Landing force reveals new form of motion-induced sound camouflage in a wild predator
- Department of Ecology and Evolution, University of Lausanne, Switzerland
- Agroecology and Environment, Agroscope, Switzerland
- Swiss Ornithological Institute, Switzerland
- Department of Biology, Lund University, Sweden
- Les Ailes de l’Urga, France
- Department of Biosciences, Swansea University, United Kingdom
Figures
Figure 1
Sequence of perching and hunting strike events throughout a barn owl foraging trip.
(A) GPS tracks (black line) during one complete foraging trip (duration = 73 min) performed by a female barn owl, with perching events (squares), unsuccessful (circles), and successful (triangles) hunting strikes. Black arrows indicate the flight direction. Successful hunting strikes were identified by the presence of self-feeding events (identified from the acceleration data), or by the direct return to the nest box (identified from the acceleration data and validated with the nest box camera footage). Inset panels show an example of the tri-axial acceleration signals corresponding to both nest-box return and self-feeding behaviours (see Appendix 1—figure 3 for detailed representations). (B) The heave acceleration and the associated force during a perching event (highlighted in orange) and a hunting strike (highlighted in dark purple). (C) Variation in peak landing force for perching events (orange dots, n=56,874) and hunting strikes (dark dots, n=27,981). White dots show the estimated mean, and data distribution is represented by both violin and box plots. The owl picture at the top left of panel A is courtesy of J. Bierer, and owl drawings are courtesy of L. Willenegger, all used with permission.
Figure 2
Sequential changes in perch type and landing force prior to hunting strikes during a sit-and-wait hunt.
(A) Landing force during perching events (n=40,305) in relation to the time until the next hunting event and perch type. Each line represents the predicted mean for each perch type (here shown for males), with the 95% confidence intervals. (B) The selection of perch type in relation to time until the next hunting strike, highlighting the change in perch type that occurs ~10 min prior to a strike. (C) A sequence of perching events (orange circles) prior to a successful strike (purple circle) for a typical sit-and-wait hunt, showing the variation in peak landing force through time. White arrows indicate flight direction and numbers under each perching events indicate the time until the next hunting attempt (i.e pre-hunt time).
Figure 3
Sexual differences in foraging behaviour, landing force, and hunting success.
(A) GPS tracks showing the foraging activities of a barn owl breeding pair during one complete night. Movement patterns of both male (yellow lines) and female (blue lines) are shown, with colour scale changing from the first trip of the night (foraging trip 1) to the last one (Male: nmax = 11; Female: nmax = 4). Perching events (squares), unsuccessful (circle) and successful (triangle) hunting attempts are shown for each foraging trip. (B) Variation in foraging flight speed for female (blues dots, n=9,223) and male (orange dots, n=18,019) barn owls (females: n=84; males: n=78). (C) Variation in peak landing force during perching events (perching force) for female (blue dots, n=30,378) and male (yellow dots, n=26,496) barn owls. (D) Variation in hunting success when barn owls hunted on the wing or used the sit-and-wait strategy for female (blue dots, non-the-wing=8,136,, nsit-and-wait=1981) and male (yellow dots, non-the-wing=16,328,, nsit-and-wait=1532) barn owls. For visualisation purposes, each dot shows the average hunting success of each individual for both hunting strategies. White dots and bars show the mean and the 95% CI around the mean, respectively, and data distribution is represented by both violin and boxplots. Owl drawings are courtesy of L. Willenegger, all used with permission.
Figure 4
Pre-hunt perching force affects hunting success during sit-and-wait hunting.
Variation in hunting success according to the pre-hunt perching force (N), depending on whether owls hunted on the wing (yellow) or using the sit-and-wait strategy (blue). Solid lines show the estimated means (averaged over both sexes), and the shaded area corresponds to the 95% confidence intervals around each mean. Blue and yellow circles show the force, recorded during the last perching event before hunting (pre-hunt perching, n=3040), pooled to the nearest integer N value for representation purposes, when hunting on the wing or using the sit-and-wait strategy, respectively. Circle size is related to the amount of data with the same value. The owl illustrations at the top right are courtesy of L. Willenegger, used with permission.
