Ecologist
Dátum 2009.11.8. 0:33:00 | Téma: Európai Méhész hírei
|
Case for Support - Expression of Interest for the INSECT POLLINATORS INITIATIVE
Investigation of a non-invasive tool for inexpensive and large scale
monitoring of honey bee health and swarming
Investigators: Martin Bencsik (Acoustician), Tom Reader (Ecologist)
Collaborators: Michael Baxter (Statistician), Ahmad Lotfi (Computer Scientist).
Background
This project is building upon extensive existing scientific evidence1 that the sounds and vibrations emitted by insects are important to understand their physiology and behaviour patterns. Honey bees (Apis mellifera) in particular make exquisitely sophisticated use of vibrations and sounds for various purposes, including e.g. conveying to others the location and abundance of remote food sources, or preparing for the swarming2,3. Honey bees are amongst important pollinator insects that have recently shown substantial health issues often leading to massive losses of colonies, leading to the expression ‘colony collapse disorder’. We propose to investigate the hypothesis that long term time series data for honey bee hive vibrations can be used to diagnose colony condition and predict swarming behaviour.
Pilot Study
We have obtained encouraging results from a pilot study which has recently been submitted for peer-reviewed publication4. In our preliminary work honey bee vibrations were sensed by two accelerometer sensors (Isotron 7259B-10, Endevco) secured onto the outer wall of two separate hives, approximately 3 meters away from one other, each comprising of a healthy colony of Apis mellifera honey bees. These sensors were connected to a dual channel conditioner (Nexus, Brüel and Kjær) residing between the two hives. The ouput channels of the conditioner were continuously digitised and stored on a hard disk from October 2008 to June 2009. By computing short frequency spectra that are averaged together, a very clear curve with pronounced peaks can be rapidly obtained. The entire time series of spectra for one hive was analysed by Principal Component Analysis5 (PCA), using a separate script (Fig. 1). We have shown that swarming behaviour is predicted several days in advance by striking changes in scores derived from PCA.
Fig.1: One-day long Spectrogram (top) and PCA scores (bottom) of the vibrational signal from a hive. The ninth score (not shown) was found to be an excellent indicator of the hive’s preparation and execution of swarming4.
Aims
We propose to further investigate our hypothesis, supported by our preliminary results, that honey bee hive vibrations can be used to diagnose colony condition and predict swarming behaviour. Our principal aim is to establish the repeatability of our results and its usefulness as a diagnostic monitoring tool. A key benefit is that this is a non-invasive method of study. Our objectives are:
1. Establish the repeatability of the results on a collection of at least 20 hives. Is swarming consistently predicted days in advance by changes in vibrational patterns?
2. Establish the sensitivity of the information extracted from the acoustic signal to colony size, hive condition and common health problems.
These aims directly tackle the fifth bullet point of the SCOPE of the pollinator's initiative and, if successful, will help on the long term the first and second bullet points, if suitable protocols are explored in later research projects.
Project description
1. A sample of 20 hives, some in poor health, others thriving, will be obtained by collaboration with local beekeeper PB. If needed, hives will be relocated near to a shed that has mains power supply. Such suitable locations already exist on the University of Nottingham (Sutton Bonington) and Nottingham Trent University (Clifton) campuses.
2. The time course of the vibration of each hive will be continuously and automatically monitored with accelerometers as described in the previous section together with time lapse photography for evidence of swarming, and digital weather logging device (temperature, humidity, rain fall, wind, atmospheric pressure).
3. The size and condition of each colony will be characterised at regular intervals (at least twice a month) during the course of the study by measuring variables such as worker abundance, hive mass, brood area, honey and pollen mass, varroa mite and moribund larvae counts using standard methods6-7.
4. PCA will be compared with more complex, non-linear, pattern recognition methods applied on the vibration data to establish whether it is possible to
o extract diagnostic information relevant to some of the monitored measures of condition/health
o predict swarming
5. Repeatability and reliability of the method will be assessed by comparing our results within the set of hives.
In the first year of the study whilst the new experiments are being set up, the existing data set collected in our preliminary study will be interrogated with non-linear clustering algorithms to compare their strengths in predicting the swarming process. One specific (healthy) hive will also be prepared differently, in that 10 accelerometers will be installed on the individual internal frames (in addition to the external one) to establish the relevance of spatially resolved measurements, in contrast with the point-measurements we have so far undertaken.
Applicants expertise
MBe has extensive experience in the logging and interpretation of acoustic noise generated by MRI coils and in data processing using Matlab®; he also led the pilot study and presented it at an international conference (Apimondia 2009). TR has a well established track record of peer-reviewed publications involving ecological field experiments and bioacoustics. MBa is a Professor of Statistics with extensive experience of, and publications in, applications of multivariate methods applied to field subjects (e.g. archaeology). AL has extensive experience in computational intelligence and data analysis.
Potential impact
As is the case when other sophisticated monitoring tools emerged for humans (EEG, fMRI, MEG etc..), this tool is likely to provide a wealth of new understanding in the physiology of honey bees in healthy and non-healthy scenarios, in addition to being simple, non-invasive and inexpensive. As a diagnostic tool, it might allow early prediction of diseases, which is known to be a crucial factor in our ability to effectively treat numerous diseases in humans.
Resources required
The development and operation of the non-invasive measurements (accelerometers, video, etc) and processing of data requires expertise at the postdoctoral level (1xPDRA based at NTU). To set up and run the hives and field sampling programme, the PDRA will require assistance from a research technician based at the UoN campus and an experienced beekeeper (we have existing good links with local beekeepers, such as Petar Bogunovic from the LRBKA). The close proximity of the two University campuses will allow close integration of the project. PI and CoI time, technician time, a beekeeper subcontract, hives, honey bees, twenty accelerometers with compatible signal conditioners, data acquisition and Uninterruptible Power Supply (UPS) systems plus small consumables are the other major costs.
References
1. Michelsen A. The Dance Language of Honeybees: Recent Findings and Problems In: The Design of Animal Communication (MD Hauser and M Konishi, editors) MIT Press, pp. 111-131 (1999).
2. Von Frisch K. The dance language and orientation of bees. Cambridge, Mass.: The Belknap Press of Harvard University Press (1967).
3. Schneider S.S. and Lewis L.L.. Apidologie 35, 117-131 (2004).
4. Bencsik M. et al, submitted to Journal of expermental biology, Sept 10th 2009.
5. Baxter, M.J, Archaeometry 48, 671-694 (2006).
6. Chauzat, M. et al. Environ. Entomol. 38(3): 514-523 (2009).
7. Meikle, W. G. et al. Exp Appl Acarol 46:105–117 (2008).
|
|