Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Ellard Hunting, [email protected]
Observations were carried out at our field station at the University of Bristol, School of Veterinary Sciences, Langford United Kingdom, which is equipped with an electric field monitor to continuously measure atmospheric PG (Boltek EFM 100 Field Mill, calibrated in a capacitor plat setup). This site features several honeybee hives that are used for research. In the event of overcrowding of a beehive, the original queen leaves the hive with a fraction of the workers (on average around 12,000 bees) (), resulting in occasional swarming events near PG measuring equipment at our study site. When honeybees swarm, they usually cluster on a limb of a tree for several days while scout bees search for suitable cavities to nest in. After an appropriate nest is found, the swarm will collectively migrate. An electric field monitor was placed near the swarm. A camera (AKASO V50×, 30 fps) was positioned next to the field mill with an upward orientation to record the swarm in flight ( Figure 1 B). A second electric field monitor was placed in an open field, 50m away from the other electric field monitor and swarm to ascertain any dynamics in the electric field monitor was caused by the presence of a swarm. For about three minutes, part of the swarm passed over the electric field monitor.
Modelling of electric fields was performed using finite element analysis within COMSOL Multiphysics® v. 5.4 (COMSOL AB, Stockholm, Sweden) utilising the “Electrostatics” interface within the “AC/DC” module. The three-dimensional geometry consisted of a 60 m × 60 m x 40 m (length, width, height) cuboid within which the model operated. The two runs made for differently sized and charged ellipsoids represented insect swarms of honeybees and locusts. The honeybee swarm was represented as an ellipsoid with semi-axes 2 m x 1 m x 0.5 m ellipsoid at a height of 3 m, and the locust swarm an ellipsoid with semi-axes 4 m x 1 m x 1 m at a height of 15 m. The swarm charge was distributed evenly as a volume charge within the ellipsoids, with the total charge calculated respectively as the sum of 500 bees each carrying +100 pC (order of magnitude from ()) and 1000 locusts carrying +850 pC each. An approximately 8 m tall deciduous tree was included in each model to provide scale and an electrical landmark for comparison (). The remainder of the model domain was assigned as air. The upper surface of this air column was given an electrical potential typical of a 40 m altitude in fair-weather conditions (+4 kV), with the bottom surface defined as zero potential, equivalent to the established surface (first meter) atmospheric potential gradient of 100 V/m (). The surface of the tree was also defined as ground (). Meshing of the geometry was physics-controlled, set to “extremely fine”. The relative permittivity, ε, was defined as ε= 12 for living trees (), ε= 1 for air, and ε= 80 inside the insect swarms (likely an overestimate, based on the permittivity of water). Model outputs presented for this study were produced by plotting data from two-dimensional slices through the centre of the three-dimensional dataset.
Quantification and statistical analysis
The video recording capturing the swarm event was cropped to a 500 by 220-pixel window to remove foreground objects obstructing the view of the swarm and was analysed using a custom script in Python 3.8.1. The video was converted to black and white pixels such that the background was white and any non-background objects, comprising the swarm, were black. The ratio of black to white pixels was calculated for each frame of the video resulting in a proxy measure for bee density, defined as relative pixel density here. At points during the swarm’s passage, flowering heads of grass entered the frame, increasing the relative pixel density. Such affected data were therefore removed. The relative pixel densities were filtered by a moving mean over 10 data points (corresponding to 0.5 s) to emphasise the long-term trend of the data. These data were compared to data collected by the electric field monitor using cross correlation and linear regression in PAST v.4. Since insect charges can be expected to influence PGs directly, data were aligned (zero lag) based on the highest cross correlation coefficient.