Appendix 1—figure 1
GPS tracks (in black) of the 163 breeding barn owls used in this study were recorded in 2019 and 2020 in western Switzerland.
Appendix 1—figure 2
Typical barn owl foraging ground and nest location in western Switzerland.
(A) Example of how pasture poles are usually located within the agricultural landscape which represents the main habitat for barn owls in western Switzerland. (B) Typical barn in which nest boxes are usually installed on the Swiss plateau.
Appendix 1—figure 3
Behaviour classification from accelerometer data.
Time series data of the different behaviour classified using Boolean approach showing changes in the raw tri-axial acceleration, body pitch angle, and the vectorial sum of the dynamic body acceleration (VeDBA) corresponding to (A) flight, (B) hunting strikes, (C) nest box visit, and (D) self-feeding events. Behaviour-specific base element used in the Boolean classification are shown in grey bands. Note that hunting strikes involves greater acceleration amplitude, VeDBA and body pitch variation between landings in context of usual perching (here at the end of flight sequence).
Appendix 1—figure 4
Complete variation in landing force according to pre-hunt time.
Graphic representation of (A) the complete variation of the landing force calculated during perching and (B) the corresponding first derivative events in relation to the time (hours) until the next hunting event depending on whether owls perched on poles (in grey), trees (in green) and buildings (in yellow). Each line represents the predicted means for each perch type (averaged over male individuals), and shades show the 95% confidence intervals around each mean.
Appendix 1—figure 5
Differences in distance between the last perch and strike location according to the hunting strategy.
Box plots of the variation in distance between perching location and hunting strikes among both perching and flying hunting strategies, showing a significant difference in distance between perch and strike location according to the hunting strategy (Wilcoxon test: W=2344881, p-value <0.001). Boxes boundaries highlight the first and the third quartile of the range distribution of the data. The line within each box marks the median and whiskers above and below boxes indicate the 10th and 90th percentiles. Owl drawings are courtesy of L. Willenegger, all used with permission.
Appendix 1—figure 6
Sexual dimorphism in body mass.
Box plots of the variation in body mass between females (blue dots, n=84) and males (orange dots, n=79). White dots and bars, respectively highlight the average and the standard deviation of body mass.
Appendix 1—figure 7
Sexual differences in food provisioning.
Box plots of the variation in the number of prey items delivered to the nest each night between females (blue dots, n=1226) and males (orange dots, n=3105). White dots and bars, respectively highlight the average and the standard deviation of the number of prey items delivered to the nest each night.
Appendix 1—figure 8
Sexual differences in hunting activity.
Box plots of the variation in the number of hunting strikes and perching events performed each night by females (blue dots, nb perching events = 22,134, nb hunting strikes = 8176) and males (orange dots, nb perching events = 19,657, nb hunting strikes = 15,158). White dots and bars respectively highlight the average, and the standard deviation of the number of hunting strikes and perching events performed each night.
Appendix 1—figure 9
Sexual difference in the use of hunting strategy.
Box plots of the variation in the proportion of use of the sit-and-wait strategy for females (blue dots) and males (orange dots). White dots and bars, respectively highlight the average and the standard deviation of the proportion of use of the sit-and-wait strategy.
Appendix 1—figure 10
Sexual comparison of landing force as a function of landing context.
Variation in peak landing force, on the log scale, involved in perching events and hunting strikes between female (blue dots, nb hunting strikes = 10,117, nb perching events = 30,378) and male (orange dots, nb hunting strikes = 17,864, nb perching events = 26,496) individuals. (A) Shows landing force per unit of body mass and (B) shows variation in peak landing force. White dots show the estimated mean, and data distribution is represented by both violin and box plots.
Appendix 1—figure 11
Influence of hunting strategy on barn owls foraging trip duration.
