The typical flight performance of blowflies: measuring the normal performance envelope of Calliphora vicina using a novel corner-cube arena

Abstract

Despite a wealth of evidence demonstrating extraordinary maximal performance, little is known about the routine flight performance of insects. We present a set of techniques for benchmarking performance characteristics of insects in free flight, demonstrated using a model species, and comment on the significance of the performance observed. Free-flying blowflies (Calliphora vicina) were filmed inside a novel mirrored arena comprising a large (1.6 m1.6 m1.6 m) corner-cube reflector using a single high-speed digital video camera (250 or 500 fps). This arrangement permitted accurate reconstruction of the flies' 3-dimensional trajectories without the need for synchronisation hardware, by virtue of the multiple reflections of a subject within the arena. Image sequences were analysed using custom-written automated tracking software, and processed using a self-calibrating bundle adjustment procedure to determine the subject's instantaneous 3-dimensional position. We illustrate our method by using these trajectory data to benchmark the routine flight performance envelope of our flies. Flight speeds were most commonly observed between 1.2 ms(-1) and 2.3 ms(-1), with a maximum of 2.5 ms(-1). Our flies tended to dive faster than they climbed, with a maximum descent rate (-2.4 ms(-1)) almost double the maximum climb rate (1.2 ms(-1)). Modal turn rate was around 240 degrees s(-1), with maximal rates in excess of 1700 degrees s(-1). We used the maximal flight performance we observed during normal flight to construct notional physical limits on the blowfly flight envelope, and used the distribution of observations within that notional envelope to postulate behavioural preferences or physiological and anatomical constraints. The flight trajectories we recorded were never steady: rather they were constantly accelerating or decelerating, with maximum tangential accelerations and maximum centripetal accelerations on the order of 3 g.

Figure 1. Diagram showing how the primary, secondary, and tertiary reflections of a fly are formed in a corner-cube reflector:

yellow circle represents the fly itself; red circles represent the apparent locations of the three primary reflections (ray shown reflecting off the Y mirror in this case); green circles represent the apparent locations of the three secondary reflections (ray shown reflecting off the XY mirror pair in this case); the blue circle represents the tertiary reflection off all three mirrors (XYZ).

Figure 2. Schematic of the mirrored corner-cube flight arena with camera position and orientation.

Figure 3. Autocorrelation functions during filtering.

Plots of the normalized autocorrelation functions of the residuals of the 500 fps x-position data (black lines) after filtering at: (A) 10 Hz, (B) 48 Hz and (C) 100 Hz. For comparison, we also plot the normalized autocorrelation function of the residuals of a sequence of Gaussian white noise of the same length passed through each of the filters (red lines). For this method we selected a cut-off frequency at which the lowest frequency at which the autocorrelation function of the residuals of the actual data still matched closely the autocorrelation function of the residuals of the random sequence - in this case, at around 48 Hz. An animation of the change with respect to cut-off frequency of the autocorrelation (and the variance of the autocorrelation) can be found in Supporting Information.

Figure 4. Histograms of translational flight performance.

Plots show translational flight performance metrics: (A–C) total speed () and its horizontal () and vertical () components; (D–F) total acceleration (a) and its absolute tangential () and centripetal () components; (G–I) total tangential acceleration and its horizontal () and vertical () components. In each case, the variables plotted in the first column are the Pythagorean sums of the variables plotted in the second and third columns within a row. Dashed lines represent 99% confidence intervals and are one-tailed in the cases where a single line is presented, and two-tailed where two lines are plotted.

Figure 5. Histograms of turning flight performance.

Plots show performance metrics related to turning: (A) turn rate (); (B) turn radius () for all <1.75 m; (C) elevation angle of centripetal acceleration vector (). Dashed lines represent one-tailed 99% confidence intervals. No confidence intervals are plotted for since the data are constrained to ±90°.

Figure 6. Flight trajectories coloured by speed from low speeds (coloured blue) to high (coloured red).

See colour bar for detail. Trajectories closer to any mirror than 14 mm have been removed.

Figure 7. Flight trajectories coloured by tangential acceleration (ms⁻²).

Near-zero accelerations are coloured yellow; positive accelerations are coloured green; decelerations are coloured red. See colour bar for details.

Figure 8. Flight performance envelopes.

Scatter plots (A–D) and density plots (E–H) of flight performance data: (A,E) relative thrust-drag () against flight path elevation angle (); (B,F) Relative lift () against its elevation angle (); (C,G) relative thrust-drag against flight speed (); (D,H) relative lift () against relative thrust-drag (). In each case, notional physical limits of the flight performance envelope are plotted on the figures as dashed lines (see body text for detail on limit line construction).