Colony defence in bumblebees (Bombus terrestris)
Data files
Oct 28, 2025 version files 151.97 MB
-
All_Defensive_Behaviours_of_Bumblebees.mp4
148.88 MB
-
call_times_for_rhythm.csv
320.41 KB
-
Dataset1_Clean.csv
254.57 KB
-
Dataset2_Clean.csv
237.71 KB
-
MFCC_for_random_forest.csv
1.88 MB
-
README.md
5.11 KB
-
spectral_features.csv
402.46 KB
Abstract
Bumblebees are key pollinators of both wild plants and agricultural crops; hence, understanding their biology is critical for conservation efforts as well as for managing domesticated colonies. While their foraging and reproductive ecology have received a lot of attention, we know little about another crucial part of their behavioural repertoire: colony defence. The first dataset examines the defensive responses of individuals in whole colonies, maintained at the laboratory, after disturbing them with a mechanical jolt. As a result, we present a detailed ethogram of the defensive behaviour of Bombus terrestris in response to mechanical disturbances, as could be induced by vertebrate attacks or handling. In addition to identifying and describing the different responses elicited by this disturbance, this data contains information about their temporal sequence, their location and the proportion of bees involved. We also determined whether individual bees produce specific associations of behaviours, which revealed that a core set of responses is exhibited by all bees, with other behaviours being randomly associated. The second dataset compares defensive response based on disturbance type (mechanical jolt, intruder breath or foreign object). Overall, we demonstrate that colony defence includes measures preparing members for a response by signalling about the threat, searching for the source of the disturbance, warning intruders to maintain distance, and finally recovering from the disturbance. This comprehensive overview provides a valuable starting point to further understand how defensive behaviour is regulated such that bumblebee colonies can survive predator attacks and thrive.
Dataset DOI: 10.5061/dryad.jh9w0vtq8
Description of the data and file structure
This is the data pertaining to the article "Colony defence in bumblebees (Bombus terrestris), published in PloS ONE in 2025. The files "Dataset1_Clean.csv" and Dataset2_Clean.csv" list behavioural responses of individual bees to a disturbance, scored from BORIS. The file "All_Defensive_Behaviour_of_Bumblebees.mp4" is a video showing examples of these behavioural responses. The 3 other files correspond to audio data extracted from recordings of individual bees being disturbed. In all files, "NA" stands for "not applicable".
Files and variables
File: Dataset1_Clean.csv
Description: Dataset corresponding to Part I (defensive responses of colonies with 20 workers and a queen to a mechanical jolt)
Variables
- Observation.id: Concateneted information about the colony observed (A, B, C or D) and the experimental day (1 or 5)
- Total.duration: total duration of the video file analysed (in s)
- Subject: Bee performing the behaviour. Note that this ID is not maintained across experimental days.
- Behavior: behaviour observed or disturbance event
- Behavior.type: state for all observed behaviours, point for the disturbance
- Start..s.: Timepoint at which the behaviour starts, relative to the start of the video, in s.
- Stop..s.: Timepoint at which the behaviour stops, relative to the start of the video, in s.
- Duration..s.: Duration of the behaviour in s.
File: Dataset2_Clean.csv
Description: Dataset corresponding to Part II (defensive responses of microcolonies of 5 bees to different disturbances)
Variables
- Colony: Colony tested (A to F and 1 to 8)
- Day: experimental day (1 to 3)
- Disturbance: stimulus used to disturb the colony (breath, object or mechanical)
- Total.duration: total duration of the video file analysed (in s)
- Subject: Bee performing the behaviour (1 to 5)
- Behavior: behaviour observed or disturbance event
- Behavior.type: state for all observed behaviours, point for the disturbance
- Start..s.: Timepoint at which the behaviour starts, relative to the start of the video, in s.
- Stop..s.: Timepoint at which the behaviour stops, relative to the start of the video, in s.
- Duration..s.: Duration of the behaviour in s.
File: All_Defensive_Behaviours_of_Bumblebees.mp4
Video showing examples of the following behaviours: Abdominal pumping, Climbing/Flying, Continuous buzzing, Fanning, Grooming, Leg raising, Lying on the back, Patrolling, Perching, Pulse Buzzing
Files containing audio data:
These files contains acoustic temporal measurements of bumblebee buzzes, organized into three main CSV files. Each file serves a different analytical purpose: rhythm analysis, Buzz type classification, and spectral feature extraction.
----------------------------------------
1. call_times_for_rhythm.csv
----------------------------------------
Description: Metadata and timing information for individual calls, used for rhythm and inter-call interval analysis.
Variables:
- uniqBee, Colony, Day, Experiment, BeeID: Identifiers of individuals and context.
- Behavior, Behavior2, Description: Behavioral annotations.
- Start_s, Stop_s, Duration_s: Buzz timing in seconds.
- time_diff, ICI: Temporal spacing and inter-call intervals.
- buzz_pair, rythm: Buzz dyads and rhythmic ratio for rhythm-related analysis.
- wave_file, UID: File references.
----------------------------------------
2. MFCC_for_random_forest.csv
----------------------------------------
Description: Mel-Frequency Cepstral Coefficients (MFCCs) for supervised classification (e.g., Random Forest).
Processing details:
- Each call was segmented into nine equal-length windows.
- For each window, 12 Mel-filters (frequency bins) were extracted.
- Standard window everagig ws used
Variables:
- class: Target label (call type / category).
- 1 ... 177: MFCC feature coefficients.
- Unnamed: 0: Row index.
----------------------------------------
3. spectral_features.csv
----------------------------------------
Description: Acoustic feature set extracted from buzzes
Spectral features were extracted using the function spectro_analysis() from the warbleR package in R.
Calls were analyzed with the following main settings:
- Window length (wl) = 512
- Overlap (ovlp) = 70%
- Band-pass filter (bp) = 0.07–4 kHz
- Threshold = 15 dB
- Window type = Hanning
- Frequency smoothing (fsmooth) = 0.1
- Harmonicity analysis enabled (harmonicity = TRUE) with 3 harmonics.
Variables:
- sound.files, selec: Recording identifiers.
- duration: Call duration.
- Frequency measures: meanfreq, sd, freq.median, freq.Q25, freq.Q75, freq.IQR.
- Temporal measures: time.median, time.Q25, time.Q75, time.IQR.
- Spectral shape: skew, kurt, sp.ent, entropy, sfm.
- Harmonic features: h1_freq, h2_freq, h3_freq, harmonics, HNR.
- Additional: SNR$Annotation, new_name.