Spatial proximity loggers for recording animal social networks: consequences of inter-logger variation in performance

Social network analysis has become an increasingly popular method to link individual behaviour to population level patterns (and vice versa). Technological advances of recent years, such as the development of spatial proximity loggers, have enhanced our abilities to record contact patterns between animals. However, loggers are often deployed without calibration which may lead to sampling biases and spurious results. In particular, loggers may differ in their performance (i.e., some loggers may over-sample and other loggers may under-sample social associations). However, the consequences of inter-logger variation in logging performance has not been thoroughly considered or quantified. In this study, proximity loggers made by Sirtrack Ltd. were fitted to 20 dairy cows over a 3-week period. Contact records resulting from field deployment demonstrated variability in logger performance when recording contact duration, which was highly consistent for each logger over time. Testing loggers under standardised conditions suggested that inter-logger variation observed in the field was due to a combination of intrinsic variation in devices, and environmental/behavioural effects. We demonstrate the potential consequences that inter-logger variation in logging performance can have for social network analysis; particularly how measures of connectivity can be biased by logging performance. Finally, we suggest some approaches to correct data generated by proximity loggers with imperfect performance, that should be used to improve the robustness of future analyses.     

Long-term recovery narratives following major disasters in Southeast Asia

Regional Environmental Change - Most studies of major disasters focus on the impacts of the event and the short-term responses. Some evaluate the underlying causes of vulnerability, but few...     

Why Use One Model? An Approach for Encompassing Model Uncertainty and Improving Best Practice

There has been some discussion within the UK modelling community regarding the preferred choice of dispersion model for assessing the air quality impact of elevated point sources, each of which is reported to perform better than the other in certain circumstances. Coupled to this is the growing acceptance within the modelling community to include some recognition of model uncertainty in its reporting. As computer processing speeds increase further, this paper considers the possibility of using more than one model and highlights some of the advantages in doing so.