphonies: a call for studies on acoustic communication in plants.
Behavioral Ecology. 2012:ars206.
21. Jung J, Kim S-K, Kim JY, Jeong M-J, Ryu C-M. Beyond Chemical Triggers: Evidence for
Sound-Evoked Physiological Reactions in Plants. Frontiers in Plant Science. 2018;9:25.
22. Miller LA, Surlykke A. How Some Insects Detect and Avoid Being Eaten by Bats: Tactics
485 and Countertactics of Prey and Predator Evolutionarily speaking, insects have responded to
selective pressure from bats with new evasive mechanisms, and these very responses in turn
put pressure on bats to “improve” their tactics. BioScience. 2001;51(7):570-81.
20
23. Spangler HG. Moth hearing, defense, and communication. Annual Review of
Entomology. 1988;33(1):59-81.
490 24. Fullard JH, Dawson JW, Jacobs DS. Auditory encoding during the last moment of a
moth's life. J Exp Biol. 2003;206(2):281-94.
25. Mishra RC, Ghosh R, Bae H. Plant acoustics: in the search of a sound mechanism for
sound signaling in plants. Journal of Experimental Botany. 2016;67(15):4483-94.
26. Ghosh R, Mishra RC, Choi B, Kwon YS, Bae DW, Park S-C, et al. Exposure to Sound
495 Vibrations Lead to Transcriptomic, Proteomic and Hormonal Changes in Arabidopsis. Scientific
Reports. 2016;6:33370. doi: 10.1038/srep33370
https://www.nature.com/articles/srep33370#supplementary-information.
27. Jeong M-J, Shim C-K, Lee J-O, Kwon H-B, Kim Y-H, Lee S-K, et al. Plant gene responses to
frequency-specific sound signals. Molecular Breeding. 2008;21(2):217-26.
500 28. Veits M, Khait I, Obolski U, Zinger E, Boonman A, Goldshtein A, et al. Flowers respond to
pollinator sound within minutes by increasing nectar sugar concentration. Ecology Letters.
2019;0(0). doi: 10.1111/ele.13331.
29. Acevedo MA, Corrada-Bravo CJ, Corrada-Bravo H, Villanueva-Rivera LJ, Aide TM.
Automated classification of bird and amphibian calls using machine learning: A comparison of
505 methods. Ecological Informatics. 2009;4(4):206-14.
30. Giannakopoulos T, Pikrakis A. Introduction to Audio Analysis: a MATLAB® approach:
Academic Press; 2014.
31. Ellis DP. {PLP} and {RASTA}(and {MFCC}, and inversion) in {M} atlab. 2005.
32. Heffner HE, Heffner RS. Hearing in two cricetid rodents: Wood rat (Neotoma floridana)
510 and grasshopper mouse (Onychomys leucogaster). Journal of Comparative Psychology.
1985;99(3):275.
33. Specht A, de Paula-Moraes SV, Sosa-Gómez DR. Host plants of Chrysodeixis includens
(Walker)(Lepidoptera, Noctuidae, Plusiinae). Revista Brasileira de Entomologia. 2015;59(4):343-
5.
21
515 34. Liu Z, Li D, Gong P, Wu K. Life table studies of the cotton bollworm, Helicoverpa
armigera (Hübner)(Lepidoptera: Noctuidae), on different host plants. Environmental
Entomology. 2004;33(6):1570-6.
35. Wilson JM, Barclay RM. Consumption of caterpillars by bats during an outbreak of
western spruce budworm. The American Midland Naturalist. 2006;155(1):244-9.
520 36. Jones G. Scaling of echolocation call parameters in bats. Journal of Experimental Biology.
1999;202(23):3359-67.
37. López-Ribera I, Vicient CM. Drought tolerance induced by sound in Arabidopsis plants.
Plant Signaling & Behavior. 2017;12(10):e1368938. doi: 10.1080/15592324.2017.1368938.
38. Jeong M-J, Cho J-I, Park S-H, Kim K-H, Lee SK, Kwon T-R, et al. Sound frequencies induce
525 drought tolerance in rice plant. Pak J Bot. 2014;46:2015-20.
39. Playán E, Mateos L. Modernization and optimization of irrigation systems to increase
water productivity. Agricultural Water Management. 2006;80(1-3):100-16.
40. Sadler E, Evans R, Stone K, Camp C. Opportunities for conservation with precision
irrigation. Journal of Soil and Water Conservation. 2005;60(6):371-8.
530 41. Allen CD, Breshears DD. Drought-induced shift of a forest–woodland ecotone: rapid
landscape response to climate variation. Proceedings of the National Academy of Sciences.
1998;95(25):14839-42.
42. Mueller ND, Gerber JS, Johnston M, Ray DK, Ramankutty N, Foley JA. Closing yield gaps
through nutrient and water management. Nature. 2012;490(7419):254.
535 43. Scott J, Jones J, Somodi G, Stall R. Screening tomato accessions for resistance to
Xanthomonas campestris pv. vesicatoria, race T3. HortScience. 1995;30(3):579-81.
44. Andén J, Mallat S. Deep scattering spectrum. IEEE Transactions on Signal Processing.
2014;62(16):4114-28.
45. Sifre L, Kapoko M, Oyallon E, Lostanlen V. Scatnet: a MATLAB toolbox for scattering
540 networks. 2013.
46. Huang C-J, Yang Y-J, Yang D-X, Chen Y-J. Frog classification using machine learning
techniques. Expert Systems with Applications. 2009;36(2):3737-43. doi:
10.1016/j.eswa.2008.02.059.
22
47. Noda J, Travieso C, Sánchez-Rodríguez D. Fusion of Linear and Mel Frequency Cepstral
545 Coefﬁcients for Automatic Classiﬁcation of Reptiles. Applied Sciences. 2017;7(2):178. doi:
10.3390/app7020178.
550 Supporting Information

Fig. S1 Examples for spectrograms of sounds which emitted by stressed plants.
Fig. S2 Comparison of different scattering network configurations.
Fig. S3 ROC curves of four neural network models.
Fig. S4 Tomato sounds distribution.
555 Fig. S5 ROC curve of plant condition decision threshold.
Fig. S6 Recorded sounds from different plants.
Table S1 Parameters used in the feature extraction phase.

Table S2 Pairs total sizes.
560 Table S3 Groups sizes.
23

Plants Emit Informative Airborne Sounds Under Stress: Authors

Uploaded by

Copyright:

Available Formats

Plants Emit Informative Airborne Sounds Under Stress: Authors

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Plants Emit Informative Airborne Sounds Under Stress: Authors

Uploaded by

Copyright:

Available Formats

bioRxiv preprint doi: https://doi.org/10.1101/507590. this version posted December 2, 2019.

The copyright holder for this preprint (which was

Plants emit informative airborne sounds under stress

*Correspondence: Email: [email protected]; Tel: +972-36409831

Plants emit airborne sounds when stressed

The sounds of plants reveal their condition

Plant stress can be identified from its sounds in a greenhouse

We continued to study the acoustic manifestation of the dehydration process, by recording

Materials and Methods

325 Data pre-processing

Acoustic-box Drought stress experiment

Acoustic-box Cut stress experiment

365 Classifying sounds recorded in the acoustic box

Classifying sounds recorded in the greenhouse

435 Competing interests: Authors declare no competing interests.

550 Supporting Information

Table S1 Parameters used in the feature extraction phase.

You might also like