Recent technological developments allow for passive acoustic monitoring of Orthoptera (grasshoppers and crickets) in research and conservation across a broad range of temporal and spatial scales

Abstract

Passive acoustic monitoring (PAM) uses stationary recorders to detect wildlife in field conditions. The method has long been valuable for surveying certain species groups, especially bats. However, PAM has been limited by resource costs and availability of automatic classifiers to assist data analysis. With recent developments of inexpensive devices, such as Audiomoth, landscape-scale monitoring has become more feasible. This also opens possibilities to apply PAM to species groups that traditionally have been studied via expert-based, labour-intensive monitoring, such as transect surveys.Utilizing recordings of Orthoptera from online databases, specialists and from our own recordings, we built a machine-learning classifier to automatically identify 17 Orthoptera species, OrthopterOSS. Assessment included the comparison of PAM to traditional transects surveys. We also compared the performance of inexpensive Audiomoth with classic Batlogger recorders for surveying Orthoptera species with PAM, at eight sites, where we also tested whether adding two additional Audiomoths in 50 m distances from the initial device towards the edge of the wildflower area would increase species detections. We also assessed how the number of species detected changed over time.In total, we detected 20 Orthoptera species during the study. Our new classifier achieved a true positive rate of 86.4 % validated against independent test data. PAM outperformed traditional sweep netting transects overall, although differences were not statistically significant. There was no difference in species composition detected by Audiomoth v1.2 or Batlogger, the species composition detected by three Audiomoths compared to one Audiomoth and no difference between hedgerow and centre species communities. There was also no significant relationship between Orthoptera richness and the percentage of permanent semi-natural habitat in the nearby landscape.Relatively inexpensive equipment allows for efficient PAM of Orthoptera. Our OrthopterOSS classifier could represent a useful tool for future PAM research in northern Europe, and serve as an extendable basis for studies elsewhere. If the species predictions are verified by an expert, the classifier could assist monitoring and conservation of Orthoptera at broad temporal and spatial scales.

Keywords

• Audiomoth
• Automatic bioacoustics
• Classification
• Batlogger
• Convolutional neural network
• OrthopterOSS