7.10.2016 1 7.10.2016 Abstract: As part of an autonomous mosquito - PDF document

7.10.2016 1

7.10.2016 Abstract: As part of an autonomous mosquito trap station, the BG-Counter automatically differentiates captured mosquitoes from other insects, counts them, and wirelessly transmits results to a cloud server for further analysis. Its sensors can also continuously collect and transmit environmental data like temperature, rH, precipitation, air movements, or light intensity. Thus, the device allows for a real-time monitoring of local mosquito populations, as well as variables that influence their development and activity. The collected data can be accessed, displayed and analysed by the end user on a cloud-based dash board, which is accessible from PCs, tablet, and smart phones. They can also be exported to Excel. Thus, vector control professionals can now follow the mosquito situation with an unprecedented data density, while overcoming labour constraints associated with manual inspection. Since mosquito abundance can be measured continuously, adulticiding can be performed when mosquitoes are the most active. The effectiveness of control measures can be validated immediately. The collected data can give insights into variables that influence mosquito activity, supporting research into the development of more efficient and environmentally friendly mosquito control techniques. In most mosquito control situations, the target species and their biology is known, and a differentiation of the mosquito species captured and counted during monitoring is normally unnecessary. In situations like the surveillance of invasive mosquitoes or the monitoring of disease vectors during epidemics, the identification of target species would however be important. In our presentation, we describe the technical background of the BG-Counter and supply examples of data sets generated in field. We also present initial results from our research into the advanced version of the BG-Counter, which is planned to be able to identify and differentiate species of special interest. (The development of the BG-Counter was partly supported by the EU's 7th Framework Programme (grant 306105, acronym MCD), the continuation of the development is being supported by the EU's Horizon 2020 programme (grant 691131, acronym REMOSIS). Reference: A. Rose, M. Weber, I. Potatmitis, P.Villlonga, C. Pruszynski, M. Doyle, M. Geismar, J. Encarnacao & M. Geier (2016) The BG-Counter, the first operative automatic mosquito counting device for online mosquito monitoring: field tests and technical outlook. E-SOVE 2016 Book of Abstracts, 177, ISBN: 978-90-8686- 291-7 2

7.10.2016 3

7.10.2016 4

7.10.2016 Exterior view of a BG-Counter prototype. The device is integrated into the intake funnel of a BG-Sentinel mosquito trap (www.bg-sentinel.com). It can also be incorporated into the intakes of other mosquito traps, eave tubes or other openings through which mosquitoes enter actively or passively. 5

7.10.2016 Passing mosquitoes are detected by an IR light barrier. 6

7.10.2016 Passing insects of different sizes and shapes lead to different signal patterns that can automaticall be distinguished and counted. Insects of the size and shape of mosquitoes can thus be seperated from larger insects such as moths or fiels and smaller insects like fruit flies or biting midges. 7

7.10.2016 These are the main capabilities the BG-Counter currently features. 8

7.10.2016 Pictues of the BG-Counter integrated into a BG-Sentinel trap with carbon dioxide. A roof above the trap keeps the catch dry during rain and the counter from counting rain drops. 9

7.10.2016 10

7.10.2016 Since the BG-Sentiel traps were baited with both skin odours and carbon dioxide, the traps are quite specific for mosquitoes. Other insects like nonbiting midges (Chironomidae) or phantom midges (Chaoboridae) are not attracted to this combination of stimuli. Traps with BG-Counters places in typical floodwater mosquito environments had an accuracy of counting the number of captured mosquitoes correctly of more than 90%. 11

7.10.2016 Using a platform available on the internet, the BG-Counter and the trap can be controlled remotely. The operator can run the traps at times when mosquitoes show to be expecually adundant, helping to save energy and carbon dioxide. 12

7.10.2016 The online dashboard also shows the location (in this case, purposely only with a small resolution) and other interesting parameters. 13

7.10.2016 Catches can be inspected and analysed online and the data can be also downloaded. These examples are froma garden near a flood area with predominately Ae. vexans and Ae. sticticus . Note the fluctuation of mosquito activity between the time, but also between days. Since adulticiding should be performed when the mosquitoes are most active, an indiscriminate treatment would have lead to a great variation in actual efficacy. With real-time data from the BG-Counter, treatment can be performed in times of high activity. 14

7.10.2016 This is from a wetland near a lake south of Munich, with mainly the same species. The fluctaution in activity between days and times of day can be seen here also. 15

7.10.2016 The BG-Counter is being further developed in a cooperative research project funded by the EU‘s Horizon 2020 program. The project is called REMOSIS, which is an accronym for „Remote Mosquito Situation and Identification System”. 16

7.10.2016 The main goal is to be able to identify individual species, especially those of medical importance, such as the invasive Asian tiger mosquito, Aedes albopictus . The REMOSIS project started in February 2016. 17

7.10.2016 Distinguishing different mosquito species will need a much closer look at the signals the mosquitoes produce in the BG-Counter. 18

7.10.2016 This is a sample read-out from BG-Counter signals collected from a population of malaria mosquitoes, Anopheles gambiae . On the upper left, one signal from a total of 444 tracks (shown on the lower left). On the upper right, the Fourier transform into the frequencies that make up this signal, with the fundamental tone and the harmonics (shown in the middle right). On the bottom right, the frequency distribution of the fundamental tones of all 444 signals. The mean of all fundamental tones was 578 Hz. In this data collection, it is unclear how many of these tracks originate from the same mosquitoes. 19

7.10.2016 This is a sample read-out from BG-Counter signals collected from a population of southern house mosquitoes, Culex quinquefasciatus . These were more active and yielded a total of 1778 tracks. Again on the bottom right, the frequency distribution of the fundamental tones of all 1778 signals. The mean of all fundamental tones was 518 Hz. 20

7.10.2016 Here, both frequency distributions are shown again, one beneath the other. Note the difference in the means. However, even if this difference would be statistically significant, it would still not help to specify the species of an individual mosquito just from such analysis. Again, it is unclear how many of these tracks originate from the same mosquitoes. 21

7.10.2016 These are different tracks from individual mosquitoes that were placed in a small cage in the center of the counter board. The fundamental tone of every singel track of each individual mosquito is listed, with rising frequncies. Different colours denominate different species. This graph shows that individual mosquitoes tend to have mostly the same fundamental frequencies. The mean differences between species can be seen here also. In our future work, we will further analyse these and other aspects of such tracks to see if additional differences will be observable between species. Through the combined analysis of these different variables and advanced machine learning, we are confident to succeed in a successful species identification. 22

7.10.2016 23