Variation in foraging trip duration (in min) as a function of the frequency of use of the sit-and-wait strategy (relative to hunting on the wing). Solid line shows the estimated mean (averaged over both sexes) and the grey shade corresponds to the 95% confidence intervals around the mean.
Tables
Appendix 1—table 1
Time series (TS) of base elements applied for the classification of flight, hunting strikes, nest return (indicating prey delivery to the nest), and self-feeding behaviours.
Each base element (BE) has a defined temporal flexibility that includes a number of events over which the conditions of the element are met (Present), as well as a range to the next element (Range), a window of flexibility (Flexibility) and a period over which the element is extended (ETNE), where the units are in events, 50 Hz having 50 events per second. See Wilson et al., 2018 for details of use of these algorithms in the Boolean approach.
| Flight | BE | Present | Range | Flexibility | ETNE |
|---|---|---|---|---|---|
| TS1 | Take-off | 25 | 26 | 50 | |
| TS2 | Flying | 5 | 6 | 60’000 | 10 |
| TS3 | Landing | 90 | 91 | 5 | |
| TS4 | Standing | 4 | |||
| Hunting strike | BE | Present | Range | Flexibility | ETNE |
| TS1 | Tilt head down | 5 | 20 | 40 | 10 |
| TS2 | Legs swinging | 3 | 5 | 110 | |
| TS3 | Impact | 1 | 1 | 160 | |
| TS4 | Standing | 5 | |||
| Nest return | BE | Present | Range | Flexibility | ETNE |
| TS1 | Nest enter | 1 | 3 | 4 | |
| TS2 | Nest in | 1 | 50 | 1500 | 5 |
| TS3 | Nest exit | 1 | |||
| Self feeding | BE | Present | Range | Flexibility | ETNE |
| TS1 | High peak | 1 | 1 | 15 | |
| TS2 | Low peak | 1 | 1 | 15 | |
| TS3 | High peak | 1 | 1 | 15 | |
| TS4 | Low peak | 1 | 1 | 15 | |
| TS5 | Low VeDBA | 10 | 10 | 150 | |
| TS6 | End | 15 |
Appendix 1—table 2
Model output for the linear mixed model (LMM) predicting variations in landing force (log-transformed) due to landing context (hunting strike vs perching) and sex (Males vs Females).
The model also included a random effect of BirdID and NightID (nested in BirdID). Intercept is reported for both landing contexts (highlighted in grey) and give information about the averaged landing force considering female individuals. Estimates for interactions give the % of change between females and males for each landing contexts. Variance (σ2), intra-class correlation coefficient (ICC), and number of observations are provided for random effects.
| Impact force (N) | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Landing context [strike] | 40.81 | 39.49–42.18 | <0.001 |
| Landing context [perching] | 9.94 | 9.63–10.27 | <0.001 |
| Landing context [strike] * sex[M] | 0.94 | 0.90–0.99 | 0.011 |
| Landing context [perching] * sex[M] | 0.80 | 0.76–0.83 | <0.001 |
| Observations | 84855 | ||
| Marginal R2 /Conditional R2 | 0.736/0.772 | ||
-
Random Effects: o2=0.15 | ICC = 0.14 | NBirdlD = 163 | NNightID = 7
Appendix 1—table 3
Model selection results using Akaike Information Criteria corrected for small sample sizes (AICc) for all possible models evaluating the effect of landing context and sex on barn owl landing force.
A ‘*’ in the variable columns indicates that the variable was included in that model. K is the number of variables included in each model. All linear mixed model (LMM) included a random effect of BirdID and NightID (nested in BirdID). Table includes all possible models, ranked by AICc. Models with Delta AICc <2 are highlighted in grey and the final model is shown in bold.
| # | Sex | Landing context | Sex: Landing context | K | AIC | ∆AIC | Model weight |
|---|---|---|---|---|---|---|---|
| 1 | * | * | * | 3 | 82517 | 0 | 1 |
| 2 | * | * | 2 | 83252.5 | 735.51 | 0 | |
| 3 | * | 1 | 83297.9 | 780.84 | 0 | ||
| 4 | * | 1 | 198495.8 | 115978.81 | 0 | ||
| 5 | 0 | 198498.4 | 115981.36 | 0 |
Appendix 1—table 4
Model output of the generalized additive mixed-effects model (GAMM) predicting variations in pre-hunt perching force in N (log-transformed) due to perch type (buildings, tree branches, and road/pasture poles), wind speed, and sex as linear predictor (lme) and time to the next strike as an additive term (gam).
