The aim of this preliminary study was to provide a better understanding of the bioacoustics of woodboring insects in order to develop a practical method for early detection. Sound emissions from larvae of the red palm weevil (Rhynchophorus ferrugineus), the asian longehorn beetle (Anoplophora glabripennis) and longhorn beetle of the genus Monochamus were analyzed. Clustering experiments showed that biting and stridulations sounds are promising elements for species recognition.
|Biting and stridulations sounds are promising elements for species recognition|
Former investigations showed that active larvae inside timber produce measurable sound emissions which can be recorded, and that specific sounds are typical for each species (Chesmore 2008). The study concept to appropriate sound emissions as a tool for an early detection method is not new. Electronic listening devices have been available since the 1920’s, but only due to recent technologic advances has it been able to hear noises under less optimal conditions (Mankin et al.2011) and separate background noises which often interfere.
With pattern recognitation algorithm it is possible to automatically detect characteristic noises, even when they are overlain with background noises. In principle a computer can automatically do the complete interception- and evaluation process.
As part of the EU-project "Q-Detect: Development of detection methods for quarantine plant pests for use by Plant Health Inspection Services", we investigated three Model organisms, which are of importance for European countries.
The main objective of our work was to get basic information for the early detection of these wood boring insects through a new practical approach. To achieve this it would be essential to record a large quantity of reference sound recordings, to chart and descriptively establish basic bioacoustic parameters for automatic species identification, and to practically test this in outdoor field trials.
Representative recordings of all relevant larvae sounds produced within wood were taken, and catalogued. This sound library is the basis for continuing empirical analyses, and it is planned to develop it further. Recordings were made in the field as well in laboratory. The laboratory recordings were taken in a soundproof box, where the activity of Monochamus sator, Monochamus sutor and Anophlophora glabripennis larvae in logs were taken.
The Anophlophora glabripennis material was quarantine infested logs stored and reared in the Bfw institute. Logs from two infestation areas were taken, one located in Italy (Cornuda, Masere) and the other Upper Austria (Braunau/Inn). Records of R. ferrugineus were made in quarantine-rooms in Israel (Volcani Center) on sugarcane as well on palm trees in Spain (Mallorca).
Reference sound records were made with Panasonic Elektret with microphone capsules fixed directly on the trees. Improved anchoring was achieved by debarking 1cm² of a stem, but this isn’t necessary for palm trees, as palm fronds have a smooth surface. The installed microphones are furthermore fixed by a sealing compound which has the advantage of reducing background noises.
The microphones distinguish by a uniform frequency response from 50 Hz – 16 kHz. The connection to a hard disc recorder was over a low-noise, double shielded preamplifier (Sejona R&D) and a USB card (BehringerUCA222).The software Amon (Sejona R&D) was used for the recordings as it has a automatic file-splitting and enables continuous recording over a several weeks. In the field a portable recorder Marantz PMD 661 was used. All recordings are produced with 44.1 Khz, 16 Bit.
The sound library now consists of 150 reference records in laboratory conditions. There are also 9.7 hours field recordings on infested palms and which include examples of background noises.
The next step was to generate a noise inventory out of the reference recordings. All characteristic noises were identified and utilize for advanced analyses. The inventory was created manually, by audio monitoring, visualization, comparing and counting. The inventory was carried out with the software Sound Recognition Lab (SejonaR&D). Six different noise of types emerged.
A good knowledge about bioacoustics parameters is necessary to evaluate which of the noise- types are relevant for infestation detection and for species identification.
|Table 1: Sound inventory for all three investigated species|
|Typ-I- bitesounds||Typ-II- bitesounds||Stridulations-sounds||Breath-sounds||Motion-sounds||Knocking-sounds|
|Rhynchophorus ferrugineus||very frequent||
|Table 2: Background noise inventory for urban areas, determined in measurements of palm trees on Mallorca|
|backgroundnoise||Impairment of Analyse||
|traffic, street cleaner||middle||frequent||≤ 90 dB||anywhere|
|Lawn-mover, chainsaw||intense||middle||≤ 102 dB||Green area|
|Aircraft sound||minor||rare||≤ 70 dB||airport|
|barking, birdsongs||middle||frequent||≤ 63 dB||anywhere|
|speech, call||minor||frequent||≤ 65 dB||anywhere|
|churchbell||minor||rare||≤ 65 dB||city|
||frequent||≤ 72 dB||anywhere|
||≤ 96 dB||anywhere|
To that end, the following six Parameters were examined:
The latter measurements were made on specific events and complete sequences, whereby the following parameters could be defined: (A) item repetition rate per second, (B) number of specific events per sequence, (C) length of sequence, (D) rhythmicity (chronology of specific events). The determination of the acoustic parameters was carried out using the Sound Recognition Lab software, which through programmed noise sequence filters search inside the sound library for species-specific noises. Located noises were then tagged and charted.
|Diagram 1: Histogram "Peak-Frequency" of the stridulating sounds of M. sutor and R. ferrugineus|
With the help of the parameters it was possible to select candidates from the noise inventory to make infestation tests. In this case, silent and diffuse noises are not suitable. Frequent and loud sounds with stabilized acoustic characteristics ( z. B stridulation) are suitable.
In all three investigated species, Typ-I-bitesounds showed the best results for detection, whereby with R. ferrugineus oftenTyp-II-bitesounds are found and recorded. Breathing, motion, and knocking sounds were seldom, therefore only the average of these parameters were determined.
