Skip to main content

Wind and route choice affect performance of bees flying above versus within a cluttered obstacle field

Cite this dataset

Burnett, Nicholas; Badger, Marc; Combes, Stacey (2022). Wind and route choice affect performance of bees flying above versus within a cluttered obstacle field [Dataset]. Dryad.


Bees flying through natural landscapes encounter physical challenges, such as wind and cluttered vegetation.  The influence of these factors on the flight performance of bees remains unknown.  We analyzed 548 videos of wild-caught honeybees (Apis mellifera) flying through an enclosure containing a field of vertical obstacles that bees could fly within (through open corridors, without maneuvering) or above.  We examined how obstacle field height, wind presence and direction (headwinds or tailwinds) affected altitude, ground speed, and side-to-side casting (lateral excursions) of bees.   When obstacle fields were short, bees flew at altitudes near the midpoint between the tunnel floor and ceiling.  When obstacle fields approached or exceeded this midpoint, bees typically, but not always, increased their altitudes to fly over the obstacles.  Bees that flew above the obstacle fields exhibited 40% faster ground speeds and 36% larger lateral excursions than bees that flew within the obstacle fields, likely due to the visual feedback from obstacles and narrow space available within the obstacle field.  Wind had a strong effect on ground speed and lateral excursions, but not altitude.  Bees flew 12-19% faster in tailwinds than in the other wind conditions, but their lateral excursions were 19% larger in any wind, regardless of its direction, than in still air.  Our results show that bees flying through complex environments display flexible flight behaviors (e.g., flying above versus within obstacles), which affect flight performance.  Similar choices in natural landscapes could have broad implications for foraging efficiency, pollination, and mortality in wild bees. 


Experimental set-up

Experiments were conducted in a laboratory flight tunnel (20.0 x 19.1 x 115.0 cm; width x height x length) with a field of obstacles (hereafter referred to as the ‘obstacle field’) in the middle of the tunnel.  The obstacle field consisted of vertical columns (diameter = 7 mm) arranged in three parallel rows of five obstacles each, running along the length of the tunnel.  Obstacles were made of dark green, cylindrical blocks (LEGO, Billund, Denmark) that contrasted with the black and white speckled pattern of the tunnel’s walls, and flight data indicated that bees were able to detect and avoid these obstacles.  All obstacles within an obstacle field were of the same height, and the obstacles extended only partway to the tunnel’s ceiling, allowing bees to fly either within or above the obstacle field.  The total height of the tunnel was 191 mm and the obstacle field heights tested were 11, 40, 69, 98, or 127 mm (Figure 1b), so a minimum of 64 mm (~1/3 of the total vertical height) between the top of the obstacle field and the ceiling remained free of obstructions.  We consider the 11-mm obstacle field as a control for the presence of obstacles because this obstacle field was too short for bees to fly within.  There were approximately 20 mm between the outer rows of the obstacle field and the walls in each arrangement.  Fans (AC Infinity, City of Industry, CA, USA) on each end of the tunnel produced a mild wind of the same mean flow speed (0.54 m s-1, measured with a Velocicalc Air Velocity Meter Model 9535, TSI, Shoreview, Minnesota, USA) above the obstacle field and within the corridors of the obstacle field (i.e., the space between the rows of obstacles); within the rows of obstacles (i.e., immediately downstream of an individual obstacle), the wind speed dropped to 0.36 m s-1.

Freely flying honeybees Apis mellifera (n = 58) were collected on the campus of the University of California, Davis.  Single bees were flown in the tunnel with an obstacle field of one of the five possible heights, and the obstacle field height used for each bee was determined by a random number generator.  Lights (26 Watts, full spectrum; Hagen, Mansfield, MA, USA) were alternately turned on and off at each end of the tunnel to motivate the bees to fly back and forth past the obstacle field (towards a light).  We filmed between 5 and 13 flights per bee (mean = 9), with approximately half the flights in still air and half the flights with wind, for a total of 548 recorded flights.  We randomly assigned each bee to begin their flights with either wind or still air.  Bees flew in headwinds (flying into the wind) or tailwinds (flying with the wind), depending on the direction they flew relative to the air flow on a given transit.  We define the two flight directions in our tunnel as ‘up-tunnel’ and ‘down-tunnel’ – bees experienced headwinds when flying in the up-tunnel direction with wind and tailwinds when flying in the down-tunnel direction with wind.

Flights were filmed with two synchronized Phantom v611 high-speed video cameras (Vision Research, Inc., Wayne, NJ, USA) sampling at 500 frames s-1, each positioned 30° from the vertical on opposing sides of the obstacle field and viewing down the length of the tunnel.  Cameras were calibrated using a standard checkerboard calibration method and built-in MATLAB functions.  This method captures lens distortion and projective geometry (using the intrinsic parameters), as well as the global positions and orientations of the cameras relative to the flight tunnel (via the extrinsic parameters).


Kinematic analysis

            We used a detection and tracking pipeline to automatically track the centroid of bees in each camera view as they transited the obstacle field.  From each frame, we subtracted the background and found one or more candidate positions of the bee using MATLAB’s built-in blob detection functions.  We associated these detections into a single trajectory over time using a Kalman filter and Munkres' assignment algorithm.  We then used DLTdv6 to check and manually correct the automatically tracked positions of bees.  We also labeled the positions of obstacles in the field using DLTdv6.  Using the camera calibration, we converted the two-dimensional locations of the objects in each view into three-dimensional coordinates of the bees and obstacles.  We analyzed bees’ trajectories from when they entered to when they exited the obstacle field, and we smoothed the trajectories with quintic spline curves.  From each flight’s position data, we calculated the median altitude of the bee across its entire flight and the range of altitudes of the bee across its entire flight.  To assess flight performance, we calculated two metrics: ground speed – the median of the bee’s speed relative to the ground, based on its movement in the horizontal plane (i.e., lateral and fore-aft motion), and lateral excursion, quantified by variation (i.e., interquartile range) in the bee’s lateral position over the entire flight.


National Science Foundation, Award: 1711980