The model also included a random effect of BirdID. Effective degrees of freedom (EDF) are shown for additive terms, providing the degree of non-linearity between pre-hunt perching force and time to hunt for each perch type.
| LME | Impact force (N) | ||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Intercept | 8.99 | 8.74–9.24 | <0.001 |
| Perch type [tree] | 1.07 | 1.07–1.08 | <0.001 |
| Perch type [buildings] | 1.09 | 1.08–1.09 | <0.001 |
| Wind speed | 1.00 | 0.99–1.00 | 0.04 |
| Sex [Male] | 0.85 | 0.81–0.88 | <0.001 |
| GAM | |||
| Predictors | EDF | F | p-value |
| Time to hunt: perch type [road/pasture] | 4.22 | 22.43 | <0.001 |
| Time to hunt: perch type [tree] | 1.50 | 5.12 | 0.005 |
| Time to hunt: perch type [buildings] | 1.00 | 12.86 | <0.001 |
| Observations | 27981 | ||
| R2 | 0.235 | ||
Appendix 1—table 5
Model selection results using Akaike Information Criteria corrected for small sample sizes (AICc) for all possible models evaluating the effect of time to hunt and perch type on landing force during perching events.
A ‘*’ in the variable columns indicates that the variable was included in that model. K is the number of variables included in each model. All linear mixed model (LMM) included a random effect of BirdID and NightID (nested in BirdID). Table includes the top five models, ranked by AICc. Models with Delta AICc <2 are highlighted in grey and the final model is shown in bold.
| # | Perch type | Wind speed | Sex | s (time to hunt: perch type) | K | AICc | ∆AICc | Model weight |
|---|---|---|---|---|---|---|---|---|
| 1 | * | * | * | * | 4 | 203.1 | 0 | 0.48 |
| 2 | * | * | * | 3 | 204.5 | 1.42 | 0.24 | |
| 3 | * | * | * | 3 | 205.0 | 1.92 | 0.18 | |
| 4 | * | * | 2 | 206.4 | 3.26 | 0.09 | ||
| 5 | * | * | * | 3 | 324.2 | 121.13 | 0 |
Appendix 1—table 6
Model output for the linear mixed model (LMM) predicting variations in hunting strike force (log-transformed) due to hunting success (0 vs 1), hunting strategy (perching vs flying), and sex.
The model also included a random effect of BirdID and NightID (nested in BirdID). Intercept provides the averaged strike force (N) considering female individuals hunting on the wing. Variance (σ2), intra-class correlation coefficient (ICC), and number of observations are provided for random effects.
| Impact force (N) | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Intercept | 41.51 | 40.31–42.75 | <0.001 |
| Hunting success [0] | 0.95 | 0.94–0.96 | <0.001 |
| Sex [M] | 0.94 | 0.91–0.98 | 0.005 |
| Hunting strategy [sit-and-wait] | 0.96 | 0.94–0.99 | 0.002 |
| Hunting success [0] * hunting strategy [sit-and-wait] | 1.04 | 1.01–1.07 | 0.004 |
| Observations | 27981 | ||
| Marginal R2 /Conditional R2 | 0.007/0.115 | ||
-
Random Effects: o2=0.13 | ICC = 0.11 | NBirdlD = 163 | NNightID = 7
Appendix 1—table 7
Model selection results using Akaike Information Criteria corrected for small sample sizes (AICc) for all possible models evaluating the effect of hunting success and hunting strategy on landing force during hunting strike.
