Data and code from: Collision with linear infrastructures reduces survival of a long-lived migratory bird
Data files
Sep 08, 2026 version files 11.84 KB
-
Collision_with_linear_infrastructures.zip
8.45 KB
-
README.md
3.39 KB
Abstract
Human infrastructure, particularly power lines and fences, poses a growing threat to birds worldwide. Quantifying the impact of collisions across different life stages and spatio-temporal contexts is critical for developing effective monitoring and conservation strategies.
We analyzed satellite-tracking data from 203 black-necked cranes (Grus nigricollis) and combined transmitter information, field surveys, carcass examinations, and satellite imagery to determine individual fates and infer causes of mortality. We estimated annual variation in survival among different age classes using Kaplan-Meier survival models. We then used Cox proportional hazards to test age-specific mortality patterns from different causes of mortality including power line collisions, fence collisions and other causes and generalized additive models to compare seasonal and diurnal collision patterns of different causes.
Immature individuals (first-year individuals and subadults) had a significantly lower probability of annual survival (0.598) than adults (0.893), with first-year individuals particularly vulnerable. Power line and fence collisions accounted for 71% of all confirmed mortality cases. Although mortality risk declined with age class, power line and fence collisions affected cranes across all age classes.
Power line collisions exhibited a bimodal diurnal pattern with peaks around midday and sunset, but no significant monthly variation. In contrast, fence collisions were concentrated in August but showed no clear diurnal pattern. The timing of peaks in mortality aligned with crane behaviour: daytime power line collisions coincided with juvenile flight practice and evening returns to roosts, whereas fence collisions occurred when juveniles were still flightless and adults or subadults were moulting flight feathers.
Synthesis and applications. Our findings demonstrate that collisions with human infrastructure represent a major threat to cranes and likely other migratory birds. The identified non-random patterns of infrastructure-specific collision peaks were consistent with variation in flight ability and visibility across different stages of the annual cycle and times of day, improving the accuracy of predictive collision-risk models and supporting more effective mitigation efforts. Based on these findings, we recommend that fences should not be constructed in key breeding and moulting areas. For power lines, monitoring and mitigation approaches, such as the installation of line-marking devices that are visible under both low-light and bright conditions, should be prioritized and applied to key movement corridors.
Dataset DOI: 10.5061/dryad.70rxwdcfh
Description of the data and file structure
Data were collected from long‑term tracking monitoring of black-necked crane. The dataset includes individual bird capture‑mark‑resighting histories, age class information, fate and survival status, and collision‑related mortality events caused by linear infrastructures (powerlines, fences and other anthropogenic obstacles). Survival‑time variables were derived from individual monitoring records. Datasets support Kaplan‑Meier survival analysis, Cox mixed‑effects survival models, and Generalized Additive Models analyses to evaluate how linear‑infrastructure collision affects survival and seasonal‑diurnal patterns of collision risk.
Files and variables
Compressed folder: Collision_with_linear_infrastructures.zip
File: Kaplan-Meier survival analysis.csv
Description: CSV file for survival analysis and Cox proportional‑hazards analysis.
Variables
NO: Identifier for each individual.
AGE: Age class of the individual bird, pre-defined as a factor with 4 levels.
EXIT: Right-censored time in days, calculated from 1 June to the mortality event or last observation each year.
sday: Start time for survival model, unit: days; fixed as day 1 (June 1 of each year) for counting‑process survival format.
eday: End time for survival model, unit: days; identical to EXIT, used for counting‑process Cox mixed‑effects model.
event: Survival outcome indicator for the individual. 0 = censored (alive at last observation), 1 = mortality event occurred.
powerline: Indicator of mortality caused by collision with power lines. 0 = not a power line collision mortality, 1= mortality due to power line collision.
fence: Indicator of mortality caused by collision with fences. 0 = not a fence collision mortality, 1 = mortality due to fence collision.
other: Indicator of mortality from other causes. 0 = mortality not attributed to other causes, 1 = mortality from other non-collision cause.
File: XData_GAM.csv
Description: Predictor variable dataset for GAM analysis.
Variables
ID: Numeric index (exactly matches the ID field in YData_GAM.csv).
Month: Calendar month when the collision event/observation occurred (1-12).
Time: Time of day when the collision event/observation occurred, unit: hours (24-hour clock).
File: YData_GAM.csv
Description: Outcome variable dataset for GAM analysis.
Variables
ID: Numeric index (exactly matches the ID field in XData_GAM.csv).
powerline: Indicator of whether the mortality event was related to power line collision. 0 = unrelated to power line collision, 1 = power line collision.
fence: Indicator of whether the mortality event was related to fence collision. 0 = unrelated to fence collision, 1 = fence collision.
Unknown or else: Indicator of mortality from unknown causes or other non-collision causes. 0 = not unknown/other-cause mortality, 1 = mortality from unknown or other non-collision cause.
Code/software
CSV file can be viewed in any text or spreadsheet editor.
The R script contains fully reproducible code for this study, including statistical analysis, and figure‑and‑table generation.
