Bird collision in high speed railways (HSRs) is a highly topical subject in recent years due to the expansion of this infrastructure. Understanding how birds interact with railway networks could improve mitigation actions. Therefore, it is urgent to develop cost-effective systems to monitor bird behaviour in HSRs. For this purpose, we compared in-situ flight behaviour of birds across a HSR with ex-situ observations extracted from videos. During three seasons, three HD video-cameras were used to record the flight and crossing behaviour of birds in a HSR in Central Spain, where an anti-bird collision screen had been installed. In each season two HSR sections of 360m each were sampled: i) one along the anti-bird screen, and ii) one control section. The section covered by each camera was 120m approx. Simultaneously, direct observation censuses were done with binoculars (10x42), covering the same area as cameras. The time and direction of passing trains in the study area were also recorded, and used as synchronization points between videos and direct observations. An independent observer visualized 18 videos per season of 10’ (lapses hereafter). For each day of recording, three lapses were randomly selected, albeit ensuring that at least one of the lapses contained a behaviour event. For each event observed in the HSR, during field observation or extracted from videos, the following parameters were noted: time, type of behaviour (crossing, flying or resting), bird species and size of the flock. Concordance between videos and direct observations were tested using the t-Student test for paired samples, comparing the number of individuals and the number of events per lapse. Pearson’s Chi-squared tests for contingency tables were used to compare differences in the total number of individuals and events detected in function of the bird size (large, medium and small) and the flight behaviour. Bird size was used instead of bird species due to very low species detection was found in videos (42.99% of unidentified observations). Within the 540’ of visualized lapses, 170 events and 321 individuals were detected in videos, whereas 201 events and 827 individuals were counted at field-work in the same lapses. We found discordance between methods in the number of individuals and events detected per lapse. In 20.37% of lapses, bird behaviours were detected only by direct observations, whereas behaviours detected only through video recordings represented only 3.70% of lapses. Results showed that the number of individuals detected by the different procedure differs and that the detection ratios vary among sizes and types of behaviour. However, the number of events detected was similar between methods and sizes, although differences were found in function of the flight behaviour. Video recordings may be useful to monitor bird crossing events, but their efficiency is conditioned by the flight behaviour of birds and thus are less reliable method to identify species or their abundance, relevant information for conservation or mitigation actions. Therefore, video recordings might be an economical method for general monitoring but direct observations and technicians are still necessary in order to obtain reliable field data.
Spain