A ‘*’ in the variable columns indicates that the variable was included in that model. K is the number of variables included in each model. All linear mixed model (LMM) included a random effect of BirdID and NightID (nested in BirdID). Table includes the top five models, ranked by AICc. Models with Delta AICc <2 are highlighted in grey and the final model is shown in bold.
| # | Sex | Hunting success | Hunting strategy | Hunting strategy: Hunting success | K | AIC | ∆AIC | Model weight |
|---|---|---|---|---|---|---|---|---|
| 1 | * | * | * | * | 4 | 23996.3 | 0 | 0.87 |
| 2 | * | * | 2 | 24002.1 | 5.82 | 0.05 | ||
| 3 | * | * | * | 3 | 24002.1 | 5.82 | 0.05 | |
| 4 | * | * | * | 3 | 24002.7 | 6.36 | 0.04 | |
| 5 | * | 1 | 24007.7 | 11.41 | 0.003 |
Appendix 1—table 8
Model output for the generalized linear mixed-effect model (GLMM) predicting variations in hunting success (binary response 0,1) due to sex and hunting strategy (on the wing vs sit-and-wait).
The model also included a random effect of BirdID and NightID (nested in BirdID). Intercept is reported for both sexes and give information about the averaged hunting success considering individuals hunting on the wing. Standardized estimates are provided for any additional terms in the model, representing % of change (odds ratio) of hunting success. The influence of landing force on hunting success is provided considering both hunting strategies. Variance (σ2), intra-class correlation coefficient (ICC), and number of observations of each group are provided for random effects.
| Hunting success | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Sex [Female] | 0.24 | 0.22–0.26 | <0.001 |
| Sex [Male] | 0.35 | 0.33–0.38 | <0.001 |
| Hunting strategy [sit-and-wait]: Females | 1.55 | 1.38–1.75 | <0.001 |
| Hunting strategy [sit-and-wait]: Males | 1.49 | 1.32–1.68 | <0.001 |
| Observations | 27981 | ||
| Marginal R2 /Conditional R2 | 0.014/0.051 | ||
-
Random Effects: o2=3.29 | ICC = 0.04 | NBirdlD = 163 | NNightID = 7
Appendix 1—table 9
Model selection results using Akaike Information Criteria corrected for small sample sizes (AICc) for all possible models evaluating the effect of sex, hunting strategy, and their interaction on hunting success.
A ‘*’ in the variable columns indicates that the variable was included in that model. K is the number of variables included in each model. All linear mixed model (LMM) included a random effect of BirdID and NightID (nested in BirdID). Table includes the top five models, ranked by AICc. Models with Delta AICc <2 are highlighted in grey and the final model is shown in bold.
| # | Sex | Hunting strategy | Sex:Hunting strategy | K | AIC | ∆AIC | Model weight |
|---|---|---|---|---|---|---|---|
| 1 | * | * | 2 | 30726.8 | 0 | 0.71 | |
| 2 | * | * | * | 3 | 30728.6 | 1.79 | 0.29 |
| 3 | * | 1 | 30768.0 | 41.22 | 0 | ||
| 4 | * | 1 | 30816.2 | 89.38 | 0 | ||
| 5 | 0 | 30848.5 | 121.68 | 0 |
Appendix 1—table 10
Modelling the effect of landing force during pre-hunt perching on hunting success.
Model output for the generalized linear mixed-effect model (GLMM) predicting variations in hunting success (binary response 0,1) due to pre-hunt perching force, sex, hunting strategy (on the wing vs sit-and-wait), and wind speed. The model also included a random effect of BirdID and NightID (nested in BirdID). Intercept is reported for both sexes (highlighted in grey) and give information about the average hunting success considering individuals hunting on the wing. Standardized estimates are provided for any additional terms in the model, representing % of change (odds ratio) of hunting success. The influence of landing force on hunting success is provided considering both hunting strategies. Variance (σ2), intra-class correlation coefficient (ICC), and number of observations of each group are provided for random effects.