For all other noises, histograms and parameters were measurable. Stridulations are particularly qualified for the species identification and detection of Monochamus sp. and Rhynchophorus ferrugineus.
The audio-material also allowed experiments for an automated distinction of different species. Here the aim was to determine the different acoustic parameter combinations inside equal noise-types, and to find out which combination is applicable for species identification.
Two different experiments were carried out, where Typ-I-bitesounds of Rhynchophorus ferrugineus against Anoplophora glabripennis and the stridulation-sounds of Rhynchophorus ferrugineus against Monochamus sutor were compared.
For this we selected sounds out of the noise-inventory and computed this into data format. The size Center-Frequency, Length, -3Db -Bandwidth and Peak-Energy were determined and in scatterplots construed.
That all points in the cluster could be categorized to a species, it was then possible the species specifity of the cluster to verify, disjointed clusters being species-specific. Identical and overlapping clusters are not species-specific. The manual classification and creation of scatter plots was made with the software Sound Recognition Lab.
|Diagram 2: Scatterplot of stridulation sounds of Rhynchophorus ferrugineus and Monochamus sutor. The sounds of Monochamus sutor form three adjacent clusters while sounds of Rhynchophorus ferrugineus are widely scattered. The Cluster experiments show that for all three species it´s possible to calculate species-specific clusters.|
Noise detection was tested under practical conditions in an urban area of Mallorca. The aim was to check whether - and under what conditions - sounds of Rhynchophorus ferrugineus were practically suited for detecting palm weevil infestations. Another aim was to confirm if in situ records would correspond to labor recordings. Concurrently, types of background noises would be identified and analysed. The latter to control if they could influence the recordings, and ultimately confirm if the resource and effort input in the detection method could be realistically implemented.
In July 2013 different acoustic activity was recorded on different palm species (Tab. 3). The measurements were done in streets and urban gardens between 6.00 am and 20.00 pm. Palms were randomly selected and investigated for visible symptoms. The main focus was on Typ-I-bitesounds and Typ-II-bitesounds, as well as stridulation sounds.
The sounds of red palm weevil larvae were very well recorded, with the good conductivity of palm wood being advantageous. Also in the field tests, all sound types were found, but only one recording contained stridulation sounds.
Most of the palm trees had both bite-types (Typ I and Typ II) present. Bite-sounds of red palm weevil could be observed in 37 of 56 palms (66%) some of which were palms without extern symptoms, or were treated with insecticides.
|Tab 3: General view about field studies on palm trees|
|Species||Number tested palms||Number of measurements||Total of Duration (h)||Palms with sounds of Rhynchophorus|
|Phoenix canariensis||42||46||8,1||30 (71,4%)|
|Phoenix dactylifera||10||10||1,0||3 (30,0%)|
Additionally to the recordings, a background noise inventory including the most frequent surrounding noises became installed (see Tab. 2). Wind caused a lot of background noises due to the rubbing of palm fonds together. These proved strong enough that they overlay bite- and stridulations sounds, so that diagnose was impossible. However in practice, to elude the problem, only sections of the records with fewer disturbances would be analysed. It is also possible with specific clustering processes able to separate wind affected bio acoustic sounds.
This means, that with little effort bioacoustic diagnoses in urban areas are possible. A total of 20-30 min per tree was needed, split in equal parts of recording and analyzing.
Although the analyse can be carried out semi-automatically, to achieve reliable results, all three sound types have to be considered in a multistage procedure.
Bioacoustic investigation methods lend themselves to a non invasive testing of wood and living plants for insect infestation, however the insects must be in an active stadium. Eggs, diapause or pupation stages are not detectable. For the red palm weevil, the asian longehorn beetle and the longhorn beetle of Monochamus it was possible to record specific bioacoustics e.g. stridulation sounds for species identification. That especially by Anoplophora glabripennis and Monochamus spp., a series of other species could also be present in the investigated plant, it is seen as imperative that further studies are necessary to secure suitable differential diagnosis.
The results to date have shown that bioacoustic offer new possibilities in checking imported timber and living plants, and also for checking infested plants in outdoor situations. Computer software has proven to increase and improve the efficiency of the diagnosis processes. Bioacoustic sounds can also -through the use of correct algorithms- be identified even when overlaid by other noise disturbances.
Chesmore, D. 2008: Automated bioacoustic identification of insects for phytosanitary and ecological applications. In Frommolt K.H., Bardeli R. and Clausen M. (Eds.) Computational bioacoustics for assessing biodiversity. BfN-scripts Nr. 234: 59-72.
Mankin, R: w., Hagstrum, D: W., Smith, M. T., Roda A.L., Kairo, M. T. K. 2011: Perspective and promise: acentury of insect acoustic detection and monitoring. American Entomologist 57: 30-44.
Victorrson, J:; Wikars, L.O. 1996: Sound production and cannibalism in larvae of pine-sawyer beetle Monochamus sutor L. (Coleoptera: Cerambycidae.) Entomol. Tidsskr. 117: 29-33.
Ing. Martin Brandstetter, Federal Research and Training Centre of Forests, Natural Hazards and Landscape Austria, Institute for Forest Protection
Seckendorff-Gudent-Weg 8, 1131 Vienna, Austria
Dr. Sebastian Hübner, Sejona R&D, Grüner Markt 31, 96047 Bamberg, Germany,