| Hunting success | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Females | 0.22 | 0.18–0.26 | <0.001 |
| Males | 0.32 | 0.28–0.37 | <0.001 |
| Hunting strategy [sit-and-wait] | 1.51 | 1.26–1.81 | <0.001 |
| Hunting strategy [on the wing]: Pre-hunt perching force | 1.08 | 0.97–1.20 | 0.178 |
| Hunting strategy [sit-and-wait]: Pre-hunt perching force | 0.85 | 0.73–0.99 | 0.037 |
| Observations | 3040 | ||
| Marginal R2 /Conditional R2 | 0.023/0.043 | ||
-
Random Effects: o2=3.29 | ICC = 0.02 | NBirdlD = 151 | NNightID = 7
Appendix 1—table 11
Model selection results using Akaike Information Criteria corrected for small sample sizes (AICc) for all possible models evaluating the effect of landing force during pre-hunt perching on hunting success.
A ‘*’ in the variable columns indicates that the variable was included in that model. K is the number of variables included in each model. All linear mixed model (LMM) included a random effect of BirdID and NightID (nested in BirdID). Table includes the top five models, ranked by AICc. Models with Delta AICc <2 are highlighted in grey and the final model is shown in bold.
| # | Sex | Wind speed | Perch type | Hunting strategy | Pre-hunt perching force | Hunting strategy: Pre-hunt perching force | K | AIC | ∆AIC | Model weight |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | * | * | * | * | 4 | 3326.5 | 0 | 0.24 | ||
| 2 | * | * | * | * | * | 5 | 3327.1 | 0.54 | 0.19 | |
| 3 | * | * | * | * | * | 5 | 3327.4 | 0.82 | 0.16 | |
| 4 | * | * | * | * | * | * | 6 | 3328.1 | 1.55 | 0.11 |
| 5 | * | * | 2 | 3329.1 | 2.56 | 0.07 |
Appendix 1—table 12
Model output of the linear mixed model (LMM) predicting variation of barn owls foraging flights speed (in ms–1) as a function of the sex.
Intercept provides the averaged foraging flight speed (in ms–1) considering female individuals. The model also included a random effect of BirdID and NightID (nested in BirdID). Variance (σ2), intra-class correlation coefficient (ICC), and number of observations of each group are provided for random effects.
| Foraging flight speed (ms–1) | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Intercept | 5.47 | 5.38–5.56 | <0.001 |
| Sex [Male] | –0.23 | –0.36 to –0.10 | <0.001 |
| Observations | 27242 | ||
| Marginal R2 /Conditional R2 | 0.006/0.141 | ||
-
Random Effects: o2=1.60 | ICC = 0.14 | NBirdlD = 163 | NNightID = 7
Appendix 1—table 13
Model output of the linear mixed model (LMM) predicting variation of barn owls foraging trip duration (in min) as a function of the total number of hunting attempts per trip, the frequency of use of the sit-and-wait strategy (relative to hunting on the wing), and sex.
The model also included a random effect of BirdID and NightID (nested in BirdID). Variance (σ2), intra-class correlation coefficient (ICC), and number of observations of each group are provided for random effects.
| Foraging trip duration (min) | |||
|---|---|---|---|
| Predictors | Estimate | Conf. Int. | p-value |
| Intercept | 9.49 | 8.43–10.67 | <0.001 |
| Nb of hunting attempts | 1.16 | 1.16–1.17 | <0.001 |
| Freq sit-and-wait | 2.26 | 1.97–2.59 | <0.001 |
| Sex [Male] | 0.79 | 0.69–0.90 | 0.001 |
| Observations | 27242 | ||
| Marginal R2 /Conditional R2 | 0.006/0.141 | ||
-
Random Effects: o2=1 | ICC = 0.15 | NBirdlD = 150 | NNightID = 7
Additional files
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Landing force reveals new form of motion-induced sound camouflage in a wild predator
eLife 12:RP87775.
https://doi.org/10.7554/eLife.87775.3
